#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS5)
#if !defined(NDEBUG) && !defined(SQLITE_DEBUG)
# define NDEBUG 1
#endif
#if defined(NDEBUG) && defined(SQLITE_DEBUG)
# undef NDEBUG
#endif
#line 1 "fts5.h"
/*
** 2014 May 31
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
** Interfaces to extend FTS5. Using the interfaces defined in this file,
** FTS5 may be extended with:
**
** * custom tokenizers, and
** * custom auxiliary functions.
*/
#ifndef _FTS5_H
#define _FTS5_H
#include "sqlite3.h"
#ifdef __cplusplus
extern "C" {
#endif
/*************************************************************************
** CUSTOM AUXILIARY FUNCTIONS
**
** Virtual table implementations may overload SQL functions by implementing
** the sqlite3_module.xFindFunction() method.
*/
typedef struct Fts5ExtensionApi Fts5ExtensionApi;
typedef struct Fts5Context Fts5Context;
typedef struct Fts5PhraseIter Fts5PhraseIter;
typedef void (*fts5_extension_function)(
const Fts5ExtensionApi *pApi, /* API offered by current FTS version */
Fts5Context *pFts, /* First arg to pass to pApi functions */
sqlite3_context *pCtx, /* Context for returning result/error */
int nVal, /* Number of values in apVal[] array */
sqlite3_value **apVal /* Array of trailing arguments */
);
struct Fts5PhraseIter {
const unsigned char *a;
const unsigned char *b;
};
/*
** EXTENSION API FUNCTIONS
**
** xUserData(pFts):
** Return a copy of the context pointer the extension function was
** registered with.
**
** xColumnTotalSize(pFts, iCol, pnToken):
** If parameter iCol is less than zero, set output variable *pnToken
** to the total number of tokens in the FTS5 table. Or, if iCol is
** non-negative but less than the number of columns in the table, return
** the total number of tokens in column iCol, considering all rows in
** the FTS5 table.
**
** If parameter iCol is greater than or equal to the number of columns
** in the table, SQLITE_RANGE is returned. Or, if an error occurs (e.g.
** an OOM condition or IO error), an appropriate SQLite error code is
** returned.
**
** xColumnCount(pFts):
** Return the number of columns in the table.
**
** xColumnSize(pFts, iCol, pnToken):
** If parameter iCol is less than zero, set output variable *pnToken
** to the total number of tokens in the current row. Or, if iCol is
** non-negative but less than the number of columns in the table, set
** *pnToken to the number of tokens in column iCol of the current row.
**
** If parameter iCol is greater than or equal to the number of columns
** in the table, SQLITE_RANGE is returned. Or, if an error occurs (e.g.
** an OOM condition or IO error), an appropriate SQLite error code is
** returned.
**
** This function may be quite inefficient if used with an FTS5 table
** created with the "columnsize=0" option.
**
** xColumnText:
** If parameter iCol is less than zero, or greater than or equal to the
** number of columns in the table, SQLITE_RANGE is returned.
**
** Otherwise, this function attempts to retrieve the text of column iCol of
** the current document. If successful, (*pz) is set to point to a buffer
** containing the text in utf-8 encoding, (*pn) is set to the size in bytes
** (not characters) of the buffer and SQLITE_OK is returned. Otherwise,
** if an error occurs, an SQLite error code is returned and the final values
** of (*pz) and (*pn) are undefined.
**
** xPhraseCount:
** Returns the number of phrases in the current query expression.
**
** xPhraseSize:
** If parameter iCol is less than zero, or greater than or equal to the
** number of phrases in the current query, as returned by xPhraseCount,
** 0 is returned. Otherwise, this function returns the number of tokens in
** phrase iPhrase of the query. Phrases are numbered starting from zero.
**
** xInstCount:
** Set *pnInst to the total number of occurrences of all phrases within
** the query within the current row. Return SQLITE_OK if successful, or
** an error code (i.e. SQLITE_NOMEM) if an error occurs.
**
** This API can be quite slow if used with an FTS5 table created with the
** "detail=none" or "detail=column" option. If the FTS5 table is created
** with either "detail=none" or "detail=column" and "content=" option
** (i.e. if it is a contentless table), then this API always returns 0.
**
** xInst:
** Query for the details of phrase match iIdx within the current row.
** Phrase matches are numbered starting from zero, so the iIdx argument
** should be greater than or equal to zero and smaller than the value
** output by xInstCount(). If iIdx is less than zero or greater than
** or equal to the value returned by xInstCount(), SQLITE_RANGE is returned.
**
** Otherwise, output parameter *piPhrase is set to the phrase number, *piCol
** to the column in which it occurs and *piOff the token offset of the
** first token of the phrase. SQLITE_OK is returned if successful, or an
** error code (i.e. SQLITE_NOMEM) if an error occurs.
**
** This API can be quite slow if used with an FTS5 table created with the
** "detail=none" or "detail=column" option.
**
** xRowid:
** Returns the rowid of the current row.
**
** xTokenize:
** Tokenize text using the tokenizer belonging to the FTS5 table.
**
** xQueryPhrase(pFts5, iPhrase, pUserData, xCallback):
** This API function is used to query the FTS table for phrase iPhrase
** of the current query. Specifically, a query equivalent to:
**
** ... FROM ftstable WHERE ftstable MATCH $p ORDER BY rowid
**
** with $p set to a phrase equivalent to the phrase iPhrase of the
** current query is executed. Any column filter that applies to
** phrase iPhrase of the current query is included in $p. For each
** row visited, the callback function passed as the fourth argument
** is invoked. The context and API objects passed to the callback
** function may be used to access the properties of each matched row.
** Invoking Api.xUserData() returns a copy of the pointer passed as
** the third argument to pUserData.
**
** If parameter iPhrase is less than zero, or greater than or equal to
** the number of phrases in the query, as returned by xPhraseCount(),
** this function returns SQLITE_RANGE.
**
** If the callback function returns any value other than SQLITE_OK, the
** query is abandoned and the xQueryPhrase function returns immediately.
** If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK.
** Otherwise, the error code is propagated upwards.
**
** If the query runs to completion without incident, SQLITE_OK is returned.
** Or, if some error occurs before the query completes or is aborted by
** the callback, an SQLite error code is returned.
**
**
** xSetAuxdata(pFts5, pAux, xDelete)
**
** Save the pointer passed as the second argument as the extension function's
** "auxiliary data". The pointer may then be retrieved by the current or any
** future invocation of the same fts5 extension function made as part of
** the same MATCH query using the xGetAuxdata() API.
**
** Each extension function is allocated a single auxiliary data slot for
** each FTS query (MATCH expression). If the extension function is invoked
** more than once for a single FTS query, then all invocations share a
** single auxiliary data context.
**
** If there is already an auxiliary data pointer when this function is
** invoked, then it is replaced by the new pointer. If an xDelete callback
** was specified along with the original pointer, it is invoked at this
** point.
**
** The xDelete callback, if one is specified, is also invoked on the
** auxiliary data pointer after the FTS5 query has finished.
**
** If an error (e.g. an OOM condition) occurs within this function,
** the auxiliary data is set to NULL and an error code returned. If the
** xDelete parameter was not NULL, it is invoked on the auxiliary data
** pointer before returning.
**
**
** xGetAuxdata(pFts5, bClear)
**
** Returns the current auxiliary data pointer for the fts5 extension
** function. See the xSetAuxdata() method for details.
**
** If the bClear argument is non-zero, then the auxiliary data is cleared
** (set to NULL) before this function returns. In this case the xDelete,
** if any, is not invoked.
**
**
** xRowCount(pFts5, pnRow)
**
** This function is used to retrieve the total number of rows in the table.
** In other words, the same value that would be returned by:
**
** SELECT count(*) FROM ftstable;
**
** xPhraseFirst()
** This function is used, along with type Fts5PhraseIter and the xPhraseNext
** method, to iterate through all instances of a single query phrase within
** the current row. This is the same information as is accessible via the
** xInstCount/xInst APIs. While the xInstCount/xInst APIs are more convenient
** to use, this API may be faster under some circumstances. To iterate
** through instances of phrase iPhrase, use the following code:
**
** Fts5PhraseIter iter;
** int iCol, iOff;
** for(pApi->xPhraseFirst(pFts, iPhrase, &iter, &iCol, &iOff);
** iCol>=0;
** pApi->xPhraseNext(pFts, &iter, &iCol, &iOff)
** ){
** // An instance of phrase iPhrase at offset iOff of column iCol
** }
**
** The Fts5PhraseIter structure is defined above. Applications should not
** modify this structure directly - it should only be used as shown above
** with the xPhraseFirst() and xPhraseNext() API methods (and by
** xPhraseFirstColumn() and xPhraseNextColumn() as illustrated below).
**
** This API can be quite slow if used with an FTS5 table created with the
** "detail=none" or "detail=column" option. If the FTS5 table is created
** with either "detail=none" or "detail=column" and "content=" option
** (i.e. if it is a contentless table), then this API always iterates
** through an empty set (all calls to xPhraseFirst() set iCol to -1).
**
** xPhraseNext()
** See xPhraseFirst above.
**
** xPhraseFirstColumn()
** This function and xPhraseNextColumn() are similar to the xPhraseFirst()
** and xPhraseNext() APIs described above. The difference is that instead
** of iterating through all instances of a phrase in the current row, these
** APIs are used to iterate through the set of columns in the current row
** that contain one or more instances of a specified phrase. For example:
**
** Fts5PhraseIter iter;
** int iCol;
** for(pApi->xPhraseFirstColumn(pFts, iPhrase, &iter, &iCol);
** iCol>=0;
** pApi->xPhraseNextColumn(pFts, &iter, &iCol)
** ){
** // Column iCol contains at least one instance of phrase iPhrase
** }
**
** This API can be quite slow if used with an FTS5 table created with the
** "detail=none" option. If the FTS5 table is created with either
** "detail=none" "content=" option (i.e. if it is a contentless table),
** then this API always iterates through an empty set (all calls to
** xPhraseFirstColumn() set iCol to -1).
**
** The information accessed using this API and its companion
** xPhraseFirstColumn() may also be obtained using xPhraseFirst/xPhraseNext
** (or xInst/xInstCount). The chief advantage of this API is that it is
** significantly more efficient than those alternatives when used with
** "detail=column" tables.
**
** xPhraseNextColumn()
** See xPhraseFirstColumn above.
**
** xQueryToken(pFts5, iPhrase, iToken, ppToken, pnToken)
** This is used to access token iToken of phrase iPhrase of the current
** query. Before returning, output parameter *ppToken is set to point
** to a buffer containing the requested token, and *pnToken to the
** size of this buffer in bytes.
**
** If iPhrase or iToken are less than zero, or if iPhrase is greater than
** or equal to the number of phrases in the query as reported by
** xPhraseCount(), or if iToken is equal to or greater than the number of
** tokens in the phrase, SQLITE_RANGE is returned and *ppToken and *pnToken
are both zeroed.
**
** The output text is not a copy of the query text that specified the
** token. It is the output of the tokenizer module. For tokendata=1
** tables, this includes any embedded 0x00 and trailing data.
**
** xInstToken(pFts5, iIdx, iToken, ppToken, pnToken)
** This is used to access token iToken of phrase hit iIdx within the
** current row. If iIdx is less than zero or greater than or equal to the
** value returned by xInstCount(), SQLITE_RANGE is returned. Otherwise,
** output variable (*ppToken) is set to point to a buffer containing the
** matching document token, and (*pnToken) to the size of that buffer in
** bytes. This API is not available if the specified token matches a
** prefix query term. In that case both output variables are always set
** to 0.
**
** The output text is not a copy of the document text that was tokenized.
** It is the output of the tokenizer module. For tokendata=1 tables, this
** includes any embedded 0x00 and trailing data.
**
** This API can be quite slow if used with an FTS5 table created with the
** "detail=none" or "detail=column" option.
*/
struct Fts5ExtensionApi {
int iVersion; /* Currently always set to 3 */
void *(*xUserData)(Fts5Context*);
int (*xColumnCount)(Fts5Context*);
int (*xRowCount)(Fts5Context*, sqlite3_int64 *pnRow);
int (*xColumnTotalSize)(Fts5Context*, int iCol, sqlite3_int64 *pnToken);
int (*xTokenize)(Fts5Context*,
const char *pText, int nText, /* Text to tokenize */
void *pCtx, /* Context passed to xToken() */
int (*xToken)(void*, int, const char*, int, int, int) /* Callback */
);
int (*xPhraseCount)(Fts5Context*);
int (*xPhraseSize)(Fts5Context*, int iPhrase);
int (*xInstCount)(Fts5Context*, int *pnInst);
int (*xInst)(Fts5Context*, int iIdx, int *piPhrase, int *piCol, int *piOff);
sqlite3_int64 (*xRowid)(Fts5Context*);
int (*xColumnText)(Fts5Context*, int iCol, const char **pz, int *pn);
int (*xColumnSize)(Fts5Context*, int iCol, int *pnToken);
int (*xQueryPhrase)(Fts5Context*, int iPhrase, void *pUserData,
int(*)(const Fts5ExtensionApi*,Fts5Context*,void*)
);
int (*xSetAuxdata)(Fts5Context*, void *pAux, void(*xDelete)(void*));
void *(*xGetAuxdata)(Fts5Context*, int bClear);
int (*xPhraseFirst)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*, int*);
void (*xPhraseNext)(Fts5Context*, Fts5PhraseIter*, int *piCol, int *piOff);
int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*);
void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol);
/* Below this point are iVersion>=3 only */
int (*xQueryToken)(Fts5Context*,
int iPhrase, int iToken,
const char **ppToken, int *pnToken
);
int (*xInstToken)(Fts5Context*, int iIdx, int iToken, const char**, int*);
};
/*
** CUSTOM AUXILIARY FUNCTIONS
*************************************************************************/
/*************************************************************************
** CUSTOM TOKENIZERS
**
** Applications may also register custom tokenizer types. A tokenizer
** is registered by providing fts5 with a populated instance of the
** following structure. All structure methods must be defined, setting
** any member of the fts5_tokenizer struct to NULL leads to undefined
** behaviour. The structure methods are expected to function as follows:
**
** xCreate:
** This function is used to allocate and initialize a tokenizer instance.
** A tokenizer instance is required to actually tokenize text.
**
** The first argument passed to this function is a copy of the (void*)
** pointer provided by the application when the fts5_tokenizer object
** was registered with FTS5 (the third argument to xCreateTokenizer()).
** The second and third arguments are an array of nul-terminated strings
** containing the tokenizer arguments, if any, specified following the
** tokenizer name as part of the CREATE VIRTUAL TABLE statement used
** to create the FTS5 table.
**
** The final argument is an output variable. If successful, (*ppOut)
** should be set to point to the new tokenizer handle and SQLITE_OK
** returned. If an error occurs, some value other than SQLITE_OK should
** be returned. In this case, fts5 assumes that the final value of *ppOut
** is undefined.
**
** xDelete:
** This function is invoked to delete a tokenizer handle previously
** allocated using xCreate(). Fts5 guarantees that this function will
** be invoked exactly once for each successful call to xCreate().
**
** xTokenize:
** This function is expected to tokenize the nText byte string indicated
** by argument pText. pText may or may not be nul-terminated. The first
** argument passed to this function is a pointer to an Fts5Tokenizer object
** returned by an earlier call to xCreate().
**
** The second argument indicates the reason that FTS5 is requesting
** tokenization of the supplied text. This is always one of the following
** four values:
**
**
- FTS5_TOKENIZE_DOCUMENT - A document is being inserted into
** or removed from the FTS table. The tokenizer is being invoked to
** determine the set of tokens to add to (or delete from) the
** FTS index.
**
**
- FTS5_TOKENIZE_QUERY - A MATCH query is being executed
** against the FTS index. The tokenizer is being called to tokenize
** a bareword or quoted string specified as part of the query.
**
**
- (FTS5_TOKENIZE_QUERY | FTS5_TOKENIZE_PREFIX) - Same as
** FTS5_TOKENIZE_QUERY, except that the bareword or quoted string is
** followed by a "*" character, indicating that the last token
** returned by the tokenizer will be treated as a token prefix.
**
**
- FTS5_TOKENIZE_AUX - The tokenizer is being invoked to
** satisfy an fts5_api.xTokenize() request made by an auxiliary
** function. Or an fts5_api.xColumnSize() request made by the same
** on a columnsize=0 database.
**
**
** For each token in the input string, the supplied callback xToken() must
** be invoked. The first argument to it should be a copy of the pointer
** passed as the second argument to xTokenize(). The third and fourth
** arguments are a pointer to a buffer containing the token text, and the
** size of the token in bytes. The 4th and 5th arguments are the byte offsets
** of the first byte of and first byte immediately following the text from
** which the token is derived within the input.
**
** The second argument passed to the xToken() callback ("tflags") should
** normally be set to 0. The exception is if the tokenizer supports
** synonyms. In this case see the discussion below for details.
**
** FTS5 assumes the xToken() callback is invoked for each token in the
** order that they occur within the input text.
**
** If an xToken() callback returns any value other than SQLITE_OK, then
** the tokenization should be abandoned and the xTokenize() method should
** immediately return a copy of the xToken() return value. Or, if the
** input buffer is exhausted, xTokenize() should return SQLITE_OK. Finally,
** if an error occurs with the xTokenize() implementation itself, it
** may abandon the tokenization and return any error code other than
** SQLITE_OK or SQLITE_DONE.
**
** SYNONYM SUPPORT
**
** Custom tokenizers may also support synonyms. Consider a case in which a
** user wishes to query for a phrase such as "first place". Using the
** built-in tokenizers, the FTS5 query 'first + place' will match instances
** of "first place" within the document set, but not alternative forms
** such as "1st place". In some applications, it would be better to match
** all instances of "first place" or "1st place" regardless of which form
** the user specified in the MATCH query text.
**
** There are several ways to approach this in FTS5:
**
** - By mapping all synonyms to a single token. In this case, using
** the above example, this means that the tokenizer returns the
** same token for inputs "first" and "1st". Say that token is in
** fact "first", so that when the user inserts the document "I won
** 1st place" entries are added to the index for tokens "i", "won",
** "first" and "place". If the user then queries for '1st + place',
** the tokenizer substitutes "first" for "1st" and the query works
** as expected.
**
**
- By querying the index for all synonyms of each query term
** separately. In this case, when tokenizing query text, the
** tokenizer may provide multiple synonyms for a single term
** within the document. FTS5 then queries the index for each
** synonym individually. For example, faced with the query:
**
**
** ... MATCH 'first place'
**
** the tokenizer offers both "1st" and "first" as synonyms for the
** first token in the MATCH query and FTS5 effectively runs a query
** similar to:
**
**
** ... MATCH '(first OR 1st) place'
**
** except that, for the purposes of auxiliary functions, the query
** still appears to contain just two phrases - "(first OR 1st)"
** being treated as a single phrase.
**
**
- By adding multiple synonyms for a single term to the FTS index.
** Using this method, when tokenizing document text, the tokenizer
** provides multiple synonyms for each token. So that when a
** document such as "I won first place" is tokenized, entries are
** added to the FTS index for "i", "won", "first", "1st" and
** "place".
**
** This way, even if the tokenizer does not provide synonyms
** when tokenizing query text (it should not - to do so would be
** inefficient), it doesn't matter if the user queries for
** 'first + place' or '1st + place', as there are entries in the
** FTS index corresponding to both forms of the first token.
**
**
** Whether it is parsing document or query text, any call to xToken that
** specifies a tflags argument with the FTS5_TOKEN_COLOCATED bit
** is considered to supply a synonym for the previous token. For example,
** when parsing the document "I won first place", a tokenizer that supports
** synonyms would call xToken() 5 times, as follows:
**
**
** xToken(pCtx, 0, "i", 1, 0, 1);
** xToken(pCtx, 0, "won", 3, 2, 5);
** xToken(pCtx, 0, "first", 5, 6, 11);
** xToken(pCtx, FTS5_TOKEN_COLOCATED, "1st", 3, 6, 11);
** xToken(pCtx, 0, "place", 5, 12, 17);
**
**
** It is an error to specify the FTS5_TOKEN_COLOCATED flag the first time
** xToken() is called. Multiple synonyms may be specified for a single token
** by making multiple calls to xToken(FTS5_TOKEN_COLOCATED) in sequence.
** There is no limit to the number of synonyms that may be provided for a
** single token.
**
** In many cases, method (1) above is the best approach. It does not add
** extra data to the FTS index or require FTS5 to query for multiple terms,
** so it is efficient in terms of disk space and query speed. However, it
** does not support prefix queries very well. If, as suggested above, the
** token "first" is substituted for "1st" by the tokenizer, then the query:
**
**
** ... MATCH '1s*'
**
** will not match documents that contain the token "1st" (as the tokenizer
** will probably not map "1s" to any prefix of "first").
**
** For full prefix support, method (3) may be preferred. In this case,
** because the index contains entries for both "first" and "1st", prefix
** queries such as 'fi*' or '1s*' will match correctly. However, because
** extra entries are added to the FTS index, this method uses more space
** within the database.
**
** Method (2) offers a midpoint between (1) and (3). Using this method,
** a query such as '1s*' will match documents that contain the literal
** token "1st", but not "first" (assuming the tokenizer is not able to
** provide synonyms for prefixes). However, a non-prefix query like '1st'
** will match against "1st" and "first". This method does not require
** extra disk space, as no extra entries are added to the FTS index.
** On the other hand, it may require more CPU cycles to run MATCH queries,
** as separate queries of the FTS index are required for each synonym.
**
** When using methods (2) or (3), it is important that the tokenizer only
** provide synonyms when tokenizing document text (method (3)) or query
** text (method (2)), not both. Doing so will not cause any errors, but is
** inefficient.
*/
typedef struct Fts5Tokenizer Fts5Tokenizer;
typedef struct fts5_tokenizer fts5_tokenizer;
struct fts5_tokenizer {
int (*xCreate)(void*, const char **azArg, int nArg, Fts5Tokenizer **ppOut);
void (*xDelete)(Fts5Tokenizer*);
int (*xTokenize)(Fts5Tokenizer*,
void *pCtx,
int flags, /* Mask of FTS5_TOKENIZE_* flags */
const char *pText, int nText,
int (*xToken)(
void *pCtx, /* Copy of 2nd argument to xTokenize() */
int tflags, /* Mask of FTS5_TOKEN_* flags */
const char *pToken, /* Pointer to buffer containing token */
int nToken, /* Size of token in bytes */
int iStart, /* Byte offset of token within input text */
int iEnd /* Byte offset of end of token within input text */
)
);
};
/* Flags that may be passed as the third argument to xTokenize() */
#define FTS5_TOKENIZE_QUERY 0x0001
#define FTS5_TOKENIZE_PREFIX 0x0002
#define FTS5_TOKENIZE_DOCUMENT 0x0004
#define FTS5_TOKENIZE_AUX 0x0008
/* Flags that may be passed by the tokenizer implementation back to FTS5
** as the third argument to the supplied xToken callback. */
#define FTS5_TOKEN_COLOCATED 0x0001 /* Same position as prev. token */
/*
** END OF CUSTOM TOKENIZERS
*************************************************************************/
/*************************************************************************
** FTS5 EXTENSION REGISTRATION API
*/
typedef struct fts5_api fts5_api;
struct fts5_api {
int iVersion; /* Currently always set to 2 */
/* Create a new tokenizer */
int (*xCreateTokenizer)(
fts5_api *pApi,
const char *zName,
void *pUserData,
fts5_tokenizer *pTokenizer,
void (*xDestroy)(void*)
);
/* Find an existing tokenizer */
int (*xFindTokenizer)(
fts5_api *pApi,
const char *zName,
void **ppUserData,
fts5_tokenizer *pTokenizer
);
/* Create a new auxiliary function */
int (*xCreateFunction)(
fts5_api *pApi,
const char *zName,
void *pUserData,
fts5_extension_function xFunction,
void (*xDestroy)(void*)
);
};
/*
** END OF REGISTRATION API
*************************************************************************/
#ifdef __cplusplus
} /* end of the 'extern "C"' block */
#endif
#endif /* _FTS5_H */
#line 1 "fts5Int.h"
/*
** 2014 May 31
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
*/
#ifndef _FTS5INT_H
#define _FTS5INT_H
/* #include "fts5.h" */
#include "sqlite3ext.h"
SQLITE_EXTENSION_INIT1
#include
#include
#ifndef SQLITE_AMALGAMATION
typedef unsigned char u8;
typedef unsigned int u32;
typedef unsigned short u16;
typedef short i16;
typedef sqlite3_int64 i64;
typedef sqlite3_uint64 u64;
#ifndef ArraySize
# define ArraySize(x) ((int)(sizeof(x) / sizeof(x[0])))
#endif
#define testcase(x)
#if defined(SQLITE_COVERAGE_TEST) || defined(SQLITE_MUTATION_TEST)
# define SQLITE_OMIT_AUXILIARY_SAFETY_CHECKS 1
#endif
#if defined(SQLITE_OMIT_AUXILIARY_SAFETY_CHECKS)
# define ALWAYS(X) (1)
# define NEVER(X) (0)
#elif !defined(NDEBUG)
# define ALWAYS(X) ((X)?1:(assert(0),0))
# define NEVER(X) ((X)?(assert(0),1):0)
#else
# define ALWAYS(X) (X)
# define NEVER(X) (X)
#endif
#define MIN(x,y) (((x) < (y)) ? (x) : (y))
#define MAX(x,y) (((x) > (y)) ? (x) : (y))
/*
** Constants for the largest and smallest possible 64-bit signed integers.
*/
# define LARGEST_INT64 (0xffffffff|(((i64)0x7fffffff)<<32))
# define SMALLEST_INT64 (((i64)-1) - LARGEST_INT64)
#endif
/* Truncate very long tokens to this many bytes. Hard limit is
** (65536-1-1-4-9)==65521 bytes. The limiting factor is the 16-bit offset
** field that occurs at the start of each leaf page (see fts5_index.c). */
#define FTS5_MAX_TOKEN_SIZE 32768
/*
** Maximum number of prefix indexes on single FTS5 table. This must be
** less than 32. If it is set to anything large than that, an #error
** directive in fts5_index.c will cause the build to fail.
*/
#define FTS5_MAX_PREFIX_INDEXES 31
/*
** Maximum segments permitted in a single index
*/
#define FTS5_MAX_SEGMENT 2000
#define FTS5_DEFAULT_NEARDIST 10
#define FTS5_DEFAULT_RANK "bm25"
/* Name of rank and rowid columns */
#define FTS5_RANK_NAME "rank"
#define FTS5_ROWID_NAME "rowid"
#ifdef SQLITE_DEBUG
# define FTS5_CORRUPT sqlite3Fts5Corrupt()
static int sqlite3Fts5Corrupt(void);
#else
# define FTS5_CORRUPT SQLITE_CORRUPT_VTAB
#endif
/*
** The assert_nc() macro is similar to the assert() macro, except that it
** is used for assert() conditions that are true only if it can be
** guranteed that the database is not corrupt.
*/
#ifdef SQLITE_DEBUG
extern int sqlite3_fts5_may_be_corrupt;
# define assert_nc(x) assert(sqlite3_fts5_may_be_corrupt || (x))
#else
# define assert_nc(x) assert(x)
#endif
/*
** A version of memcmp() that does not cause asan errors if one of the pointer
** parameters is NULL and the number of bytes to compare is zero.
*/
#define fts5Memcmp(s1, s2, n) ((n)<=0 ? 0 : memcmp((s1), (s2), (n)))
/* Mark a function parameter as unused, to suppress nuisance compiler
** warnings. */
#ifndef UNUSED_PARAM
# define UNUSED_PARAM(X) (void)(X)
#endif
#ifndef UNUSED_PARAM2
# define UNUSED_PARAM2(X, Y) (void)(X), (void)(Y)
#endif
typedef struct Fts5Global Fts5Global;
typedef struct Fts5Colset Fts5Colset;
/* If a NEAR() clump or phrase may only match a specific set of columns,
** then an object of the following type is used to record the set of columns.
** Each entry in the aiCol[] array is a column that may be matched.
**
** This object is used by fts5_expr.c and fts5_index.c.
*/
struct Fts5Colset {
int nCol;
int aiCol[1];
};
/**************************************************************************
** Interface to code in fts5_config.c. fts5_config.c contains contains code
** to parse the arguments passed to the CREATE VIRTUAL TABLE statement.
*/
typedef struct Fts5Config Fts5Config;
/*
** An instance of the following structure encodes all information that can
** be gleaned from the CREATE VIRTUAL TABLE statement.
**
** And all information loaded from the %_config table.
**
** nAutomerge:
** The minimum number of segments that an auto-merge operation should
** attempt to merge together. A value of 1 sets the object to use the
** compile time default. Zero disables auto-merge altogether.
**
** bContentlessDelete:
** True if the contentless_delete option was present in the CREATE
** VIRTUAL TABLE statement.
**
** zContent:
**
** zContentRowid:
** The value of the content_rowid= option, if one was specified. Or
** the string "rowid" otherwise. This text is not quoted - if it is
** used as part of an SQL statement it needs to be quoted appropriately.
**
** zContentExprlist:
**
** pzErrmsg:
** This exists in order to allow the fts5_index.c module to return a
** decent error message if it encounters a file-format version it does
** not understand.
**
** bColumnsize:
** True if the %_docsize table is created.
**
** bPrefixIndex:
** This is only used for debugging. If set to false, any prefix indexes
** are ignored. This value is configured using:
**
** INSERT INTO tbl(tbl, rank) VALUES('prefix-index', $bPrefixIndex);
**
*/
struct Fts5Config {
sqlite3 *db; /* Database handle */
char *zDb; /* Database holding FTS index (e.g. "main") */
char *zName; /* Name of FTS index */
int nCol; /* Number of columns */
char **azCol; /* Column names */
u8 *abUnindexed; /* True for unindexed columns */
int nPrefix; /* Number of prefix indexes */
int *aPrefix; /* Sizes in bytes of nPrefix prefix indexes */
int eContent; /* An FTS5_CONTENT value */
int bContentlessDelete; /* "contentless_delete=" option (dflt==0) */
char *zContent; /* content table */
char *zContentRowid; /* "content_rowid=" option value */
int bColumnsize; /* "columnsize=" option value (dflt==1) */
int bTokendata; /* "tokendata=" option value (dflt==0) */
int eDetail; /* FTS5_DETAIL_XXX value */
char *zContentExprlist;
Fts5Tokenizer *pTok;
fts5_tokenizer *pTokApi;
int bLock; /* True when table is preparing statement */
int ePattern; /* FTS_PATTERN_XXX constant */
/* Values loaded from the %_config table */
int iVersion; /* fts5 file format 'version' */
int iCookie; /* Incremented when %_config is modified */
int pgsz; /* Approximate page size used in %_data */
int nAutomerge; /* 'automerge' setting */
int nCrisisMerge; /* Maximum allowed segments per level */
int nUsermerge; /* 'usermerge' setting */
int nHashSize; /* Bytes of memory for in-memory hash */
char *zRank; /* Name of rank function */
char *zRankArgs; /* Arguments to rank function */
int bSecureDelete; /* 'secure-delete' */
int nDeleteMerge; /* 'deletemerge' */
/* If non-NULL, points to sqlite3_vtab.base.zErrmsg. Often NULL. */
char **pzErrmsg;
#ifdef SQLITE_DEBUG
int bPrefixIndex; /* True to use prefix-indexes */
#endif
};
/* Current expected value of %_config table 'version' field. And
** the expected version if the 'secure-delete' option has ever been
** set on the table. */
#define FTS5_CURRENT_VERSION 4
#define FTS5_CURRENT_VERSION_SECUREDELETE 5
#define FTS5_CONTENT_NORMAL 0
#define FTS5_CONTENT_NONE 1
#define FTS5_CONTENT_EXTERNAL 2
#define FTS5_DETAIL_FULL 0
#define FTS5_DETAIL_NONE 1
#define FTS5_DETAIL_COLUMNS 2
#define FTS5_PATTERN_NONE 0
#define FTS5_PATTERN_LIKE 65 /* matches SQLITE_INDEX_CONSTRAINT_LIKE */
#define FTS5_PATTERN_GLOB 66 /* matches SQLITE_INDEX_CONSTRAINT_GLOB */
static int sqlite3Fts5ConfigParse(
Fts5Global*, sqlite3*, int, const char **, Fts5Config**, char**
);
static void sqlite3Fts5ConfigFree(Fts5Config*);
static int sqlite3Fts5ConfigDeclareVtab(Fts5Config *pConfig);
static int sqlite3Fts5Tokenize(
Fts5Config *pConfig, /* FTS5 Configuration object */
int flags, /* FTS5_TOKENIZE_* flags */
const char *pText, int nText, /* Text to tokenize */
void *pCtx, /* Context passed to xToken() */
int (*xToken)(void*, int, const char*, int, int, int) /* Callback */
);
static void sqlite3Fts5Dequote(char *z);
/* Load the contents of the %_config table */
static int sqlite3Fts5ConfigLoad(Fts5Config*, int);
/* Set the value of a single config attribute */
static int sqlite3Fts5ConfigSetValue(Fts5Config*, const char*, sqlite3_value*, int*);
static int sqlite3Fts5ConfigParseRank(const char*, char**, char**);
/*
** End of interface to code in fts5_config.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_buffer.c.
*/
/*
** Buffer object for the incremental building of string data.
*/
typedef struct Fts5Buffer Fts5Buffer;
struct Fts5Buffer {
u8 *p;
int n;
int nSpace;
};
static int sqlite3Fts5BufferSize(int*, Fts5Buffer*, u32);
static void sqlite3Fts5BufferAppendVarint(int*, Fts5Buffer*, i64);
static void sqlite3Fts5BufferAppendBlob(int*, Fts5Buffer*, u32, const u8*);
static void sqlite3Fts5BufferAppendString(int *, Fts5Buffer*, const char*);
static void sqlite3Fts5BufferFree(Fts5Buffer*);
static void sqlite3Fts5BufferZero(Fts5Buffer*);
static void sqlite3Fts5BufferSet(int*, Fts5Buffer*, int, const u8*);
static void sqlite3Fts5BufferAppendPrintf(int *, Fts5Buffer*, char *zFmt, ...);
static char *sqlite3Fts5Mprintf(int *pRc, const char *zFmt, ...);
#define fts5BufferZero(x) sqlite3Fts5BufferZero(x)
#define fts5BufferAppendVarint(a,b,c) sqlite3Fts5BufferAppendVarint(a,b,(i64)c)
#define fts5BufferFree(a) sqlite3Fts5BufferFree(a)
#define fts5BufferAppendBlob(a,b,c,d) sqlite3Fts5BufferAppendBlob(a,b,c,d)
#define fts5BufferSet(a,b,c,d) sqlite3Fts5BufferSet(a,b,c,d)
#define fts5BufferGrow(pRc,pBuf,nn) ( \
(u32)((pBuf)->n) + (u32)(nn) <= (u32)((pBuf)->nSpace) ? 0 : \
sqlite3Fts5BufferSize((pRc),(pBuf),(nn)+(pBuf)->n) \
)
/* Write and decode big-endian 32-bit integer values */
static void sqlite3Fts5Put32(u8*, int);
static int sqlite3Fts5Get32(const u8*);
#define FTS5_POS2COLUMN(iPos) (int)(iPos >> 32)
#define FTS5_POS2OFFSET(iPos) (int)(iPos & 0x7FFFFFFF)
typedef struct Fts5PoslistReader Fts5PoslistReader;
struct Fts5PoslistReader {
/* Variables used only by sqlite3Fts5PoslistIterXXX() functions. */
const u8 *a; /* Position list to iterate through */
int n; /* Size of buffer at a[] in bytes */
int i; /* Current offset in a[] */
u8 bFlag; /* For client use (any custom purpose) */
/* Output variables */
u8 bEof; /* Set to true at EOF */
i64 iPos; /* (iCol<<32) + iPos */
};
static int sqlite3Fts5PoslistReaderInit(
const u8 *a, int n, /* Poslist buffer to iterate through */
Fts5PoslistReader *pIter /* Iterator object to initialize */
);
static int sqlite3Fts5PoslistReaderNext(Fts5PoslistReader*);
typedef struct Fts5PoslistWriter Fts5PoslistWriter;
struct Fts5PoslistWriter {
i64 iPrev;
};
static int sqlite3Fts5PoslistWriterAppend(Fts5Buffer*, Fts5PoslistWriter*, i64);
static void sqlite3Fts5PoslistSafeAppend(Fts5Buffer*, i64*, i64);
static int sqlite3Fts5PoslistNext64(
const u8 *a, int n, /* Buffer containing poslist */
int *pi, /* IN/OUT: Offset within a[] */
i64 *piOff /* IN/OUT: Current offset */
);
/* Malloc utility */
static void *sqlite3Fts5MallocZero(int *pRc, sqlite3_int64 nByte);
static char *sqlite3Fts5Strndup(int *pRc, const char *pIn, int nIn);
/* Character set tests (like isspace(), isalpha() etc.) */
static int sqlite3Fts5IsBareword(char t);
/* Bucket of terms object used by the integrity-check in offsets=0 mode. */
typedef struct Fts5Termset Fts5Termset;
static int sqlite3Fts5TermsetNew(Fts5Termset**);
static int sqlite3Fts5TermsetAdd(Fts5Termset*, int, const char*, int, int *pbPresent);
static void sqlite3Fts5TermsetFree(Fts5Termset*);
/*
** End of interface to code in fts5_buffer.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_index.c. fts5_index.c contains contains code
** to access the data stored in the %_data table.
*/
typedef struct Fts5Index Fts5Index;
typedef struct Fts5IndexIter Fts5IndexIter;
struct Fts5IndexIter {
i64 iRowid;
const u8 *pData;
int nData;
u8 bEof;
};
#define sqlite3Fts5IterEof(x) ((x)->bEof)
/*
** Values used as part of the flags argument passed to IndexQuery().
*/
#define FTS5INDEX_QUERY_PREFIX 0x0001 /* Prefix query */
#define FTS5INDEX_QUERY_DESC 0x0002 /* Docs in descending rowid order */
#define FTS5INDEX_QUERY_TEST_NOIDX 0x0004 /* Do not use prefix index */
#define FTS5INDEX_QUERY_SCAN 0x0008 /* Scan query (fts5vocab) */
/* The following are used internally by the fts5_index.c module. They are
** defined here only to make it easier to avoid clashes with the flags
** above. */
#define FTS5INDEX_QUERY_SKIPEMPTY 0x0010
#define FTS5INDEX_QUERY_NOOUTPUT 0x0020
#define FTS5INDEX_QUERY_SKIPHASH 0x0040
#define FTS5INDEX_QUERY_NOTOKENDATA 0x0080
#define FTS5INDEX_QUERY_SCANONETERM 0x0100
/*
** Create/destroy an Fts5Index object.
*/
static int sqlite3Fts5IndexOpen(Fts5Config *pConfig, int bCreate, Fts5Index**, char**);
static int sqlite3Fts5IndexClose(Fts5Index *p);
/*
** Return a simple checksum value based on the arguments.
*/
static u64 sqlite3Fts5IndexEntryCksum(
i64 iRowid,
int iCol,
int iPos,
int iIdx,
const char *pTerm,
int nTerm
);
/*
** Argument p points to a buffer containing utf-8 text that is n bytes in
** size. Return the number of bytes in the nChar character prefix of the
** buffer, or 0 if there are less than nChar characters in total.
*/
static int sqlite3Fts5IndexCharlenToBytelen(
const char *p,
int nByte,
int nChar
);
/*
** Open a new iterator to iterate though all rowids that match the
** specified token or token prefix.
*/
static int sqlite3Fts5IndexQuery(
Fts5Index *p, /* FTS index to query */
const char *pToken, int nToken, /* Token (or prefix) to query for */
int flags, /* Mask of FTS5INDEX_QUERY_X flags */
Fts5Colset *pColset, /* Match these columns only */
Fts5IndexIter **ppIter /* OUT: New iterator object */
);
/*
** The various operations on open token or token prefix iterators opened
** using sqlite3Fts5IndexQuery().
*/
static int sqlite3Fts5IterNext(Fts5IndexIter*);
static int sqlite3Fts5IterNextFrom(Fts5IndexIter*, i64 iMatch);
/*
** Close an iterator opened by sqlite3Fts5IndexQuery().
*/
static void sqlite3Fts5IterClose(Fts5IndexIter*);
/*
** Close the reader blob handle, if it is open.
*/
static void sqlite3Fts5IndexCloseReader(Fts5Index*);
/*
** This interface is used by the fts5vocab module.
*/
static const char *sqlite3Fts5IterTerm(Fts5IndexIter*, int*);
static int sqlite3Fts5IterNextScan(Fts5IndexIter*);
static void *sqlite3Fts5StructureRef(Fts5Index*);
static void sqlite3Fts5StructureRelease(void*);
static int sqlite3Fts5StructureTest(Fts5Index*, void*);
/*
** Used by xInstToken():
*/
static int sqlite3Fts5IterToken(Fts5IndexIter*, i64, int, int, const char**, int*);
/*
** Insert or remove data to or from the index. Each time a document is
** added to or removed from the index, this function is called one or more
** times.
**
** For an insert, it must be called once for each token in the new document.
** If the operation is a delete, it must be called (at least) once for each
** unique token in the document with an iCol value less than zero. The iPos
** argument is ignored for a delete.
*/
static int sqlite3Fts5IndexWrite(
Fts5Index *p, /* Index to write to */
int iCol, /* Column token appears in (-ve -> delete) */
int iPos, /* Position of token within column */
const char *pToken, int nToken /* Token to add or remove to or from index */
);
/*
** Indicate that subsequent calls to sqlite3Fts5IndexWrite() pertain to
** document iDocid.
*/
static int sqlite3Fts5IndexBeginWrite(
Fts5Index *p, /* Index to write to */
int bDelete, /* True if current operation is a delete */
i64 iDocid /* Docid to add or remove data from */
);
/*
** Flush any data stored in the in-memory hash tables to the database.
** Also close any open blob handles.
*/
static int sqlite3Fts5IndexSync(Fts5Index *p);
/*
** Discard any data stored in the in-memory hash tables. Do not write it
** to the database. Additionally, assume that the contents of the %_data
** table may have changed on disk. So any in-memory caches of %_data
** records must be invalidated.
*/
static int sqlite3Fts5IndexRollback(Fts5Index *p);
/*
** Get or set the "averages" values.
*/
static int sqlite3Fts5IndexGetAverages(Fts5Index *p, i64 *pnRow, i64 *anSize);
static int sqlite3Fts5IndexSetAverages(Fts5Index *p, const u8*, int);
/*
** Functions called by the storage module as part of integrity-check.
*/
static int sqlite3Fts5IndexIntegrityCheck(Fts5Index*, u64 cksum, int bUseCksum);
/*
** Called during virtual module initialization to register UDF
** fts5_decode() with SQLite
*/
static int sqlite3Fts5IndexInit(sqlite3*);
static int sqlite3Fts5IndexSetCookie(Fts5Index*, int);
/*
** Return the total number of entries read from the %_data table by
** this connection since it was created.
*/
static int sqlite3Fts5IndexReads(Fts5Index *p);
static int sqlite3Fts5IndexReinit(Fts5Index *p);
static int sqlite3Fts5IndexOptimize(Fts5Index *p);
static int sqlite3Fts5IndexMerge(Fts5Index *p, int nMerge);
static int sqlite3Fts5IndexReset(Fts5Index *p);
static int sqlite3Fts5IndexLoadConfig(Fts5Index *p);
static int sqlite3Fts5IndexGetOrigin(Fts5Index *p, i64 *piOrigin);
static int sqlite3Fts5IndexContentlessDelete(Fts5Index *p, i64 iOrigin, i64 iRowid);
static void sqlite3Fts5IndexIterClearTokendata(Fts5IndexIter*);
/* Used to populate hash tables for xInstToken in detail=none/column mode. */
static int sqlite3Fts5IndexIterWriteTokendata(
Fts5IndexIter*, const char*, int, i64 iRowid, int iCol, int iOff
);
/*
** End of interface to code in fts5_index.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_varint.c.
*/
static int sqlite3Fts5GetVarint32(const unsigned char *p, u32 *v);
static int sqlite3Fts5GetVarintLen(u32 iVal);
static u8 sqlite3Fts5GetVarint(const unsigned char*, u64*);
static int sqlite3Fts5PutVarint(unsigned char *p, u64 v);
#define fts5GetVarint32(a,b) sqlite3Fts5GetVarint32(a,(u32*)&(b))
#define fts5GetVarint sqlite3Fts5GetVarint
#define fts5FastGetVarint32(a, iOff, nVal) { \
nVal = (a)[iOff++]; \
if( nVal & 0x80 ){ \
iOff--; \
iOff += fts5GetVarint32(&(a)[iOff], nVal); \
} \
}
/*
** End of interface to code in fts5_varint.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_main.c.
*/
/*
** Virtual-table object.
*/
typedef struct Fts5Table Fts5Table;
struct Fts5Table {
sqlite3_vtab base; /* Base class used by SQLite core */
Fts5Config *pConfig; /* Virtual table configuration */
Fts5Index *pIndex; /* Full-text index */
};
static int sqlite3Fts5GetTokenizer(
Fts5Global*,
const char **azArg,
int nArg,
Fts5Config*,
char **pzErr
);
static Fts5Table *sqlite3Fts5TableFromCsrid(Fts5Global*, i64);
static int sqlite3Fts5FlushToDisk(Fts5Table*);
/*
** End of interface to code in fts5.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_hash.c.
*/
typedef struct Fts5Hash Fts5Hash;
/*
** Create a hash table, free a hash table.
*/
static int sqlite3Fts5HashNew(Fts5Config*, Fts5Hash**, int *pnSize);
static void sqlite3Fts5HashFree(Fts5Hash*);
static int sqlite3Fts5HashWrite(
Fts5Hash*,
i64 iRowid, /* Rowid for this entry */
int iCol, /* Column token appears in (-ve -> delete) */
int iPos, /* Position of token within column */
char bByte,
const char *pToken, int nToken /* Token to add or remove to or from index */
);
/*
** Empty (but do not delete) a hash table.
*/
static void sqlite3Fts5HashClear(Fts5Hash*);
/*
** Return true if the hash is empty, false otherwise.
*/
static int sqlite3Fts5HashIsEmpty(Fts5Hash*);
static int sqlite3Fts5HashQuery(
Fts5Hash*, /* Hash table to query */
int nPre,
const char *pTerm, int nTerm, /* Query term */
void **ppObj, /* OUT: Pointer to doclist for pTerm */
int *pnDoclist /* OUT: Size of doclist in bytes */
);
static int sqlite3Fts5HashScanInit(
Fts5Hash*, /* Hash table to query */
const char *pTerm, int nTerm /* Query prefix */
);
static void sqlite3Fts5HashScanNext(Fts5Hash*);
static int sqlite3Fts5HashScanEof(Fts5Hash*);
static void sqlite3Fts5HashScanEntry(Fts5Hash *,
const char **pzTerm, /* OUT: term (nul-terminated) */
int *pnTerm, /* OUT: Size of term in bytes */
const u8 **ppDoclist, /* OUT: pointer to doclist */
int *pnDoclist /* OUT: size of doclist in bytes */
);
/*
** End of interface to code in fts5_hash.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_storage.c. fts5_storage.c contains contains
** code to access the data stored in the %_content and %_docsize tables.
*/
#define FTS5_STMT_SCAN_ASC 0 /* SELECT rowid, * FROM ... ORDER BY 1 ASC */
#define FTS5_STMT_SCAN_DESC 1 /* SELECT rowid, * FROM ... ORDER BY 1 DESC */
#define FTS5_STMT_LOOKUP 2 /* SELECT rowid, * FROM ... WHERE rowid=? */
typedef struct Fts5Storage Fts5Storage;
static int sqlite3Fts5StorageOpen(Fts5Config*, Fts5Index*, int, Fts5Storage**, char**);
static int sqlite3Fts5StorageClose(Fts5Storage *p);
static int sqlite3Fts5StorageRename(Fts5Storage*, const char *zName);
static int sqlite3Fts5DropAll(Fts5Config*);
static int sqlite3Fts5CreateTable(Fts5Config*, const char*, const char*, int, char **);
static int sqlite3Fts5StorageDelete(Fts5Storage *p, i64, sqlite3_value**);
static int sqlite3Fts5StorageContentInsert(Fts5Storage *p, sqlite3_value**, i64*);
static int sqlite3Fts5StorageIndexInsert(Fts5Storage *p, sqlite3_value**, i64);
static int sqlite3Fts5StorageIntegrity(Fts5Storage *p, int iArg);
static int sqlite3Fts5StorageStmt(Fts5Storage *p, int eStmt, sqlite3_stmt**, char**);
static void sqlite3Fts5StorageStmtRelease(Fts5Storage *p, int eStmt, sqlite3_stmt*);
static int sqlite3Fts5StorageDocsize(Fts5Storage *p, i64 iRowid, int *aCol);
static int sqlite3Fts5StorageSize(Fts5Storage *p, int iCol, i64 *pnAvg);
static int sqlite3Fts5StorageRowCount(Fts5Storage *p, i64 *pnRow);
static int sqlite3Fts5StorageSync(Fts5Storage *p);
static int sqlite3Fts5StorageRollback(Fts5Storage *p);
static int sqlite3Fts5StorageConfigValue(
Fts5Storage *p, const char*, sqlite3_value*, int
);
static int sqlite3Fts5StorageDeleteAll(Fts5Storage *p);
static int sqlite3Fts5StorageRebuild(Fts5Storage *p);
static int sqlite3Fts5StorageOptimize(Fts5Storage *p);
static int sqlite3Fts5StorageMerge(Fts5Storage *p, int nMerge);
static int sqlite3Fts5StorageReset(Fts5Storage *p);
/*
** End of interface to code in fts5_storage.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_expr.c.
*/
typedef struct Fts5Expr Fts5Expr;
typedef struct Fts5ExprNode Fts5ExprNode;
typedef struct Fts5Parse Fts5Parse;
typedef struct Fts5Token Fts5Token;
typedef struct Fts5ExprPhrase Fts5ExprPhrase;
typedef struct Fts5ExprNearset Fts5ExprNearset;
struct Fts5Token {
const char *p; /* Token text (not NULL terminated) */
int n; /* Size of buffer p in bytes */
};
/* Parse a MATCH expression. */
static int sqlite3Fts5ExprNew(
Fts5Config *pConfig,
int bPhraseToAnd,
int iCol, /* Column on LHS of MATCH operator */
const char *zExpr,
Fts5Expr **ppNew,
char **pzErr
);
static int sqlite3Fts5ExprPattern(
Fts5Config *pConfig,
int bGlob,
int iCol,
const char *zText,
Fts5Expr **pp
);
/*
** for(rc = sqlite3Fts5ExprFirst(pExpr, pIdx, bDesc);
** rc==SQLITE_OK && 0==sqlite3Fts5ExprEof(pExpr);
** rc = sqlite3Fts5ExprNext(pExpr)
** ){
** // The document with rowid iRowid matches the expression!
** i64 iRowid = sqlite3Fts5ExprRowid(pExpr);
** }
*/
static int sqlite3Fts5ExprFirst(Fts5Expr*, Fts5Index *pIdx, i64 iMin, int bDesc);
static int sqlite3Fts5ExprNext(Fts5Expr*, i64 iMax);
static int sqlite3Fts5ExprEof(Fts5Expr*);
static i64 sqlite3Fts5ExprRowid(Fts5Expr*);
static void sqlite3Fts5ExprFree(Fts5Expr*);
static int sqlite3Fts5ExprAnd(Fts5Expr **pp1, Fts5Expr *p2);
/* Called during startup to register a UDF with SQLite */
static int sqlite3Fts5ExprInit(Fts5Global*, sqlite3*);
static int sqlite3Fts5ExprPhraseCount(Fts5Expr*);
static int sqlite3Fts5ExprPhraseSize(Fts5Expr*, int iPhrase);
static int sqlite3Fts5ExprPoslist(Fts5Expr*, int, const u8 **);
typedef struct Fts5PoslistPopulator Fts5PoslistPopulator;
static Fts5PoslistPopulator *sqlite3Fts5ExprClearPoslists(Fts5Expr*, int);
static int sqlite3Fts5ExprPopulatePoslists(
Fts5Config*, Fts5Expr*, Fts5PoslistPopulator*, int, const char*, int
);
static void sqlite3Fts5ExprCheckPoslists(Fts5Expr*, i64);
static int sqlite3Fts5ExprClonePhrase(Fts5Expr*, int, Fts5Expr**);
static int sqlite3Fts5ExprPhraseCollist(Fts5Expr *, int, const u8 **, int *);
static int sqlite3Fts5ExprQueryToken(Fts5Expr*, int, int, const char**, int*);
static int sqlite3Fts5ExprInstToken(Fts5Expr*, i64, int, int, int, int, const char**, int*);
static void sqlite3Fts5ExprClearTokens(Fts5Expr*);
/*******************************************
** The fts5_expr.c API above this point is used by the other hand-written
** C code in this module. The interfaces below this point are called by
** the parser code in fts5parse.y. */
static void sqlite3Fts5ParseError(Fts5Parse *pParse, const char *zFmt, ...);
static Fts5ExprNode *sqlite3Fts5ParseNode(
Fts5Parse *pParse,
int eType,
Fts5ExprNode *pLeft,
Fts5ExprNode *pRight,
Fts5ExprNearset *pNear
);
static Fts5ExprNode *sqlite3Fts5ParseImplicitAnd(
Fts5Parse *pParse,
Fts5ExprNode *pLeft,
Fts5ExprNode *pRight
);
static Fts5ExprPhrase *sqlite3Fts5ParseTerm(
Fts5Parse *pParse,
Fts5ExprPhrase *pPhrase,
Fts5Token *pToken,
int bPrefix
);
static void sqlite3Fts5ParseSetCaret(Fts5ExprPhrase*);
static Fts5ExprNearset *sqlite3Fts5ParseNearset(
Fts5Parse*,
Fts5ExprNearset*,
Fts5ExprPhrase*
);
static Fts5Colset *sqlite3Fts5ParseColset(
Fts5Parse*,
Fts5Colset*,
Fts5Token *
);
static void sqlite3Fts5ParsePhraseFree(Fts5ExprPhrase*);
static void sqlite3Fts5ParseNearsetFree(Fts5ExprNearset*);
static void sqlite3Fts5ParseNodeFree(Fts5ExprNode*);
static void sqlite3Fts5ParseSetDistance(Fts5Parse*, Fts5ExprNearset*, Fts5Token*);
static void sqlite3Fts5ParseSetColset(Fts5Parse*, Fts5ExprNode*, Fts5Colset*);
static Fts5Colset *sqlite3Fts5ParseColsetInvert(Fts5Parse*, Fts5Colset*);
static void sqlite3Fts5ParseFinished(Fts5Parse *pParse, Fts5ExprNode *p);
static void sqlite3Fts5ParseNear(Fts5Parse *pParse, Fts5Token*);
/*
** End of interface to code in fts5_expr.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_aux.c.
*/
static int sqlite3Fts5AuxInit(fts5_api*);
/*
** End of interface to code in fts5_aux.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_tokenizer.c.
*/
static int sqlite3Fts5TokenizerInit(fts5_api*);
static int sqlite3Fts5TokenizerPattern(
int (*xCreate)(void*, const char**, int, Fts5Tokenizer**),
Fts5Tokenizer *pTok
);
/*
** End of interface to code in fts5_tokenizer.c.
**************************************************************************/
/**************************************************************************
** Interface to code in fts5_vocab.c.
*/
static int sqlite3Fts5VocabInit(Fts5Global*, sqlite3*);
/*
** End of interface to code in fts5_vocab.c.
**************************************************************************/
/**************************************************************************
** Interface to automatically generated code in fts5_unicode2.c.
*/
static int sqlite3Fts5UnicodeIsdiacritic(int c);
static int sqlite3Fts5UnicodeFold(int c, int bRemoveDiacritic);
static int sqlite3Fts5UnicodeCatParse(const char*, u8*);
static int sqlite3Fts5UnicodeCategory(u32 iCode);
static void sqlite3Fts5UnicodeAscii(u8*, u8*);
/*
** End of interface to code in fts5_unicode2.c.
**************************************************************************/
#endif
#line 1 "fts5parse.h"
#define FTS5_OR 1
#define FTS5_AND 2
#define FTS5_NOT 3
#define FTS5_TERM 4
#define FTS5_COLON 5
#define FTS5_MINUS 6
#define FTS5_LCP 7
#define FTS5_RCP 8
#define FTS5_STRING 9
#define FTS5_LP 10
#define FTS5_RP 11
#define FTS5_CARET 12
#define FTS5_COMMA 13
#define FTS5_PLUS 14
#define FTS5_STAR 15
#line 1 "fts5parse.c"
/* This file is automatically generated by Lemon from input grammar
** source file "fts5parse.y".
*/
/*
** 2000-05-29
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
*************************************************************************
** Driver template for the LEMON parser generator.
**
** The "lemon" program processes an LALR(1) input grammar file, then uses
** this template to construct a parser. The "lemon" program inserts text
** at each "%%" line. Also, any "P-a-r-s-e" identifer prefix (without the
** interstitial "-" characters) contained in this template is changed into
** the value of the %name directive from the grammar. Otherwise, the content
** of this template is copied straight through into the generate parser
** source file.
**
** The following is the concatenation of all %include directives from the
** input grammar file:
*/
/************ Begin %include sections from the grammar ************************/
#line 47 "fts5parse.y"
/* #include "fts5Int.h" */
/* #include "fts5parse.h" */
/*
** Disable all error recovery processing in the parser push-down
** automaton.
*/
#define fts5YYNOERRORRECOVERY 1
/*
** Make fts5yytestcase() the same as testcase()
*/
#define fts5yytestcase(X) testcase(X)
/*
** Indicate that sqlite3ParserFree() will never be called with a null
** pointer.
*/
#define fts5YYPARSEFREENOTNULL 1
/*
** Alternative datatype for the argument to the malloc() routine passed
** into sqlite3ParserAlloc(). The default is size_t.
*/
#define fts5YYMALLOCARGTYPE u64
#line 58 "fts5parse.sql"
/**************** End of %include directives **********************************/
/* These constants specify the various numeric values for terminal symbols.
***************** Begin token definitions *************************************/
#ifndef FTS5_OR
#define FTS5_OR 1
#define FTS5_AND 2
#define FTS5_NOT 3
#define FTS5_TERM 4
#define FTS5_COLON 5
#define FTS5_MINUS 6
#define FTS5_LCP 7
#define FTS5_RCP 8
#define FTS5_STRING 9
#define FTS5_LP 10
#define FTS5_RP 11
#define FTS5_CARET 12
#define FTS5_COMMA 13
#define FTS5_PLUS 14
#define FTS5_STAR 15
#endif
/**************** End token definitions ***************************************/
/* The next sections is a series of control #defines.
** various aspects of the generated parser.
** fts5YYCODETYPE is the data type used to store the integer codes
** that represent terminal and non-terminal symbols.
** "unsigned char" is used if there are fewer than
** 256 symbols. Larger types otherwise.
** fts5YYNOCODE is a number of type fts5YYCODETYPE that is not used for
** any terminal or nonterminal symbol.
** fts5YYFALLBACK If defined, this indicates that one or more tokens
** (also known as: "terminal symbols") have fall-back
** values which should be used if the original symbol
** would not parse. This permits keywords to sometimes
** be used as identifiers, for example.
** fts5YYACTIONTYPE is the data type used for "action codes" - numbers
** that indicate what to do in response to the next
** token.
** sqlite3Fts5ParserFTS5TOKENTYPE is the data type used for minor type for terminal
** symbols. Background: A "minor type" is a semantic
** value associated with a terminal or non-terminal
** symbols. For example, for an "ID" terminal symbol,
** the minor type might be the name of the identifier.
** Each non-terminal can have a different minor type.
** Terminal symbols all have the same minor type, though.
** This macros defines the minor type for terminal
** symbols.
** fts5YYMINORTYPE is the data type used for all minor types.
** This is typically a union of many types, one of
** which is sqlite3Fts5ParserFTS5TOKENTYPE. The entry in the union
** for terminal symbols is called "fts5yy0".
** fts5YYSTACKDEPTH is the maximum depth of the parser's stack. If
** zero the stack is dynamically sized using realloc()
** sqlite3Fts5ParserARG_SDECL A static variable declaration for the %extra_argument
** sqlite3Fts5ParserARG_PDECL A parameter declaration for the %extra_argument
** sqlite3Fts5ParserARG_PARAM Code to pass %extra_argument as a subroutine parameter
** sqlite3Fts5ParserARG_STORE Code to store %extra_argument into fts5yypParser
** sqlite3Fts5ParserARG_FETCH Code to extract %extra_argument from fts5yypParser
** sqlite3Fts5ParserCTX_* As sqlite3Fts5ParserARG_ except for %extra_context
** fts5YYERRORSYMBOL is the code number of the error symbol. If not
** defined, then do no error processing.
** fts5YYNSTATE the combined number of states.
** fts5YYNRULE the number of rules in the grammar
** fts5YYNFTS5TOKEN Number of terminal symbols
** fts5YY_MAX_SHIFT Maximum value for shift actions
** fts5YY_MIN_SHIFTREDUCE Minimum value for shift-reduce actions
** fts5YY_MAX_SHIFTREDUCE Maximum value for shift-reduce actions
** fts5YY_ERROR_ACTION The fts5yy_action[] code for syntax error
** fts5YY_ACCEPT_ACTION The fts5yy_action[] code for accept
** fts5YY_NO_ACTION The fts5yy_action[] code for no-op
** fts5YY_MIN_REDUCE Minimum value for reduce actions
** fts5YY_MAX_REDUCE Maximum value for reduce actions
*/
#ifndef INTERFACE
# define INTERFACE 1
#endif
/************* Begin control #defines *****************************************/
#define fts5YYCODETYPE unsigned char
#define fts5YYNOCODE 27
#define fts5YYACTIONTYPE unsigned char
#define sqlite3Fts5ParserFTS5TOKENTYPE Fts5Token
typedef union {
int fts5yyinit;
sqlite3Fts5ParserFTS5TOKENTYPE fts5yy0;
int fts5yy4;
Fts5Colset* fts5yy11;
Fts5ExprNode* fts5yy24;
Fts5ExprNearset* fts5yy46;
Fts5ExprPhrase* fts5yy53;
} fts5YYMINORTYPE;
#ifndef fts5YYSTACKDEPTH
#define fts5YYSTACKDEPTH 100
#endif
#define sqlite3Fts5ParserARG_SDECL Fts5Parse *pParse;
#define sqlite3Fts5ParserARG_PDECL ,Fts5Parse *pParse
#define sqlite3Fts5ParserARG_PARAM ,pParse
#define sqlite3Fts5ParserARG_FETCH Fts5Parse *pParse=fts5yypParser->pParse;
#define sqlite3Fts5ParserARG_STORE fts5yypParser->pParse=pParse;
#define sqlite3Fts5ParserCTX_SDECL
#define sqlite3Fts5ParserCTX_PDECL
#define sqlite3Fts5ParserCTX_PARAM
#define sqlite3Fts5ParserCTX_FETCH
#define sqlite3Fts5ParserCTX_STORE
#define fts5YYNSTATE 35
#define fts5YYNRULE 28
#define fts5YYNRULE_WITH_ACTION 28
#define fts5YYNFTS5TOKEN 16
#define fts5YY_MAX_SHIFT 34
#define fts5YY_MIN_SHIFTREDUCE 52
#define fts5YY_MAX_SHIFTREDUCE 79
#define fts5YY_ERROR_ACTION 80
#define fts5YY_ACCEPT_ACTION 81
#define fts5YY_NO_ACTION 82
#define fts5YY_MIN_REDUCE 83
#define fts5YY_MAX_REDUCE 110
/************* End control #defines *******************************************/
#define fts5YY_NLOOKAHEAD ((int)(sizeof(fts5yy_lookahead)/sizeof(fts5yy_lookahead[0])))
/* Define the fts5yytestcase() macro to be a no-op if is not already defined
** otherwise.
**
** Applications can choose to define fts5yytestcase() in the %include section
** to a macro that can assist in verifying code coverage. For production
** code the fts5yytestcase() macro should be turned off. But it is useful
** for testing.
*/
#ifndef fts5yytestcase
# define fts5yytestcase(X)
#endif
/* Next are the tables used to determine what action to take based on the
** current state and lookahead token. These tables are used to implement
** functions that take a state number and lookahead value and return an
** action integer.
**
** Suppose the action integer is N. Then the action is determined as
** follows
**
** 0 <= N <= fts5YY_MAX_SHIFT Shift N. That is, push the lookahead
** token onto the stack and goto state N.
**
** N between fts5YY_MIN_SHIFTREDUCE Shift to an arbitrary state then
** and fts5YY_MAX_SHIFTREDUCE reduce by rule N-fts5YY_MIN_SHIFTREDUCE.
**
** N == fts5YY_ERROR_ACTION A syntax error has occurred.
**
** N == fts5YY_ACCEPT_ACTION The parser accepts its input.
**
** N == fts5YY_NO_ACTION No such action. Denotes unused
** slots in the fts5yy_action[] table.
**
** N between fts5YY_MIN_REDUCE Reduce by rule N-fts5YY_MIN_REDUCE
** and fts5YY_MAX_REDUCE
**
** The action table is constructed as a single large table named fts5yy_action[].
** Given state S and lookahead X, the action is computed as either:
**
** (A) N = fts5yy_action[ fts5yy_shift_ofst[S] + X ]
** (B) N = fts5yy_default[S]
**
** The (A) formula is preferred. The B formula is used instead if
** fts5yy_lookahead[fts5yy_shift_ofst[S]+X] is not equal to X.
**
** The formulas above are for computing the action when the lookahead is
** a terminal symbol. If the lookahead is a non-terminal (as occurs after
** a reduce action) then the fts5yy_reduce_ofst[] array is used in place of
** the fts5yy_shift_ofst[] array.
**
** The following are the tables generated in this section:
**
** fts5yy_action[] A single table containing all actions.
** fts5yy_lookahead[] A table containing the lookahead for each entry in
** fts5yy_action. Used to detect hash collisions.
** fts5yy_shift_ofst[] For each state, the offset into fts5yy_action for
** shifting terminals.
** fts5yy_reduce_ofst[] For each state, the offset into fts5yy_action for
** shifting non-terminals after a reduce.
** fts5yy_default[] Default action for each state.
**
*********** Begin parsing tables **********************************************/
#define fts5YY_ACTTAB_COUNT (105)
static const fts5YYACTIONTYPE fts5yy_action[] = {
/* 0 */ 81, 20, 96, 6, 28, 99, 98, 26, 26, 18,
/* 10 */ 96, 6, 28, 17, 98, 56, 26, 19, 96, 6,
/* 20 */ 28, 14, 98, 14, 26, 31, 92, 96, 6, 28,
/* 30 */ 108, 98, 25, 26, 21, 96, 6, 28, 78, 98,
/* 40 */ 58, 26, 29, 96, 6, 28, 107, 98, 22, 26,
/* 50 */ 24, 16, 12, 11, 1, 13, 13, 24, 16, 23,
/* 60 */ 11, 33, 34, 13, 97, 8, 27, 32, 98, 7,
/* 70 */ 26, 3, 4, 5, 3, 4, 5, 3, 83, 4,
/* 80 */ 5, 3, 63, 5, 3, 62, 12, 2, 86, 13,
/* 90 */ 9, 30, 10, 10, 54, 57, 75, 78, 78, 53,
/* 100 */ 57, 15, 82, 82, 71,
};
static const fts5YYCODETYPE fts5yy_lookahead[] = {
/* 0 */ 16, 17, 18, 19, 20, 22, 22, 24, 24, 17,
/* 10 */ 18, 19, 20, 7, 22, 9, 24, 17, 18, 19,
/* 20 */ 20, 9, 22, 9, 24, 13, 17, 18, 19, 20,
/* 30 */ 26, 22, 24, 24, 17, 18, 19, 20, 15, 22,
/* 40 */ 9, 24, 17, 18, 19, 20, 26, 22, 21, 24,
/* 50 */ 6, 7, 9, 9, 10, 12, 12, 6, 7, 21,
/* 60 */ 9, 24, 25, 12, 18, 5, 20, 14, 22, 5,
/* 70 */ 24, 3, 1, 2, 3, 1, 2, 3, 0, 1,
/* 80 */ 2, 3, 11, 2, 3, 11, 9, 10, 5, 12,
/* 90 */ 23, 24, 10, 10, 8, 9, 9, 15, 15, 8,
/* 100 */ 9, 9, 27, 27, 11, 27, 27, 27, 27, 27,
/* 110 */ 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
/* 120 */ 27,
};
#define fts5YY_SHIFT_COUNT (34)
#define fts5YY_SHIFT_MIN (0)
#define fts5YY_SHIFT_MAX (93)
static const unsigned char fts5yy_shift_ofst[] = {
/* 0 */ 44, 44, 44, 44, 44, 44, 51, 77, 43, 12,
/* 10 */ 14, 83, 82, 14, 23, 23, 31, 31, 71, 74,
/* 20 */ 78, 81, 86, 91, 6, 53, 53, 60, 64, 68,
/* 30 */ 53, 87, 92, 53, 93,
};
#define fts5YY_REDUCE_COUNT (17)
#define fts5YY_REDUCE_MIN (-17)
#define fts5YY_REDUCE_MAX (67)
static const signed char fts5yy_reduce_ofst[] = {
/* 0 */ -16, -8, 0, 9, 17, 25, 46, -17, -17, 37,
/* 10 */ 67, 4, 4, 8, 4, 20, 27, 38,
};
static const fts5YYACTIONTYPE fts5yy_default[] = {
/* 0 */ 80, 80, 80, 80, 80, 80, 95, 80, 80, 105,
/* 10 */ 80, 110, 110, 80, 110, 110, 80, 80, 80, 80,
/* 20 */ 80, 91, 80, 80, 80, 101, 100, 80, 80, 90,
/* 30 */ 103, 80, 80, 104, 80,
};
/********** End of lemon-generated parsing tables *****************************/
/* The next table maps tokens (terminal symbols) into fallback tokens.
** If a construct like the following:
**
** %fallback ID X Y Z.
**
** appears in the grammar, then ID becomes a fallback token for X, Y,
** and Z. Whenever one of the tokens X, Y, or Z is input to the parser
** but it does not parse, the type of the token is changed to ID and
** the parse is retried before an error is thrown.
**
** This feature can be used, for example, to cause some keywords in a language
** to revert to identifiers if they keyword does not apply in the context where
** it appears.
*/
#ifdef fts5YYFALLBACK
static const fts5YYCODETYPE fts5yyFallback[] = {
};
#endif /* fts5YYFALLBACK */
/* The following structure represents a single element of the
** parser's stack. Information stored includes:
**
** + The state number for the parser at this level of the stack.
**
** + The value of the token stored at this level of the stack.
** (In other words, the "major" token.)
**
** + The semantic value stored at this level of the stack. This is
** the information used by the action routines in the grammar.
** It is sometimes called the "minor" token.
**
** After the "shift" half of a SHIFTREDUCE action, the stateno field
** actually contains the reduce action for the second half of the
** SHIFTREDUCE.
*/
struct fts5yyStackEntry {
fts5YYACTIONTYPE stateno; /* The state-number, or reduce action in SHIFTREDUCE */
fts5YYCODETYPE major; /* The major token value. This is the code
** number for the token at this stack level */
fts5YYMINORTYPE minor; /* The user-supplied minor token value. This
** is the value of the token */
};
typedef struct fts5yyStackEntry fts5yyStackEntry;
/* The state of the parser is completely contained in an instance of
** the following structure */
struct fts5yyParser {
fts5yyStackEntry *fts5yytos; /* Pointer to top element of the stack */
#ifdef fts5YYTRACKMAXSTACKDEPTH
int fts5yyhwm; /* High-water mark of the stack */
#endif
#ifndef fts5YYNOERRORRECOVERY
int fts5yyerrcnt; /* Shifts left before out of the error */
#endif
sqlite3Fts5ParserARG_SDECL /* A place to hold %extra_argument */
sqlite3Fts5ParserCTX_SDECL /* A place to hold %extra_context */
#if fts5YYSTACKDEPTH<=0
int fts5yystksz; /* Current side of the stack */
fts5yyStackEntry *fts5yystack; /* The parser's stack */
fts5yyStackEntry fts5yystk0; /* First stack entry */
#else
fts5yyStackEntry fts5yystack[fts5YYSTACKDEPTH]; /* The parser's stack */
fts5yyStackEntry *fts5yystackEnd; /* Last entry in the stack */
#endif
};
typedef struct fts5yyParser fts5yyParser;
#include
#ifndef NDEBUG
#include
static FILE *fts5yyTraceFILE = 0;
static char *fts5yyTracePrompt = 0;
#endif /* NDEBUG */
#ifndef NDEBUG
/*
** Turn parser tracing on by giving a stream to which to write the trace
** and a prompt to preface each trace message. Tracing is turned off
** by making either argument NULL
**
** Inputs:
**
** - A FILE* to which trace output should be written.
** If NULL, then tracing is turned off.
**
- A prefix string written at the beginning of every
** line of trace output. If NULL, then tracing is
** turned off.
**
**
** Outputs:
** None.
*/
static void sqlite3Fts5ParserTrace(FILE *TraceFILE, char *zTracePrompt){
fts5yyTraceFILE = TraceFILE;
fts5yyTracePrompt = zTracePrompt;
if( fts5yyTraceFILE==0 ) fts5yyTracePrompt = 0;
else if( fts5yyTracePrompt==0 ) fts5yyTraceFILE = 0;
}
#endif /* NDEBUG */
#if defined(fts5YYCOVERAGE) || !defined(NDEBUG)
/* For tracing shifts, the names of all terminals and nonterminals
** are required. The following table supplies these names */
static const char *const fts5yyTokenName[] = {
/* 0 */ "$",
/* 1 */ "OR",
/* 2 */ "AND",
/* 3 */ "NOT",
/* 4 */ "TERM",
/* 5 */ "COLON",
/* 6 */ "MINUS",
/* 7 */ "LCP",
/* 8 */ "RCP",
/* 9 */ "STRING",
/* 10 */ "LP",
/* 11 */ "RP",
/* 12 */ "CARET",
/* 13 */ "COMMA",
/* 14 */ "PLUS",
/* 15 */ "STAR",
/* 16 */ "input",
/* 17 */ "expr",
/* 18 */ "cnearset",
/* 19 */ "exprlist",
/* 20 */ "colset",
/* 21 */ "colsetlist",
/* 22 */ "nearset",
/* 23 */ "nearphrases",
/* 24 */ "phrase",
/* 25 */ "neardist_opt",
/* 26 */ "star_opt",
};
#endif /* defined(fts5YYCOVERAGE) || !defined(NDEBUG) */
#ifndef NDEBUG
/* For tracing reduce actions, the names of all rules are required.
*/
static const char *const fts5yyRuleName[] = {
/* 0 */ "input ::= expr",
/* 1 */ "colset ::= MINUS LCP colsetlist RCP",
/* 2 */ "colset ::= LCP colsetlist RCP",
/* 3 */ "colset ::= STRING",
/* 4 */ "colset ::= MINUS STRING",
/* 5 */ "colsetlist ::= colsetlist STRING",
/* 6 */ "colsetlist ::= STRING",
/* 7 */ "expr ::= expr AND expr",
/* 8 */ "expr ::= expr OR expr",
/* 9 */ "expr ::= expr NOT expr",
/* 10 */ "expr ::= colset COLON LP expr RP",
/* 11 */ "expr ::= LP expr RP",
/* 12 */ "expr ::= exprlist",
/* 13 */ "exprlist ::= cnearset",
/* 14 */ "exprlist ::= exprlist cnearset",
/* 15 */ "cnearset ::= nearset",
/* 16 */ "cnearset ::= colset COLON nearset",
/* 17 */ "nearset ::= phrase",
/* 18 */ "nearset ::= CARET phrase",
/* 19 */ "nearset ::= STRING LP nearphrases neardist_opt RP",
/* 20 */ "nearphrases ::= phrase",
/* 21 */ "nearphrases ::= nearphrases phrase",
/* 22 */ "neardist_opt ::=",
/* 23 */ "neardist_opt ::= COMMA STRING",
/* 24 */ "phrase ::= phrase PLUS STRING star_opt",
/* 25 */ "phrase ::= STRING star_opt",
/* 26 */ "star_opt ::= STAR",
/* 27 */ "star_opt ::=",
};
#endif /* NDEBUG */
#if fts5YYSTACKDEPTH<=0
/*
** Try to increase the size of the parser stack. Return the number
** of errors. Return 0 on success.
*/
static int fts5yyGrowStack(fts5yyParser *p){
int newSize;
int idx;
fts5yyStackEntry *pNew;
newSize = p->fts5yystksz*2 + 100;
idx = p->fts5yytos ? (int)(p->fts5yytos - p->fts5yystack) : 0;
if( p->fts5yystack==&p->fts5yystk0 ){
pNew = malloc(newSize*sizeof(pNew[0]));
if( pNew ) pNew[0] = p->fts5yystk0;
}else{
pNew = realloc(p->fts5yystack, newSize*sizeof(pNew[0]));
}
if( pNew ){
p->fts5yystack = pNew;
p->fts5yytos = &p->fts5yystack[idx];
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE,"%sStack grows from %d to %d entries.\n",
fts5yyTracePrompt, p->fts5yystksz, newSize);
}
#endif
p->fts5yystksz = newSize;
}
return pNew==0;
}
#endif
/* Datatype of the argument to the memory allocated passed as the
** second argument to sqlite3Fts5ParserAlloc() below. This can be changed by
** putting an appropriate #define in the %include section of the input
** grammar.
*/
#ifndef fts5YYMALLOCARGTYPE
# define fts5YYMALLOCARGTYPE size_t
#endif
/* Initialize a new parser that has already been allocated.
*/
static void sqlite3Fts5ParserInit(void *fts5yypRawParser sqlite3Fts5ParserCTX_PDECL){
fts5yyParser *fts5yypParser = (fts5yyParser*)fts5yypRawParser;
sqlite3Fts5ParserCTX_STORE
#ifdef fts5YYTRACKMAXSTACKDEPTH
fts5yypParser->fts5yyhwm = 0;
#endif
#if fts5YYSTACKDEPTH<=0
fts5yypParser->fts5yytos = NULL;
fts5yypParser->fts5yystack = NULL;
fts5yypParser->fts5yystksz = 0;
if( fts5yyGrowStack(fts5yypParser) ){
fts5yypParser->fts5yystack = &fts5yypParser->fts5yystk0;
fts5yypParser->fts5yystksz = 1;
}
#endif
#ifndef fts5YYNOERRORRECOVERY
fts5yypParser->fts5yyerrcnt = -1;
#endif
fts5yypParser->fts5yytos = fts5yypParser->fts5yystack;
fts5yypParser->fts5yystack[0].stateno = 0;
fts5yypParser->fts5yystack[0].major = 0;
#if fts5YYSTACKDEPTH>0
fts5yypParser->fts5yystackEnd = &fts5yypParser->fts5yystack[fts5YYSTACKDEPTH-1];
#endif
}
#ifndef sqlite3Fts5Parser_ENGINEALWAYSONSTACK
/*
** This function allocates a new parser.
** The only argument is a pointer to a function which works like
** malloc.
**
** Inputs:
** A pointer to the function used to allocate memory.
**
** Outputs:
** A pointer to a parser. This pointer is used in subsequent calls
** to sqlite3Fts5Parser and sqlite3Fts5ParserFree.
*/
static void *sqlite3Fts5ParserAlloc(void *(*mallocProc)(fts5YYMALLOCARGTYPE) sqlite3Fts5ParserCTX_PDECL){
fts5yyParser *fts5yypParser;
fts5yypParser = (fts5yyParser*)(*mallocProc)( (fts5YYMALLOCARGTYPE)sizeof(fts5yyParser) );
if( fts5yypParser ){
sqlite3Fts5ParserCTX_STORE
sqlite3Fts5ParserInit(fts5yypParser sqlite3Fts5ParserCTX_PARAM);
}
return (void*)fts5yypParser;
}
#endif /* sqlite3Fts5Parser_ENGINEALWAYSONSTACK */
/* The following function deletes the "minor type" or semantic value
** associated with a symbol. The symbol can be either a terminal
** or nonterminal. "fts5yymajor" is the symbol code, and "fts5yypminor" is
** a pointer to the value to be deleted. The code used to do the
** deletions is derived from the %destructor and/or %token_destructor
** directives of the input grammar.
*/
static void fts5yy_destructor(
fts5yyParser *fts5yypParser, /* The parser */
fts5YYCODETYPE fts5yymajor, /* Type code for object to destroy */
fts5YYMINORTYPE *fts5yypminor /* The object to be destroyed */
){
sqlite3Fts5ParserARG_FETCH
sqlite3Fts5ParserCTX_FETCH
switch( fts5yymajor ){
/* Here is inserted the actions which take place when a
** terminal or non-terminal is destroyed. This can happen
** when the symbol is popped from the stack during a
** reduce or during error processing or when a parser is
** being destroyed before it is finished parsing.
**
** Note: during a reduce, the only symbols destroyed are those
** which appear on the RHS of the rule, but which are *not* used
** inside the C code.
*/
/********* Begin destructor definitions ***************************************/
case 16: /* input */
{
#line 83 "fts5parse.y"
(void)pParse;
#line 587 "fts5parse.sql"
}
break;
case 17: /* expr */
case 18: /* cnearset */
case 19: /* exprlist */
{
#line 89 "fts5parse.y"
sqlite3Fts5ParseNodeFree((fts5yypminor->fts5yy24));
#line 596 "fts5parse.sql"
}
break;
case 20: /* colset */
case 21: /* colsetlist */
{
#line 93 "fts5parse.y"
sqlite3_free((fts5yypminor->fts5yy11));
#line 604 "fts5parse.sql"
}
break;
case 22: /* nearset */
case 23: /* nearphrases */
{
#line 148 "fts5parse.y"
sqlite3Fts5ParseNearsetFree((fts5yypminor->fts5yy46));
#line 612 "fts5parse.sql"
}
break;
case 24: /* phrase */
{
#line 183 "fts5parse.y"
sqlite3Fts5ParsePhraseFree((fts5yypminor->fts5yy53));
#line 619 "fts5parse.sql"
}
break;
/********* End destructor definitions *****************************************/
default: break; /* If no destructor action specified: do nothing */
}
}
/*
** Pop the parser's stack once.
**
** If there is a destructor routine associated with the token which
** is popped from the stack, then call it.
*/
static void fts5yy_pop_parser_stack(fts5yyParser *pParser){
fts5yyStackEntry *fts5yytos;
assert( pParser->fts5yytos!=0 );
assert( pParser->fts5yytos > pParser->fts5yystack );
fts5yytos = pParser->fts5yytos--;
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE,"%sPopping %s\n",
fts5yyTracePrompt,
fts5yyTokenName[fts5yytos->major]);
}
#endif
fts5yy_destructor(pParser, fts5yytos->major, &fts5yytos->minor);
}
/*
** Clear all secondary memory allocations from the parser
*/
static void sqlite3Fts5ParserFinalize(void *p){
fts5yyParser *pParser = (fts5yyParser*)p;
while( pParser->fts5yytos>pParser->fts5yystack ) fts5yy_pop_parser_stack(pParser);
#if fts5YYSTACKDEPTH<=0
if( pParser->fts5yystack!=&pParser->fts5yystk0 ) free(pParser->fts5yystack);
#endif
}
#ifndef sqlite3Fts5Parser_ENGINEALWAYSONSTACK
/*
** Deallocate and destroy a parser. Destructors are called for
** all stack elements before shutting the parser down.
**
** If the fts5YYPARSEFREENEVERNULL macro exists (for example because it
** is defined in a %include section of the input grammar) then it is
** assumed that the input pointer is never NULL.
*/
static void sqlite3Fts5ParserFree(
void *p, /* The parser to be deleted */
void (*freeProc)(void*) /* Function used to reclaim memory */
){
#ifndef fts5YYPARSEFREENEVERNULL
if( p==0 ) return;
#endif
sqlite3Fts5ParserFinalize(p);
(*freeProc)(p);
}
#endif /* sqlite3Fts5Parser_ENGINEALWAYSONSTACK */
/*
** Return the peak depth of the stack for a parser.
*/
#ifdef fts5YYTRACKMAXSTACKDEPTH
static int sqlite3Fts5ParserStackPeak(void *p){
fts5yyParser *pParser = (fts5yyParser*)p;
return pParser->fts5yyhwm;
}
#endif
/* This array of booleans keeps track of the parser statement
** coverage. The element fts5yycoverage[X][Y] is set when the parser
** is in state X and has a lookahead token Y. In a well-tested
** systems, every element of this matrix should end up being set.
*/
#if defined(fts5YYCOVERAGE)
static unsigned char fts5yycoverage[fts5YYNSTATE][fts5YYNFTS5TOKEN];
#endif
/*
** Write into out a description of every state/lookahead combination that
**
** (1) has not been used by the parser, and
** (2) is not a syntax error.
**
** Return the number of missed state/lookahead combinations.
*/
#if defined(fts5YYCOVERAGE)
static int sqlite3Fts5ParserCoverage(FILE *out){
int stateno, iLookAhead, i;
int nMissed = 0;
for(stateno=0; statenofts5YY_MAX_SHIFT ) return stateno;
assert( stateno <= fts5YY_SHIFT_COUNT );
#if defined(fts5YYCOVERAGE)
fts5yycoverage[stateno][iLookAhead] = 1;
#endif
do{
i = fts5yy_shift_ofst[stateno];
assert( i>=0 );
assert( i<=fts5YY_ACTTAB_COUNT );
assert( i+fts5YYNFTS5TOKEN<=(int)fts5YY_NLOOKAHEAD );
assert( iLookAhead!=fts5YYNOCODE );
assert( iLookAhead < fts5YYNFTS5TOKEN );
i += iLookAhead;
assert( i<(int)fts5YY_NLOOKAHEAD );
if( fts5yy_lookahead[i]!=iLookAhead ){
#ifdef fts5YYFALLBACK
fts5YYCODETYPE iFallback; /* Fallback token */
assert( iLookAhead %s\n",
fts5yyTracePrompt, fts5yyTokenName[iLookAhead], fts5yyTokenName[iFallback]);
}
#endif
assert( fts5yyFallback[iFallback]==0 ); /* Fallback loop must terminate */
iLookAhead = iFallback;
continue;
}
#endif
#ifdef fts5YYWILDCARD
{
int j = i - iLookAhead + fts5YYWILDCARD;
assert( j<(int)(sizeof(fts5yy_lookahead)/sizeof(fts5yy_lookahead[0])) );
if( fts5yy_lookahead[j]==fts5YYWILDCARD && iLookAhead>0 ){
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE, "%sWILDCARD %s => %s\n",
fts5yyTracePrompt, fts5yyTokenName[iLookAhead],
fts5yyTokenName[fts5YYWILDCARD]);
}
#endif /* NDEBUG */
return fts5yy_action[j];
}
}
#endif /* fts5YYWILDCARD */
return fts5yy_default[stateno];
}else{
assert( i>=0 && i<(int)(sizeof(fts5yy_action)/sizeof(fts5yy_action[0])) );
return fts5yy_action[i];
}
}while(1);
}
/*
** Find the appropriate action for a parser given the non-terminal
** look-ahead token iLookAhead.
*/
static fts5YYACTIONTYPE fts5yy_find_reduce_action(
fts5YYACTIONTYPE stateno, /* Current state number */
fts5YYCODETYPE iLookAhead /* The look-ahead token */
){
int i;
#ifdef fts5YYERRORSYMBOL
if( stateno>fts5YY_REDUCE_COUNT ){
return fts5yy_default[stateno];
}
#else
assert( stateno<=fts5YY_REDUCE_COUNT );
#endif
i = fts5yy_reduce_ofst[stateno];
assert( iLookAhead!=fts5YYNOCODE );
i += iLookAhead;
#ifdef fts5YYERRORSYMBOL
if( i<0 || i>=fts5YY_ACTTAB_COUNT || fts5yy_lookahead[i]!=iLookAhead ){
return fts5yy_default[stateno];
}
#else
assert( i>=0 && ifts5yytos>fts5yypParser->fts5yystack ) fts5yy_pop_parser_stack(fts5yypParser);
/* Here code is inserted which will execute if the parser
** stack every overflows */
/******** Begin %stack_overflow code ******************************************/
#line 36 "fts5parse.y"
sqlite3Fts5ParseError(pParse, "fts5: parser stack overflow");
#line 840 "fts5parse.sql"
/******** End %stack_overflow code ********************************************/
sqlite3Fts5ParserARG_STORE /* Suppress warning about unused %extra_argument var */
sqlite3Fts5ParserCTX_STORE
}
/*
** Print tracing information for a SHIFT action
*/
#ifndef NDEBUG
static void fts5yyTraceShift(fts5yyParser *fts5yypParser, int fts5yyNewState, const char *zTag){
if( fts5yyTraceFILE ){
if( fts5yyNewStatefts5yytos->major],
fts5yyNewState);
}else{
fprintf(fts5yyTraceFILE,"%s%s '%s', pending reduce %d\n",
fts5yyTracePrompt, zTag, fts5yyTokenName[fts5yypParser->fts5yytos->major],
fts5yyNewState - fts5YY_MIN_REDUCE);
}
}
}
#else
# define fts5yyTraceShift(X,Y,Z)
#endif
/*
** Perform a shift action.
*/
static void fts5yy_shift(
fts5yyParser *fts5yypParser, /* The parser to be shifted */
fts5YYACTIONTYPE fts5yyNewState, /* The new state to shift in */
fts5YYCODETYPE fts5yyMajor, /* The major token to shift in */
sqlite3Fts5ParserFTS5TOKENTYPE fts5yyMinor /* The minor token to shift in */
){
fts5yyStackEntry *fts5yytos;
fts5yypParser->fts5yytos++;
#ifdef fts5YYTRACKMAXSTACKDEPTH
if( (int)(fts5yypParser->fts5yytos - fts5yypParser->fts5yystack)>fts5yypParser->fts5yyhwm ){
fts5yypParser->fts5yyhwm++;
assert( fts5yypParser->fts5yyhwm == (int)(fts5yypParser->fts5yytos - fts5yypParser->fts5yystack) );
}
#endif
#if fts5YYSTACKDEPTH>0
if( fts5yypParser->fts5yytos>fts5yypParser->fts5yystackEnd ){
fts5yypParser->fts5yytos--;
fts5yyStackOverflow(fts5yypParser);
return;
}
#else
if( fts5yypParser->fts5yytos>=&fts5yypParser->fts5yystack[fts5yypParser->fts5yystksz] ){
if( fts5yyGrowStack(fts5yypParser) ){
fts5yypParser->fts5yytos--;
fts5yyStackOverflow(fts5yypParser);
return;
}
}
#endif
if( fts5yyNewState > fts5YY_MAX_SHIFT ){
fts5yyNewState += fts5YY_MIN_REDUCE - fts5YY_MIN_SHIFTREDUCE;
}
fts5yytos = fts5yypParser->fts5yytos;
fts5yytos->stateno = fts5yyNewState;
fts5yytos->major = fts5yyMajor;
fts5yytos->minor.fts5yy0 = fts5yyMinor;
fts5yyTraceShift(fts5yypParser, fts5yyNewState, "Shift");
}
/* For rule J, fts5yyRuleInfoLhs[J] contains the symbol on the left-hand side
** of that rule */
static const fts5YYCODETYPE fts5yyRuleInfoLhs[] = {
16, /* (0) input ::= expr */
20, /* (1) colset ::= MINUS LCP colsetlist RCP */
20, /* (2) colset ::= LCP colsetlist RCP */
20, /* (3) colset ::= STRING */
20, /* (4) colset ::= MINUS STRING */
21, /* (5) colsetlist ::= colsetlist STRING */
21, /* (6) colsetlist ::= STRING */
17, /* (7) expr ::= expr AND expr */
17, /* (8) expr ::= expr OR expr */
17, /* (9) expr ::= expr NOT expr */
17, /* (10) expr ::= colset COLON LP expr RP */
17, /* (11) expr ::= LP expr RP */
17, /* (12) expr ::= exprlist */
19, /* (13) exprlist ::= cnearset */
19, /* (14) exprlist ::= exprlist cnearset */
18, /* (15) cnearset ::= nearset */
18, /* (16) cnearset ::= colset COLON nearset */
22, /* (17) nearset ::= phrase */
22, /* (18) nearset ::= CARET phrase */
22, /* (19) nearset ::= STRING LP nearphrases neardist_opt RP */
23, /* (20) nearphrases ::= phrase */
23, /* (21) nearphrases ::= nearphrases phrase */
25, /* (22) neardist_opt ::= */
25, /* (23) neardist_opt ::= COMMA STRING */
24, /* (24) phrase ::= phrase PLUS STRING star_opt */
24, /* (25) phrase ::= STRING star_opt */
26, /* (26) star_opt ::= STAR */
26, /* (27) star_opt ::= */
};
/* For rule J, fts5yyRuleInfoNRhs[J] contains the negative of the number
** of symbols on the right-hand side of that rule. */
static const signed char fts5yyRuleInfoNRhs[] = {
-1, /* (0) input ::= expr */
-4, /* (1) colset ::= MINUS LCP colsetlist RCP */
-3, /* (2) colset ::= LCP colsetlist RCP */
-1, /* (3) colset ::= STRING */
-2, /* (4) colset ::= MINUS STRING */
-2, /* (5) colsetlist ::= colsetlist STRING */
-1, /* (6) colsetlist ::= STRING */
-3, /* (7) expr ::= expr AND expr */
-3, /* (8) expr ::= expr OR expr */
-3, /* (9) expr ::= expr NOT expr */
-5, /* (10) expr ::= colset COLON LP expr RP */
-3, /* (11) expr ::= LP expr RP */
-1, /* (12) expr ::= exprlist */
-1, /* (13) exprlist ::= cnearset */
-2, /* (14) exprlist ::= exprlist cnearset */
-1, /* (15) cnearset ::= nearset */
-3, /* (16) cnearset ::= colset COLON nearset */
-1, /* (17) nearset ::= phrase */
-2, /* (18) nearset ::= CARET phrase */
-5, /* (19) nearset ::= STRING LP nearphrases neardist_opt RP */
-1, /* (20) nearphrases ::= phrase */
-2, /* (21) nearphrases ::= nearphrases phrase */
0, /* (22) neardist_opt ::= */
-2, /* (23) neardist_opt ::= COMMA STRING */
-4, /* (24) phrase ::= phrase PLUS STRING star_opt */
-2, /* (25) phrase ::= STRING star_opt */
-1, /* (26) star_opt ::= STAR */
0, /* (27) star_opt ::= */
};
static void fts5yy_accept(fts5yyParser*); /* Forward Declaration */
/*
** Perform a reduce action and the shift that must immediately
** follow the reduce.
**
** The fts5yyLookahead and fts5yyLookaheadToken parameters provide reduce actions
** access to the lookahead token (if any). The fts5yyLookahead will be fts5YYNOCODE
** if the lookahead token has already been consumed. As this procedure is
** only called from one place, optimizing compilers will in-line it, which
** means that the extra parameters have no performance impact.
*/
static fts5YYACTIONTYPE fts5yy_reduce(
fts5yyParser *fts5yypParser, /* The parser */
unsigned int fts5yyruleno, /* Number of the rule by which to reduce */
int fts5yyLookahead, /* Lookahead token, or fts5YYNOCODE if none */
sqlite3Fts5ParserFTS5TOKENTYPE fts5yyLookaheadToken /* Value of the lookahead token */
sqlite3Fts5ParserCTX_PDECL /* %extra_context */
){
int fts5yygoto; /* The next state */
fts5YYACTIONTYPE fts5yyact; /* The next action */
fts5yyStackEntry *fts5yymsp; /* The top of the parser's stack */
int fts5yysize; /* Amount to pop the stack */
sqlite3Fts5ParserARG_FETCH
(void)fts5yyLookahead;
(void)fts5yyLookaheadToken;
fts5yymsp = fts5yypParser->fts5yytos;
switch( fts5yyruleno ){
/* Beginning here are the reduction cases. A typical example
** follows:
** case 0:
** #line
** { ... } // User supplied code
** #line
** break;
*/
/********** Begin reduce actions **********************************************/
fts5YYMINORTYPE fts5yylhsminor;
case 0: /* input ::= expr */
#line 82 "fts5parse.y"
{ sqlite3Fts5ParseFinished(pParse, fts5yymsp[0].minor.fts5yy24); }
#line 1017 "fts5parse.sql"
break;
case 1: /* colset ::= MINUS LCP colsetlist RCP */
#line 97 "fts5parse.y"
{
fts5yymsp[-3].minor.fts5yy11 = sqlite3Fts5ParseColsetInvert(pParse, fts5yymsp[-1].minor.fts5yy11);
}
#line 1024 "fts5parse.sql"
break;
case 2: /* colset ::= LCP colsetlist RCP */
#line 100 "fts5parse.y"
{ fts5yymsp[-2].minor.fts5yy11 = fts5yymsp[-1].minor.fts5yy11; }
#line 1029 "fts5parse.sql"
break;
case 3: /* colset ::= STRING */
#line 101 "fts5parse.y"
{
fts5yylhsminor.fts5yy11 = sqlite3Fts5ParseColset(pParse, 0, &fts5yymsp[0].minor.fts5yy0);
}
#line 1036 "fts5parse.sql"
fts5yymsp[0].minor.fts5yy11 = fts5yylhsminor.fts5yy11;
break;
case 4: /* colset ::= MINUS STRING */
#line 104 "fts5parse.y"
{
fts5yymsp[-1].minor.fts5yy11 = sqlite3Fts5ParseColset(pParse, 0, &fts5yymsp[0].minor.fts5yy0);
fts5yymsp[-1].minor.fts5yy11 = sqlite3Fts5ParseColsetInvert(pParse, fts5yymsp[-1].minor.fts5yy11);
}
#line 1045 "fts5parse.sql"
break;
case 5: /* colsetlist ::= colsetlist STRING */
#line 109 "fts5parse.y"
{
fts5yylhsminor.fts5yy11 = sqlite3Fts5ParseColset(pParse, fts5yymsp[-1].minor.fts5yy11, &fts5yymsp[0].minor.fts5yy0); }
#line 1051 "fts5parse.sql"
fts5yymsp[-1].minor.fts5yy11 = fts5yylhsminor.fts5yy11;
break;
case 6: /* colsetlist ::= STRING */
#line 111 "fts5parse.y"
{
fts5yylhsminor.fts5yy11 = sqlite3Fts5ParseColset(pParse, 0, &fts5yymsp[0].minor.fts5yy0);
}
#line 1059 "fts5parse.sql"
fts5yymsp[0].minor.fts5yy11 = fts5yylhsminor.fts5yy11;
break;
case 7: /* expr ::= expr AND expr */
#line 115 "fts5parse.y"
{
fts5yylhsminor.fts5yy24 = sqlite3Fts5ParseNode(pParse, FTS5_AND, fts5yymsp[-2].minor.fts5yy24, fts5yymsp[0].minor.fts5yy24, 0);
}
#line 1067 "fts5parse.sql"
fts5yymsp[-2].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 8: /* expr ::= expr OR expr */
#line 118 "fts5parse.y"
{
fts5yylhsminor.fts5yy24 = sqlite3Fts5ParseNode(pParse, FTS5_OR, fts5yymsp[-2].minor.fts5yy24, fts5yymsp[0].minor.fts5yy24, 0);
}
#line 1075 "fts5parse.sql"
fts5yymsp[-2].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 9: /* expr ::= expr NOT expr */
#line 121 "fts5parse.y"
{
fts5yylhsminor.fts5yy24 = sqlite3Fts5ParseNode(pParse, FTS5_NOT, fts5yymsp[-2].minor.fts5yy24, fts5yymsp[0].minor.fts5yy24, 0);
}
#line 1083 "fts5parse.sql"
fts5yymsp[-2].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 10: /* expr ::= colset COLON LP expr RP */
#line 125 "fts5parse.y"
{
sqlite3Fts5ParseSetColset(pParse, fts5yymsp[-1].minor.fts5yy24, fts5yymsp[-4].minor.fts5yy11);
fts5yylhsminor.fts5yy24 = fts5yymsp[-1].minor.fts5yy24;
}
#line 1092 "fts5parse.sql"
fts5yymsp[-4].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 11: /* expr ::= LP expr RP */
#line 129 "fts5parse.y"
{fts5yymsp[-2].minor.fts5yy24 = fts5yymsp[-1].minor.fts5yy24;}
#line 1098 "fts5parse.sql"
break;
case 12: /* expr ::= exprlist */
case 13: /* exprlist ::= cnearset */ fts5yytestcase(fts5yyruleno==13);
#line 130 "fts5parse.y"
{fts5yylhsminor.fts5yy24 = fts5yymsp[0].minor.fts5yy24;}
#line 1104 "fts5parse.sql"
fts5yymsp[0].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 14: /* exprlist ::= exprlist cnearset */
#line 133 "fts5parse.y"
{
fts5yylhsminor.fts5yy24 = sqlite3Fts5ParseImplicitAnd(pParse, fts5yymsp[-1].minor.fts5yy24, fts5yymsp[0].minor.fts5yy24);
}
#line 1112 "fts5parse.sql"
fts5yymsp[-1].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 15: /* cnearset ::= nearset */
#line 137 "fts5parse.y"
{
fts5yylhsminor.fts5yy24 = sqlite3Fts5ParseNode(pParse, FTS5_STRING, 0, 0, fts5yymsp[0].minor.fts5yy46);
}
#line 1120 "fts5parse.sql"
fts5yymsp[0].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 16: /* cnearset ::= colset COLON nearset */
#line 140 "fts5parse.y"
{
fts5yylhsminor.fts5yy24 = sqlite3Fts5ParseNode(pParse, FTS5_STRING, 0, 0, fts5yymsp[0].minor.fts5yy46);
sqlite3Fts5ParseSetColset(pParse, fts5yylhsminor.fts5yy24, fts5yymsp[-2].minor.fts5yy11);
}
#line 1129 "fts5parse.sql"
fts5yymsp[-2].minor.fts5yy24 = fts5yylhsminor.fts5yy24;
break;
case 17: /* nearset ::= phrase */
#line 151 "fts5parse.y"
{ fts5yylhsminor.fts5yy46 = sqlite3Fts5ParseNearset(pParse, 0, fts5yymsp[0].minor.fts5yy53); }
#line 1135 "fts5parse.sql"
fts5yymsp[0].minor.fts5yy46 = fts5yylhsminor.fts5yy46;
break;
case 18: /* nearset ::= CARET phrase */
#line 152 "fts5parse.y"
{
sqlite3Fts5ParseSetCaret(fts5yymsp[0].minor.fts5yy53);
fts5yymsp[-1].minor.fts5yy46 = sqlite3Fts5ParseNearset(pParse, 0, fts5yymsp[0].minor.fts5yy53);
}
#line 1144 "fts5parse.sql"
break;
case 19: /* nearset ::= STRING LP nearphrases neardist_opt RP */
#line 156 "fts5parse.y"
{
sqlite3Fts5ParseNear(pParse, &fts5yymsp[-4].minor.fts5yy0);
sqlite3Fts5ParseSetDistance(pParse, fts5yymsp[-2].minor.fts5yy46, &fts5yymsp[-1].minor.fts5yy0);
fts5yylhsminor.fts5yy46 = fts5yymsp[-2].minor.fts5yy46;
}
#line 1153 "fts5parse.sql"
fts5yymsp[-4].minor.fts5yy46 = fts5yylhsminor.fts5yy46;
break;
case 20: /* nearphrases ::= phrase */
#line 162 "fts5parse.y"
{
fts5yylhsminor.fts5yy46 = sqlite3Fts5ParseNearset(pParse, 0, fts5yymsp[0].minor.fts5yy53);
}
#line 1161 "fts5parse.sql"
fts5yymsp[0].minor.fts5yy46 = fts5yylhsminor.fts5yy46;
break;
case 21: /* nearphrases ::= nearphrases phrase */
#line 165 "fts5parse.y"
{
fts5yylhsminor.fts5yy46 = sqlite3Fts5ParseNearset(pParse, fts5yymsp[-1].minor.fts5yy46, fts5yymsp[0].minor.fts5yy53);
}
#line 1169 "fts5parse.sql"
fts5yymsp[-1].minor.fts5yy46 = fts5yylhsminor.fts5yy46;
break;
case 22: /* neardist_opt ::= */
#line 172 "fts5parse.y"
{ fts5yymsp[1].minor.fts5yy0.p = 0; fts5yymsp[1].minor.fts5yy0.n = 0; }
#line 1175 "fts5parse.sql"
break;
case 23: /* neardist_opt ::= COMMA STRING */
#line 173 "fts5parse.y"
{ fts5yymsp[-1].minor.fts5yy0 = fts5yymsp[0].minor.fts5yy0; }
#line 1180 "fts5parse.sql"
break;
case 24: /* phrase ::= phrase PLUS STRING star_opt */
#line 185 "fts5parse.y"
{
fts5yylhsminor.fts5yy53 = sqlite3Fts5ParseTerm(pParse, fts5yymsp[-3].minor.fts5yy53, &fts5yymsp[-1].minor.fts5yy0, fts5yymsp[0].minor.fts5yy4);
}
#line 1187 "fts5parse.sql"
fts5yymsp[-3].minor.fts5yy53 = fts5yylhsminor.fts5yy53;
break;
case 25: /* phrase ::= STRING star_opt */
#line 188 "fts5parse.y"
{
fts5yylhsminor.fts5yy53 = sqlite3Fts5ParseTerm(pParse, 0, &fts5yymsp[-1].minor.fts5yy0, fts5yymsp[0].minor.fts5yy4);
}
#line 1195 "fts5parse.sql"
fts5yymsp[-1].minor.fts5yy53 = fts5yylhsminor.fts5yy53;
break;
case 26: /* star_opt ::= STAR */
#line 196 "fts5parse.y"
{ fts5yymsp[0].minor.fts5yy4 = 1; }
#line 1201 "fts5parse.sql"
break;
case 27: /* star_opt ::= */
#line 197 "fts5parse.y"
{ fts5yymsp[1].minor.fts5yy4 = 0; }
#line 1206 "fts5parse.sql"
break;
default:
break;
/********** End reduce actions ************************************************/
};
assert( fts5yyrulenofts5YY_MAX_SHIFT && fts5yyact<=fts5YY_MAX_SHIFTREDUCE) );
/* It is not possible for a REDUCE to be followed by an error */
assert( fts5yyact!=fts5YY_ERROR_ACTION );
fts5yymsp += fts5yysize+1;
fts5yypParser->fts5yytos = fts5yymsp;
fts5yymsp->stateno = (fts5YYACTIONTYPE)fts5yyact;
fts5yymsp->major = (fts5YYCODETYPE)fts5yygoto;
fts5yyTraceShift(fts5yypParser, fts5yyact, "... then shift");
return fts5yyact;
}
/*
** The following code executes when the parse fails
*/
#ifndef fts5YYNOERRORRECOVERY
static void fts5yy_parse_failed(
fts5yyParser *fts5yypParser /* The parser */
){
sqlite3Fts5ParserARG_FETCH
sqlite3Fts5ParserCTX_FETCH
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE,"%sFail!\n",fts5yyTracePrompt);
}
#endif
while( fts5yypParser->fts5yytos>fts5yypParser->fts5yystack ) fts5yy_pop_parser_stack(fts5yypParser);
/* Here code is inserted which will be executed whenever the
** parser fails */
/************ Begin %parse_failure code ***************************************/
/************ End %parse_failure code *****************************************/
sqlite3Fts5ParserARG_STORE /* Suppress warning about unused %extra_argument variable */
sqlite3Fts5ParserCTX_STORE
}
#endif /* fts5YYNOERRORRECOVERY */
/*
** The following code executes when a syntax error first occurs.
*/
static void fts5yy_syntax_error(
fts5yyParser *fts5yypParser, /* The parser */
int fts5yymajor, /* The major type of the error token */
sqlite3Fts5ParserFTS5TOKENTYPE fts5yyminor /* The minor type of the error token */
){
sqlite3Fts5ParserARG_FETCH
sqlite3Fts5ParserCTX_FETCH
#define FTS5TOKEN fts5yyminor
/************ Begin %syntax_error code ****************************************/
#line 30 "fts5parse.y"
UNUSED_PARAM(fts5yymajor); /* Silence a compiler warning */
sqlite3Fts5ParseError(
pParse, "fts5: syntax error near \"%.*s\"",FTS5TOKEN.n,FTS5TOKEN.p
);
#line 1274 "fts5parse.sql"
/************ End %syntax_error code ******************************************/
sqlite3Fts5ParserARG_STORE /* Suppress warning about unused %extra_argument variable */
sqlite3Fts5ParserCTX_STORE
}
/*
** The following is executed when the parser accepts
*/
static void fts5yy_accept(
fts5yyParser *fts5yypParser /* The parser */
){
sqlite3Fts5ParserARG_FETCH
sqlite3Fts5ParserCTX_FETCH
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE,"%sAccept!\n",fts5yyTracePrompt);
}
#endif
#ifndef fts5YYNOERRORRECOVERY
fts5yypParser->fts5yyerrcnt = -1;
#endif
assert( fts5yypParser->fts5yytos==fts5yypParser->fts5yystack );
/* Here code is inserted which will be executed whenever the
** parser accepts */
/*********** Begin %parse_accept code *****************************************/
/*********** End %parse_accept code *******************************************/
sqlite3Fts5ParserARG_STORE /* Suppress warning about unused %extra_argument variable */
sqlite3Fts5ParserCTX_STORE
}
/* The main parser program.
** The first argument is a pointer to a structure obtained from
** "sqlite3Fts5ParserAlloc" which describes the current state of the parser.
** The second argument is the major token number. The third is
** the minor token. The fourth optional argument is whatever the
** user wants (and specified in the grammar) and is available for
** use by the action routines.
**
** Inputs:
**
** - A pointer to the parser (an opaque structure.)
**
- The major token number.
**
- The minor token number.
**
- An option argument of a grammar-specified type.
**
**
** Outputs:
** None.
*/
static void sqlite3Fts5Parser(
void *fts5yyp, /* The parser */
int fts5yymajor, /* The major token code number */
sqlite3Fts5ParserFTS5TOKENTYPE fts5yyminor /* The value for the token */
sqlite3Fts5ParserARG_PDECL /* Optional %extra_argument parameter */
){
fts5YYMINORTYPE fts5yyminorunion;
fts5YYACTIONTYPE fts5yyact; /* The parser action. */
#if !defined(fts5YYERRORSYMBOL) && !defined(fts5YYNOERRORRECOVERY)
int fts5yyendofinput; /* True if we are at the end of input */
#endif
#ifdef fts5YYERRORSYMBOL
int fts5yyerrorhit = 0; /* True if fts5yymajor has invoked an error */
#endif
fts5yyParser *fts5yypParser = (fts5yyParser*)fts5yyp; /* The parser */
sqlite3Fts5ParserCTX_FETCH
sqlite3Fts5ParserARG_STORE
assert( fts5yypParser->fts5yytos!=0 );
#if !defined(fts5YYERRORSYMBOL) && !defined(fts5YYNOERRORRECOVERY)
fts5yyendofinput = (fts5yymajor==0);
#endif
fts5yyact = fts5yypParser->fts5yytos->stateno;
#ifndef NDEBUG
if( fts5yyTraceFILE ){
if( fts5yyact < fts5YY_MIN_REDUCE ){
fprintf(fts5yyTraceFILE,"%sInput '%s' in state %d\n",
fts5yyTracePrompt,fts5yyTokenName[fts5yymajor],fts5yyact);
}else{
fprintf(fts5yyTraceFILE,"%sInput '%s' with pending reduce %d\n",
fts5yyTracePrompt,fts5yyTokenName[fts5yymajor],fts5yyact-fts5YY_MIN_REDUCE);
}
}
#endif
while(1){ /* Exit by "break" */
assert( fts5yypParser->fts5yytos>=fts5yypParser->fts5yystack );
assert( fts5yyact==fts5yypParser->fts5yytos->stateno );
fts5yyact = fts5yy_find_shift_action((fts5YYCODETYPE)fts5yymajor,fts5yyact);
if( fts5yyact >= fts5YY_MIN_REDUCE ){
unsigned int fts5yyruleno = fts5yyact - fts5YY_MIN_REDUCE; /* Reduce by this rule */
#ifndef NDEBUG
assert( fts5yyruleno<(int)(sizeof(fts5yyRuleName)/sizeof(fts5yyRuleName[0])) );
if( fts5yyTraceFILE ){
int fts5yysize = fts5yyRuleInfoNRhs[fts5yyruleno];
if( fts5yysize ){
fprintf(fts5yyTraceFILE, "%sReduce %d [%s]%s, pop back to state %d.\n",
fts5yyTracePrompt,
fts5yyruleno, fts5yyRuleName[fts5yyruleno],
fts5yyrulenofts5yytos[fts5yysize].stateno);
}else{
fprintf(fts5yyTraceFILE, "%sReduce %d [%s]%s.\n",
fts5yyTracePrompt, fts5yyruleno, fts5yyRuleName[fts5yyruleno],
fts5yyrulenofts5yytos - fts5yypParser->fts5yystack)>fts5yypParser->fts5yyhwm ){
fts5yypParser->fts5yyhwm++;
assert( fts5yypParser->fts5yyhwm ==
(int)(fts5yypParser->fts5yytos - fts5yypParser->fts5yystack));
}
#endif
#if fts5YYSTACKDEPTH>0
if( fts5yypParser->fts5yytos>=fts5yypParser->fts5yystackEnd ){
fts5yyStackOverflow(fts5yypParser);
break;
}
#else
if( fts5yypParser->fts5yytos>=&fts5yypParser->fts5yystack[fts5yypParser->fts5yystksz-1] ){
if( fts5yyGrowStack(fts5yypParser) ){
fts5yyStackOverflow(fts5yypParser);
break;
}
}
#endif
}
fts5yyact = fts5yy_reduce(fts5yypParser,fts5yyruleno,fts5yymajor,fts5yyminor sqlite3Fts5ParserCTX_PARAM);
}else if( fts5yyact <= fts5YY_MAX_SHIFTREDUCE ){
fts5yy_shift(fts5yypParser,fts5yyact,(fts5YYCODETYPE)fts5yymajor,fts5yyminor);
#ifndef fts5YYNOERRORRECOVERY
fts5yypParser->fts5yyerrcnt--;
#endif
break;
}else if( fts5yyact==fts5YY_ACCEPT_ACTION ){
fts5yypParser->fts5yytos--;
fts5yy_accept(fts5yypParser);
return;
}else{
assert( fts5yyact == fts5YY_ERROR_ACTION );
fts5yyminorunion.fts5yy0 = fts5yyminor;
#ifdef fts5YYERRORSYMBOL
int fts5yymx;
#endif
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE,"%sSyntax Error!\n",fts5yyTracePrompt);
}
#endif
#ifdef fts5YYERRORSYMBOL
/* A syntax error has occurred.
** The response to an error depends upon whether or not the
** grammar defines an error token "ERROR".
**
** This is what we do if the grammar does define ERROR:
**
** * Call the %syntax_error function.
**
** * Begin popping the stack until we enter a state where
** it is legal to shift the error symbol, then shift
** the error symbol.
**
** * Set the error count to three.
**
** * Begin accepting and shifting new tokens. No new error
** processing will occur until three tokens have been
** shifted successfully.
**
*/
if( fts5yypParser->fts5yyerrcnt<0 ){
fts5yy_syntax_error(fts5yypParser,fts5yymajor,fts5yyminor);
}
fts5yymx = fts5yypParser->fts5yytos->major;
if( fts5yymx==fts5YYERRORSYMBOL || fts5yyerrorhit ){
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fprintf(fts5yyTraceFILE,"%sDiscard input token %s\n",
fts5yyTracePrompt,fts5yyTokenName[fts5yymajor]);
}
#endif
fts5yy_destructor(fts5yypParser, (fts5YYCODETYPE)fts5yymajor, &fts5yyminorunion);
fts5yymajor = fts5YYNOCODE;
}else{
while( fts5yypParser->fts5yytos > fts5yypParser->fts5yystack ){
fts5yyact = fts5yy_find_reduce_action(fts5yypParser->fts5yytos->stateno,
fts5YYERRORSYMBOL);
if( fts5yyact<=fts5YY_MAX_SHIFTREDUCE ) break;
fts5yy_pop_parser_stack(fts5yypParser);
}
if( fts5yypParser->fts5yytos <= fts5yypParser->fts5yystack || fts5yymajor==0 ){
fts5yy_destructor(fts5yypParser,(fts5YYCODETYPE)fts5yymajor,&fts5yyminorunion);
fts5yy_parse_failed(fts5yypParser);
#ifndef fts5YYNOERRORRECOVERY
fts5yypParser->fts5yyerrcnt = -1;
#endif
fts5yymajor = fts5YYNOCODE;
}else if( fts5yymx!=fts5YYERRORSYMBOL ){
fts5yy_shift(fts5yypParser,fts5yyact,fts5YYERRORSYMBOL,fts5yyminor);
}
}
fts5yypParser->fts5yyerrcnt = 3;
fts5yyerrorhit = 1;
if( fts5yymajor==fts5YYNOCODE ) break;
fts5yyact = fts5yypParser->fts5yytos->stateno;
#elif defined(fts5YYNOERRORRECOVERY)
/* If the fts5YYNOERRORRECOVERY macro is defined, then do not attempt to
** do any kind of error recovery. Instead, simply invoke the syntax
** error routine and continue going as if nothing had happened.
**
** Applications can set this macro (for example inside %include) if
** they intend to abandon the parse upon the first syntax error seen.
*/
fts5yy_syntax_error(fts5yypParser,fts5yymajor, fts5yyminor);
fts5yy_destructor(fts5yypParser,(fts5YYCODETYPE)fts5yymajor,&fts5yyminorunion);
break;
#else /* fts5YYERRORSYMBOL is not defined */
/* This is what we do if the grammar does not define ERROR:
**
** * Report an error message, and throw away the input token.
**
** * If the input token is $, then fail the parse.
**
** As before, subsequent error messages are suppressed until
** three input tokens have been successfully shifted.
*/
if( fts5yypParser->fts5yyerrcnt<=0 ){
fts5yy_syntax_error(fts5yypParser,fts5yymajor, fts5yyminor);
}
fts5yypParser->fts5yyerrcnt = 3;
fts5yy_destructor(fts5yypParser,(fts5YYCODETYPE)fts5yymajor,&fts5yyminorunion);
if( fts5yyendofinput ){
fts5yy_parse_failed(fts5yypParser);
#ifndef fts5YYNOERRORRECOVERY
fts5yypParser->fts5yyerrcnt = -1;
#endif
}
break;
#endif
}
}
#ifndef NDEBUG
if( fts5yyTraceFILE ){
fts5yyStackEntry *i;
char cDiv = '[';
fprintf(fts5yyTraceFILE,"%sReturn. Stack=",fts5yyTracePrompt);
for(i=&fts5yypParser->fts5yystack[1]; i<=fts5yypParser->fts5yytos; i++){
fprintf(fts5yyTraceFILE,"%c%s", cDiv, fts5yyTokenName[i->major]);
cDiv = ' ';
}
fprintf(fts5yyTraceFILE,"]\n");
}
#endif
return;
}
/*
** Return the fallback token corresponding to canonical token iToken, or
** 0 if iToken has no fallback.
*/
static int sqlite3Fts5ParserFallback(int iToken){
#ifdef fts5YYFALLBACK
assert( iToken<(int)(sizeof(fts5yyFallback)/sizeof(fts5yyFallback[0])) );
return fts5yyFallback[iToken];
#else
(void)iToken;
return 0;
#endif
}
#line 1 "fts5_aux.c"
/*
** 2014 May 31
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
*/
/* #include "fts5Int.h" */
#include /* amalgamator: keep */
/*
** Object used to iterate through all "coalesced phrase instances" in
** a single column of the current row. If the phrase instances in the
** column being considered do not overlap, this object simply iterates
** through them. Or, if they do overlap (share one or more tokens in
** common), each set of overlapping instances is treated as a single
** match. See documentation for the highlight() auxiliary function for
** details.
**
** Usage is:
**
** for(rc = fts5CInstIterNext(pApi, pFts, iCol, &iter);
** (rc==SQLITE_OK && 0==fts5CInstIterEof(&iter);
** rc = fts5CInstIterNext(&iter)
** ){
** printf("instance starts at %d, ends at %d\n", iter.iStart, iter.iEnd);
** }
**
*/
typedef struct CInstIter CInstIter;
struct CInstIter {
const Fts5ExtensionApi *pApi; /* API offered by current FTS version */
Fts5Context *pFts; /* First arg to pass to pApi functions */
int iCol; /* Column to search */
int iInst; /* Next phrase instance index */
int nInst; /* Total number of phrase instances */
/* Output variables */
int iStart; /* First token in coalesced phrase instance */
int iEnd; /* Last token in coalesced phrase instance */
};
/*
** Advance the iterator to the next coalesced phrase instance. Return
** an SQLite error code if an error occurs, or SQLITE_OK otherwise.
*/
static int fts5CInstIterNext(CInstIter *pIter){
int rc = SQLITE_OK;
pIter->iStart = -1;
pIter->iEnd = -1;
while( rc==SQLITE_OK && pIter->iInstnInst ){
int ip; int ic; int io;
rc = pIter->pApi->xInst(pIter->pFts, pIter->iInst, &ip, &ic, &io);
if( rc==SQLITE_OK ){
if( ic==pIter->iCol ){
int iEnd = io - 1 + pIter->pApi->xPhraseSize(pIter->pFts, ip);
if( pIter->iStart<0 ){
pIter->iStart = io;
pIter->iEnd = iEnd;
}else if( io<=pIter->iEnd ){
if( iEnd>pIter->iEnd ) pIter->iEnd = iEnd;
}else{
break;
}
}
pIter->iInst++;
}
}
return rc;
}
/*
** Initialize the iterator object indicated by the final parameter to
** iterate through coalesced phrase instances in column iCol.
*/
static int fts5CInstIterInit(
const Fts5ExtensionApi *pApi,
Fts5Context *pFts,
int iCol,
CInstIter *pIter
){
int rc;
memset(pIter, 0, sizeof(CInstIter));
pIter->pApi = pApi;
pIter->pFts = pFts;
pIter->iCol = iCol;
rc = pApi->xInstCount(pFts, &pIter->nInst);
if( rc==SQLITE_OK ){
rc = fts5CInstIterNext(pIter);
}
return rc;
}
/*************************************************************************
** Start of highlight() implementation.
*/
typedef struct HighlightContext HighlightContext;
struct HighlightContext {
/* Constant parameters to fts5HighlightCb() */
int iRangeStart; /* First token to include */
int iRangeEnd; /* If non-zero, last token to include */
const char *zOpen; /* Opening highlight */
const char *zClose; /* Closing highlight */
const char *zIn; /* Input text */
int nIn; /* Size of input text in bytes */
/* Variables modified by fts5HighlightCb() */
CInstIter iter; /* Coalesced Instance Iterator */
int iPos; /* Current token offset in zIn[] */
int iOff; /* Have copied up to this offset in zIn[] */
int bOpen; /* True if highlight is open */
char *zOut; /* Output value */
};
/*
** Append text to the HighlightContext output string - p->zOut. Argument
** z points to a buffer containing n bytes of text to append. If n is
** negative, everything up until the first '\0' is appended to the output.
**
** If *pRc is set to any value other than SQLITE_OK when this function is
** called, it is a no-op. If an error (i.e. an OOM condition) is encountered,
** *pRc is set to an error code before returning.
*/
static void fts5HighlightAppend(
int *pRc,
HighlightContext *p,
const char *z, int n
){
if( *pRc==SQLITE_OK && z ){
if( n<0 ) n = (int)strlen(z);
p->zOut = sqlite3_mprintf("%z%.*s", p->zOut, n, z);
if( p->zOut==0 ) *pRc = SQLITE_NOMEM;
}
}
/*
** Tokenizer callback used by implementation of highlight() function.
*/
static int fts5HighlightCb(
void *pContext, /* Pointer to HighlightContext object */
int tflags, /* Mask of FTS5_TOKEN_* flags */
const char *pToken, /* Buffer containing token */
int nToken, /* Size of token in bytes */
int iStartOff, /* Start byte offset of token */
int iEndOff /* End byte offset of token */
){
HighlightContext *p = (HighlightContext*)pContext;
int rc = SQLITE_OK;
int iPos;
UNUSED_PARAM2(pToken, nToken);
if( tflags & FTS5_TOKEN_COLOCATED ) return SQLITE_OK;
iPos = p->iPos++;
if( p->iRangeEnd>=0 ){
if( iPosiRangeStart || iPos>p->iRangeEnd ) return SQLITE_OK;
if( p->iRangeStart && iPos==p->iRangeStart ) p->iOff = iStartOff;
}
/* If the parenthesis is open, and this token is not part of the current
** phrase, and the starting byte offset of this token is past the point
** that has currently been copied into the output buffer, close the
** parenthesis. */
if( p->bOpen
&& (iPos<=p->iter.iStart || p->iter.iStart<0)
&& iStartOff>p->iOff
){
fts5HighlightAppend(&rc, p, p->zClose, -1);
p->bOpen = 0;
}
/* If this is the start of a new phrase, and the highlight is not open:
**
** * copy text from the input up to the start of the phrase, and
** * open the highlight.
*/
if( iPos==p->iter.iStart && p->bOpen==0 ){
fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iStartOff - p->iOff);
fts5HighlightAppend(&rc, p, p->zOpen, -1);
p->iOff = iStartOff;
p->bOpen = 1;
}
if( iPos==p->iter.iEnd ){
if( p->bOpen==0 ){
assert( p->iRangeEnd>=0 );
fts5HighlightAppend(&rc, p, p->zOpen, -1);
p->bOpen = 1;
}
fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff);
p->iOff = iEndOff;
if( rc==SQLITE_OK ){
rc = fts5CInstIterNext(&p->iter);
}
}
if( iPos==p->iRangeEnd ){
if( p->bOpen ){
if( p->iter.iStart>=0 && iPos>=p->iter.iStart ){
fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff);
p->iOff = iEndOff;
}
fts5HighlightAppend(&rc, p, p->zClose, -1);
p->bOpen = 0;
}
fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff);
p->iOff = iEndOff;
}
return rc;
}
/*
** Implementation of highlight() function.
*/
static void fts5HighlightFunction(
const Fts5ExtensionApi *pApi, /* API offered by current FTS version */
Fts5Context *pFts, /* First arg to pass to pApi functions */
sqlite3_context *pCtx, /* Context for returning result/error */
int nVal, /* Number of values in apVal[] array */
sqlite3_value **apVal /* Array of trailing arguments */
){
HighlightContext ctx;
int rc;
int iCol;
if( nVal!=3 ){
const char *zErr = "wrong number of arguments to function highlight()";
sqlite3_result_error(pCtx, zErr, -1);
return;
}
iCol = sqlite3_value_int(apVal[0]);
memset(&ctx, 0, sizeof(HighlightContext));
ctx.zOpen = (const char*)sqlite3_value_text(apVal[1]);
ctx.zClose = (const char*)sqlite3_value_text(apVal[2]);
ctx.iRangeEnd = -1;
rc = pApi->xColumnText(pFts, iCol, &ctx.zIn, &ctx.nIn);
if( rc==SQLITE_RANGE ){
sqlite3_result_text(pCtx, "", -1, SQLITE_STATIC);
rc = SQLITE_OK;
}else if( ctx.zIn ){
if( rc==SQLITE_OK ){
rc = fts5CInstIterInit(pApi, pFts, iCol, &ctx.iter);
}
if( rc==SQLITE_OK ){
rc = pApi->xTokenize(pFts, ctx.zIn, ctx.nIn, (void*)&ctx,fts5HighlightCb);
}
if( ctx.bOpen ){
fts5HighlightAppend(&rc, &ctx, ctx.zClose, -1);
}
fts5HighlightAppend(&rc, &ctx, &ctx.zIn[ctx.iOff], ctx.nIn - ctx.iOff);
if( rc==SQLITE_OK ){
sqlite3_result_text(pCtx, (const char*)ctx.zOut, -1, SQLITE_TRANSIENT);
}
sqlite3_free(ctx.zOut);
}
if( rc!=SQLITE_OK ){
sqlite3_result_error_code(pCtx, rc);
}
}
/*
** End of highlight() implementation.
**************************************************************************/
/*
** Context object passed to the fts5SentenceFinderCb() function.
*/
typedef struct Fts5SFinder Fts5SFinder;
struct Fts5SFinder {
int iPos; /* Current token position */
int nFirstAlloc; /* Allocated size of aFirst[] */
int nFirst; /* Number of entries in aFirst[] */
int *aFirst; /* Array of first token in each sentence */
const char *zDoc; /* Document being tokenized */
};
/*
** Add an entry to the Fts5SFinder.aFirst[] array. Grow the array if
** necessary. Return SQLITE_OK if successful, or SQLITE_NOMEM if an
** error occurs.
*/
static int fts5SentenceFinderAdd(Fts5SFinder *p, int iAdd){
if( p->nFirstAlloc==p->nFirst ){
int nNew = p->nFirstAlloc ? p->nFirstAlloc*2 : 64;
int *aNew;
aNew = (int*)sqlite3_realloc64(p->aFirst, nNew*sizeof(int));
if( aNew==0 ) return SQLITE_NOMEM;
p->aFirst = aNew;
p->nFirstAlloc = nNew;
}
p->aFirst[p->nFirst++] = iAdd;
return SQLITE_OK;
}
/*
** This function is an xTokenize() callback used by the auxiliary snippet()
** function. Its job is to identify tokens that are the first in a sentence.
** For each such token, an entry is added to the SFinder.aFirst[] array.
*/
static int fts5SentenceFinderCb(
void *pContext, /* Pointer to HighlightContext object */
int tflags, /* Mask of FTS5_TOKEN_* flags */
const char *pToken, /* Buffer containing token */
int nToken, /* Size of token in bytes */
int iStartOff, /* Start offset of token */
int iEndOff /* End offset of token */
){
int rc = SQLITE_OK;
UNUSED_PARAM2(pToken, nToken);
UNUSED_PARAM(iEndOff);
if( (tflags & FTS5_TOKEN_COLOCATED)==0 ){
Fts5SFinder *p = (Fts5SFinder*)pContext;
if( p->iPos>0 ){
int i;
char c = 0;
for(i=iStartOff-1; i>=0; i--){
c = p->zDoc[i];
if( c!=' ' && c!='\t' && c!='\n' && c!='\r' ) break;
}
if( i!=iStartOff-1 && (c=='.' || c==':') ){
rc = fts5SentenceFinderAdd(p, p->iPos);
}
}else{
rc = fts5SentenceFinderAdd(p, 0);
}
p->iPos++;
}
return rc;
}
static int fts5SnippetScore(
const Fts5ExtensionApi *pApi, /* API offered by current FTS version */
Fts5Context *pFts, /* First arg to pass to pApi functions */
int nDocsize, /* Size of column in tokens */
unsigned char *aSeen, /* Array with one element per query phrase */
int iCol, /* Column to score */
int iPos, /* Starting offset to score */
int nToken, /* Max tokens per snippet */
int *pnScore, /* OUT: Score */
int *piPos /* OUT: Adjusted offset */
){
int rc;
int i;
int ip = 0;
int ic = 0;
int iOff = 0;
int iFirst = -1;
int nInst;
int nScore = 0;
int iLast = 0;
sqlite3_int64 iEnd = (sqlite3_int64)iPos + nToken;
rc = pApi->xInstCount(pFts, &nInst);
for(i=0; ixInst(pFts, i, &ip, &ic, &iOff);
if( rc==SQLITE_OK && ic==iCol && iOff>=iPos && iOffxPhraseSize(pFts, ip);
}
}
*pnScore = nScore;
if( piPos ){
sqlite3_int64 iAdj = iFirst - (nToken - (iLast-iFirst)) / 2;
if( (iAdj+nToken)>nDocsize ) iAdj = nDocsize - nToken;
if( iAdj<0 ) iAdj = 0;
*piPos = (int)iAdj;
}
return rc;
}
/*
** Return the value in pVal interpreted as utf-8 text. Except, if pVal
** contains a NULL value, return a pointer to a static string zero
** bytes in length instead of a NULL pointer.
*/
static const char *fts5ValueToText(sqlite3_value *pVal){
const char *zRet = (const char*)sqlite3_value_text(pVal);
return zRet ? zRet : "";
}
/*
** Implementation of snippet() function.
*/
static void fts5SnippetFunction(
const Fts5ExtensionApi *pApi, /* API offered by current FTS version */
Fts5Context *pFts, /* First arg to pass to pApi functions */
sqlite3_context *pCtx, /* Context for returning result/error */
int nVal, /* Number of values in apVal[] array */
sqlite3_value **apVal /* Array of trailing arguments */
){
HighlightContext ctx;
int rc = SQLITE_OK; /* Return code */
int iCol; /* 1st argument to snippet() */
const char *zEllips; /* 4th argument to snippet() */
int nToken; /* 5th argument to snippet() */
int nInst = 0; /* Number of instance matches this row */
int i; /* Used to iterate through instances */
int nPhrase; /* Number of phrases in query */
unsigned char *aSeen; /* Array of "seen instance" flags */
int iBestCol; /* Column containing best snippet */
int iBestStart = 0; /* First token of best snippet */
int nBestScore = 0; /* Score of best snippet */
int nColSize = 0; /* Total size of iBestCol in tokens */
Fts5SFinder sFinder; /* Used to find the beginnings of sentences */
int nCol;
if( nVal!=5 ){
const char *zErr = "wrong number of arguments to function snippet()";
sqlite3_result_error(pCtx, zErr, -1);
return;
}
nCol = pApi->xColumnCount(pFts);
memset(&ctx, 0, sizeof(HighlightContext));
iCol = sqlite3_value_int(apVal[0]);
ctx.zOpen = fts5ValueToText(apVal[1]);
ctx.zClose = fts5ValueToText(apVal[2]);
ctx.iRangeEnd = -1;
zEllips = fts5ValueToText(apVal[3]);
nToken = sqlite3_value_int(apVal[4]);
iBestCol = (iCol>=0 ? iCol : 0);
nPhrase = pApi->xPhraseCount(pFts);
aSeen = sqlite3_malloc(nPhrase);
if( aSeen==0 ){
rc = SQLITE_NOMEM;
}
if( rc==SQLITE_OK ){
rc = pApi->xInstCount(pFts, &nInst);
}
memset(&sFinder, 0, sizeof(Fts5SFinder));
for(i=0; ixColumnText(pFts, i, &sFinder.zDoc, &nDoc);
if( rc!=SQLITE_OK ) break;
rc = pApi->xTokenize(pFts,
sFinder.zDoc, nDoc, (void*)&sFinder,fts5SentenceFinderCb
);
if( rc!=SQLITE_OK ) break;
rc = pApi->xColumnSize(pFts, i, &nDocsize);
if( rc!=SQLITE_OK ) break;
for(ii=0; rc==SQLITE_OK && iixInst(pFts, ii, &ip, &ic, &io);
if( ic!=i ) continue;
if( io>nDocsize ) rc = FTS5_CORRUPT;
if( rc!=SQLITE_OK ) continue;
memset(aSeen, 0, nPhrase);
rc = fts5SnippetScore(pApi, pFts, nDocsize, aSeen, i,
io, nToken, &nScore, &iAdj
);
if( rc==SQLITE_OK && nScore>nBestScore ){
nBestScore = nScore;
iBestCol = i;
iBestStart = iAdj;
nColSize = nDocsize;
}
if( rc==SQLITE_OK && sFinder.nFirst && nDocsize>nToken ){
for(jj=0; jj<(sFinder.nFirst-1); jj++){
if( sFinder.aFirst[jj+1]>io ) break;
}
if( sFinder.aFirst[jj]nBestScore ){
nBestScore = nScore;
iBestCol = i;
iBestStart = sFinder.aFirst[jj];
nColSize = nDocsize;
}
}
}
}
}
}
if( rc==SQLITE_OK ){
rc = pApi->xColumnText(pFts, iBestCol, &ctx.zIn, &ctx.nIn);
}
if( rc==SQLITE_OK && nColSize==0 ){
rc = pApi->xColumnSize(pFts, iBestCol, &nColSize);
}
if( ctx.zIn ){
if( rc==SQLITE_OK ){
rc = fts5CInstIterInit(pApi, pFts, iBestCol, &ctx.iter);
}
ctx.iRangeStart = iBestStart;
ctx.iRangeEnd = iBestStart + nToken - 1;
if( iBestStart>0 ){
fts5HighlightAppend(&rc, &ctx, zEllips, -1);
}
/* Advance iterator ctx.iter so that it points to the first coalesced
** phrase instance at or following position iBestStart. */
while( ctx.iter.iStart>=0 && ctx.iter.iStartxTokenize(pFts, ctx.zIn, ctx.nIn, (void*)&ctx,fts5HighlightCb);
}
if( ctx.bOpen ){
fts5HighlightAppend(&rc, &ctx, ctx.zClose, -1);
}
if( ctx.iRangeEnd>=(nColSize-1) ){
fts5HighlightAppend(&rc, &ctx, &ctx.zIn[ctx.iOff], ctx.nIn - ctx.iOff);
}else{
fts5HighlightAppend(&rc, &ctx, zEllips, -1);
}
}
if( rc==SQLITE_OK ){
sqlite3_result_text(pCtx, (const char*)ctx.zOut, -1, SQLITE_TRANSIENT);
}else{
sqlite3_result_error_code(pCtx, rc);
}
sqlite3_free(ctx.zOut);
sqlite3_free(aSeen);
sqlite3_free(sFinder.aFirst);
}
/************************************************************************/
/*
** The first time the bm25() function is called for a query, an instance
** of the following structure is allocated and populated.
*/
typedef struct Fts5Bm25Data Fts5Bm25Data;
struct Fts5Bm25Data {
int nPhrase; /* Number of phrases in query */
double avgdl; /* Average number of tokens in each row */
double *aIDF; /* IDF for each phrase */
double *aFreq; /* Array used to calculate phrase freq. */
};
/*
** Callback used by fts5Bm25GetData() to count the number of rows in the
** table matched by each individual phrase within the query.
*/
static int fts5CountCb(
const Fts5ExtensionApi *pApi,
Fts5Context *pFts,
void *pUserData /* Pointer to sqlite3_int64 variable */
){
sqlite3_int64 *pn = (sqlite3_int64*)pUserData;
UNUSED_PARAM2(pApi, pFts);
(*pn)++;
return SQLITE_OK;
}
/*
** Set *ppData to point to the Fts5Bm25Data object for the current query.
** If the object has not already been allocated, allocate and populate it
** now.
*/
static int fts5Bm25GetData(
const Fts5ExtensionApi *pApi,
Fts5Context *pFts,
Fts5Bm25Data **ppData /* OUT: bm25-data object for this query */
){
int rc = SQLITE_OK; /* Return code */
Fts5Bm25Data *p; /* Object to return */
p = (Fts5Bm25Data*)pApi->xGetAuxdata(pFts, 0);
if( p==0 ){
int nPhrase; /* Number of phrases in query */
sqlite3_int64 nRow = 0; /* Number of rows in table */
sqlite3_int64 nToken = 0; /* Number of tokens in table */
sqlite3_int64 nByte; /* Bytes of space to allocate */
int i;
/* Allocate the Fts5Bm25Data object */
nPhrase = pApi->xPhraseCount(pFts);
nByte = sizeof(Fts5Bm25Data) + nPhrase*2*sizeof(double);
p = (Fts5Bm25Data*)sqlite3_malloc64(nByte);
if( p==0 ){
rc = SQLITE_NOMEM;
}else{
memset(p, 0, (size_t)nByte);
p->nPhrase = nPhrase;
p->aIDF = (double*)&p[1];
p->aFreq = &p->aIDF[nPhrase];
}
/* Calculate the average document length for this FTS5 table */
if( rc==SQLITE_OK ) rc = pApi->xRowCount(pFts, &nRow);
assert( rc!=SQLITE_OK || nRow>0 );
if( rc==SQLITE_OK ) rc = pApi->xColumnTotalSize(pFts, -1, &nToken);
if( rc==SQLITE_OK ) p->avgdl = (double)nToken / (double)nRow;
/* Calculate an IDF for each phrase in the query */
for(i=0; rc==SQLITE_OK && ixQueryPhrase(pFts, i, (void*)&nHit, fts5CountCb);
if( rc==SQLITE_OK ){
/* Calculate the IDF (Inverse Document Frequency) for phrase i.
** This is done using the standard BM25 formula as found on wikipedia:
**
** IDF = log( (N - nHit + 0.5) / (nHit + 0.5) )
**
** where "N" is the total number of documents in the set and nHit
** is the number that contain at least one instance of the phrase
** under consideration.
**
** The problem with this is that if (N < 2*nHit), the IDF is
** negative. Which is undesirable. So the mimimum allowable IDF is
** (1e-6) - roughly the same as a term that appears in just over
** half of set of 5,000,000 documents. */
double idf = log( (nRow - nHit + 0.5) / (nHit + 0.5) );
if( idf<=0.0 ) idf = 1e-6;
p->aIDF[i] = idf;
}
}
if( rc!=SQLITE_OK ){
sqlite3_free(p);
}else{
rc = pApi->xSetAuxdata(pFts, p, sqlite3_free);
}
if( rc!=SQLITE_OK ) p = 0;
}
*ppData = p;
return rc;
}
/*
** Implementation of bm25() function.
*/
static void fts5Bm25Function(
const Fts5ExtensionApi *pApi, /* API offered by current FTS version */
Fts5Context *pFts, /* First arg to pass to pApi functions */
sqlite3_context *pCtx, /* Context for returning result/error */
int nVal, /* Number of values in apVal[] array */
sqlite3_value **apVal /* Array of trailing arguments */
){
const double k1 = 1.2; /* Constant "k1" from BM25 formula */
const double b = 0.75; /* Constant "b" from BM25 formula */
int rc; /* Error code */
double score = 0.0; /* SQL function return value */
Fts5Bm25Data *pData; /* Values allocated/calculated once only */
int i; /* Iterator variable */
int nInst = 0; /* Value returned by xInstCount() */
double D = 0.0; /* Total number of tokens in row */
double *aFreq = 0; /* Array of phrase freq. for current row */
/* Calculate the phrase frequency (symbol "f(qi,D)" in the documentation)
** for each phrase in the query for the current row. */
rc = fts5Bm25GetData(pApi, pFts, &pData);
if( rc==SQLITE_OK ){
aFreq = pData->aFreq;
memset(aFreq, 0, sizeof(double) * pData->nPhrase);
rc = pApi->xInstCount(pFts, &nInst);
}
for(i=0; rc==SQLITE_OK && ixInst(pFts, i, &ip, &ic, &io);
if( rc==SQLITE_OK ){
double w = (nVal > ic) ? sqlite3_value_double(apVal[ic]) : 1.0;
aFreq[ip] += w;
}
}
/* Figure out the total size of the current row in tokens. */
if( rc==SQLITE_OK ){
int nTok;
rc = pApi->xColumnSize(pFts, -1, &nTok);
D = (double)nTok;
}
/* Determine and return the BM25 score for the current row. Or, if an
** error has occurred, throw an exception. */
if( rc==SQLITE_OK ){
for(i=0; inPhrase; i++){
score += pData->aIDF[i] * (
( aFreq[i] * (k1 + 1.0) ) /
( aFreq[i] + k1 * (1 - b + b * D / pData->avgdl) )
);
}
sqlite3_result_double(pCtx, -1.0 * score);
}else{
sqlite3_result_error_code(pCtx, rc);
}
}
static int sqlite3Fts5AuxInit(fts5_api *pApi){
struct Builtin {
const char *zFunc; /* Function name (nul-terminated) */
void *pUserData; /* User-data pointer */
fts5_extension_function xFunc;/* Callback function */
void (*xDestroy)(void*); /* Destructor function */
} aBuiltin [] = {
{ "snippet", 0, fts5SnippetFunction, 0 },
{ "highlight", 0, fts5HighlightFunction, 0 },
{ "bm25", 0, fts5Bm25Function, 0 },
};
int rc = SQLITE_OK; /* Return code */
int i; /* To iterate through builtin functions */
for(i=0; rc==SQLITE_OK && ixCreateFunction(pApi,
aBuiltin[i].zFunc,
aBuiltin[i].pUserData,
aBuiltin[i].xFunc,
aBuiltin[i].xDestroy
);
}
return rc;
}
#line 1 "fts5_buffer.c"
/*
** 2014 May 31
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
*/
/* #include "fts5Int.h" */
static int sqlite3Fts5BufferSize(int *pRc, Fts5Buffer *pBuf, u32 nByte){
if( (u32)pBuf->nSpacenSpace ? pBuf->nSpace : 64;
u8 *pNew;
while( nNewp, nNew);
if( pNew==0 ){
*pRc = SQLITE_NOMEM;
return 1;
}else{
pBuf->nSpace = (int)nNew;
pBuf->p = pNew;
}
}
return 0;
}
/*
** Encode value iVal as an SQLite varint and append it to the buffer object
** pBuf. If an OOM error occurs, set the error code in p.
*/
static void sqlite3Fts5BufferAppendVarint(int *pRc, Fts5Buffer *pBuf, i64 iVal){
if( fts5BufferGrow(pRc, pBuf, 9) ) return;
pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], iVal);
}
static void sqlite3Fts5Put32(u8 *aBuf, int iVal){
aBuf[0] = (iVal>>24) & 0x00FF;
aBuf[1] = (iVal>>16) & 0x00FF;
aBuf[2] = (iVal>> 8) & 0x00FF;
aBuf[3] = (iVal>> 0) & 0x00FF;
}
static int sqlite3Fts5Get32(const u8 *aBuf){
return (int)((((u32)aBuf[0])<<24) + (aBuf[1]<<16) + (aBuf[2]<<8) + aBuf[3]);
}
/*
** Append buffer nData/pData to buffer pBuf. If an OOM error occurs, set
** the error code in p. If an error has already occurred when this function
** is called, it is a no-op.
*/
static void sqlite3Fts5BufferAppendBlob(
int *pRc,
Fts5Buffer *pBuf,
u32 nData,
const u8 *pData
){
if( nData ){
if( fts5BufferGrow(pRc, pBuf, nData) ) return;
assert( pBuf->p!=0 );
memcpy(&pBuf->p[pBuf->n], pData, nData);
pBuf->n += nData;
}
}
/*
** Append the nul-terminated string zStr to the buffer pBuf. This function
** ensures that the byte following the buffer data is set to 0x00, even
** though this byte is not included in the pBuf->n count.
*/
static void sqlite3Fts5BufferAppendString(
int *pRc,
Fts5Buffer *pBuf,
const char *zStr
){
int nStr = (int)strlen(zStr);
sqlite3Fts5BufferAppendBlob(pRc, pBuf, nStr+1, (const u8*)zStr);
pBuf->n--;
}
/*
** Argument zFmt is a printf() style format string. This function performs
** the printf() style processing, then appends the results to buffer pBuf.
**
** Like sqlite3Fts5BufferAppendString(), this function ensures that the byte
** following the buffer data is set to 0x00, even though this byte is not
** included in the pBuf->n count.
*/
static void sqlite3Fts5BufferAppendPrintf(
int *pRc,
Fts5Buffer *pBuf,
char *zFmt, ...
){
if( *pRc==SQLITE_OK ){
char *zTmp;
va_list ap;
va_start(ap, zFmt);
zTmp = sqlite3_vmprintf(zFmt, ap);
va_end(ap);
if( zTmp==0 ){
*pRc = SQLITE_NOMEM;
}else{
sqlite3Fts5BufferAppendString(pRc, pBuf, zTmp);
sqlite3_free(zTmp);
}
}
}
static char *sqlite3Fts5Mprintf(int *pRc, const char *zFmt, ...){
char *zRet = 0;
if( *pRc==SQLITE_OK ){
va_list ap;
va_start(ap, zFmt);
zRet = sqlite3_vmprintf(zFmt, ap);
va_end(ap);
if( zRet==0 ){
*pRc = SQLITE_NOMEM;
}
}
return zRet;
}
/*
** Free any buffer allocated by pBuf. Zero the structure before returning.
*/
static void sqlite3Fts5BufferFree(Fts5Buffer *pBuf){
sqlite3_free(pBuf->p);
memset(pBuf, 0, sizeof(Fts5Buffer));
}
/*
** Zero the contents of the buffer object. But do not free the associated
** memory allocation.
*/
static void sqlite3Fts5BufferZero(Fts5Buffer *pBuf){
pBuf->n = 0;
}
/*
** Set the buffer to contain nData/pData. If an OOM error occurs, leave an
** the error code in p. If an error has already occurred when this function
** is called, it is a no-op.
*/
static void sqlite3Fts5BufferSet(
int *pRc,
Fts5Buffer *pBuf,
int nData,
const u8 *pData
){
pBuf->n = 0;
sqlite3Fts5BufferAppendBlob(pRc, pBuf, nData, pData);
}
static int sqlite3Fts5PoslistNext64(
const u8 *a, int n, /* Buffer containing poslist */
int *pi, /* IN/OUT: Offset within a[] */
i64 *piOff /* IN/OUT: Current offset */
){
int i = *pi;
assert( a!=0 || i==0 );
if( i>=n ){
/* EOF */
*piOff = -1;
return 1;
}else{
i64 iOff = *piOff;
u32 iVal;
assert( a!=0 );
fts5FastGetVarint32(a, i, iVal);
if( iVal<=1 ){
if( iVal==0 ){
*pi = i;
return 0;
}
fts5FastGetVarint32(a, i, iVal);
iOff = ((i64)iVal) << 32;
assert( iOff>=0 );
fts5FastGetVarint32(a, i, iVal);
if( iVal<2 ){
/* This is a corrupt record. So stop parsing it here. */
*piOff = -1;
return 1;
}
*piOff = iOff + ((iVal-2) & 0x7FFFFFFF);
}else{
*piOff = (iOff & (i64)0x7FFFFFFF<<32)+((iOff + (iVal-2)) & 0x7FFFFFFF);
}
*pi = i;
assert_nc( *piOff>=iOff );
return 0;
}
}
/*
** Advance the iterator object passed as the only argument. Return true
** if the iterator reaches EOF, or false otherwise.
*/
static int sqlite3Fts5PoslistReaderNext(Fts5PoslistReader *pIter){
if( sqlite3Fts5PoslistNext64(pIter->a, pIter->n, &pIter->i, &pIter->iPos) ){
pIter->bEof = 1;
}
return pIter->bEof;
}
static int sqlite3Fts5PoslistReaderInit(
const u8 *a, int n, /* Poslist buffer to iterate through */
Fts5PoslistReader *pIter /* Iterator object to initialize */
){
memset(pIter, 0, sizeof(*pIter));
pIter->a = a;
pIter->n = n;
sqlite3Fts5PoslistReaderNext(pIter);
return pIter->bEof;
}
/*
** Append position iPos to the position list being accumulated in buffer
** pBuf, which must be already be large enough to hold the new data.
** The previous position written to this list is *piPrev. *piPrev is set
** to iPos before returning.
*/
static void sqlite3Fts5PoslistSafeAppend(
Fts5Buffer *pBuf,
i64 *piPrev,
i64 iPos
){
if( iPos>=*piPrev ){
static const i64 colmask = ((i64)(0x7FFFFFFF)) << 32;
if( (iPos & colmask) != (*piPrev & colmask) ){
pBuf->p[pBuf->n++] = 1;
pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], (iPos>>32));
*piPrev = (iPos & colmask);
}
pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], (iPos-*piPrev)+2);
*piPrev = iPos;
}
}
static int sqlite3Fts5PoslistWriterAppend(
Fts5Buffer *pBuf,
Fts5PoslistWriter *pWriter,
i64 iPos
){
int rc = 0; /* Initialized only to suppress erroneous warning from Clang */
if( fts5BufferGrow(&rc, pBuf, 5+5+5) ) return rc;
sqlite3Fts5PoslistSafeAppend(pBuf, &pWriter->iPrev, iPos);
return SQLITE_OK;
}
static void *sqlite3Fts5MallocZero(int *pRc, sqlite3_int64 nByte){
void *pRet = 0;
if( *pRc==SQLITE_OK ){
pRet = sqlite3_malloc64(nByte);
if( pRet==0 ){
if( nByte>0 ) *pRc = SQLITE_NOMEM;
}else{
memset(pRet, 0, (size_t)nByte);
}
}
return pRet;
}
/*
** Return a nul-terminated copy of the string indicated by pIn. If nIn
** is non-negative, then it is the length of the string in bytes. Otherwise,
** the length of the string is determined using strlen().
**
** It is the responsibility of the caller to eventually free the returned
** buffer using sqlite3_free(). If an OOM error occurs, NULL is returned.
*/
static char *sqlite3Fts5Strndup(int *pRc, const char *pIn, int nIn){
char *zRet = 0;
if( *pRc==SQLITE_OK ){
if( nIn<0 ){
nIn = (int)strlen(pIn);
}
zRet = (char*)sqlite3_malloc(nIn+1);
if( zRet ){
memcpy(zRet, pIn, nIn);
zRet[nIn] = '\0';
}else{
*pRc = SQLITE_NOMEM;
}
}
return zRet;
}
/*
** Return true if character 't' may be part of an FTS5 bareword, or false
** otherwise. Characters that may be part of barewords:
**
** * All non-ASCII characters,
** * The 52 upper and lower case ASCII characters, and
** * The 10 integer ASCII characters.
** * The underscore character "_" (0x5F).
** * The unicode "subsitute" character (0x1A).
*/
static int sqlite3Fts5IsBareword(char t){
u8 aBareword[128] = {
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x00 .. 0x0F */
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, /* 0x10 .. 0x1F */
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0x20 .. 0x2F */
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, /* 0x30 .. 0x3F */
0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x40 .. 0x4F */
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, /* 0x50 .. 0x5F */
0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 0x60 .. 0x6F */
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 /* 0x70 .. 0x7F */
};
return (t & 0x80) || aBareword[(int)t];
}
/*************************************************************************
*/
typedef struct Fts5TermsetEntry Fts5TermsetEntry;
struct Fts5TermsetEntry {
char *pTerm;
int nTerm;
int iIdx; /* Index (main or aPrefix[] entry) */
Fts5TermsetEntry *pNext;
};
struct Fts5Termset {
Fts5TermsetEntry *apHash[512];
};
static int sqlite3Fts5TermsetNew(Fts5Termset **pp){
int rc = SQLITE_OK;
*pp = sqlite3Fts5MallocZero(&rc, sizeof(Fts5Termset));
return rc;
}
static int sqlite3Fts5TermsetAdd(
Fts5Termset *p,
int iIdx,
const char *pTerm, int nTerm,
int *pbPresent
){
int rc = SQLITE_OK;
*pbPresent = 0;
if( p ){
int i;
u32 hash = 13;
Fts5TermsetEntry *pEntry;
/* Calculate a hash value for this term. This is the same hash checksum
** used by the fts5_hash.c module. This is not important for correct
** operation of the module, but is necessary to ensure that some tests
** designed to produce hash table collisions really do work. */
for(i=nTerm-1; i>=0; i--){
hash = (hash << 3) ^ hash ^ pTerm[i];
}
hash = (hash << 3) ^ hash ^ iIdx;
hash = hash % ArraySize(p->apHash);
for(pEntry=p->apHash[hash]; pEntry; pEntry=pEntry->pNext){
if( pEntry->iIdx==iIdx
&& pEntry->nTerm==nTerm
&& memcmp(pEntry->pTerm, pTerm, nTerm)==0
){
*pbPresent = 1;
break;
}
}
if( pEntry==0 ){
pEntry = sqlite3Fts5MallocZero(&rc, sizeof(Fts5TermsetEntry) + nTerm);
if( pEntry ){
pEntry->pTerm = (char*)&pEntry[1];
pEntry->nTerm = nTerm;
pEntry->iIdx = iIdx;
memcpy(pEntry->pTerm, pTerm, nTerm);
pEntry->pNext = p->apHash[hash];
p->apHash[hash] = pEntry;
}
}
}
return rc;
}
static void sqlite3Fts5TermsetFree(Fts5Termset *p){
if( p ){
u32 i;
for(i=0; iapHash); i++){
Fts5TermsetEntry *pEntry = p->apHash[i];
while( pEntry ){
Fts5TermsetEntry *pDel = pEntry;
pEntry = pEntry->pNext;
sqlite3_free(pDel);
}
}
sqlite3_free(p);
}
}
#line 1 "fts5_config.c"
/*
** 2014 Jun 09
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
** This is an SQLite module implementing full-text search.
*/
/* #include "fts5Int.h" */
#define FTS5_DEFAULT_PAGE_SIZE 4050
#define FTS5_DEFAULT_AUTOMERGE 4
#define FTS5_DEFAULT_USERMERGE 4
#define FTS5_DEFAULT_CRISISMERGE 16
#define FTS5_DEFAULT_HASHSIZE (1024*1024)
#define FTS5_DEFAULT_DELETE_AUTOMERGE 10 /* default 10% */
/* Maximum allowed page size */
#define FTS5_MAX_PAGE_SIZE (64*1024)
static int fts5_iswhitespace(char x){
return (x==' ');
}
static int fts5_isopenquote(char x){
return (x=='"' || x=='\'' || x=='[' || x=='`');
}
/*
** Argument pIn points to a character that is part of a nul-terminated
** string. Return a pointer to the first character following *pIn in
** the string that is not a white-space character.
*/
static const char *fts5ConfigSkipWhitespace(const char *pIn){
const char *p = pIn;
if( p ){
while( fts5_iswhitespace(*p) ){ p++; }
}
return p;
}
/*
** Argument pIn points to a character that is part of a nul-terminated
** string. Return a pointer to the first character following *pIn in
** the string that is not a "bareword" character.
*/
static const char *fts5ConfigSkipBareword(const char *pIn){
const char *p = pIn;
while ( sqlite3Fts5IsBareword(*p) ) p++;
if( p==pIn ) p = 0;
return p;
}
static int fts5_isdigit(char a){
return (a>='0' && a<='9');
}
static const char *fts5ConfigSkipLiteral(const char *pIn){
const char *p = pIn;
switch( *p ){
case 'n': case 'N':
if( sqlite3_strnicmp("null", p, 4)==0 ){
p = &p[4];
}else{
p = 0;
}
break;
case 'x': case 'X':
p++;
if( *p=='\'' ){
p++;
while( (*p>='a' && *p<='f')
|| (*p>='A' && *p<='F')
|| (*p>='0' && *p<='9')
){
p++;
}
if( *p=='\'' && 0==((p-pIn)%2) ){
p++;
}else{
p = 0;
}
}else{
p = 0;
}
break;
case '\'':
p++;
while( p ){
if( *p=='\'' ){
p++;
if( *p!='\'' ) break;
}
p++;
if( *p==0 ) p = 0;
}
break;
default:
/* maybe a number */
if( *p=='+' || *p=='-' ) p++;
while( fts5_isdigit(*p) ) p++;
/* At this point, if the literal was an integer, the parse is
** finished. Or, if it is a floating point value, it may continue
** with either a decimal point or an 'E' character. */
if( *p=='.' && fts5_isdigit(p[1]) ){
p += 2;
while( fts5_isdigit(*p) ) p++;
}
if( p==pIn ) p = 0;
break;
}
return p;
}
/*
** The first character of the string pointed to by argument z is guaranteed
** to be an open-quote character (see function fts5_isopenquote()).
**
** This function searches for the corresponding close-quote character within
** the string and, if found, dequotes the string in place and adds a new
** nul-terminator byte.
**
** If the close-quote is found, the value returned is the byte offset of
** the character immediately following it. Or, if the close-quote is not
** found, -1 is returned. If -1 is returned, the buffer is left in an
** undefined state.
*/
static int fts5Dequote(char *z){
char q;
int iIn = 1;
int iOut = 0;
q = z[0];
/* Set stack variable q to the close-quote character */
assert( q=='[' || q=='\'' || q=='"' || q=='`' );
if( q=='[' ) q = ']';
while( z[iIn] ){
if( z[iIn]==q ){
if( z[iIn+1]!=q ){
/* Character iIn was the close quote. */
iIn++;
break;
}else{
/* Character iIn and iIn+1 form an escaped quote character. Skip
** the input cursor past both and copy a single quote character
** to the output buffer. */
iIn += 2;
z[iOut++] = q;
}
}else{
z[iOut++] = z[iIn++];
}
}
z[iOut] = '\0';
return iIn;
}
/*
** Convert an SQL-style quoted string into a normal string by removing
** the quote characters. The conversion is done in-place. If the
** input does not begin with a quote character, then this routine
** is a no-op.
**
** Examples:
**
** "abc" becomes abc
** 'xyz' becomes xyz
** [pqr] becomes pqr
** `mno` becomes mno
*/
static void sqlite3Fts5Dequote(char *z){
char quote; /* Quote character (if any ) */
assert( 0==fts5_iswhitespace(z[0]) );
quote = z[0];
if( quote=='[' || quote=='\'' || quote=='"' || quote=='`' ){
fts5Dequote(z);
}
}
struct Fts5Enum {
const char *zName;
int eVal;
};
typedef struct Fts5Enum Fts5Enum;
static int fts5ConfigSetEnum(
const Fts5Enum *aEnum,
const char *zEnum,
int *peVal
){
int nEnum = (int)strlen(zEnum);
int i;
int iVal = -1;
for(i=0; aEnum[i].zName; i++){
if( sqlite3_strnicmp(aEnum[i].zName, zEnum, nEnum)==0 ){
if( iVal>=0 ) return SQLITE_ERROR;
iVal = aEnum[i].eVal;
}
}
*peVal = iVal;
return iVal<0 ? SQLITE_ERROR : SQLITE_OK;
}
/*
** Parse a "special" CREATE VIRTUAL TABLE directive and update
** configuration object pConfig as appropriate.
**
** If successful, object pConfig is updated and SQLITE_OK returned. If
** an error occurs, an SQLite error code is returned and an error message
** may be left in *pzErr. It is the responsibility of the caller to
** eventually free any such error message using sqlite3_free().
*/
static int fts5ConfigParseSpecial(
Fts5Global *pGlobal,
Fts5Config *pConfig, /* Configuration object to update */
const char *zCmd, /* Special command to parse */
const char *zArg, /* Argument to parse */
char **pzErr /* OUT: Error message */
){
int rc = SQLITE_OK;
int nCmd = (int)strlen(zCmd);
if( sqlite3_strnicmp("prefix", zCmd, nCmd)==0 ){
const int nByte = sizeof(int) * FTS5_MAX_PREFIX_INDEXES;
const char *p;
int bFirst = 1;
if( pConfig->aPrefix==0 ){
pConfig->aPrefix = sqlite3Fts5MallocZero(&rc, nByte);
if( rc ) return rc;
}
p = zArg;
while( 1 ){
int nPre = 0;
while( p[0]==' ' ) p++;
if( bFirst==0 && p[0]==',' ){
p++;
while( p[0]==' ' ) p++;
}else if( p[0]=='\0' ){
break;
}
if( p[0]<'0' || p[0]>'9' ){
*pzErr = sqlite3_mprintf("malformed prefix=... directive");
rc = SQLITE_ERROR;
break;
}
if( pConfig->nPrefix==FTS5_MAX_PREFIX_INDEXES ){
*pzErr = sqlite3_mprintf(
"too many prefix indexes (max %d)", FTS5_MAX_PREFIX_INDEXES
);
rc = SQLITE_ERROR;
break;
}
while( p[0]>='0' && p[0]<='9' && nPre<1000 ){
nPre = nPre*10 + (p[0] - '0');
p++;
}
if( nPre<=0 || nPre>=1000 ){
*pzErr = sqlite3_mprintf("prefix length out of range (max 999)");
rc = SQLITE_ERROR;
break;
}
pConfig->aPrefix[pConfig->nPrefix] = nPre;
pConfig->nPrefix++;
bFirst = 0;
}
assert( pConfig->nPrefix<=FTS5_MAX_PREFIX_INDEXES );
return rc;
}
if( sqlite3_strnicmp("tokenize", zCmd, nCmd)==0 ){
const char *p = (const char*)zArg;
sqlite3_int64 nArg = strlen(zArg) + 1;
char **azArg = sqlite3Fts5MallocZero(&rc, sizeof(char*) * nArg);
char *pDel = sqlite3Fts5MallocZero(&rc, nArg * 2);
char *pSpace = pDel;
if( azArg && pSpace ){
if( pConfig->pTok ){
*pzErr = sqlite3_mprintf("multiple tokenize=... directives");
rc = SQLITE_ERROR;
}else{
for(nArg=0; p && *p; nArg++){
const char *p2 = fts5ConfigSkipWhitespace(p);
if( *p2=='\'' ){
p = fts5ConfigSkipLiteral(p2);
}else{
p = fts5ConfigSkipBareword(p2);
}
if( p ){
memcpy(pSpace, p2, p-p2);
azArg[nArg] = pSpace;
sqlite3Fts5Dequote(pSpace);
pSpace += (p - p2) + 1;
p = fts5ConfigSkipWhitespace(p);
}
}
if( p==0 ){
*pzErr = sqlite3_mprintf("parse error in tokenize directive");
rc = SQLITE_ERROR;
}else{
rc = sqlite3Fts5GetTokenizer(pGlobal,
(const char**)azArg, (int)nArg, pConfig,
pzErr
);
}
}
}
sqlite3_free(azArg);
sqlite3_free(pDel);
return rc;
}
if( sqlite3_strnicmp("content", zCmd, nCmd)==0 ){
if( pConfig->eContent!=FTS5_CONTENT_NORMAL ){
*pzErr = sqlite3_mprintf("multiple content=... directives");
rc = SQLITE_ERROR;
}else{
if( zArg[0] ){
pConfig->eContent = FTS5_CONTENT_EXTERNAL;
pConfig->zContent = sqlite3Fts5Mprintf(&rc, "%Q.%Q", pConfig->zDb,zArg);
}else{
pConfig->eContent = FTS5_CONTENT_NONE;
}
}
return rc;
}
if( sqlite3_strnicmp("contentless_delete", zCmd, nCmd)==0 ){
if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){
*pzErr = sqlite3_mprintf("malformed contentless_delete=... directive");
rc = SQLITE_ERROR;
}else{
pConfig->bContentlessDelete = (zArg[0]=='1');
}
return rc;
}
if( sqlite3_strnicmp("content_rowid", zCmd, nCmd)==0 ){
if( pConfig->zContentRowid ){
*pzErr = sqlite3_mprintf("multiple content_rowid=... directives");
rc = SQLITE_ERROR;
}else{
pConfig->zContentRowid = sqlite3Fts5Strndup(&rc, zArg, -1);
}
return rc;
}
if( sqlite3_strnicmp("columnsize", zCmd, nCmd)==0 ){
if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){
*pzErr = sqlite3_mprintf("malformed columnsize=... directive");
rc = SQLITE_ERROR;
}else{
pConfig->bColumnsize = (zArg[0]=='1');
}
return rc;
}
if( sqlite3_strnicmp("detail", zCmd, nCmd)==0 ){
const Fts5Enum aDetail[] = {
{ "none", FTS5_DETAIL_NONE },
{ "full", FTS5_DETAIL_FULL },
{ "columns", FTS5_DETAIL_COLUMNS },
{ 0, 0 }
};
if( (rc = fts5ConfigSetEnum(aDetail, zArg, &pConfig->eDetail)) ){
*pzErr = sqlite3_mprintf("malformed detail=... directive");
}
return rc;
}
if( sqlite3_strnicmp("tokendata", zCmd, nCmd)==0 ){
if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){
*pzErr = sqlite3_mprintf("malformed tokendata=... directive");
rc = SQLITE_ERROR;
}else{
pConfig->bTokendata = (zArg[0]=='1');
}
return rc;
}
*pzErr = sqlite3_mprintf("unrecognized option: \"%.*s\"", nCmd, zCmd);
return SQLITE_ERROR;
}
/*
** Allocate an instance of the default tokenizer ("simple") at
** Fts5Config.pTokenizer. Return SQLITE_OK if successful, or an SQLite error
** code if an error occurs.
*/
static int fts5ConfigDefaultTokenizer(Fts5Global *pGlobal, Fts5Config *pConfig){
assert( pConfig->pTok==0 && pConfig->pTokApi==0 );
return sqlite3Fts5GetTokenizer(pGlobal, 0, 0, pConfig, 0);
}
/*
** Gobble up the first bareword or quoted word from the input buffer zIn.
** Return a pointer to the character immediately following the last in
** the gobbled word if successful, or a NULL pointer otherwise (failed
** to find close-quote character).
**
** Before returning, set pzOut to point to a new buffer containing a
** nul-terminated, dequoted copy of the gobbled word. If the word was
** quoted, *pbQuoted is also set to 1 before returning.
**
** If *pRc is other than SQLITE_OK when this function is called, it is
** a no-op (NULL is returned). Otherwise, if an OOM occurs within this
** function, *pRc is set to SQLITE_NOMEM before returning. *pRc is *not*
** set if a parse error (failed to find close quote) occurs.
*/
static const char *fts5ConfigGobbleWord(
int *pRc, /* IN/OUT: Error code */
const char *zIn, /* Buffer to gobble string/bareword from */
char **pzOut, /* OUT: malloc'd buffer containing str/bw */
int *pbQuoted /* OUT: Set to true if dequoting required */
){
const char *zRet = 0;
sqlite3_int64 nIn = strlen(zIn);
char *zOut = sqlite3_malloc64(nIn+1);
assert( *pRc==SQLITE_OK );
*pbQuoted = 0;
*pzOut = 0;
if( zOut==0 ){
*pRc = SQLITE_NOMEM;
}else{
memcpy(zOut, zIn, (size_t)(nIn+1));
if( fts5_isopenquote(zOut[0]) ){
int ii = fts5Dequote(zOut);
zRet = &zIn[ii];
*pbQuoted = 1;
}else{
zRet = fts5ConfigSkipBareword(zIn);
if( zRet ){
zOut[zRet-zIn] = '\0';
}
}
}
if( zRet==0 ){
sqlite3_free(zOut);
}else{
*pzOut = zOut;
}
return zRet;
}
static int fts5ConfigParseColumn(
Fts5Config *p,
char *zCol,
char *zArg,
char **pzErr
){
int rc = SQLITE_OK;
if( 0==sqlite3_stricmp(zCol, FTS5_RANK_NAME)
|| 0==sqlite3_stricmp(zCol, FTS5_ROWID_NAME)
){
*pzErr = sqlite3_mprintf("reserved fts5 column name: %s", zCol);
rc = SQLITE_ERROR;
}else if( zArg ){
if( 0==sqlite3_stricmp(zArg, "unindexed") ){
p->abUnindexed[p->nCol] = 1;
}else{
*pzErr = sqlite3_mprintf("unrecognized column option: %s", zArg);
rc = SQLITE_ERROR;
}
}
p->azCol[p->nCol++] = zCol;
return rc;
}
/*
** Populate the Fts5Config.zContentExprlist string.
*/
static int fts5ConfigMakeExprlist(Fts5Config *p){
int i;
int rc = SQLITE_OK;
Fts5Buffer buf = {0, 0, 0};
sqlite3Fts5BufferAppendPrintf(&rc, &buf, "T.%Q", p->zContentRowid);
if( p->eContent!=FTS5_CONTENT_NONE ){
for(i=0; inCol; i++){
if( p->eContent==FTS5_CONTENT_EXTERNAL ){
sqlite3Fts5BufferAppendPrintf(&rc, &buf, ", T.%Q", p->azCol[i]);
}else{
sqlite3Fts5BufferAppendPrintf(&rc, &buf, ", T.c%d", i);
}
}
}
assert( p->zContentExprlist==0 );
p->zContentExprlist = (char*)buf.p;
return rc;
}
/*
** Arguments nArg/azArg contain the string arguments passed to the xCreate
** or xConnect method of the virtual table. This function attempts to
** allocate an instance of Fts5Config containing the results of parsing
** those arguments.
**
** If successful, SQLITE_OK is returned and *ppOut is set to point to the
** new Fts5Config object. If an error occurs, an SQLite error code is
** returned, *ppOut is set to NULL and an error message may be left in
** *pzErr. It is the responsibility of the caller to eventually free any
** such error message using sqlite3_free().
*/
static int sqlite3Fts5ConfigParse(
Fts5Global *pGlobal,
sqlite3 *db,
int nArg, /* Number of arguments */
const char **azArg, /* Array of nArg CREATE VIRTUAL TABLE args */
Fts5Config **ppOut, /* OUT: Results of parse */
char **pzErr /* OUT: Error message */
){
int rc = SQLITE_OK; /* Return code */
Fts5Config *pRet; /* New object to return */
int i;
sqlite3_int64 nByte;
*ppOut = pRet = (Fts5Config*)sqlite3_malloc(sizeof(Fts5Config));
if( pRet==0 ) return SQLITE_NOMEM;
memset(pRet, 0, sizeof(Fts5Config));
pRet->db = db;
pRet->iCookie = -1;
nByte = nArg * (sizeof(char*) + sizeof(u8));
pRet->azCol = (char**)sqlite3Fts5MallocZero(&rc, nByte);
pRet->abUnindexed = pRet->azCol ? (u8*)&pRet->azCol[nArg] : 0;
pRet->zDb = sqlite3Fts5Strndup(&rc, azArg[1], -1);
pRet->zName = sqlite3Fts5Strndup(&rc, azArg[2], -1);
pRet->bColumnsize = 1;
pRet->eDetail = FTS5_DETAIL_FULL;
#ifdef SQLITE_DEBUG
pRet->bPrefixIndex = 1;
#endif
if( rc==SQLITE_OK && sqlite3_stricmp(pRet->zName, FTS5_RANK_NAME)==0 ){
*pzErr = sqlite3_mprintf("reserved fts5 table name: %s", pRet->zName);
rc = SQLITE_ERROR;
}
assert( (pRet->abUnindexed && pRet->azCol) || rc!=SQLITE_OK );
for(i=3; rc==SQLITE_OK && ibContentlessDelete
&& pRet->eContent!=FTS5_CONTENT_NONE
){
*pzErr = sqlite3_mprintf(
"contentless_delete=1 requires a contentless table"
);
rc = SQLITE_ERROR;
}
/* We only allow contentless_delete=1 if columnsize=0 is not present.
**
** This restriction may be removed at some point.
*/
if( rc==SQLITE_OK && pRet->bContentlessDelete && pRet->bColumnsize==0 ){
*pzErr = sqlite3_mprintf(
"contentless_delete=1 is incompatible with columnsize=0"
);
rc = SQLITE_ERROR;
}
/* If a tokenizer= option was successfully parsed, the tokenizer has
** already been allocated. Otherwise, allocate an instance of the default
** tokenizer (unicode61) now. */
if( rc==SQLITE_OK && pRet->pTok==0 ){
rc = fts5ConfigDefaultTokenizer(pGlobal, pRet);
}
/* If no zContent option was specified, fill in the default values. */
if( rc==SQLITE_OK && pRet->zContent==0 ){
const char *zTail = 0;
assert( pRet->eContent==FTS5_CONTENT_NORMAL
|| pRet->eContent==FTS5_CONTENT_NONE
);
if( pRet->eContent==FTS5_CONTENT_NORMAL ){
zTail = "content";
}else if( pRet->bColumnsize ){
zTail = "docsize";
}
if( zTail ){
pRet->zContent = sqlite3Fts5Mprintf(
&rc, "%Q.'%q_%s'", pRet->zDb, pRet->zName, zTail
);
}
}
if( rc==SQLITE_OK && pRet->zContentRowid==0 ){
pRet->zContentRowid = sqlite3Fts5Strndup(&rc, "rowid", -1);
}
/* Formulate the zContentExprlist text */
if( rc==SQLITE_OK ){
rc = fts5ConfigMakeExprlist(pRet);
}
if( rc!=SQLITE_OK ){
sqlite3Fts5ConfigFree(pRet);
*ppOut = 0;
}
return rc;
}
/*
** Free the configuration object passed as the only argument.
*/
static void sqlite3Fts5ConfigFree(Fts5Config *pConfig){
if( pConfig ){
int i;
if( pConfig->pTok ){
pConfig->pTokApi->xDelete(pConfig->pTok);
}
sqlite3_free(pConfig->zDb);
sqlite3_free(pConfig->zName);
for(i=0; inCol; i++){
sqlite3_free(pConfig->azCol[i]);
}
sqlite3_free(pConfig->azCol);
sqlite3_free(pConfig->aPrefix);
sqlite3_free(pConfig->zRank);
sqlite3_free(pConfig->zRankArgs);
sqlite3_free(pConfig->zContent);
sqlite3_free(pConfig->zContentRowid);
sqlite3_free(pConfig->zContentExprlist);
sqlite3_free(pConfig);
}
}
/*
** Call sqlite3_declare_vtab() based on the contents of the configuration
** object passed as the only argument. Return SQLITE_OK if successful, or
** an SQLite error code if an error occurs.
*/
static int sqlite3Fts5ConfigDeclareVtab(Fts5Config *pConfig){
int i;
int rc = SQLITE_OK;
char *zSql;
zSql = sqlite3Fts5Mprintf(&rc, "CREATE TABLE x(");
for(i=0; zSql && inCol; i++){
const char *zSep = (i==0?"":", ");
zSql = sqlite3Fts5Mprintf(&rc, "%z%s%Q", zSql, zSep, pConfig->azCol[i]);
}
zSql = sqlite3Fts5Mprintf(&rc, "%z, %Q HIDDEN, %s HIDDEN)",
zSql, pConfig->zName, FTS5_RANK_NAME
);
assert( zSql || rc==SQLITE_NOMEM );
if( zSql ){
rc = sqlite3_declare_vtab(pConfig->db, zSql);
sqlite3_free(zSql);
}
return rc;
}
/*
** Tokenize the text passed via the second and third arguments.
**
** The callback is invoked once for each token in the input text. The
** arguments passed to it are, in order:
**
** void *pCtx // Copy of 4th argument to sqlite3Fts5Tokenize()
** const char *pToken // Pointer to buffer containing token
** int nToken // Size of token in bytes
** int iStart // Byte offset of start of token within input text
** int iEnd // Byte offset of end of token within input text
** int iPos // Position of token in input (first token is 0)
**
** If the callback returns a non-zero value the tokenization is abandoned
** and no further callbacks are issued.
**
** This function returns SQLITE_OK if successful or an SQLite error code
** if an error occurs. If the tokenization was abandoned early because
** the callback returned SQLITE_DONE, this is not an error and this function
** still returns SQLITE_OK. Or, if the tokenization was abandoned early
** because the callback returned another non-zero value, it is assumed
** to be an SQLite error code and returned to the caller.
*/
static int sqlite3Fts5Tokenize(
Fts5Config *pConfig, /* FTS5 Configuration object */
int flags, /* FTS5_TOKENIZE_* flags */
const char *pText, int nText, /* Text to tokenize */
void *pCtx, /* Context passed to xToken() */
int (*xToken)(void*, int, const char*, int, int, int) /* Callback */
){
if( pText==0 ) return SQLITE_OK;
return pConfig->pTokApi->xTokenize(
pConfig->pTok, pCtx, flags, pText, nText, xToken
);
}
/*
** Argument pIn points to the first character in what is expected to be
** a comma-separated list of SQL literals followed by a ')' character.
** If it actually is this, return a pointer to the ')'. Otherwise, return
** NULL to indicate a parse error.
*/
static const char *fts5ConfigSkipArgs(const char *pIn){
const char *p = pIn;
while( 1 ){
p = fts5ConfigSkipWhitespace(p);
p = fts5ConfigSkipLiteral(p);
p = fts5ConfigSkipWhitespace(p);
if( p==0 || *p==')' ) break;
if( *p!=',' ){
p = 0;
break;
}
p++;
}
return p;
}
/*
** Parameter zIn contains a rank() function specification. The format of
** this is:
**
** + Bareword (function name)
** + Open parenthesis - "("
** + Zero or more SQL literals in a comma separated list
** + Close parenthesis - ")"
*/
static int sqlite3Fts5ConfigParseRank(
const char *zIn, /* Input string */
char **pzRank, /* OUT: Rank function name */
char **pzRankArgs /* OUT: Rank function arguments */
){
const char *p = zIn;
const char *pRank;
char *zRank = 0;
char *zRankArgs = 0;
int rc = SQLITE_OK;
*pzRank = 0;
*pzRankArgs = 0;
if( p==0 ){
rc = SQLITE_ERROR;
}else{
p = fts5ConfigSkipWhitespace(p);
pRank = p;
p = fts5ConfigSkipBareword(p);
if( p ){
zRank = sqlite3Fts5MallocZero(&rc, 1 + p - pRank);
if( zRank ) memcpy(zRank, pRank, p-pRank);
}else{
rc = SQLITE_ERROR;
}
if( rc==SQLITE_OK ){
p = fts5ConfigSkipWhitespace(p);
if( *p!='(' ) rc = SQLITE_ERROR;
p++;
}
if( rc==SQLITE_OK ){
const char *pArgs;
p = fts5ConfigSkipWhitespace(p);
pArgs = p;
if( *p!=')' ){
p = fts5ConfigSkipArgs(p);
if( p==0 ){
rc = SQLITE_ERROR;
}else{
zRankArgs = sqlite3Fts5MallocZero(&rc, 1 + p - pArgs);
if( zRankArgs ) memcpy(zRankArgs, pArgs, p-pArgs);
}
}
}
}
if( rc!=SQLITE_OK ){
sqlite3_free(zRank);
assert( zRankArgs==0 );
}else{
*pzRank = zRank;
*pzRankArgs = zRankArgs;
}
return rc;
}
static int sqlite3Fts5ConfigSetValue(
Fts5Config *pConfig,
const char *zKey,
sqlite3_value *pVal,
int *pbBadkey
){
int rc = SQLITE_OK;
if( 0==sqlite3_stricmp(zKey, "pgsz") ){
int pgsz = 0;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
pgsz = sqlite3_value_int(pVal);
}
if( pgsz<32 || pgsz>FTS5_MAX_PAGE_SIZE ){
*pbBadkey = 1;
}else{
pConfig->pgsz = pgsz;
}
}
else if( 0==sqlite3_stricmp(zKey, "hashsize") ){
int nHashSize = -1;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
nHashSize = sqlite3_value_int(pVal);
}
if( nHashSize<=0 ){
*pbBadkey = 1;
}else{
pConfig->nHashSize = nHashSize;
}
}
else if( 0==sqlite3_stricmp(zKey, "automerge") ){
int nAutomerge = -1;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
nAutomerge = sqlite3_value_int(pVal);
}
if( nAutomerge<0 || nAutomerge>64 ){
*pbBadkey = 1;
}else{
if( nAutomerge==1 ) nAutomerge = FTS5_DEFAULT_AUTOMERGE;
pConfig->nAutomerge = nAutomerge;
}
}
else if( 0==sqlite3_stricmp(zKey, "usermerge") ){
int nUsermerge = -1;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
nUsermerge = sqlite3_value_int(pVal);
}
if( nUsermerge<2 || nUsermerge>16 ){
*pbBadkey = 1;
}else{
pConfig->nUsermerge = nUsermerge;
}
}
else if( 0==sqlite3_stricmp(zKey, "crisismerge") ){
int nCrisisMerge = -1;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
nCrisisMerge = sqlite3_value_int(pVal);
}
if( nCrisisMerge<0 ){
*pbBadkey = 1;
}else{
if( nCrisisMerge<=1 ) nCrisisMerge = FTS5_DEFAULT_CRISISMERGE;
if( nCrisisMerge>=FTS5_MAX_SEGMENT ) nCrisisMerge = FTS5_MAX_SEGMENT-1;
pConfig->nCrisisMerge = nCrisisMerge;
}
}
else if( 0==sqlite3_stricmp(zKey, "deletemerge") ){
int nVal = -1;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
nVal = sqlite3_value_int(pVal);
}else{
*pbBadkey = 1;
}
if( nVal<0 ) nVal = FTS5_DEFAULT_DELETE_AUTOMERGE;
if( nVal>100 ) nVal = 0;
pConfig->nDeleteMerge = nVal;
}
else if( 0==sqlite3_stricmp(zKey, "rank") ){
const char *zIn = (const char*)sqlite3_value_text(pVal);
char *zRank;
char *zRankArgs;
rc = sqlite3Fts5ConfigParseRank(zIn, &zRank, &zRankArgs);
if( rc==SQLITE_OK ){
sqlite3_free(pConfig->zRank);
sqlite3_free(pConfig->zRankArgs);
pConfig->zRank = zRank;
pConfig->zRankArgs = zRankArgs;
}else if( rc==SQLITE_ERROR ){
rc = SQLITE_OK;
*pbBadkey = 1;
}
}
else if( 0==sqlite3_stricmp(zKey, "secure-delete") ){
int bVal = -1;
if( SQLITE_INTEGER==sqlite3_value_numeric_type(pVal) ){
bVal = sqlite3_value_int(pVal);
}
if( bVal<0 ){
*pbBadkey = 1;
}else{
pConfig->bSecureDelete = (bVal ? 1 : 0);
}
}else{
*pbBadkey = 1;
}
return rc;
}
/*
** Load the contents of the %_config table into memory.
*/
static int sqlite3Fts5ConfigLoad(Fts5Config *pConfig, int iCookie){
const char *zSelect = "SELECT k, v FROM %Q.'%q_config'";
char *zSql;
sqlite3_stmt *p = 0;
int rc = SQLITE_OK;
int iVersion = 0;
/* Set default values */
pConfig->pgsz = FTS5_DEFAULT_PAGE_SIZE;
pConfig->nAutomerge = FTS5_DEFAULT_AUTOMERGE;
pConfig->nUsermerge = FTS5_DEFAULT_USERMERGE;
pConfig->nCrisisMerge = FTS5_DEFAULT_CRISISMERGE;
pConfig->nHashSize = FTS5_DEFAULT_HASHSIZE;
pConfig->nDeleteMerge = FTS5_DEFAULT_DELETE_AUTOMERGE;
zSql = sqlite3Fts5Mprintf(&rc, zSelect, pConfig->zDb, pConfig->zName);
if( zSql ){
rc = sqlite3_prepare_v2(pConfig->db, zSql, -1, &p, 0);
sqlite3_free(zSql);
}
assert( rc==SQLITE_OK || p==0 );
if( rc==SQLITE_OK ){
while( SQLITE_ROW==sqlite3_step(p) ){
const char *zK = (const char*)sqlite3_column_text(p, 0);
sqlite3_value *pVal = sqlite3_column_value(p, 1);
if( 0==sqlite3_stricmp(zK, "version") ){
iVersion = sqlite3_value_int(pVal);
}else{
int bDummy = 0;
sqlite3Fts5ConfigSetValue(pConfig, zK, pVal, &bDummy);
}
}
rc = sqlite3_finalize(p);
}
if( rc==SQLITE_OK
&& iVersion!=FTS5_CURRENT_VERSION
&& iVersion!=FTS5_CURRENT_VERSION_SECUREDELETE
){
rc = SQLITE_ERROR;
if( pConfig->pzErrmsg ){
assert( 0==*pConfig->pzErrmsg );
*pConfig->pzErrmsg = sqlite3_mprintf("invalid fts5 file format "
"(found %d, expected %d or %d) - run 'rebuild'",
iVersion, FTS5_CURRENT_VERSION, FTS5_CURRENT_VERSION_SECUREDELETE
);
}
}else{
pConfig->iVersion = iVersion;
}
if( rc==SQLITE_OK ){
pConfig->iCookie = iCookie;
}
return rc;
}
#line 1 "fts5_expr.c"
/*
** 2014 May 31
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
*/
/* #include "fts5Int.h" */
/* #include "fts5parse.h" */
#ifndef SQLITE_FTS5_MAX_EXPR_DEPTH
# define SQLITE_FTS5_MAX_EXPR_DEPTH 256
#endif
/*
** All token types in the generated fts5parse.h file are greater than 0.
*/
#define FTS5_EOF 0
#define FTS5_LARGEST_INT64 (0xffffffff|(((i64)0x7fffffff)<<32))
typedef struct Fts5ExprTerm Fts5ExprTerm;
/*
** Functions generated by lemon from fts5parse.y.
*/
static void *sqlite3Fts5ParserAlloc(void *(*mallocProc)(u64));
static void sqlite3Fts5ParserFree(void*, void (*freeProc)(void*));
static void sqlite3Fts5Parser(void*, int, Fts5Token, Fts5Parse*);
#ifndef NDEBUG
#include
static void sqlite3Fts5ParserTrace(FILE*, char*);
#endif
static int sqlite3Fts5ParserFallback(int);
struct Fts5Expr {
Fts5Index *pIndex;
Fts5Config *pConfig;
Fts5ExprNode *pRoot;
int bDesc; /* Iterate in descending rowid order */
int nPhrase; /* Number of phrases in expression */
Fts5ExprPhrase **apExprPhrase; /* Pointers to phrase objects */
};
/*
** eType:
** Expression node type. Always one of:
**
** FTS5_AND (nChild, apChild valid)
** FTS5_OR (nChild, apChild valid)
** FTS5_NOT (nChild, apChild valid)
** FTS5_STRING (pNear valid)
** FTS5_TERM (pNear valid)
**
** iHeight:
** Distance from this node to furthest leaf. This is always 0 for nodes
** of type FTS5_STRING and FTS5_TERM. For all other nodes it is one
** greater than the largest child value.
*/
struct Fts5ExprNode {
int eType; /* Node type */
int bEof; /* True at EOF */
int bNomatch; /* True if entry is not a match */
int iHeight; /* Distance to tree leaf nodes */
/* Next method for this node. */
int (*xNext)(Fts5Expr*, Fts5ExprNode*, int, i64);
i64 iRowid; /* Current rowid */
Fts5ExprNearset *pNear; /* For FTS5_STRING - cluster of phrases */
/* Child nodes. For a NOT node, this array always contains 2 entries. For
** AND or OR nodes, it contains 2 or more entries. */
int nChild; /* Number of child nodes */
Fts5ExprNode *apChild[1]; /* Array of child nodes */
};
#define Fts5NodeIsString(p) ((p)->eType==FTS5_TERM || (p)->eType==FTS5_STRING)
/*
** Invoke the xNext method of an Fts5ExprNode object. This macro should be
** used as if it has the same signature as the xNext() methods themselves.
*/
#define fts5ExprNodeNext(a,b,c,d) (b)->xNext((a), (b), (c), (d))
/*
** An instance of the following structure represents a single search term
** or term prefix.
*/
struct Fts5ExprTerm {
u8 bPrefix; /* True for a prefix term */
u8 bFirst; /* True if token must be first in column */
char *pTerm; /* Term data */
int nQueryTerm; /* Effective size of term in bytes */
int nFullTerm; /* Size of term in bytes incl. tokendata */
Fts5IndexIter *pIter; /* Iterator for this term */
Fts5ExprTerm *pSynonym; /* Pointer to first in list of synonyms */
};
/*
** A phrase. One or more terms that must appear in a contiguous sequence
** within a document for it to match.
*/
struct Fts5ExprPhrase {
Fts5ExprNode *pNode; /* FTS5_STRING node this phrase is part of */
Fts5Buffer poslist; /* Current position list */
int nTerm; /* Number of entries in aTerm[] */
Fts5ExprTerm aTerm[1]; /* Terms that make up this phrase */
};
/*
** One or more phrases that must appear within a certain token distance of
** each other within each matching document.
*/
struct Fts5ExprNearset {
int nNear; /* NEAR parameter */
Fts5Colset *pColset; /* Columns to search (NULL -> all columns) */
int nPhrase; /* Number of entries in aPhrase[] array */
Fts5ExprPhrase *apPhrase[1]; /* Array of phrase pointers */
};
/*
** Parse context.
*/
struct Fts5Parse {
Fts5Config *pConfig;
char *zErr;
int rc;
int nPhrase; /* Size of apPhrase array */
Fts5ExprPhrase **apPhrase; /* Array of all phrases */
Fts5ExprNode *pExpr; /* Result of a successful parse */
int bPhraseToAnd; /* Convert "a+b" to "a AND b" */
};
/*
** Check that the Fts5ExprNode.iHeight variables are set correctly in
** the expression tree passed as the only argument.
*/
#ifndef NDEBUG
static void assert_expr_depth_ok(int rc, Fts5ExprNode *p){
if( rc==SQLITE_OK ){
if( p->eType==FTS5_TERM || p->eType==FTS5_STRING || p->eType==0 ){
assert( p->iHeight==0 );
}else{
int ii;
int iMaxChild = 0;
for(ii=0; iinChild; ii++){
Fts5ExprNode *pChild = p->apChild[ii];
iMaxChild = MAX(iMaxChild, pChild->iHeight);
assert_expr_depth_ok(SQLITE_OK, pChild);
}
assert( p->iHeight==iMaxChild+1 );
}
}
}
#else
# define assert_expr_depth_ok(rc, p)
#endif
static void sqlite3Fts5ParseError(Fts5Parse *pParse, const char *zFmt, ...){
va_list ap;
va_start(ap, zFmt);
if( pParse->rc==SQLITE_OK ){
assert( pParse->zErr==0 );
pParse->zErr = sqlite3_vmprintf(zFmt, ap);
pParse->rc = SQLITE_ERROR;
}
va_end(ap);
}
static int fts5ExprIsspace(char t){
return t==' ' || t=='\t' || t=='\n' || t=='\r';
}
/*
** Read the first token from the nul-terminated string at *pz.
*/
static int fts5ExprGetToken(
Fts5Parse *pParse,
const char **pz, /* IN/OUT: Pointer into buffer */
Fts5Token *pToken
){
const char *z = *pz;
int tok;
/* Skip past any whitespace */
while( fts5ExprIsspace(*z) ) z++;
pToken->p = z;
pToken->n = 1;
switch( *z ){
case '(': tok = FTS5_LP; break;
case ')': tok = FTS5_RP; break;
case '{': tok = FTS5_LCP; break;
case '}': tok = FTS5_RCP; break;
case ':': tok = FTS5_COLON; break;
case ',': tok = FTS5_COMMA; break;
case '+': tok = FTS5_PLUS; break;
case '*': tok = FTS5_STAR; break;
case '-': tok = FTS5_MINUS; break;
case '^': tok = FTS5_CARET; break;
case '\0': tok = FTS5_EOF; break;
case '"': {
const char *z2;
tok = FTS5_STRING;
for(z2=&z[1]; 1; z2++){
if( z2[0]=='"' ){
z2++;
if( z2[0]!='"' ) break;
}
if( z2[0]=='\0' ){
sqlite3Fts5ParseError(pParse, "unterminated string");
return FTS5_EOF;
}
}
pToken->n = (z2 - z);
break;
}
default: {
const char *z2;
if( sqlite3Fts5IsBareword(z[0])==0 ){
sqlite3Fts5ParseError(pParse, "fts5: syntax error near \"%.1s\"", z);
return FTS5_EOF;
}
tok = FTS5_STRING;
for(z2=&z[1]; sqlite3Fts5IsBareword(*z2); z2++);
pToken->n = (z2 - z);
if( pToken->n==2 && memcmp(pToken->p, "OR", 2)==0 ) tok = FTS5_OR;
if( pToken->n==3 && memcmp(pToken->p, "NOT", 3)==0 ) tok = FTS5_NOT;
if( pToken->n==3 && memcmp(pToken->p, "AND", 3)==0 ) tok = FTS5_AND;
break;
}
}
*pz = &pToken->p[pToken->n];
return tok;
}
static void *fts5ParseAlloc(u64 t){ return sqlite3_malloc64((sqlite3_int64)t);}
static void fts5ParseFree(void *p){ sqlite3_free(p); }
static int sqlite3Fts5ExprNew(
Fts5Config *pConfig, /* FTS5 Configuration */
int bPhraseToAnd,
int iCol,
const char *zExpr, /* Expression text */
Fts5Expr **ppNew,
char **pzErr
){
Fts5Parse sParse;
Fts5Token token;
const char *z = zExpr;
int t; /* Next token type */
void *pEngine;
Fts5Expr *pNew;
*ppNew = 0;
*pzErr = 0;
memset(&sParse, 0, sizeof(sParse));
sParse.bPhraseToAnd = bPhraseToAnd;
pEngine = sqlite3Fts5ParserAlloc(fts5ParseAlloc);
if( pEngine==0 ){ return SQLITE_NOMEM; }
sParse.pConfig = pConfig;
do {
t = fts5ExprGetToken(&sParse, &z, &token);
sqlite3Fts5Parser(pEngine, t, token, &sParse);
}while( sParse.rc==SQLITE_OK && t!=FTS5_EOF );
sqlite3Fts5ParserFree(pEngine, fts5ParseFree);
assert_expr_depth_ok(sParse.rc, sParse.pExpr);
/* If the LHS of the MATCH expression was a user column, apply the
** implicit column-filter. */
if( iColnCol && sParse.pExpr && sParse.rc==SQLITE_OK ){
int n = sizeof(Fts5Colset);
Fts5Colset *pColset = (Fts5Colset*)sqlite3Fts5MallocZero(&sParse.rc, n);
if( pColset ){
pColset->nCol = 1;
pColset->aiCol[0] = iCol;
sqlite3Fts5ParseSetColset(&sParse, sParse.pExpr, pColset);
}
}
assert( sParse.rc!=SQLITE_OK || sParse.zErr==0 );
if( sParse.rc==SQLITE_OK ){
*ppNew = pNew = sqlite3_malloc(sizeof(Fts5Expr));
if( pNew==0 ){
sParse.rc = SQLITE_NOMEM;
sqlite3Fts5ParseNodeFree(sParse.pExpr);
}else{
if( !sParse.pExpr ){
const int nByte = sizeof(Fts5ExprNode);
pNew->pRoot = (Fts5ExprNode*)sqlite3Fts5MallocZero(&sParse.rc, nByte);
if( pNew->pRoot ){
pNew->pRoot->bEof = 1;
}
}else{
pNew->pRoot = sParse.pExpr;
}
pNew->pIndex = 0;
pNew->pConfig = pConfig;
pNew->apExprPhrase = sParse.apPhrase;
pNew->nPhrase = sParse.nPhrase;
pNew->bDesc = 0;
sParse.apPhrase = 0;
}
}else{
sqlite3Fts5ParseNodeFree(sParse.pExpr);
}
sqlite3_free(sParse.apPhrase);
*pzErr = sParse.zErr;
return sParse.rc;
}
/*
** Assuming that buffer z is at least nByte bytes in size and contains a
** valid utf-8 string, return the number of characters in the string.
*/
static int fts5ExprCountChar(const char *z, int nByte){
int nRet = 0;
int ii;
for(ii=0; ii=3 ){
int jj;
zExpr[iOut++] = '"';
for(jj=iFirst; jj0 ){
int bAnd = 0;
if( pConfig->eDetail!=FTS5_DETAIL_FULL ){
bAnd = 1;
if( pConfig->eDetail==FTS5_DETAIL_NONE ){
iCol = pConfig->nCol;
}
}
zExpr[iOut] = '\0';
rc = sqlite3Fts5ExprNew(pConfig, bAnd, iCol, zExpr, pp,pConfig->pzErrmsg);
}else{
*pp = 0;
}
sqlite3_free(zExpr);
}
return rc;
}
/*
** Free the expression node object passed as the only argument.
*/
static void sqlite3Fts5ParseNodeFree(Fts5ExprNode *p){
if( p ){
int i;
for(i=0; inChild; i++){
sqlite3Fts5ParseNodeFree(p->apChild[i]);
}
sqlite3Fts5ParseNearsetFree(p->pNear);
sqlite3_free(p);
}
}
/*
** Free the expression object passed as the only argument.
*/
static void sqlite3Fts5ExprFree(Fts5Expr *p){
if( p ){
sqlite3Fts5ParseNodeFree(p->pRoot);
sqlite3_free(p->apExprPhrase);
sqlite3_free(p);
}
}
static int sqlite3Fts5ExprAnd(Fts5Expr **pp1, Fts5Expr *p2){
Fts5Parse sParse;
memset(&sParse, 0, sizeof(sParse));
if( *pp1 && p2 ){
Fts5Expr *p1 = *pp1;
int nPhrase = p1->nPhrase + p2->nPhrase;
p1->pRoot = sqlite3Fts5ParseNode(&sParse, FTS5_AND, p1->pRoot, p2->pRoot,0);
p2->pRoot = 0;
if( sParse.rc==SQLITE_OK ){
Fts5ExprPhrase **ap = (Fts5ExprPhrase**)sqlite3_realloc(
p1->apExprPhrase, nPhrase * sizeof(Fts5ExprPhrase*)
);
if( ap==0 ){
sParse.rc = SQLITE_NOMEM;
}else{
int i;
memmove(&ap[p2->nPhrase], ap, p1->nPhrase*sizeof(Fts5ExprPhrase*));
for(i=0; inPhrase; i++){
ap[i] = p2->apExprPhrase[i];
}
p1->nPhrase = nPhrase;
p1->apExprPhrase = ap;
}
}
sqlite3_free(p2->apExprPhrase);
sqlite3_free(p2);
}else if( p2 ){
*pp1 = p2;
}
return sParse.rc;
}
/*
** Argument pTerm must be a synonym iterator. Return the current rowid
** that it points to.
*/
static i64 fts5ExprSynonymRowid(Fts5ExprTerm *pTerm, int bDesc, int *pbEof){
i64 iRet = 0;
int bRetValid = 0;
Fts5ExprTerm *p;
assert( pTerm );
assert( pTerm->pSynonym );
assert( bDesc==0 || bDesc==1 );
for(p=pTerm; p; p=p->pSynonym){
if( 0==sqlite3Fts5IterEof(p->pIter) ){
i64 iRowid = p->pIter->iRowid;
if( bRetValid==0 || (bDesc!=(iRowidpSynonym );
for(p=pTerm; p; p=p->pSynonym){
Fts5IndexIter *pIter = p->pIter;
if( sqlite3Fts5IterEof(pIter)==0 && pIter->iRowid==iRowid ){
if( pIter->nData==0 ) continue;
if( nIter==nAlloc ){
sqlite3_int64 nByte = sizeof(Fts5PoslistReader) * nAlloc * 2;
Fts5PoslistReader *aNew = (Fts5PoslistReader*)sqlite3_malloc64(nByte);
if( aNew==0 ){
rc = SQLITE_NOMEM;
goto synonym_poslist_out;
}
memcpy(aNew, aIter, sizeof(Fts5PoslistReader) * nIter);
nAlloc = nAlloc*2;
if( aIter!=aStatic ) sqlite3_free(aIter);
aIter = aNew;
}
sqlite3Fts5PoslistReaderInit(pIter->pData, pIter->nData, &aIter[nIter]);
assert( aIter[nIter].bEof==0 );
nIter++;
}
}
if( nIter==1 ){
*pa = (u8*)aIter[0].a;
*pn = aIter[0].n;
}else{
Fts5PoslistWriter writer = {0};
i64 iPrev = -1;
fts5BufferZero(pBuf);
while( 1 ){
int i;
i64 iMin = FTS5_LARGEST_INT64;
for(i=0; ip;
*pn = pBuf->n;
}
}
synonym_poslist_out:
if( aIter!=aStatic ) sqlite3_free(aIter);
return rc;
}
/*
** All individual term iterators in pPhrase are guaranteed to be valid and
** pointing to the same rowid when this function is called. This function
** checks if the current rowid really is a match, and if so populates
** the pPhrase->poslist buffer accordingly. Output parameter *pbMatch
** is set to true if this is really a match, or false otherwise.
**
** SQLITE_OK is returned if an error occurs, or an SQLite error code
** otherwise. It is not considered an error code if the current rowid is
** not a match.
*/
static int fts5ExprPhraseIsMatch(
Fts5ExprNode *pNode, /* Node pPhrase belongs to */
Fts5ExprPhrase *pPhrase, /* Phrase object to initialize */
int *pbMatch /* OUT: Set to true if really a match */
){
Fts5PoslistWriter writer = {0};
Fts5PoslistReader aStatic[4];
Fts5PoslistReader *aIter = aStatic;
int i;
int rc = SQLITE_OK;
int bFirst = pPhrase->aTerm[0].bFirst;
fts5BufferZero(&pPhrase->poslist);
/* If the aStatic[] array is not large enough, allocate a large array
** using sqlite3_malloc(). This approach could be improved upon. */
if( pPhrase->nTerm>ArraySize(aStatic) ){
sqlite3_int64 nByte = sizeof(Fts5PoslistReader) * pPhrase->nTerm;
aIter = (Fts5PoslistReader*)sqlite3_malloc64(nByte);
if( !aIter ) return SQLITE_NOMEM;
}
memset(aIter, 0, sizeof(Fts5PoslistReader) * pPhrase->nTerm);
/* Initialize a term iterator for each term in the phrase */
for(i=0; inTerm; i++){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[i];
int n = 0;
int bFlag = 0;
u8 *a = 0;
if( pTerm->pSynonym ){
Fts5Buffer buf = {0, 0, 0};
rc = fts5ExprSynonymList(pTerm, pNode->iRowid, &buf, &a, &n);
if( rc ){
sqlite3_free(a);
goto ismatch_out;
}
if( a==buf.p ) bFlag = 1;
}else{
a = (u8*)pTerm->pIter->pData;
n = pTerm->pIter->nData;
}
sqlite3Fts5PoslistReaderInit(a, n, &aIter[i]);
aIter[i].bFlag = (u8)bFlag;
if( aIter[i].bEof ) goto ismatch_out;
}
while( 1 ){
int bMatch;
i64 iPos = aIter[0].iPos;
do {
bMatch = 1;
for(i=0; inTerm; i++){
Fts5PoslistReader *pPos = &aIter[i];
i64 iAdj = iPos + i;
if( pPos->iPos!=iAdj ){
bMatch = 0;
while( pPos->iPosiPos>iAdj ) iPos = pPos->iPos-i;
}
}
}while( bMatch==0 );
/* Append position iPos to the output */
if( bFirst==0 || FTS5_POS2OFFSET(iPos)==0 ){
rc = sqlite3Fts5PoslistWriterAppend(&pPhrase->poslist, &writer, iPos);
if( rc!=SQLITE_OK ) goto ismatch_out;
}
for(i=0; inTerm; i++){
if( sqlite3Fts5PoslistReaderNext(&aIter[i]) ) goto ismatch_out;
}
}
ismatch_out:
*pbMatch = (pPhrase->poslist.n>0);
for(i=0; inTerm; i++){
if( aIter[i].bFlag ) sqlite3_free((u8*)aIter[i].a);
}
if( aIter!=aStatic ) sqlite3_free(aIter);
return rc;
}
typedef struct Fts5LookaheadReader Fts5LookaheadReader;
struct Fts5LookaheadReader {
const u8 *a; /* Buffer containing position list */
int n; /* Size of buffer a[] in bytes */
int i; /* Current offset in position list */
i64 iPos; /* Current position */
i64 iLookahead; /* Next position */
};
#define FTS5_LOOKAHEAD_EOF (((i64)1) << 62)
static int fts5LookaheadReaderNext(Fts5LookaheadReader *p){
p->iPos = p->iLookahead;
if( sqlite3Fts5PoslistNext64(p->a, p->n, &p->i, &p->iLookahead) ){
p->iLookahead = FTS5_LOOKAHEAD_EOF;
}
return (p->iPos==FTS5_LOOKAHEAD_EOF);
}
static int fts5LookaheadReaderInit(
const u8 *a, int n, /* Buffer to read position list from */
Fts5LookaheadReader *p /* Iterator object to initialize */
){
memset(p, 0, sizeof(Fts5LookaheadReader));
p->a = a;
p->n = n;
fts5LookaheadReaderNext(p);
return fts5LookaheadReaderNext(p);
}
typedef struct Fts5NearTrimmer Fts5NearTrimmer;
struct Fts5NearTrimmer {
Fts5LookaheadReader reader; /* Input iterator */
Fts5PoslistWriter writer; /* Writer context */
Fts5Buffer *pOut; /* Output poslist */
};
/*
** The near-set object passed as the first argument contains more than
** one phrase. All phrases currently point to the same row. The
** Fts5ExprPhrase.poslist buffers are populated accordingly. This function
** tests if the current row contains instances of each phrase sufficiently
** close together to meet the NEAR constraint. Non-zero is returned if it
** does, or zero otherwise.
**
** If in/out parameter (*pRc) is set to other than SQLITE_OK when this
** function is called, it is a no-op. Or, if an error (e.g. SQLITE_NOMEM)
** occurs within this function (*pRc) is set accordingly before returning.
** The return value is undefined in both these cases.
**
** If no error occurs and non-zero (a match) is returned, the position-list
** of each phrase object is edited to contain only those entries that
** meet the constraint before returning.
*/
static int fts5ExprNearIsMatch(int *pRc, Fts5ExprNearset *pNear){
Fts5NearTrimmer aStatic[4];
Fts5NearTrimmer *a = aStatic;
Fts5ExprPhrase **apPhrase = pNear->apPhrase;
int i;
int rc = *pRc;
int bMatch;
assert( pNear->nPhrase>1 );
/* If the aStatic[] array is not large enough, allocate a large array
** using sqlite3_malloc(). This approach could be improved upon. */
if( pNear->nPhrase>ArraySize(aStatic) ){
sqlite3_int64 nByte = sizeof(Fts5NearTrimmer) * pNear->nPhrase;
a = (Fts5NearTrimmer*)sqlite3Fts5MallocZero(&rc, nByte);
}else{
memset(aStatic, 0, sizeof(aStatic));
}
if( rc!=SQLITE_OK ){
*pRc = rc;
return 0;
}
/* Initialize a lookahead iterator for each phrase. After passing the
** buffer and buffer size to the lookaside-reader init function, zero
** the phrase poslist buffer. The new poslist for the phrase (containing
** the same entries as the original with some entries removed on account
** of the NEAR constraint) is written over the original even as it is
** being read. This is safe as the entries for the new poslist are a
** subset of the old, so it is not possible for data yet to be read to
** be overwritten. */
for(i=0; inPhrase; i++){
Fts5Buffer *pPoslist = &apPhrase[i]->poslist;
fts5LookaheadReaderInit(pPoslist->p, pPoslist->n, &a[i].reader);
pPoslist->n = 0;
a[i].pOut = pPoslist;
}
while( 1 ){
int iAdv;
i64 iMin;
i64 iMax;
/* This block advances the phrase iterators until they point to a set of
** entries that together comprise a match. */
iMax = a[0].reader.iPos;
do {
bMatch = 1;
for(i=0; inPhrase; i++){
Fts5LookaheadReader *pPos = &a[i].reader;
iMin = iMax - pNear->apPhrase[i]->nTerm - pNear->nNear;
if( pPos->iPosiPos>iMax ){
bMatch = 0;
while( pPos->iPosiPos>iMax ) iMax = pPos->iPos;
}
}
}while( bMatch==0 );
/* Add an entry to each output position list */
for(i=0; inPhrase; i++){
i64 iPos = a[i].reader.iPos;
Fts5PoslistWriter *pWriter = &a[i].writer;
if( a[i].pOut->n==0 || iPos!=pWriter->iPrev ){
sqlite3Fts5PoslistWriterAppend(a[i].pOut, pWriter, iPos);
}
}
iAdv = 0;
iMin = a[0].reader.iLookahead;
for(i=0; inPhrase; i++){
if( a[i].reader.iLookahead < iMin ){
iMin = a[i].reader.iLookahead;
iAdv = i;
}
}
if( fts5LookaheadReaderNext(&a[iAdv].reader) ) goto ismatch_out;
}
ismatch_out: {
int bRet = a[0].pOut->n>0;
*pRc = rc;
if( a!=aStatic ) sqlite3_free(a);
return bRet;
}
}
/*
** Advance iterator pIter until it points to a value equal to or laster
** than the initial value of *piLast. If this means the iterator points
** to a value laster than *piLast, update *piLast to the new lastest value.
**
** If the iterator reaches EOF, set *pbEof to true before returning. If
** an error occurs, set *pRc to an error code. If either *pbEof or *pRc
** are set, return a non-zero value. Otherwise, return zero.
*/
static int fts5ExprAdvanceto(
Fts5IndexIter *pIter, /* Iterator to advance */
int bDesc, /* True if iterator is "rowid DESC" */
i64 *piLast, /* IN/OUT: Lastest rowid seen so far */
int *pRc, /* OUT: Error code */
int *pbEof /* OUT: Set to true if EOF */
){
i64 iLast = *piLast;
i64 iRowid;
iRowid = pIter->iRowid;
if( (bDesc==0 && iLast>iRowid) || (bDesc && iLastiRowid;
assert( (bDesc==0 && iRowid>=iLast) || (bDesc==1 && iRowid<=iLast) );
}
*piLast = iRowid;
return 0;
}
static int fts5ExprSynonymAdvanceto(
Fts5ExprTerm *pTerm, /* Term iterator to advance */
int bDesc, /* True if iterator is "rowid DESC" */
i64 *piLast, /* IN/OUT: Lastest rowid seen so far */
int *pRc /* OUT: Error code */
){
int rc = SQLITE_OK;
i64 iLast = *piLast;
Fts5ExprTerm *p;
int bEof = 0;
for(p=pTerm; rc==SQLITE_OK && p; p=p->pSynonym){
if( sqlite3Fts5IterEof(p->pIter)==0 ){
i64 iRowid = p->pIter->iRowid;
if( (bDesc==0 && iLast>iRowid) || (bDesc && iLastpIter, iLast);
}
}
}
if( rc!=SQLITE_OK ){
*pRc = rc;
bEof = 1;
}else{
*piLast = fts5ExprSynonymRowid(pTerm, bDesc, &bEof);
}
return bEof;
}
static int fts5ExprNearTest(
int *pRc,
Fts5Expr *pExpr, /* Expression that pNear is a part of */
Fts5ExprNode *pNode /* The "NEAR" node (FTS5_STRING) */
){
Fts5ExprNearset *pNear = pNode->pNear;
int rc = *pRc;
if( pExpr->pConfig->eDetail!=FTS5_DETAIL_FULL ){
Fts5ExprTerm *pTerm;
Fts5ExprPhrase *pPhrase = pNear->apPhrase[0];
pPhrase->poslist.n = 0;
for(pTerm=&pPhrase->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){
Fts5IndexIter *pIter = pTerm->pIter;
if( sqlite3Fts5IterEof(pIter)==0 ){
if( pIter->iRowid==pNode->iRowid && pIter->nData>0 ){
pPhrase->poslist.n = 1;
}
}
}
return pPhrase->poslist.n;
}else{
int i;
/* Check that each phrase in the nearset matches the current row.
** Populate the pPhrase->poslist buffers at the same time. If any
** phrase is not a match, break out of the loop early. */
for(i=0; rc==SQLITE_OK && inPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
if( pPhrase->nTerm>1 || pPhrase->aTerm[0].pSynonym
|| pNear->pColset || pPhrase->aTerm[0].bFirst
){
int bMatch = 0;
rc = fts5ExprPhraseIsMatch(pNode, pPhrase, &bMatch);
if( bMatch==0 ) break;
}else{
Fts5IndexIter *pIter = pPhrase->aTerm[0].pIter;
fts5BufferSet(&rc, &pPhrase->poslist, pIter->nData, pIter->pData);
}
}
*pRc = rc;
if( i==pNear->nPhrase && (i==1 || fts5ExprNearIsMatch(pRc, pNear)) ){
return 1;
}
return 0;
}
}
/*
** Initialize all term iterators in the pNear object. If any term is found
** to match no documents at all, return immediately without initializing any
** further iterators.
**
** If an error occurs, return an SQLite error code. Otherwise, return
** SQLITE_OK. It is not considered an error if some term matches zero
** documents.
*/
static int fts5ExprNearInitAll(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
Fts5ExprNearset *pNear = pNode->pNear;
int i;
assert( pNode->bNomatch==0 );
for(i=0; inPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
if( pPhrase->nTerm==0 ){
pNode->bEof = 1;
return SQLITE_OK;
}else{
int j;
for(j=0; jnTerm; j++){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[j];
Fts5ExprTerm *p;
int bHit = 0;
for(p=pTerm; p; p=p->pSynonym){
int rc;
if( p->pIter ){
sqlite3Fts5IterClose(p->pIter);
p->pIter = 0;
}
rc = sqlite3Fts5IndexQuery(
pExpr->pIndex, p->pTerm, p->nQueryTerm,
(pTerm->bPrefix ? FTS5INDEX_QUERY_PREFIX : 0) |
(pExpr->bDesc ? FTS5INDEX_QUERY_DESC : 0),
pNear->pColset,
&p->pIter
);
assert( (rc==SQLITE_OK)==(p->pIter!=0) );
if( rc!=SQLITE_OK ) return rc;
if( 0==sqlite3Fts5IterEof(p->pIter) ){
bHit = 1;
}
}
if( bHit==0 ){
pNode->bEof = 1;
return SQLITE_OK;
}
}
}
}
pNode->bEof = 0;
return SQLITE_OK;
}
/*
** If pExpr is an ASC iterator, this function returns a value with the
** same sign as:
**
** (iLhs - iRhs)
**
** Otherwise, if this is a DESC iterator, the opposite is returned:
**
** (iRhs - iLhs)
*/
static int fts5RowidCmp(
Fts5Expr *pExpr,
i64 iLhs,
i64 iRhs
){
assert( pExpr->bDesc==0 || pExpr->bDesc==1 );
if( pExpr->bDesc==0 ){
if( iLhs iRhs);
}else{
if( iLhs>iRhs ) return -1;
return (iLhs < iRhs);
}
}
static void fts5ExprSetEof(Fts5ExprNode *pNode){
int i;
pNode->bEof = 1;
pNode->bNomatch = 0;
for(i=0; inChild; i++){
fts5ExprSetEof(pNode->apChild[i]);
}
}
static void fts5ExprNodeZeroPoslist(Fts5ExprNode *pNode){
if( pNode->eType==FTS5_STRING || pNode->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pNode->pNear;
int i;
for(i=0; inPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
pPhrase->poslist.n = 0;
}
}else{
int i;
for(i=0; inChild; i++){
fts5ExprNodeZeroPoslist(pNode->apChild[i]);
}
}
}
/*
** Compare the values currently indicated by the two nodes as follows:
**
** res = (*p1) - (*p2)
**
** Nodes that point to values that come later in the iteration order are
** considered to be larger. Nodes at EOF are the largest of all.
**
** This means that if the iteration order is ASC, then numerically larger
** rowids are considered larger. Or if it is the default DESC, numerically
** smaller rowids are larger.
*/
static int fts5NodeCompare(
Fts5Expr *pExpr,
Fts5ExprNode *p1,
Fts5ExprNode *p2
){
if( p2->bEof ) return -1;
if( p1->bEof ) return +1;
return fts5RowidCmp(pExpr, p1->iRowid, p2->iRowid);
}
/*
** All individual term iterators in pNear are guaranteed to be valid when
** this function is called. This function checks if all term iterators
** point to the same rowid, and if not, advances them until they do.
** If an EOF is reached before this happens, *pbEof is set to true before
** returning.
**
** SQLITE_OK is returned if an error occurs, or an SQLite error code
** otherwise. It is not considered an error code if an iterator reaches
** EOF.
*/
static int fts5ExprNodeTest_STRING(
Fts5Expr *pExpr, /* Expression pPhrase belongs to */
Fts5ExprNode *pNode
){
Fts5ExprNearset *pNear = pNode->pNear;
Fts5ExprPhrase *pLeft = pNear->apPhrase[0];
int rc = SQLITE_OK;
i64 iLast; /* Lastest rowid any iterator points to */
int i, j; /* Phrase and token index, respectively */
int bMatch; /* True if all terms are at the same rowid */
const int bDesc = pExpr->bDesc;
/* Check that this node should not be FTS5_TERM */
assert( pNear->nPhrase>1
|| pNear->apPhrase[0]->nTerm>1
|| pNear->apPhrase[0]->aTerm[0].pSynonym
|| pNear->apPhrase[0]->aTerm[0].bFirst
);
/* Initialize iLast, the "lastest" rowid any iterator points to. If the
** iterator skips through rowids in the default ascending order, this means
** the maximum rowid. Or, if the iterator is "ORDER BY rowid DESC", then it
** means the minimum rowid. */
if( pLeft->aTerm[0].pSynonym ){
iLast = fts5ExprSynonymRowid(&pLeft->aTerm[0], bDesc, 0);
}else{
iLast = pLeft->aTerm[0].pIter->iRowid;
}
do {
bMatch = 1;
for(i=0; inPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
for(j=0; jnTerm; j++){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[j];
if( pTerm->pSynonym ){
i64 iRowid = fts5ExprSynonymRowid(pTerm, bDesc, 0);
if( iRowid==iLast ) continue;
bMatch = 0;
if( fts5ExprSynonymAdvanceto(pTerm, bDesc, &iLast, &rc) ){
pNode->bNomatch = 0;
pNode->bEof = 1;
return rc;
}
}else{
Fts5IndexIter *pIter = pPhrase->aTerm[j].pIter;
if( pIter->iRowid==iLast || pIter->bEof ) continue;
bMatch = 0;
if( fts5ExprAdvanceto(pIter, bDesc, &iLast, &rc, &pNode->bEof) ){
return rc;
}
}
}
}
}while( bMatch==0 );
pNode->iRowid = iLast;
pNode->bNomatch = ((0==fts5ExprNearTest(&rc, pExpr, pNode)) && rc==SQLITE_OK);
assert( pNode->bEof==0 || pNode->bNomatch==0 );
return rc;
}
/*
** Advance the first term iterator in the first phrase of pNear. Set output
** variable *pbEof to true if it reaches EOF or if an error occurs.
**
** Return SQLITE_OK if successful, or an SQLite error code if an error
** occurs.
*/
static int fts5ExprNodeNext_STRING(
Fts5Expr *pExpr, /* Expression pPhrase belongs to */
Fts5ExprNode *pNode, /* FTS5_STRING or FTS5_TERM node */
int bFromValid,
i64 iFrom
){
Fts5ExprTerm *pTerm = &pNode->pNear->apPhrase[0]->aTerm[0];
int rc = SQLITE_OK;
pNode->bNomatch = 0;
if( pTerm->pSynonym ){
int bEof = 1;
Fts5ExprTerm *p;
/* Find the firstest rowid any synonym points to. */
i64 iRowid = fts5ExprSynonymRowid(pTerm, pExpr->bDesc, 0);
/* Advance each iterator that currently points to iRowid. Or, if iFrom
** is valid - each iterator that points to a rowid before iFrom. */
for(p=pTerm; p; p=p->pSynonym){
if( sqlite3Fts5IterEof(p->pIter)==0 ){
i64 ii = p->pIter->iRowid;
if( ii==iRowid
|| (bFromValid && ii!=iFrom && (ii>iFrom)==pExpr->bDesc)
){
if( bFromValid ){
rc = sqlite3Fts5IterNextFrom(p->pIter, iFrom);
}else{
rc = sqlite3Fts5IterNext(p->pIter);
}
if( rc!=SQLITE_OK ) break;
if( sqlite3Fts5IterEof(p->pIter)==0 ){
bEof = 0;
}
}else{
bEof = 0;
}
}
}
/* Set the EOF flag if either all synonym iterators are at EOF or an
** error has occurred. */
pNode->bEof = (rc || bEof);
}else{
Fts5IndexIter *pIter = pTerm->pIter;
assert( Fts5NodeIsString(pNode) );
if( bFromValid ){
rc = sqlite3Fts5IterNextFrom(pIter, iFrom);
}else{
rc = sqlite3Fts5IterNext(pIter);
}
pNode->bEof = (rc || sqlite3Fts5IterEof(pIter));
}
if( pNode->bEof==0 ){
assert( rc==SQLITE_OK );
rc = fts5ExprNodeTest_STRING(pExpr, pNode);
}
return rc;
}
static int fts5ExprNodeTest_TERM(
Fts5Expr *pExpr, /* Expression that pNear is a part of */
Fts5ExprNode *pNode /* The "NEAR" node (FTS5_TERM) */
){
/* As this "NEAR" object is actually a single phrase that consists
** of a single term only, grab pointers into the poslist managed by the
** fts5_index.c iterator object. This is much faster than synthesizing
** a new poslist the way we have to for more complicated phrase or NEAR
** expressions. */
Fts5ExprPhrase *pPhrase = pNode->pNear->apPhrase[0];
Fts5IndexIter *pIter = pPhrase->aTerm[0].pIter;
assert( pNode->eType==FTS5_TERM );
assert( pNode->pNear->nPhrase==1 && pPhrase->nTerm==1 );
assert( pPhrase->aTerm[0].pSynonym==0 );
pPhrase->poslist.n = pIter->nData;
if( pExpr->pConfig->eDetail==FTS5_DETAIL_FULL ){
pPhrase->poslist.p = (u8*)pIter->pData;
}
pNode->iRowid = pIter->iRowid;
pNode->bNomatch = (pPhrase->poslist.n==0);
return SQLITE_OK;
}
/*
** xNext() method for a node of type FTS5_TERM.
*/
static int fts5ExprNodeNext_TERM(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int rc;
Fts5IndexIter *pIter = pNode->pNear->apPhrase[0]->aTerm[0].pIter;
assert( pNode->bEof==0 );
if( bFromValid ){
rc = sqlite3Fts5IterNextFrom(pIter, iFrom);
}else{
rc = sqlite3Fts5IterNext(pIter);
}
if( rc==SQLITE_OK && sqlite3Fts5IterEof(pIter)==0 ){
rc = fts5ExprNodeTest_TERM(pExpr, pNode);
}else{
pNode->bEof = 1;
pNode->bNomatch = 0;
}
return rc;
}
static void fts5ExprNodeTest_OR(
Fts5Expr *pExpr, /* Expression of which pNode is a part */
Fts5ExprNode *pNode /* Expression node to test */
){
Fts5ExprNode *pNext = pNode->apChild[0];
int i;
for(i=1; inChild; i++){
Fts5ExprNode *pChild = pNode->apChild[i];
int cmp = fts5NodeCompare(pExpr, pNext, pChild);
if( cmp>0 || (cmp==0 && pChild->bNomatch==0) ){
pNext = pChild;
}
}
pNode->iRowid = pNext->iRowid;
pNode->bEof = pNext->bEof;
pNode->bNomatch = pNext->bNomatch;
}
static int fts5ExprNodeNext_OR(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int i;
i64 iLast = pNode->iRowid;
for(i=0; inChild; i++){
Fts5ExprNode *p1 = pNode->apChild[i];
assert( p1->bEof || fts5RowidCmp(pExpr, p1->iRowid, iLast)>=0 );
if( p1->bEof==0 ){
if( (p1->iRowid==iLast)
|| (bFromValid && fts5RowidCmp(pExpr, p1->iRowid, iFrom)<0)
){
int rc = fts5ExprNodeNext(pExpr, p1, bFromValid, iFrom);
if( rc!=SQLITE_OK ){
pNode->bNomatch = 0;
return rc;
}
}
}
}
fts5ExprNodeTest_OR(pExpr, pNode);
return SQLITE_OK;
}
/*
** Argument pNode is an FTS5_AND node.
*/
static int fts5ExprNodeTest_AND(
Fts5Expr *pExpr, /* Expression pPhrase belongs to */
Fts5ExprNode *pAnd /* FTS5_AND node to advance */
){
int iChild;
i64 iLast = pAnd->iRowid;
int rc = SQLITE_OK;
int bMatch;
assert( pAnd->bEof==0 );
do {
pAnd->bNomatch = 0;
bMatch = 1;
for(iChild=0; iChildnChild; iChild++){
Fts5ExprNode *pChild = pAnd->apChild[iChild];
int cmp = fts5RowidCmp(pExpr, iLast, pChild->iRowid);
if( cmp>0 ){
/* Advance pChild until it points to iLast or laster */
rc = fts5ExprNodeNext(pExpr, pChild, 1, iLast);
if( rc!=SQLITE_OK ){
pAnd->bNomatch = 0;
return rc;
}
}
/* If the child node is now at EOF, so is the parent AND node. Otherwise,
** the child node is guaranteed to have advanced at least as far as
** rowid iLast. So if it is not at exactly iLast, pChild->iRowid is the
** new lastest rowid seen so far. */
assert( pChild->bEof || fts5RowidCmp(pExpr, iLast, pChild->iRowid)<=0 );
if( pChild->bEof ){
fts5ExprSetEof(pAnd);
bMatch = 1;
break;
}else if( iLast!=pChild->iRowid ){
bMatch = 0;
iLast = pChild->iRowid;
}
if( pChild->bNomatch ){
pAnd->bNomatch = 1;
}
}
}while( bMatch==0 );
if( pAnd->bNomatch && pAnd!=pExpr->pRoot ){
fts5ExprNodeZeroPoslist(pAnd);
}
pAnd->iRowid = iLast;
return SQLITE_OK;
}
static int fts5ExprNodeNext_AND(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int rc = fts5ExprNodeNext(pExpr, pNode->apChild[0], bFromValid, iFrom);
if( rc==SQLITE_OK ){
rc = fts5ExprNodeTest_AND(pExpr, pNode);
}else{
pNode->bNomatch = 0;
}
return rc;
}
static int fts5ExprNodeTest_NOT(
Fts5Expr *pExpr, /* Expression pPhrase belongs to */
Fts5ExprNode *pNode /* FTS5_NOT node to advance */
){
int rc = SQLITE_OK;
Fts5ExprNode *p1 = pNode->apChild[0];
Fts5ExprNode *p2 = pNode->apChild[1];
assert( pNode->nChild==2 );
while( rc==SQLITE_OK && p1->bEof==0 ){
int cmp = fts5NodeCompare(pExpr, p1, p2);
if( cmp>0 ){
rc = fts5ExprNodeNext(pExpr, p2, 1, p1->iRowid);
cmp = fts5NodeCompare(pExpr, p1, p2);
}
assert( rc!=SQLITE_OK || cmp<=0 );
if( cmp || p2->bNomatch ) break;
rc = fts5ExprNodeNext(pExpr, p1, 0, 0);
}
pNode->bEof = p1->bEof;
pNode->bNomatch = p1->bNomatch;
pNode->iRowid = p1->iRowid;
if( p1->bEof ){
fts5ExprNodeZeroPoslist(p2);
}
return rc;
}
static int fts5ExprNodeNext_NOT(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int rc = fts5ExprNodeNext(pExpr, pNode->apChild[0], bFromValid, iFrom);
if( rc==SQLITE_OK ){
rc = fts5ExprNodeTest_NOT(pExpr, pNode);
}
if( rc!=SQLITE_OK ){
pNode->bNomatch = 0;
}
return rc;
}
/*
** If pNode currently points to a match, this function returns SQLITE_OK
** without modifying it. Otherwise, pNode is advanced until it does point
** to a match or EOF is reached.
*/
static int fts5ExprNodeTest(
Fts5Expr *pExpr, /* Expression of which pNode is a part */
Fts5ExprNode *pNode /* Expression node to test */
){
int rc = SQLITE_OK;
if( pNode->bEof==0 ){
switch( pNode->eType ){
case FTS5_STRING: {
rc = fts5ExprNodeTest_STRING(pExpr, pNode);
break;
}
case FTS5_TERM: {
rc = fts5ExprNodeTest_TERM(pExpr, pNode);
break;
}
case FTS5_AND: {
rc = fts5ExprNodeTest_AND(pExpr, pNode);
break;
}
case FTS5_OR: {
fts5ExprNodeTest_OR(pExpr, pNode);
break;
}
default: assert( pNode->eType==FTS5_NOT ); {
rc = fts5ExprNodeTest_NOT(pExpr, pNode);
break;
}
}
}
return rc;
}
/*
** Set node pNode, which is part of expression pExpr, to point to the first
** match. If there are no matches, set the Node.bEof flag to indicate EOF.
**
** Return an SQLite error code if an error occurs, or SQLITE_OK otherwise.
** It is not an error if there are no matches.
*/
static int fts5ExprNodeFirst(Fts5Expr *pExpr, Fts5ExprNode *pNode){
int rc = SQLITE_OK;
pNode->bEof = 0;
pNode->bNomatch = 0;
if( Fts5NodeIsString(pNode) ){
/* Initialize all term iterators in the NEAR object. */
rc = fts5ExprNearInitAll(pExpr, pNode);
}else if( pNode->xNext==0 ){
pNode->bEof = 1;
}else{
int i;
int nEof = 0;
for(i=0; inChild && rc==SQLITE_OK; i++){
Fts5ExprNode *pChild = pNode->apChild[i];
rc = fts5ExprNodeFirst(pExpr, pNode->apChild[i]);
assert( pChild->bEof==0 || pChild->bEof==1 );
nEof += pChild->bEof;
}
pNode->iRowid = pNode->apChild[0]->iRowid;
switch( pNode->eType ){
case FTS5_AND:
if( nEof>0 ) fts5ExprSetEof(pNode);
break;
case FTS5_OR:
if( pNode->nChild==nEof ) fts5ExprSetEof(pNode);
break;
default:
assert( pNode->eType==FTS5_NOT );
pNode->bEof = pNode->apChild[0]->bEof;
break;
}
}
if( rc==SQLITE_OK ){
rc = fts5ExprNodeTest(pExpr, pNode);
}
return rc;
}
/*
** Begin iterating through the set of documents in index pIdx matched by
** the MATCH expression passed as the first argument. If the "bDesc"
** parameter is passed a non-zero value, iteration is in descending rowid
** order. Or, if it is zero, in ascending order.
**
** If iterating in ascending rowid order (bDesc==0), the first document
** visited is that with the smallest rowid that is larger than or equal
** to parameter iFirst. Or, if iterating in ascending order (bDesc==1),
** then the first document visited must have a rowid smaller than or
** equal to iFirst.
**
** Return SQLITE_OK if successful, or an SQLite error code otherwise. It
** is not considered an error if the query does not match any documents.
*/
static int sqlite3Fts5ExprFirst(Fts5Expr *p, Fts5Index *pIdx, i64 iFirst, int bDesc){
Fts5ExprNode *pRoot = p->pRoot;
int rc; /* Return code */
p->pIndex = pIdx;
p->bDesc = bDesc;
rc = fts5ExprNodeFirst(p, pRoot);
/* If not at EOF but the current rowid occurs earlier than iFirst in
** the iteration order, move to document iFirst or later. */
if( rc==SQLITE_OK
&& 0==pRoot->bEof
&& fts5RowidCmp(p, pRoot->iRowid, iFirst)<0
){
rc = fts5ExprNodeNext(p, pRoot, 1, iFirst);
}
/* If the iterator is not at a real match, skip forward until it is. */
while( pRoot->bNomatch && rc==SQLITE_OK ){
assert( pRoot->bEof==0 );
rc = fts5ExprNodeNext(p, pRoot, 0, 0);
}
return rc;
}
/*
** Move to the next document
**
** Return SQLITE_OK if successful, or an SQLite error code otherwise. It
** is not considered an error if the query does not match any documents.
*/
static int sqlite3Fts5ExprNext(Fts5Expr *p, i64 iLast){
int rc;
Fts5ExprNode *pRoot = p->pRoot;
assert( pRoot->bEof==0 && pRoot->bNomatch==0 );
do {
rc = fts5ExprNodeNext(p, pRoot, 0, 0);
assert( pRoot->bNomatch==0 || (rc==SQLITE_OK && pRoot->bEof==0) );
}while( pRoot->bNomatch );
if( fts5RowidCmp(p, pRoot->iRowid, iLast)>0 ){
pRoot->bEof = 1;
}
return rc;
}
static int sqlite3Fts5ExprEof(Fts5Expr *p){
return p->pRoot->bEof;
}
static i64 sqlite3Fts5ExprRowid(Fts5Expr *p){
return p->pRoot->iRowid;
}
static int fts5ParseStringFromToken(Fts5Token *pToken, char **pz){
int rc = SQLITE_OK;
*pz = sqlite3Fts5Strndup(&rc, pToken->p, pToken->n);
return rc;
}
/*
** Free the phrase object passed as the only argument.
*/
static void fts5ExprPhraseFree(Fts5ExprPhrase *pPhrase){
if( pPhrase ){
int i;
for(i=0; inTerm; i++){
Fts5ExprTerm *pSyn;
Fts5ExprTerm *pNext;
Fts5ExprTerm *pTerm = &pPhrase->aTerm[i];
sqlite3_free(pTerm->pTerm);
sqlite3Fts5IterClose(pTerm->pIter);
for(pSyn=pTerm->pSynonym; pSyn; pSyn=pNext){
pNext = pSyn->pSynonym;
sqlite3Fts5IterClose(pSyn->pIter);
fts5BufferFree((Fts5Buffer*)&pSyn[1]);
sqlite3_free(pSyn);
}
}
if( pPhrase->poslist.nSpace>0 ) fts5BufferFree(&pPhrase->poslist);
sqlite3_free(pPhrase);
}
}
/*
** Set the "bFirst" flag on the first token of the phrase passed as the
** only argument.
*/
static void sqlite3Fts5ParseSetCaret(Fts5ExprPhrase *pPhrase){
if( pPhrase && pPhrase->nTerm ){
pPhrase->aTerm[0].bFirst = 1;
}
}
/*
** If argument pNear is NULL, then a new Fts5ExprNearset object is allocated
** and populated with pPhrase. Or, if pNear is not NULL, phrase pPhrase is
** appended to it and the results returned.
**
** If an OOM error occurs, both the pNear and pPhrase objects are freed and
** NULL returned.
*/
static Fts5ExprNearset *sqlite3Fts5ParseNearset(
Fts5Parse *pParse, /* Parse context */
Fts5ExprNearset *pNear, /* Existing nearset, or NULL */
Fts5ExprPhrase *pPhrase /* Recently parsed phrase */
){
const int SZALLOC = 8;
Fts5ExprNearset *pRet = 0;
if( pParse->rc==SQLITE_OK ){
if( pPhrase==0 ){
return pNear;
}
if( pNear==0 ){
sqlite3_int64 nByte;
nByte = sizeof(Fts5ExprNearset) + SZALLOC * sizeof(Fts5ExprPhrase*);
pRet = sqlite3_malloc64(nByte);
if( pRet==0 ){
pParse->rc = SQLITE_NOMEM;
}else{
memset(pRet, 0, (size_t)nByte);
}
}else if( (pNear->nPhrase % SZALLOC)==0 ){
int nNew = pNear->nPhrase + SZALLOC;
sqlite3_int64 nByte;
nByte = sizeof(Fts5ExprNearset) + nNew * sizeof(Fts5ExprPhrase*);
pRet = (Fts5ExprNearset*)sqlite3_realloc64(pNear, nByte);
if( pRet==0 ){
pParse->rc = SQLITE_NOMEM;
}
}else{
pRet = pNear;
}
}
if( pRet==0 ){
assert( pParse->rc!=SQLITE_OK );
sqlite3Fts5ParseNearsetFree(pNear);
sqlite3Fts5ParsePhraseFree(pPhrase);
}else{
if( pRet->nPhrase>0 ){
Fts5ExprPhrase *pLast = pRet->apPhrase[pRet->nPhrase-1];
assert( pParse!=0 );
assert( pParse->apPhrase!=0 );
assert( pParse->nPhrase>=2 );
assert( pLast==pParse->apPhrase[pParse->nPhrase-2] );
if( pPhrase->nTerm==0 ){
fts5ExprPhraseFree(pPhrase);
pRet->nPhrase--;
pParse->nPhrase--;
pPhrase = pLast;
}else if( pLast->nTerm==0 ){
fts5ExprPhraseFree(pLast);
pParse->apPhrase[pParse->nPhrase-2] = pPhrase;
pParse->nPhrase--;
pRet->nPhrase--;
}
}
pRet->apPhrase[pRet->nPhrase++] = pPhrase;
}
return pRet;
}
typedef struct TokenCtx TokenCtx;
struct TokenCtx {
Fts5ExprPhrase *pPhrase;
Fts5Config *pConfig;
int rc;
};
/*
** Callback for tokenizing terms used by ParseTerm().
*/
static int fts5ParseTokenize(
void *pContext, /* Pointer to Fts5InsertCtx object */
int tflags, /* Mask of FTS5_TOKEN_* flags */
const char *pToken, /* Buffer containing token */
int nToken, /* Size of token in bytes */
int iUnused1, /* Start offset of token */
int iUnused2 /* End offset of token */
){
int rc = SQLITE_OK;
const int SZALLOC = 8;
TokenCtx *pCtx = (TokenCtx*)pContext;
Fts5ExprPhrase *pPhrase = pCtx->pPhrase;
UNUSED_PARAM2(iUnused1, iUnused2);
/* If an error has already occurred, this is a no-op */
if( pCtx->rc!=SQLITE_OK ) return pCtx->rc;
if( nToken>FTS5_MAX_TOKEN_SIZE ) nToken = FTS5_MAX_TOKEN_SIZE;
if( pPhrase && pPhrase->nTerm>0 && (tflags & FTS5_TOKEN_COLOCATED) ){
Fts5ExprTerm *pSyn;
sqlite3_int64 nByte = sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer) + nToken+1;
pSyn = (Fts5ExprTerm*)sqlite3_malloc64(nByte);
if( pSyn==0 ){
rc = SQLITE_NOMEM;
}else{
memset(pSyn, 0, (size_t)nByte);
pSyn->pTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer);
pSyn->nFullTerm = pSyn->nQueryTerm = nToken;
if( pCtx->pConfig->bTokendata ){
pSyn->nQueryTerm = (int)strlen(pSyn->pTerm);
}
memcpy(pSyn->pTerm, pToken, nToken);
pSyn->pSynonym = pPhrase->aTerm[pPhrase->nTerm-1].pSynonym;
pPhrase->aTerm[pPhrase->nTerm-1].pSynonym = pSyn;
}
}else{
Fts5ExprTerm *pTerm;
if( pPhrase==0 || (pPhrase->nTerm % SZALLOC)==0 ){
Fts5ExprPhrase *pNew;
int nNew = SZALLOC + (pPhrase ? pPhrase->nTerm : 0);
pNew = (Fts5ExprPhrase*)sqlite3_realloc64(pPhrase,
sizeof(Fts5ExprPhrase) + sizeof(Fts5ExprTerm) * nNew
);
if( pNew==0 ){
rc = SQLITE_NOMEM;
}else{
if( pPhrase==0 ) memset(pNew, 0, sizeof(Fts5ExprPhrase));
pCtx->pPhrase = pPhrase = pNew;
pNew->nTerm = nNew - SZALLOC;
}
}
if( rc==SQLITE_OK ){
pTerm = &pPhrase->aTerm[pPhrase->nTerm++];
memset(pTerm, 0, sizeof(Fts5ExprTerm));
pTerm->pTerm = sqlite3Fts5Strndup(&rc, pToken, nToken);
pTerm->nFullTerm = pTerm->nQueryTerm = nToken;
if( pCtx->pConfig->bTokendata && rc==SQLITE_OK ){
pTerm->nQueryTerm = (int)strlen(pTerm->pTerm);
}
}
}
pCtx->rc = rc;
return rc;
}
/*
** Free the phrase object passed as the only argument.
*/
static void sqlite3Fts5ParsePhraseFree(Fts5ExprPhrase *pPhrase){
fts5ExprPhraseFree(pPhrase);
}
/*
** Free the phrase object passed as the second argument.
*/
static void sqlite3Fts5ParseNearsetFree(Fts5ExprNearset *pNear){
if( pNear ){
int i;
for(i=0; inPhrase; i++){
fts5ExprPhraseFree(pNear->apPhrase[i]);
}
sqlite3_free(pNear->pColset);
sqlite3_free(pNear);
}
}
static void sqlite3Fts5ParseFinished(Fts5Parse *pParse, Fts5ExprNode *p){
assert( pParse->pExpr==0 );
pParse->pExpr = p;
}
static int parseGrowPhraseArray(Fts5Parse *pParse){
if( (pParse->nPhrase % 8)==0 ){
sqlite3_int64 nByte = sizeof(Fts5ExprPhrase*) * (pParse->nPhrase + 8);
Fts5ExprPhrase **apNew;
apNew = (Fts5ExprPhrase**)sqlite3_realloc64(pParse->apPhrase, nByte);
if( apNew==0 ){
pParse->rc = SQLITE_NOMEM;
return SQLITE_NOMEM;
}
pParse->apPhrase = apNew;
}
return SQLITE_OK;
}
/*
** This function is called by the parser to process a string token. The
** string may or may not be quoted. In any case it is tokenized and a
** phrase object consisting of all tokens returned.
*/
static Fts5ExprPhrase *sqlite3Fts5ParseTerm(
Fts5Parse *pParse, /* Parse context */
Fts5ExprPhrase *pAppend, /* Phrase to append to */
Fts5Token *pToken, /* String to tokenize */
int bPrefix /* True if there is a trailing "*" */
){
Fts5Config *pConfig = pParse->pConfig;
TokenCtx sCtx; /* Context object passed to callback */
int rc; /* Tokenize return code */
char *z = 0;
memset(&sCtx, 0, sizeof(TokenCtx));
sCtx.pPhrase = pAppend;
sCtx.pConfig = pConfig;
rc = fts5ParseStringFromToken(pToken, &z);
if( rc==SQLITE_OK ){
int flags = FTS5_TOKENIZE_QUERY | (bPrefix ? FTS5_TOKENIZE_PREFIX : 0);
int n;
sqlite3Fts5Dequote(z);
n = (int)strlen(z);
rc = sqlite3Fts5Tokenize(pConfig, flags, z, n, &sCtx, fts5ParseTokenize);
}
sqlite3_free(z);
if( rc || (rc = sCtx.rc) ){
pParse->rc = rc;
fts5ExprPhraseFree(sCtx.pPhrase);
sCtx.pPhrase = 0;
}else{
if( pAppend==0 ){
if( parseGrowPhraseArray(pParse) ){
fts5ExprPhraseFree(sCtx.pPhrase);
return 0;
}
pParse->nPhrase++;
}
if( sCtx.pPhrase==0 ){
/* This happens when parsing a token or quoted phrase that contains
** no token characters at all. (e.g ... MATCH '""'). */
sCtx.pPhrase = sqlite3Fts5MallocZero(&pParse->rc, sizeof(Fts5ExprPhrase));
}else if( sCtx.pPhrase->nTerm ){
sCtx.pPhrase->aTerm[sCtx.pPhrase->nTerm-1].bPrefix = (u8)bPrefix;
}
pParse->apPhrase[pParse->nPhrase-1] = sCtx.pPhrase;
}
return sCtx.pPhrase;
}
/*
** Create a new FTS5 expression by cloning phrase iPhrase of the
** expression passed as the second argument.
*/
static int sqlite3Fts5ExprClonePhrase(
Fts5Expr *pExpr,
int iPhrase,
Fts5Expr **ppNew
){
int rc = SQLITE_OK; /* Return code */
Fts5ExprPhrase *pOrig = 0; /* The phrase extracted from pExpr */
Fts5Expr *pNew = 0; /* Expression to return via *ppNew */
TokenCtx sCtx = {0,0,0}; /* Context object for fts5ParseTokenize */
if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){
rc = SQLITE_RANGE;
}else{
pOrig = pExpr->apExprPhrase[iPhrase];
pNew = (Fts5Expr*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Expr));
}
if( rc==SQLITE_OK ){
pNew->apExprPhrase = (Fts5ExprPhrase**)sqlite3Fts5MallocZero(&rc,
sizeof(Fts5ExprPhrase*));
}
if( rc==SQLITE_OK ){
pNew->pRoot = (Fts5ExprNode*)sqlite3Fts5MallocZero(&rc,
sizeof(Fts5ExprNode));
}
if( rc==SQLITE_OK ){
pNew->pRoot->pNear = (Fts5ExprNearset*)sqlite3Fts5MallocZero(&rc,
sizeof(Fts5ExprNearset) + sizeof(Fts5ExprPhrase*));
}
if( rc==SQLITE_OK && ALWAYS(pOrig!=0) ){
Fts5Colset *pColsetOrig = pOrig->pNode->pNear->pColset;
if( pColsetOrig ){
sqlite3_int64 nByte;
Fts5Colset *pColset;
nByte = sizeof(Fts5Colset) + (pColsetOrig->nCol-1) * sizeof(int);
pColset = (Fts5Colset*)sqlite3Fts5MallocZero(&rc, nByte);
if( pColset ){
memcpy(pColset, pColsetOrig, (size_t)nByte);
}
pNew->pRoot->pNear->pColset = pColset;
}
}
if( rc==SQLITE_OK ){
if( pOrig->nTerm ){
int i; /* Used to iterate through phrase terms */
sCtx.pConfig = pExpr->pConfig;
for(i=0; rc==SQLITE_OK && inTerm; i++){
int tflags = 0;
Fts5ExprTerm *p;
for(p=&pOrig->aTerm[i]; p && rc==SQLITE_OK; p=p->pSynonym){
rc = fts5ParseTokenize((void*)&sCtx,tflags,p->pTerm,p->nFullTerm,0,0);
tflags = FTS5_TOKEN_COLOCATED;
}
if( rc==SQLITE_OK ){
sCtx.pPhrase->aTerm[i].bPrefix = pOrig->aTerm[i].bPrefix;
sCtx.pPhrase->aTerm[i].bFirst = pOrig->aTerm[i].bFirst;
}
}
}else{
/* This happens when parsing a token or quoted phrase that contains
** no token characters at all. (e.g ... MATCH '""'). */
sCtx.pPhrase = sqlite3Fts5MallocZero(&rc, sizeof(Fts5ExprPhrase));
}
}
if( rc==SQLITE_OK && ALWAYS(sCtx.pPhrase) ){
/* All the allocations succeeded. Put the expression object together. */
pNew->pIndex = pExpr->pIndex;
pNew->pConfig = pExpr->pConfig;
pNew->nPhrase = 1;
pNew->apExprPhrase[0] = sCtx.pPhrase;
pNew->pRoot->pNear->apPhrase[0] = sCtx.pPhrase;
pNew->pRoot->pNear->nPhrase = 1;
sCtx.pPhrase->pNode = pNew->pRoot;
if( pOrig->nTerm==1
&& pOrig->aTerm[0].pSynonym==0
&& pOrig->aTerm[0].bFirst==0
){
pNew->pRoot->eType = FTS5_TERM;
pNew->pRoot->xNext = fts5ExprNodeNext_TERM;
}else{
pNew->pRoot->eType = FTS5_STRING;
pNew->pRoot->xNext = fts5ExprNodeNext_STRING;
}
}else{
sqlite3Fts5ExprFree(pNew);
fts5ExprPhraseFree(sCtx.pPhrase);
pNew = 0;
}
*ppNew = pNew;
return rc;
}
/*
** Token pTok has appeared in a MATCH expression where the NEAR operator
** is expected. If token pTok does not contain "NEAR", store an error
** in the pParse object.
*/
static void sqlite3Fts5ParseNear(Fts5Parse *pParse, Fts5Token *pTok){
if( pTok->n!=4 || memcmp("NEAR", pTok->p, 4) ){
sqlite3Fts5ParseError(
pParse, "fts5: syntax error near \"%.*s\"", pTok->n, pTok->p
);
}
}
static void sqlite3Fts5ParseSetDistance(
Fts5Parse *pParse,
Fts5ExprNearset *pNear,
Fts5Token *p
){
if( pNear ){
int nNear = 0;
int i;
if( p->n ){
for(i=0; in; i++){
char c = (char)p->p[i];
if( c<'0' || c>'9' ){
sqlite3Fts5ParseError(
pParse, "expected integer, got \"%.*s\"", p->n, p->p
);
return;
}
nNear = nNear * 10 + (p->p[i] - '0');
}
}else{
nNear = FTS5_DEFAULT_NEARDIST;
}
pNear->nNear = nNear;
}
}
/*
** The second argument passed to this function may be NULL, or it may be
** an existing Fts5Colset object. This function returns a pointer to
** a new colset object containing the contents of (p) with new value column
** number iCol appended.
**
** If an OOM error occurs, store an error code in pParse and return NULL.
** The old colset object (if any) is not freed in this case.
*/
static Fts5Colset *fts5ParseColset(
Fts5Parse *pParse, /* Store SQLITE_NOMEM here if required */
Fts5Colset *p, /* Existing colset object */
int iCol /* New column to add to colset object */
){
int nCol = p ? p->nCol : 0; /* Num. columns already in colset object */
Fts5Colset *pNew; /* New colset object to return */
assert( pParse->rc==SQLITE_OK );
assert( iCol>=0 && iColpConfig->nCol );
pNew = sqlite3_realloc64(p, sizeof(Fts5Colset) + sizeof(int)*nCol);
if( pNew==0 ){
pParse->rc = SQLITE_NOMEM;
}else{
int *aiCol = pNew->aiCol;
int i, j;
for(i=0; iiCol ) break;
}
for(j=nCol; j>i; j--){
aiCol[j] = aiCol[j-1];
}
aiCol[i] = iCol;
pNew->nCol = nCol+1;
#ifndef NDEBUG
/* Check that the array is in order and contains no duplicate entries. */
for(i=1; inCol; i++) assert( pNew->aiCol[i]>pNew->aiCol[i-1] );
#endif
}
return pNew;
}
/*
** Allocate and return an Fts5Colset object specifying the inverse of
** the colset passed as the second argument. Free the colset passed
** as the second argument before returning.
*/
static Fts5Colset *sqlite3Fts5ParseColsetInvert(Fts5Parse *pParse, Fts5Colset *p){
Fts5Colset *pRet;
int nCol = pParse->pConfig->nCol;
pRet = (Fts5Colset*)sqlite3Fts5MallocZero(&pParse->rc,
sizeof(Fts5Colset) + sizeof(int)*nCol
);
if( pRet ){
int i;
int iOld = 0;
for(i=0; i=p->nCol || p->aiCol[iOld]!=i ){
pRet->aiCol[pRet->nCol++] = i;
}else{
iOld++;
}
}
}
sqlite3_free(p);
return pRet;
}
static Fts5Colset *sqlite3Fts5ParseColset(
Fts5Parse *pParse, /* Store SQLITE_NOMEM here if required */
Fts5Colset *pColset, /* Existing colset object */
Fts5Token *p
){
Fts5Colset *pRet = 0;
int iCol;
char *z; /* Dequoted copy of token p */
z = sqlite3Fts5Strndup(&pParse->rc, p->p, p->n);
if( pParse->rc==SQLITE_OK ){
Fts5Config *pConfig = pParse->pConfig;
sqlite3Fts5Dequote(z);
for(iCol=0; iColnCol; iCol++){
if( 0==sqlite3_stricmp(pConfig->azCol[iCol], z) ) break;
}
if( iCol==pConfig->nCol ){
sqlite3Fts5ParseError(pParse, "no such column: %s", z);
}else{
pRet = fts5ParseColset(pParse, pColset, iCol);
}
sqlite3_free(z);
}
if( pRet==0 ){
assert( pParse->rc!=SQLITE_OK );
sqlite3_free(pColset);
}
return pRet;
}
/*
** If argument pOrig is NULL, or if (*pRc) is set to anything other than
** SQLITE_OK when this function is called, NULL is returned.
**
** Otherwise, a copy of (*pOrig) is made into memory obtained from
** sqlite3Fts5MallocZero() and a pointer to it returned. If the allocation
** fails, (*pRc) is set to SQLITE_NOMEM and NULL is returned.
*/
static Fts5Colset *fts5CloneColset(int *pRc, Fts5Colset *pOrig){
Fts5Colset *pRet;
if( pOrig ){
sqlite3_int64 nByte = sizeof(Fts5Colset) + (pOrig->nCol-1) * sizeof(int);
pRet = (Fts5Colset*)sqlite3Fts5MallocZero(pRc, nByte);
if( pRet ){
memcpy(pRet, pOrig, (size_t)nByte);
}
}else{
pRet = 0;
}
return pRet;
}
/*
** Remove from colset pColset any columns that are not also in colset pMerge.
*/
static void fts5MergeColset(Fts5Colset *pColset, Fts5Colset *pMerge){
int iIn = 0; /* Next input in pColset */
int iMerge = 0; /* Next input in pMerge */
int iOut = 0; /* Next output slot in pColset */
while( iInnCol && iMergenCol ){
int iDiff = pColset->aiCol[iIn] - pMerge->aiCol[iMerge];
if( iDiff==0 ){
pColset->aiCol[iOut++] = pMerge->aiCol[iMerge];
iMerge++;
iIn++;
}else if( iDiff>0 ){
iMerge++;
}else{
iIn++;
}
}
pColset->nCol = iOut;
}
/*
** Recursively apply colset pColset to expression node pNode and all of
** its decendents. If (*ppFree) is not NULL, it contains a spare copy
** of pColset. This function may use the spare copy and set (*ppFree) to
** zero, or it may create copies of pColset using fts5CloneColset().
*/
static void fts5ParseSetColset(
Fts5Parse *pParse,
Fts5ExprNode *pNode,
Fts5Colset *pColset,
Fts5Colset **ppFree
){
if( pParse->rc==SQLITE_OK ){
assert( pNode->eType==FTS5_TERM || pNode->eType==FTS5_STRING
|| pNode->eType==FTS5_AND || pNode->eType==FTS5_OR
|| pNode->eType==FTS5_NOT || pNode->eType==FTS5_EOF
);
if( pNode->eType==FTS5_STRING || pNode->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pNode->pNear;
if( pNear->pColset ){
fts5MergeColset(pNear->pColset, pColset);
if( pNear->pColset->nCol==0 ){
pNode->eType = FTS5_EOF;
pNode->xNext = 0;
}
}else if( *ppFree ){
pNear->pColset = pColset;
*ppFree = 0;
}else{
pNear->pColset = fts5CloneColset(&pParse->rc, pColset);
}
}else{
int i;
assert( pNode->eType!=FTS5_EOF || pNode->nChild==0 );
for(i=0; inChild; i++){
fts5ParseSetColset(pParse, pNode->apChild[i], pColset, ppFree);
}
}
}
}
/*
** Apply colset pColset to expression node pExpr and all of its descendents.
*/
static void sqlite3Fts5ParseSetColset(
Fts5Parse *pParse,
Fts5ExprNode *pExpr,
Fts5Colset *pColset
){
Fts5Colset *pFree = pColset;
if( pParse->pConfig->eDetail==FTS5_DETAIL_NONE ){
sqlite3Fts5ParseError(pParse,
"fts5: column queries are not supported (detail=none)"
);
}else{
fts5ParseSetColset(pParse, pExpr, pColset, &pFree);
}
sqlite3_free(pFree);
}
static void fts5ExprAssignXNext(Fts5ExprNode *pNode){
switch( pNode->eType ){
case FTS5_STRING: {
Fts5ExprNearset *pNear = pNode->pNear;
if( pNear->nPhrase==1 && pNear->apPhrase[0]->nTerm==1
&& pNear->apPhrase[0]->aTerm[0].pSynonym==0
&& pNear->apPhrase[0]->aTerm[0].bFirst==0
){
pNode->eType = FTS5_TERM;
pNode->xNext = fts5ExprNodeNext_TERM;
}else{
pNode->xNext = fts5ExprNodeNext_STRING;
}
break;
};
case FTS5_OR: {
pNode->xNext = fts5ExprNodeNext_OR;
break;
};
case FTS5_AND: {
pNode->xNext = fts5ExprNodeNext_AND;
break;
};
default: assert( pNode->eType==FTS5_NOT ); {
pNode->xNext = fts5ExprNodeNext_NOT;
break;
};
}
}
static void fts5ExprAddChildren(Fts5ExprNode *p, Fts5ExprNode *pSub){
int ii = p->nChild;
if( p->eType!=FTS5_NOT && pSub->eType==p->eType ){
int nByte = sizeof(Fts5ExprNode*) * pSub->nChild;
memcpy(&p->apChild[p->nChild], pSub->apChild, nByte);
p->nChild += pSub->nChild;
sqlite3_free(pSub);
}else{
p->apChild[p->nChild++] = pSub;
}
for( ; iinChild; ii++){
p->iHeight = MAX(p->iHeight, p->apChild[ii]->iHeight + 1);
}
}
/*
** This function is used when parsing LIKE or GLOB patterns against
** trigram indexes that specify either detail=column or detail=none.
** It converts a phrase:
**
** abc + def + ghi
**
** into an AND tree:
**
** abc AND def AND ghi
*/
static Fts5ExprNode *fts5ParsePhraseToAnd(
Fts5Parse *pParse,
Fts5ExprNearset *pNear
){
int nTerm = pNear->apPhrase[0]->nTerm;
int ii;
int nByte;
Fts5ExprNode *pRet;
assert( pNear->nPhrase==1 );
assert( pParse->bPhraseToAnd );
nByte = sizeof(Fts5ExprNode) + nTerm*sizeof(Fts5ExprNode*);
pRet = (Fts5ExprNode*)sqlite3Fts5MallocZero(&pParse->rc, nByte);
if( pRet ){
pRet->eType = FTS5_AND;
pRet->nChild = nTerm;
pRet->iHeight = 1;
fts5ExprAssignXNext(pRet);
pParse->nPhrase--;
for(ii=0; iirc, sizeof(Fts5ExprPhrase)
);
if( pPhrase ){
if( parseGrowPhraseArray(pParse) ){
fts5ExprPhraseFree(pPhrase);
}else{
Fts5ExprTerm *p = &pNear->apPhrase[0]->aTerm[ii];
Fts5ExprTerm *pTo = &pPhrase->aTerm[0];
pParse->apPhrase[pParse->nPhrase++] = pPhrase;
pPhrase->nTerm = 1;
pTo->pTerm = sqlite3Fts5Strndup(&pParse->rc, p->pTerm, p->nFullTerm);
pTo->nQueryTerm = p->nQueryTerm;
pTo->nFullTerm = p->nFullTerm;
pRet->apChild[ii] = sqlite3Fts5ParseNode(pParse, FTS5_STRING,
0, 0, sqlite3Fts5ParseNearset(pParse, 0, pPhrase)
);
}
}
}
if( pParse->rc ){
sqlite3Fts5ParseNodeFree(pRet);
pRet = 0;
}else{
sqlite3Fts5ParseNearsetFree(pNear);
}
}
return pRet;
}
/*
** Allocate and return a new expression object. If anything goes wrong (i.e.
** OOM error), leave an error code in pParse and return NULL.
*/
static Fts5ExprNode *sqlite3Fts5ParseNode(
Fts5Parse *pParse, /* Parse context */
int eType, /* FTS5_STRING, AND, OR or NOT */
Fts5ExprNode *pLeft, /* Left hand child expression */
Fts5ExprNode *pRight, /* Right hand child expression */
Fts5ExprNearset *pNear /* For STRING expressions, the near cluster */
){
Fts5ExprNode *pRet = 0;
if( pParse->rc==SQLITE_OK ){
int nChild = 0; /* Number of children of returned node */
sqlite3_int64 nByte; /* Bytes of space to allocate for this node */
assert( (eType!=FTS5_STRING && !pNear)
|| (eType==FTS5_STRING && !pLeft && !pRight)
);
if( eType==FTS5_STRING && pNear==0 ) return 0;
if( eType!=FTS5_STRING && pLeft==0 ) return pRight;
if( eType!=FTS5_STRING && pRight==0 ) return pLeft;
if( eType==FTS5_STRING
&& pParse->bPhraseToAnd
&& pNear->apPhrase[0]->nTerm>1
){
pRet = fts5ParsePhraseToAnd(pParse, pNear);
}else{
if( eType==FTS5_NOT ){
nChild = 2;
}else if( eType==FTS5_AND || eType==FTS5_OR ){
nChild = 2;
if( pLeft->eType==eType ) nChild += pLeft->nChild-1;
if( pRight->eType==eType ) nChild += pRight->nChild-1;
}
nByte = sizeof(Fts5ExprNode) + sizeof(Fts5ExprNode*)*(nChild-1);
pRet = (Fts5ExprNode*)sqlite3Fts5MallocZero(&pParse->rc, nByte);
if( pRet ){
pRet->eType = eType;
pRet->pNear = pNear;
fts5ExprAssignXNext(pRet);
if( eType==FTS5_STRING ){
int iPhrase;
for(iPhrase=0; iPhrasenPhrase; iPhrase++){
pNear->apPhrase[iPhrase]->pNode = pRet;
if( pNear->apPhrase[iPhrase]->nTerm==0 ){
pRet->xNext = 0;
pRet->eType = FTS5_EOF;
}
}
if( pParse->pConfig->eDetail!=FTS5_DETAIL_FULL ){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[0];
if( pNear->nPhrase!=1
|| pPhrase->nTerm>1
|| (pPhrase->nTerm>0 && pPhrase->aTerm[0].bFirst)
){
sqlite3Fts5ParseError(pParse,
"fts5: %s queries are not supported (detail!=full)",
pNear->nPhrase==1 ? "phrase": "NEAR"
);
sqlite3_free(pRet);
pRet = 0;
}
}
}else{
fts5ExprAddChildren(pRet, pLeft);
fts5ExprAddChildren(pRet, pRight);
if( pRet->iHeight>SQLITE_FTS5_MAX_EXPR_DEPTH ){
sqlite3Fts5ParseError(pParse,
"fts5 expression tree is too large (maximum depth %d)",
SQLITE_FTS5_MAX_EXPR_DEPTH
);
sqlite3_free(pRet);
pRet = 0;
}
}
}
}
}
if( pRet==0 ){
assert( pParse->rc!=SQLITE_OK );
sqlite3Fts5ParseNodeFree(pLeft);
sqlite3Fts5ParseNodeFree(pRight);
sqlite3Fts5ParseNearsetFree(pNear);
}
return pRet;
}
static Fts5ExprNode *sqlite3Fts5ParseImplicitAnd(
Fts5Parse *pParse, /* Parse context */
Fts5ExprNode *pLeft, /* Left hand child expression */
Fts5ExprNode *pRight /* Right hand child expression */
){
Fts5ExprNode *pRet = 0;
Fts5ExprNode *pPrev;
if( pParse->rc ){
sqlite3Fts5ParseNodeFree(pLeft);
sqlite3Fts5ParseNodeFree(pRight);
}else{
assert( pLeft->eType==FTS5_STRING
|| pLeft->eType==FTS5_TERM
|| pLeft->eType==FTS5_EOF
|| pLeft->eType==FTS5_AND
);
assert( pRight->eType==FTS5_STRING
|| pRight->eType==FTS5_TERM
|| pRight->eType==FTS5_EOF
);
if( pLeft->eType==FTS5_AND ){
pPrev = pLeft->apChild[pLeft->nChild-1];
}else{
pPrev = pLeft;
}
assert( pPrev->eType==FTS5_STRING
|| pPrev->eType==FTS5_TERM
|| pPrev->eType==FTS5_EOF
);
if( pRight->eType==FTS5_EOF ){
assert( pParse->apPhrase[pParse->nPhrase-1]==pRight->pNear->apPhrase[0] );
sqlite3Fts5ParseNodeFree(pRight);
pRet = pLeft;
pParse->nPhrase--;
}
else if( pPrev->eType==FTS5_EOF ){
Fts5ExprPhrase **ap;
if( pPrev==pLeft ){
pRet = pRight;
}else{
pLeft->apChild[pLeft->nChild-1] = pRight;
pRet = pLeft;
}
ap = &pParse->apPhrase[pParse->nPhrase-1-pRight->pNear->nPhrase];
assert( ap[0]==pPrev->pNear->apPhrase[0] );
memmove(ap, &ap[1], sizeof(Fts5ExprPhrase*)*pRight->pNear->nPhrase);
pParse->nPhrase--;
sqlite3Fts5ParseNodeFree(pPrev);
}
else{
pRet = sqlite3Fts5ParseNode(pParse, FTS5_AND, pLeft, pRight, 0);
}
}
return pRet;
}
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
static char *fts5ExprTermPrint(Fts5ExprTerm *pTerm){
sqlite3_int64 nByte = 0;
Fts5ExprTerm *p;
char *zQuoted;
/* Determine the maximum amount of space required. */
for(p=pTerm; p; p=p->pSynonym){
nByte += pTerm->nQueryTerm * 2 + 3 + 2;
}
zQuoted = sqlite3_malloc64(nByte);
if( zQuoted ){
int i = 0;
for(p=pTerm; p; p=p->pSynonym){
char *zIn = p->pTerm;
char *zEnd = &zIn[p->nQueryTerm];
zQuoted[i++] = '"';
while( zInpSynonym ) zQuoted[i++] = '|';
}
if( pTerm->bPrefix ){
zQuoted[i++] = ' ';
zQuoted[i++] = '*';
}
zQuoted[i++] = '\0';
}
return zQuoted;
}
static char *fts5PrintfAppend(char *zApp, const char *zFmt, ...){
char *zNew;
va_list ap;
va_start(ap, zFmt);
zNew = sqlite3_vmprintf(zFmt, ap);
va_end(ap);
if( zApp && zNew ){
char *zNew2 = sqlite3_mprintf("%s%s", zApp, zNew);
sqlite3_free(zNew);
zNew = zNew2;
}
sqlite3_free(zApp);
return zNew;
}
/*
** Compose a tcl-readable representation of expression pExpr. Return a
** pointer to a buffer containing that representation. It is the
** responsibility of the caller to at some point free the buffer using
** sqlite3_free().
*/
static char *fts5ExprPrintTcl(
Fts5Config *pConfig,
const char *zNearsetCmd,
Fts5ExprNode *pExpr
){
char *zRet = 0;
if( pExpr->eType==FTS5_STRING || pExpr->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pExpr->pNear;
int i;
int iTerm;
zRet = fts5PrintfAppend(zRet, "%s ", zNearsetCmd);
if( zRet==0 ) return 0;
if( pNear->pColset ){
int *aiCol = pNear->pColset->aiCol;
int nCol = pNear->pColset->nCol;
if( nCol==1 ){
zRet = fts5PrintfAppend(zRet, "-col %d ", aiCol[0]);
}else{
zRet = fts5PrintfAppend(zRet, "-col {%d", aiCol[0]);
for(i=1; ipColset->nCol; i++){
zRet = fts5PrintfAppend(zRet, " %d", aiCol[i]);
}
zRet = fts5PrintfAppend(zRet, "} ");
}
if( zRet==0 ) return 0;
}
if( pNear->nPhrase>1 ){
zRet = fts5PrintfAppend(zRet, "-near %d ", pNear->nNear);
if( zRet==0 ) return 0;
}
zRet = fts5PrintfAppend(zRet, "--");
if( zRet==0 ) return 0;
for(i=0; inPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
zRet = fts5PrintfAppend(zRet, " {");
for(iTerm=0; zRet && iTermnTerm; iTerm++){
Fts5ExprTerm *p = &pPhrase->aTerm[iTerm];
zRet = fts5PrintfAppend(zRet, "%s%.*s", iTerm==0?"":" ",
p->nQueryTerm, p->pTerm
);
if( pPhrase->aTerm[iTerm].bPrefix ){
zRet = fts5PrintfAppend(zRet, "*");
}
}
if( zRet ) zRet = fts5PrintfAppend(zRet, "}");
if( zRet==0 ) return 0;
}
}else if( pExpr->eType==0 ){
zRet = sqlite3_mprintf("{}");
}else{
char const *zOp = 0;
int i;
switch( pExpr->eType ){
case FTS5_AND: zOp = "AND"; break;
case FTS5_NOT: zOp = "NOT"; break;
default:
assert( pExpr->eType==FTS5_OR );
zOp = "OR";
break;
}
zRet = sqlite3_mprintf("%s", zOp);
for(i=0; zRet && inChild; i++){
char *z = fts5ExprPrintTcl(pConfig, zNearsetCmd, pExpr->apChild[i]);
if( !z ){
sqlite3_free(zRet);
zRet = 0;
}else{
zRet = fts5PrintfAppend(zRet, " [%z]", z);
}
}
}
return zRet;
}
static char *fts5ExprPrint(Fts5Config *pConfig, Fts5ExprNode *pExpr){
char *zRet = 0;
if( pExpr->eType==0 ){
return sqlite3_mprintf("\"\"");
}else
if( pExpr->eType==FTS5_STRING || pExpr->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pExpr->pNear;
int i;
int iTerm;
if( pNear->pColset ){
int ii;
Fts5Colset *pColset = pNear->pColset;
if( pColset->nCol>1 ) zRet = fts5PrintfAppend(zRet, "{");
for(ii=0; iinCol; ii++){
zRet = fts5PrintfAppend(zRet, "%s%s",
pConfig->azCol[pColset->aiCol[ii]], ii==pColset->nCol-1 ? "" : " "
);
}
if( zRet ){
zRet = fts5PrintfAppend(zRet, "%s : ", pColset->nCol>1 ? "}" : "");
}
if( zRet==0 ) return 0;
}
if( pNear->nPhrase>1 ){
zRet = fts5PrintfAppend(zRet, "NEAR(");
if( zRet==0 ) return 0;
}
for(i=0; inPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
if( i!=0 ){
zRet = fts5PrintfAppend(zRet, " ");
if( zRet==0 ) return 0;
}
for(iTerm=0; iTermnTerm; iTerm++){
char *zTerm = fts5ExprTermPrint(&pPhrase->aTerm[iTerm]);
if( zTerm ){
zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" + ", zTerm);
sqlite3_free(zTerm);
}
if( zTerm==0 || zRet==0 ){
sqlite3_free(zRet);
return 0;
}
}
}
if( pNear->nPhrase>1 ){
zRet = fts5PrintfAppend(zRet, ", %d)", pNear->nNear);
if( zRet==0 ) return 0;
}
}else{
char const *zOp = 0;
int i;
switch( pExpr->eType ){
case FTS5_AND: zOp = " AND "; break;
case FTS5_NOT: zOp = " NOT "; break;
default:
assert( pExpr->eType==FTS5_OR );
zOp = " OR ";
break;
}
for(i=0; inChild; i++){
char *z = fts5ExprPrint(pConfig, pExpr->apChild[i]);
if( z==0 ){
sqlite3_free(zRet);
zRet = 0;
}else{
int e = pExpr->apChild[i]->eType;
int b = (e!=FTS5_STRING && e!=FTS5_TERM && e!=FTS5_EOF);
zRet = fts5PrintfAppend(zRet, "%s%s%z%s",
(i==0 ? "" : zOp),
(b?"(":""), z, (b?")":"")
);
}
if( zRet==0 ) break;
}
}
return zRet;
}
/*
** The implementation of user-defined scalar functions fts5_expr() (bTcl==0)
** and fts5_expr_tcl() (bTcl!=0).
*/
static void fts5ExprFunction(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args */
sqlite3_value **apVal, /* Function arguments */
int bTcl
){
Fts5Global *pGlobal = (Fts5Global*)sqlite3_user_data(pCtx);
sqlite3 *db = sqlite3_context_db_handle(pCtx);
const char *zExpr = 0;
char *zErr = 0;
Fts5Expr *pExpr = 0;
int rc;
int i;
const char **azConfig; /* Array of arguments for Fts5Config */
const char *zNearsetCmd = "nearset";
int nConfig; /* Size of azConfig[] */
Fts5Config *pConfig = 0;
int iArg = 1;
if( nArg<1 ){
zErr = sqlite3_mprintf("wrong number of arguments to function %s",
bTcl ? "fts5_expr_tcl" : "fts5_expr"
);
sqlite3_result_error(pCtx, zErr, -1);
sqlite3_free(zErr);
return;
}
if( bTcl && nArg>1 ){
zNearsetCmd = (const char*)sqlite3_value_text(apVal[1]);
iArg = 2;
}
nConfig = 3 + (nArg-iArg);
azConfig = (const char**)sqlite3_malloc64(sizeof(char*) * nConfig);
if( azConfig==0 ){
sqlite3_result_error_nomem(pCtx);
return;
}
azConfig[0] = 0;
azConfig[1] = "main";
azConfig[2] = "tbl";
for(i=3; iArgnCol, zExpr, &pExpr, &zErr);
}
if( rc==SQLITE_OK ){
char *zText;
if( pExpr->pRoot->xNext==0 ){
zText = sqlite3_mprintf("");
}else if( bTcl ){
zText = fts5ExprPrintTcl(pConfig, zNearsetCmd, pExpr->pRoot);
}else{
zText = fts5ExprPrint(pConfig, pExpr->pRoot);
}
if( zText==0 ){
rc = SQLITE_NOMEM;
}else{
sqlite3_result_text(pCtx, zText, -1, SQLITE_TRANSIENT);
sqlite3_free(zText);
}
}
if( rc!=SQLITE_OK ){
if( zErr ){
sqlite3_result_error(pCtx, zErr, -1);
sqlite3_free(zErr);
}else{
sqlite3_result_error_code(pCtx, rc);
}
}
sqlite3_free((void *)azConfig);
sqlite3Fts5ConfigFree(pConfig);
sqlite3Fts5ExprFree(pExpr);
}
static void fts5ExprFunctionHr(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args */
sqlite3_value **apVal /* Function arguments */
){
fts5ExprFunction(pCtx, nArg, apVal, 0);
}
static void fts5ExprFunctionTcl(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args */
sqlite3_value **apVal /* Function arguments */
){
fts5ExprFunction(pCtx, nArg, apVal, 1);
}
/*
** The implementation of an SQLite user-defined-function that accepts a
** single integer as an argument. If the integer is an alpha-numeric
** unicode code point, 1 is returned. Otherwise 0.
*/
static void fts5ExprIsAlnum(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args */
sqlite3_value **apVal /* Function arguments */
){
int iCode;
u8 aArr[32];
if( nArg!=1 ){
sqlite3_result_error(pCtx,
"wrong number of arguments to function fts5_isalnum", -1
);
return;
}
memset(aArr, 0, sizeof(aArr));
sqlite3Fts5UnicodeCatParse("L*", aArr);
sqlite3Fts5UnicodeCatParse("N*", aArr);
sqlite3Fts5UnicodeCatParse("Co", aArr);
iCode = sqlite3_value_int(apVal[0]);
sqlite3_result_int(pCtx, aArr[sqlite3Fts5UnicodeCategory((u32)iCode)]);
}
static void fts5ExprFold(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args */
sqlite3_value **apVal /* Function arguments */
){
if( nArg!=1 && nArg!=2 ){
sqlite3_result_error(pCtx,
"wrong number of arguments to function fts5_fold", -1
);
}else{
int iCode;
int bRemoveDiacritics = 0;
iCode = sqlite3_value_int(apVal[0]);
if( nArg==2 ) bRemoveDiacritics = sqlite3_value_int(apVal[1]);
sqlite3_result_int(pCtx, sqlite3Fts5UnicodeFold(iCode, bRemoveDiacritics));
}
}
#endif /* if SQLITE_TEST || SQLITE_FTS5_DEBUG */
/*
** This is called during initialization to register the fts5_expr() scalar
** UDF with the SQLite handle passed as the only argument.
*/
static int sqlite3Fts5ExprInit(Fts5Global *pGlobal, sqlite3 *db){
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
struct Fts5ExprFunc {
const char *z;
void (*x)(sqlite3_context*,int,sqlite3_value**);
} aFunc[] = {
{ "fts5_expr", fts5ExprFunctionHr },
{ "fts5_expr_tcl", fts5ExprFunctionTcl },
{ "fts5_isalnum", fts5ExprIsAlnum },
{ "fts5_fold", fts5ExprFold },
};
int i;
int rc = SQLITE_OK;
void *pCtx = (void*)pGlobal;
for(i=0; rc==SQLITE_OK && iz, -1, SQLITE_UTF8, pCtx, p->x, 0, 0);
}
#else
int rc = SQLITE_OK;
UNUSED_PARAM2(pGlobal,db);
#endif
/* Avoid warnings indicating that sqlite3Fts5ParserTrace() and
** sqlite3Fts5ParserFallback() are unused */
#ifndef NDEBUG
(void)sqlite3Fts5ParserTrace;
#endif
(void)sqlite3Fts5ParserFallback;
return rc;
}
/*
** Return the number of phrases in expression pExpr.
*/
static int sqlite3Fts5ExprPhraseCount(Fts5Expr *pExpr){
return (pExpr ? pExpr->nPhrase : 0);
}
/*
** Return the number of terms in the iPhrase'th phrase in pExpr.
*/
static int sqlite3Fts5ExprPhraseSize(Fts5Expr *pExpr, int iPhrase){
if( iPhrase<0 || iPhrase>=pExpr->nPhrase ) return 0;
return pExpr->apExprPhrase[iPhrase]->nTerm;
}
/*
** This function is used to access the current position list for phrase
** iPhrase.
*/
static int sqlite3Fts5ExprPoslist(Fts5Expr *pExpr, int iPhrase, const u8 **pa){
int nRet;
Fts5ExprPhrase *pPhrase = pExpr->apExprPhrase[iPhrase];
Fts5ExprNode *pNode = pPhrase->pNode;
if( pNode->bEof==0 && pNode->iRowid==pExpr->pRoot->iRowid ){
*pa = pPhrase->poslist.p;
nRet = pPhrase->poslist.n;
}else{
*pa = 0;
nRet = 0;
}
return nRet;
}
struct Fts5PoslistPopulator {
Fts5PoslistWriter writer;
int bOk; /* True if ok to populate */
int bMiss;
};
/*
** Clear the position lists associated with all phrases in the expression
** passed as the first argument. Argument bLive is true if the expression
** might be pointing to a real entry, otherwise it has just been reset.
**
** At present this function is only used for detail=col and detail=none
** fts5 tables. This implies that all phrases must be at most 1 token
** in size, as phrase matches are not supported without detail=full.
*/
static Fts5PoslistPopulator *sqlite3Fts5ExprClearPoslists(Fts5Expr *pExpr, int bLive){
Fts5PoslistPopulator *pRet;
pRet = sqlite3_malloc64(sizeof(Fts5PoslistPopulator)*pExpr->nPhrase);
if( pRet ){
int i;
memset(pRet, 0, sizeof(Fts5PoslistPopulator)*pExpr->nPhrase);
for(i=0; inPhrase; i++){
Fts5Buffer *pBuf = &pExpr->apExprPhrase[i]->poslist;
Fts5ExprNode *pNode = pExpr->apExprPhrase[i]->pNode;
assert( pExpr->apExprPhrase[i]->nTerm<=1 );
if( bLive &&
(pBuf->n==0 || pNode->iRowid!=pExpr->pRoot->iRowid || pNode->bEof)
){
pRet[i].bMiss = 1;
}else{
pBuf->n = 0;
}
}
}
return pRet;
}
struct Fts5ExprCtx {
Fts5Expr *pExpr;
Fts5PoslistPopulator *aPopulator;
i64 iOff;
};
typedef struct Fts5ExprCtx Fts5ExprCtx;
/*
** TODO: Make this more efficient!
*/
static int fts5ExprColsetTest(Fts5Colset *pColset, int iCol){
int i;
for(i=0; inCol; i++){
if( pColset->aiCol[i]==iCol ) return 1;
}
return 0;
}
/*
** pToken is a buffer nToken bytes in size that may or may not contain
** an embedded 0x00 byte. If it does, return the number of bytes in
** the buffer before the 0x00. If it does not, return nToken.
*/
static int fts5QueryTerm(const char *pToken, int nToken){
int ii;
for(ii=0; iipExpr;
int i;
int nQuery = nToken;
i64 iRowid = pExpr->pRoot->iRowid;
UNUSED_PARAM2(iUnused1, iUnused2);
if( nQuery>FTS5_MAX_TOKEN_SIZE ) nQuery = FTS5_MAX_TOKEN_SIZE;
if( pExpr->pConfig->bTokendata ){
nQuery = fts5QueryTerm(pToken, nQuery);
}
if( (tflags & FTS5_TOKEN_COLOCATED)==0 ) p->iOff++;
for(i=0; inPhrase; i++){
Fts5ExprTerm *pT;
if( p->aPopulator[i].bOk==0 ) continue;
for(pT=&pExpr->apExprPhrase[i]->aTerm[0]; pT; pT=pT->pSynonym){
if( (pT->nQueryTerm==nQuery || (pT->nQueryTermbPrefix))
&& memcmp(pT->pTerm, pToken, pT->nQueryTerm)==0
){
int rc = sqlite3Fts5PoslistWriterAppend(
&pExpr->apExprPhrase[i]->poslist, &p->aPopulator[i].writer, p->iOff
);
if( rc==SQLITE_OK && pExpr->pConfig->bTokendata && !pT->bPrefix ){
int iCol = p->iOff>>32;
int iTokOff = p->iOff & 0x7FFFFFFF;
rc = sqlite3Fts5IndexIterWriteTokendata(
pT->pIter, pToken, nToken, iRowid, iCol, iTokOff
);
}
if( rc ) return rc;
break;
}
}
}
return SQLITE_OK;
}
static int sqlite3Fts5ExprPopulatePoslists(
Fts5Config *pConfig,
Fts5Expr *pExpr,
Fts5PoslistPopulator *aPopulator,
int iCol,
const char *z, int n
){
int i;
Fts5ExprCtx sCtx;
sCtx.pExpr = pExpr;
sCtx.aPopulator = aPopulator;
sCtx.iOff = (((i64)iCol) << 32) - 1;
for(i=0; inPhrase; i++){
Fts5ExprNode *pNode = pExpr->apExprPhrase[i]->pNode;
Fts5Colset *pColset = pNode->pNear->pColset;
if( (pColset && 0==fts5ExprColsetTest(pColset, iCol))
|| aPopulator[i].bMiss
){
aPopulator[i].bOk = 0;
}else{
aPopulator[i].bOk = 1;
}
}
return sqlite3Fts5Tokenize(pConfig,
FTS5_TOKENIZE_DOCUMENT, z, n, (void*)&sCtx, fts5ExprPopulatePoslistsCb
);
}
static void fts5ExprClearPoslists(Fts5ExprNode *pNode){
if( pNode->eType==FTS5_TERM || pNode->eType==FTS5_STRING ){
pNode->pNear->apPhrase[0]->poslist.n = 0;
}else{
int i;
for(i=0; inChild; i++){
fts5ExprClearPoslists(pNode->apChild[i]);
}
}
}
static int fts5ExprCheckPoslists(Fts5ExprNode *pNode, i64 iRowid){
pNode->iRowid = iRowid;
pNode->bEof = 0;
switch( pNode->eType ){
case FTS5_TERM:
case FTS5_STRING:
return (pNode->pNear->apPhrase[0]->poslist.n>0);
case FTS5_AND: {
int i;
for(i=0; inChild; i++){
if( fts5ExprCheckPoslists(pNode->apChild[i], iRowid)==0 ){
fts5ExprClearPoslists(pNode);
return 0;
}
}
break;
}
case FTS5_OR: {
int i;
int bRet = 0;
for(i=0; inChild; i++){
if( fts5ExprCheckPoslists(pNode->apChild[i], iRowid) ){
bRet = 1;
}
}
return bRet;
}
default: {
assert( pNode->eType==FTS5_NOT );
if( 0==fts5ExprCheckPoslists(pNode->apChild[0], iRowid)
|| 0!=fts5ExprCheckPoslists(pNode->apChild[1], iRowid)
){
fts5ExprClearPoslists(pNode);
return 0;
}
break;
}
}
return 1;
}
static void sqlite3Fts5ExprCheckPoslists(Fts5Expr *pExpr, i64 iRowid){
fts5ExprCheckPoslists(pExpr->pRoot, iRowid);
}
/*
** This function is only called for detail=columns tables.
*/
static int sqlite3Fts5ExprPhraseCollist(
Fts5Expr *pExpr,
int iPhrase,
const u8 **ppCollist,
int *pnCollist
){
Fts5ExprPhrase *pPhrase = pExpr->apExprPhrase[iPhrase];
Fts5ExprNode *pNode = pPhrase->pNode;
int rc = SQLITE_OK;
assert( iPhrase>=0 && iPhrasenPhrase );
assert( pExpr->pConfig->eDetail==FTS5_DETAIL_COLUMNS );
if( pNode->bEof==0
&& pNode->iRowid==pExpr->pRoot->iRowid
&& pPhrase->poslist.n>0
){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[0];
if( pTerm->pSynonym ){
Fts5Buffer *pBuf = (Fts5Buffer*)&pTerm->pSynonym[1];
rc = fts5ExprSynonymList(
pTerm, pNode->iRowid, pBuf, (u8**)ppCollist, pnCollist
);
}else{
*ppCollist = pPhrase->aTerm[0].pIter->pData;
*pnCollist = pPhrase->aTerm[0].pIter->nData;
}
}else{
*ppCollist = 0;
*pnCollist = 0;
}
return rc;
}
/*
** Does the work of the fts5_api.xQueryToken() API method.
*/
static int sqlite3Fts5ExprQueryToken(
Fts5Expr *pExpr,
int iPhrase,
int iToken,
const char **ppOut,
int *pnOut
){
Fts5ExprPhrase *pPhrase = 0;
if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){
return SQLITE_RANGE;
}
pPhrase = pExpr->apExprPhrase[iPhrase];
if( iToken<0 || iToken>=pPhrase->nTerm ){
return SQLITE_RANGE;
}
*ppOut = pPhrase->aTerm[iToken].pTerm;
*pnOut = pPhrase->aTerm[iToken].nFullTerm;
return SQLITE_OK;
}
/*
** Does the work of the fts5_api.xInstToken() API method.
*/
static int sqlite3Fts5ExprInstToken(
Fts5Expr *pExpr,
i64 iRowid,
int iPhrase,
int iCol,
int iOff,
int iToken,
const char **ppOut,
int *pnOut
){
Fts5ExprPhrase *pPhrase = 0;
Fts5ExprTerm *pTerm = 0;
int rc = SQLITE_OK;
if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){
return SQLITE_RANGE;
}
pPhrase = pExpr->apExprPhrase[iPhrase];
if( iToken<0 || iToken>=pPhrase->nTerm ){
return SQLITE_RANGE;
}
pTerm = &pPhrase->aTerm[iToken];
if( pTerm->bPrefix==0 ){
if( pExpr->pConfig->bTokendata ){
rc = sqlite3Fts5IterToken(
pTerm->pIter, iRowid, iCol, iOff+iToken, ppOut, pnOut
);
}else{
*ppOut = pTerm->pTerm;
*pnOut = pTerm->nFullTerm;
}
}
return rc;
}
/*
** Clear the token mappings for all Fts5IndexIter objects mannaged by
** the expression passed as the only argument.
*/
static void sqlite3Fts5ExprClearTokens(Fts5Expr *pExpr){
int ii;
for(ii=0; iinPhrase; ii++){
Fts5ExprTerm *pT;
for(pT=&pExpr->apExprPhrase[ii]->aTerm[0]; pT; pT=pT->pSynonym){
sqlite3Fts5IndexIterClearTokendata(pT->pIter);
}
}
}
#line 1 "fts5_hash.c"
/*
** 2014 August 11
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
*/
/* #include "fts5Int.h" */
typedef struct Fts5HashEntry Fts5HashEntry;
/*
** This file contains the implementation of an in-memory hash table used
** to accumuluate "term -> doclist" content before it is flused to a level-0
** segment.
*/
struct Fts5Hash {
int eDetail; /* Copy of Fts5Config.eDetail */
int *pnByte; /* Pointer to bytes counter */
int nEntry; /* Number of entries currently in hash */
int nSlot; /* Size of aSlot[] array */
Fts5HashEntry *pScan; /* Current ordered scan item */
Fts5HashEntry **aSlot; /* Array of hash slots */
};
/*
** Each entry in the hash table is represented by an object of the
** following type. Each object, its key, and its current data are stored
** in a single memory allocation. The key immediately follows the object
** in memory. The position list data immediately follows the key data
** in memory.
**
** The key is Fts5HashEntry.nKey bytes in size. It consists of a single
** byte identifying the index (either the main term index or a prefix-index),
** followed by the term data. For example: "0token". There is no
** nul-terminator - in this case nKey=6.
**
** The data that follows the key is in a similar, but not identical format
** to the doclist data stored in the database. It is:
**
** * Rowid, as a varint
** * Position list, without 0x00 terminator.
** * Size of previous position list and rowid, as a 4 byte
** big-endian integer.
**
** iRowidOff:
** Offset of last rowid written to data area. Relative to first byte of
** structure.
**
** nData:
** Bytes of data written since iRowidOff.
*/
struct Fts5HashEntry {
Fts5HashEntry *pHashNext; /* Next hash entry with same hash-key */
Fts5HashEntry *pScanNext; /* Next entry in sorted order */
int nAlloc; /* Total size of allocation */
int iSzPoslist; /* Offset of space for 4-byte poslist size */
int nData; /* Total bytes of data (incl. structure) */
int nKey; /* Length of key in bytes */
u8 bDel; /* Set delete-flag @ iSzPoslist */
u8 bContent; /* Set content-flag (detail=none mode) */
i16 iCol; /* Column of last value written */
int iPos; /* Position of last value written */
i64 iRowid; /* Rowid of last value written */
};
/*
** Eqivalent to:
**
** char *fts5EntryKey(Fts5HashEntry *pEntry){ return zKey; }
*/
#define fts5EntryKey(p) ( ((char *)(&(p)[1])) )
/*
** Allocate a new hash table.
*/
static int sqlite3Fts5HashNew(Fts5Config *pConfig, Fts5Hash **ppNew, int *pnByte){
int rc = SQLITE_OK;
Fts5Hash *pNew;
*ppNew = pNew = (Fts5Hash*)sqlite3_malloc(sizeof(Fts5Hash));
if( pNew==0 ){
rc = SQLITE_NOMEM;
}else{
sqlite3_int64 nByte;
memset(pNew, 0, sizeof(Fts5Hash));
pNew->pnByte = pnByte;
pNew->eDetail = pConfig->eDetail;
pNew->nSlot = 1024;
nByte = sizeof(Fts5HashEntry*) * pNew->nSlot;
pNew->aSlot = (Fts5HashEntry**)sqlite3_malloc64(nByte);
if( pNew->aSlot==0 ){
sqlite3_free(pNew);
*ppNew = 0;
rc = SQLITE_NOMEM;
}else{
memset(pNew->aSlot, 0, (size_t)nByte);
}
}
return rc;
}
/*
** Free a hash table object.
*/
static void sqlite3Fts5HashFree(Fts5Hash *pHash){
if( pHash ){
sqlite3Fts5HashClear(pHash);
sqlite3_free(pHash->aSlot);
sqlite3_free(pHash);
}
}
/*
** Empty (but do not delete) a hash table.
*/
static void sqlite3Fts5HashClear(Fts5Hash *pHash){
int i;
for(i=0; inSlot; i++){
Fts5HashEntry *pNext;
Fts5HashEntry *pSlot;
for(pSlot=pHash->aSlot[i]; pSlot; pSlot=pNext){
pNext = pSlot->pHashNext;
sqlite3_free(pSlot);
}
}
memset(pHash->aSlot, 0, pHash->nSlot * sizeof(Fts5HashEntry*));
pHash->nEntry = 0;
}
static unsigned int fts5HashKey(int nSlot, const u8 *p, int n){
int i;
unsigned int h = 13;
for(i=n-1; i>=0; i--){
h = (h << 3) ^ h ^ p[i];
}
return (h % nSlot);
}
static unsigned int fts5HashKey2(int nSlot, u8 b, const u8 *p, int n){
int i;
unsigned int h = 13;
for(i=n-1; i>=0; i--){
h = (h << 3) ^ h ^ p[i];
}
h = (h << 3) ^ h ^ b;
return (h % nSlot);
}
/*
** Resize the hash table by doubling the number of slots.
*/
static int fts5HashResize(Fts5Hash *pHash){
int nNew = pHash->nSlot*2;
int i;
Fts5HashEntry **apNew;
Fts5HashEntry **apOld = pHash->aSlot;
apNew = (Fts5HashEntry**)sqlite3_malloc64(nNew*sizeof(Fts5HashEntry*));
if( !apNew ) return SQLITE_NOMEM;
memset(apNew, 0, nNew*sizeof(Fts5HashEntry*));
for(i=0; inSlot; i++){
while( apOld[i] ){
unsigned int iHash;
Fts5HashEntry *p = apOld[i];
apOld[i] = p->pHashNext;
iHash = fts5HashKey(nNew, (u8*)fts5EntryKey(p), p->nKey);
p->pHashNext = apNew[iHash];
apNew[iHash] = p;
}
}
sqlite3_free(apOld);
pHash->nSlot = nNew;
pHash->aSlot = apNew;
return SQLITE_OK;
}
static int fts5HashAddPoslistSize(
Fts5Hash *pHash,
Fts5HashEntry *p,
Fts5HashEntry *p2
){
int nRet = 0;
if( p->iSzPoslist ){
u8 *pPtr = p2 ? (u8*)p2 : (u8*)p;
int nData = p->nData;
if( pHash->eDetail==FTS5_DETAIL_NONE ){
assert( nData==p->iSzPoslist );
if( p->bDel ){
pPtr[nData++] = 0x00;
if( p->bContent ){
pPtr[nData++] = 0x00;
}
}
}else{
int nSz = (nData - p->iSzPoslist - 1); /* Size in bytes */
int nPos = nSz*2 + p->bDel; /* Value of nPos field */
assert( p->bDel==0 || p->bDel==1 );
if( nPos<=127 ){
pPtr[p->iSzPoslist] = (u8)nPos;
}else{
int nByte = sqlite3Fts5GetVarintLen((u32)nPos);
memmove(&pPtr[p->iSzPoslist + nByte], &pPtr[p->iSzPoslist + 1], nSz);
sqlite3Fts5PutVarint(&pPtr[p->iSzPoslist], nPos);
nData += (nByte-1);
}
}
nRet = nData - p->nData;
if( p2==0 ){
p->iSzPoslist = 0;
p->bDel = 0;
p->bContent = 0;
p->nData = nData;
}
}
return nRet;
}
/*
** Add an entry to the in-memory hash table. The key is the concatenation
** of bByte and (pToken/nToken). The value is (iRowid/iCol/iPos).
**
** (bByte || pToken) -> (iRowid,iCol,iPos)
**
** Or, if iCol is negative, then the value is a delete marker.
*/
static int sqlite3Fts5HashWrite(
Fts5Hash *pHash,
i64 iRowid, /* Rowid for this entry */
int iCol, /* Column token appears in (-ve -> delete) */
int iPos, /* Position of token within column */
char bByte, /* First byte of token */
const char *pToken, int nToken /* Token to add or remove to or from index */
){
unsigned int iHash;
Fts5HashEntry *p;
u8 *pPtr;
int nIncr = 0; /* Amount to increment (*pHash->pnByte) by */
int bNew; /* If non-delete entry should be written */
bNew = (pHash->eDetail==FTS5_DETAIL_FULL);
/* Attempt to locate an existing hash entry */
iHash = fts5HashKey2(pHash->nSlot, (u8)bByte, (const u8*)pToken, nToken);
for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){
char *zKey = fts5EntryKey(p);
if( zKey[0]==bByte
&& p->nKey==nToken+1
&& memcmp(&zKey[1], pToken, nToken)==0
){
break;
}
}
/* If an existing hash entry cannot be found, create a new one. */
if( p==0 ){
/* Figure out how much space to allocate */
char *zKey;
sqlite3_int64 nByte = sizeof(Fts5HashEntry) + (nToken+1) + 1 + 64;
if( nByte<128 ) nByte = 128;
/* Grow the Fts5Hash.aSlot[] array if necessary. */
if( (pHash->nEntry*2)>=pHash->nSlot ){
int rc = fts5HashResize(pHash);
if( rc!=SQLITE_OK ) return rc;
iHash = fts5HashKey2(pHash->nSlot, (u8)bByte, (const u8*)pToken, nToken);
}
/* Allocate new Fts5HashEntry and add it to the hash table. */
p = (Fts5HashEntry*)sqlite3_malloc64(nByte);
if( !p ) return SQLITE_NOMEM;
memset(p, 0, sizeof(Fts5HashEntry));
p->nAlloc = (int)nByte;
zKey = fts5EntryKey(p);
zKey[0] = bByte;
memcpy(&zKey[1], pToken, nToken);
assert( iHash==fts5HashKey(pHash->nSlot, (u8*)zKey, nToken+1) );
p->nKey = nToken+1;
zKey[nToken+1] = '\0';
p->nData = nToken+1 + sizeof(Fts5HashEntry);
p->pHashNext = pHash->aSlot[iHash];
pHash->aSlot[iHash] = p;
pHash->nEntry++;
/* Add the first rowid field to the hash-entry */
p->nData += sqlite3Fts5PutVarint(&((u8*)p)[p->nData], iRowid);
p->iRowid = iRowid;
p->iSzPoslist = p->nData;
if( pHash->eDetail!=FTS5_DETAIL_NONE ){
p->nData += 1;
p->iCol = (pHash->eDetail==FTS5_DETAIL_FULL ? 0 : -1);
}
}else{
/* Appending to an existing hash-entry. Check that there is enough
** space to append the largest possible new entry. Worst case scenario
** is:
**
** + 9 bytes for a new rowid,
** + 4 byte reserved for the "poslist size" varint.
** + 1 byte for a "new column" byte,
** + 3 bytes for a new column number (16-bit max) as a varint,
** + 5 bytes for the new position offset (32-bit max).
*/
if( (p->nAlloc - p->nData) < (9 + 4 + 1 + 3 + 5) ){
sqlite3_int64 nNew = p->nAlloc * 2;
Fts5HashEntry *pNew;
Fts5HashEntry **pp;
pNew = (Fts5HashEntry*)sqlite3_realloc64(p, nNew);
if( pNew==0 ) return SQLITE_NOMEM;
pNew->nAlloc = (int)nNew;
for(pp=&pHash->aSlot[iHash]; *pp!=p; pp=&(*pp)->pHashNext);
*pp = pNew;
p = pNew;
}
nIncr -= p->nData;
}
assert( (p->nAlloc - p->nData) >= (9 + 4 + 1 + 3 + 5) );
pPtr = (u8*)p;
/* If this is a new rowid, append the 4-byte size field for the previous
** entry, and the new rowid for this entry. */
if( iRowid!=p->iRowid ){
u64 iDiff = (u64)iRowid - (u64)p->iRowid;
fts5HashAddPoslistSize(pHash, p, 0);
p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iDiff);
p->iRowid = iRowid;
bNew = 1;
p->iSzPoslist = p->nData;
if( pHash->eDetail!=FTS5_DETAIL_NONE ){
p->nData += 1;
p->iCol = (pHash->eDetail==FTS5_DETAIL_FULL ? 0 : -1);
p->iPos = 0;
}
}
if( iCol>=0 ){
if( pHash->eDetail==FTS5_DETAIL_NONE ){
p->bContent = 1;
}else{
/* Append a new column value, if necessary */
assert_nc( iCol>=p->iCol );
if( iCol!=p->iCol ){
if( pHash->eDetail==FTS5_DETAIL_FULL ){
pPtr[p->nData++] = 0x01;
p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iCol);
p->iCol = (i16)iCol;
p->iPos = 0;
}else{
bNew = 1;
p->iCol = (i16)(iPos = iCol);
}
}
/* Append the new position offset, if necessary */
if( bNew ){
p->nData += sqlite3Fts5PutVarint(&pPtr[p->nData], iPos - p->iPos + 2);
p->iPos = iPos;
}
}
}else{
/* This is a delete. Set the delete flag. */
p->bDel = 1;
}
nIncr += p->nData;
*pHash->pnByte += nIncr;
return SQLITE_OK;
}
/*
** Arguments pLeft and pRight point to linked-lists of hash-entry objects,
** each sorted in key order. This function merges the two lists into a
** single list and returns a pointer to its first element.
*/
static Fts5HashEntry *fts5HashEntryMerge(
Fts5HashEntry *pLeft,
Fts5HashEntry *pRight
){
Fts5HashEntry *p1 = pLeft;
Fts5HashEntry *p2 = pRight;
Fts5HashEntry *pRet = 0;
Fts5HashEntry **ppOut = &pRet;
while( p1 || p2 ){
if( p1==0 ){
*ppOut = p2;
p2 = 0;
}else if( p2==0 ){
*ppOut = p1;
p1 = 0;
}else{
char *zKey1 = fts5EntryKey(p1);
char *zKey2 = fts5EntryKey(p2);
int nMin = MIN(p1->nKey, p2->nKey);
int cmp = memcmp(zKey1, zKey2, nMin);
if( cmp==0 ){
cmp = p1->nKey - p2->nKey;
}
assert( cmp!=0 );
if( cmp>0 ){
/* p2 is smaller */
*ppOut = p2;
ppOut = &p2->pScanNext;
p2 = p2->pScanNext;
}else{
/* p1 is smaller */
*ppOut = p1;
ppOut = &p1->pScanNext;
p1 = p1->pScanNext;
}
*ppOut = 0;
}
}
return pRet;
}
/*
** Link all tokens from hash table iHash into a list in sorted order. The
** tokens are not removed from the hash table.
*/
static int fts5HashEntrySort(
Fts5Hash *pHash,
const char *pTerm, int nTerm, /* Query prefix, if any */
Fts5HashEntry **ppSorted
){
const int nMergeSlot = 32;
Fts5HashEntry **ap;
Fts5HashEntry *pList;
int iSlot;
int i;
*ppSorted = 0;
ap = sqlite3_malloc64(sizeof(Fts5HashEntry*) * nMergeSlot);
if( !ap ) return SQLITE_NOMEM;
memset(ap, 0, sizeof(Fts5HashEntry*) * nMergeSlot);
for(iSlot=0; iSlotnSlot; iSlot++){
Fts5HashEntry *pIter;
for(pIter=pHash->aSlot[iSlot]; pIter; pIter=pIter->pHashNext){
if( pTerm==0
|| (pIter->nKey>=nTerm && 0==memcmp(fts5EntryKey(pIter), pTerm, nTerm))
){
Fts5HashEntry *pEntry = pIter;
pEntry->pScanNext = 0;
for(i=0; ap[i]; i++){
pEntry = fts5HashEntryMerge(pEntry, ap[i]);
ap[i] = 0;
}
ap[i] = pEntry;
}
}
}
pList = 0;
for(i=0; inSlot, (const u8*)pTerm, nTerm);
char *zKey = 0;
Fts5HashEntry *p;
for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){
zKey = fts5EntryKey(p);
if( nTerm==p->nKey && memcmp(zKey, pTerm, nTerm)==0 ) break;
}
if( p ){
int nHashPre = sizeof(Fts5HashEntry) + nTerm;
int nList = p->nData - nHashPre;
u8 *pRet = (u8*)(*ppOut = sqlite3_malloc64(nPre + nList + 10));
if( pRet ){
Fts5HashEntry *pFaux = (Fts5HashEntry*)&pRet[nPre-nHashPre];
memcpy(&pRet[nPre], &((u8*)p)[nHashPre], nList);
nList += fts5HashAddPoslistSize(pHash, p, pFaux);
*pnDoclist = nList;
}else{
*pnDoclist = 0;
return SQLITE_NOMEM;
}
}else{
*ppOut = 0;
*pnDoclist = 0;
}
return SQLITE_OK;
}
static int sqlite3Fts5HashScanInit(
Fts5Hash *p, /* Hash table to query */
const char *pTerm, int nTerm /* Query prefix */
){
return fts5HashEntrySort(p, pTerm, nTerm, &p->pScan);
}
#ifdef SQLITE_DEBUG
static int fts5HashCount(Fts5Hash *pHash){
int nEntry = 0;
int ii;
for(ii=0; iinSlot; ii++){
Fts5HashEntry *p = 0;
for(p=pHash->aSlot[ii]; p; p=p->pHashNext){
nEntry++;
}
}
return nEntry;
}
#endif
/*
** Return true if the hash table is empty, false otherwise.
*/
static int sqlite3Fts5HashIsEmpty(Fts5Hash *pHash){
assert( pHash->nEntry==fts5HashCount(pHash) );
return pHash->nEntry==0;
}
static void sqlite3Fts5HashScanNext(Fts5Hash *p){
assert( !sqlite3Fts5HashScanEof(p) );
p->pScan = p->pScan->pScanNext;
}
static int sqlite3Fts5HashScanEof(Fts5Hash *p){
return (p->pScan==0);
}
static void sqlite3Fts5HashScanEntry(
Fts5Hash *pHash,
const char **pzTerm, /* OUT: term (nul-terminated) */
int *pnTerm, /* OUT: Size of term in bytes */
const u8 **ppDoclist, /* OUT: pointer to doclist */
int *pnDoclist /* OUT: size of doclist in bytes */
){
Fts5HashEntry *p;
if( (p = pHash->pScan) ){
char *zKey = fts5EntryKey(p);
int nTerm = p->nKey;
fts5HashAddPoslistSize(pHash, p, 0);
*pzTerm = zKey;
*pnTerm = nTerm;
*ppDoclist = (const u8*)&zKey[nTerm];
*pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm);
}else{
*pzTerm = 0;
*pnTerm = 0;
*ppDoclist = 0;
*pnDoclist = 0;
}
}
#line 1 "fts5_index.c"
/*
** 2014 May 31
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
** Low level access to the FTS index stored in the database file. The
** routines in this file file implement all read and write access to the
** %_data table. Other parts of the system access this functionality via
** the interface defined in fts5Int.h.
*/
/* #include "fts5Int.h" */
/*
** Overview:
**
** The %_data table contains all the FTS indexes for an FTS5 virtual table.
** As well as the main term index, there may be up to 31 prefix indexes.
** The format is similar to FTS3/4, except that:
**
** * all segment b-tree leaf data is stored in fixed size page records
** (e.g. 1000 bytes). A single doclist may span multiple pages. Care is
** taken to ensure it is possible to iterate in either direction through
** the entries in a doclist, or to seek to a specific entry within a
** doclist, without loading it into memory.
**
** * large doclists that span many pages have associated "doclist index"
** records that contain a copy of the first rowid on each page spanned by
** the doclist. This is used to speed up seek operations, and merges of
** large doclists with very small doclists.
**
** * extra fields in the "structure record" record the state of ongoing
** incremental merge operations.
**
*/
#define FTS5_OPT_WORK_UNIT 1000 /* Number of leaf pages per optimize step */
#define FTS5_WORK_UNIT 64 /* Number of leaf pages in unit of work */
#define FTS5_MIN_DLIDX_SIZE 4 /* Add dlidx if this many empty pages */
#define FTS5_MAIN_PREFIX '0'
#if FTS5_MAX_PREFIX_INDEXES > 31
# error "FTS5_MAX_PREFIX_INDEXES is too large"
#endif
#define FTS5_MAX_LEVEL 64
/*
** There are two versions of the format used for the structure record:
**
** 1. the legacy format, that may be read by all fts5 versions, and
**
** 2. the V2 format, which is used by contentless_delete=1 databases.
**
** Both begin with a 4-byte "configuration cookie" value. Then, a legacy
** format structure record contains a varint - the number of levels in
** the structure. Whereas a V2 structure record contains the constant
** 4 bytes [0xff 0x00 0x00 0x01]. This is unambiguous as the value of a
** varint has to be at least 16256 to begin with "0xFF". And the default
** maximum number of levels is 64.
**
** See below for more on structure record formats.
*/
#define FTS5_STRUCTURE_V2 "\xFF\x00\x00\x01"
/*
** Details:
**
** The %_data table managed by this module,
**
** CREATE TABLE %_data(id INTEGER PRIMARY KEY, block BLOB);
**
** , contains the following 6 types of records. See the comments surrounding
** the FTS5_*_ROWID macros below for a description of how %_data rowids are
** assigned to each fo them.
**
** 1. Structure Records:
**
** The set of segments that make up an index - the index structure - are
** recorded in a single record within the %_data table. The record consists
** of a single 32-bit configuration cookie value followed by a list of
** SQLite varints.
**
** If the structure record is a V2 record, the configuration cookie is
** followed by the following 4 bytes: [0xFF 0x00 0x00 0x01].
**
** Next, the record continues with three varints:
**
** + number of levels,
** + total number of segments on all levels,
** + value of write counter.
**
** Then, for each level from 0 to nMax:
**
** + number of input segments in ongoing merge.
** + total number of segments in level.
** + for each segment from oldest to newest:
** + segment id (always > 0)
** + first leaf page number (often 1, always greater than 0)
** + final leaf page number
**
** Then, for V2 structures only:
**
** + lower origin counter value,
** + upper origin counter value,
** + the number of tombstone hash pages.
**
** 2. The Averages Record:
**
** A single record within the %_data table. The data is a list of varints.
** The first value is the number of rows in the index. Then, for each column
** from left to right, the total number of tokens in the column for all
** rows of the table.
**
** 3. Segment leaves:
**
** TERM/DOCLIST FORMAT:
**
** Most of each segment leaf is taken up by term/doclist data. The
** general format of term/doclist, starting with the first term
** on the leaf page, is:
**
** varint : size of first term
** blob: first term data
** doclist: first doclist
** zero-or-more {
** varint: number of bytes in common with previous term
** varint: number of bytes of new term data (nNew)
** blob: nNew bytes of new term data
** doclist: next doclist
** }
**
** doclist format:
**
** varint: first rowid
** poslist: first poslist
** zero-or-more {
** varint: rowid delta (always > 0)
** poslist: next poslist
** }
**
** poslist format:
**
** varint: size of poslist in bytes multiplied by 2, not including
** this field. Plus 1 if this entry carries the "delete" flag.
** collist: collist for column 0
** zero-or-more {
** 0x01 byte
** varint: column number (I)
** collist: collist for column I
** }
**
** collist format:
**
** varint: first offset + 2
** zero-or-more {
** varint: offset delta + 2
** }
**
** PAGE FORMAT
**
** Each leaf page begins with a 4-byte header containing 2 16-bit
** unsigned integer fields in big-endian format. They are:
**
** * The byte offset of the first rowid on the page, if it exists
** and occurs before the first term (otherwise 0).
**
** * The byte offset of the start of the page footer. If the page
** footer is 0 bytes in size, then this field is the same as the
** size of the leaf page in bytes.
**
** The page footer consists of a single varint for each term located
** on the page. Each varint is the byte offset of the current term
** within the page, delta-compressed against the previous value. In
** other words, the first varint in the footer is the byte offset of
** the first term, the second is the byte offset of the second less that
** of the first, and so on.
**
** The term/doclist format described above is accurate if the entire
** term/doclist data fits on a single leaf page. If this is not the case,
** the format is changed in two ways:
**
** + if the first rowid on a page occurs before the first term, it
** is stored as a literal value:
**
** varint: first rowid
**
** + the first term on each page is stored in the same way as the
** very first term of the segment:
**
** varint : size of first term
** blob: first term data
**
** 5. Segment doclist indexes:
**
** Doclist indexes are themselves b-trees, however they usually consist of
** a single leaf record only. The format of each doclist index leaf page
** is:
**
** * Flags byte. Bits are:
** 0x01: Clear if leaf is also the root page, otherwise set.
**
** * Page number of fts index leaf page. As a varint.
**
** * First rowid on page indicated by previous field. As a varint.
**
** * A list of varints, one for each subsequent termless page. A
** positive delta if the termless page contains at least one rowid,
** or an 0x00 byte otherwise.
**
** Internal doclist index nodes are:
**
** * Flags byte. Bits are:
** 0x01: Clear for root page, otherwise set.
**
** * Page number of first child page. As a varint.
**
** * Copy of first rowid on page indicated by previous field. As a varint.
**
** * A list of delta-encoded varints - the first rowid on each subsequent
** child page.
**
** 6. Tombstone Hash Page
**
** These records are only ever present in contentless_delete=1 tables.
** There are zero or more of these associated with each segment. They
** are used to store the tombstone rowids for rows contained in the
** associated segments.
**
** The set of nHashPg tombstone hash pages associated with a single
** segment together form a single hash table containing tombstone rowids.
** To find the page of the hash on which a key might be stored:
**
** iPg = (rowid % nHashPg)
**
** Then, within page iPg, which has nSlot slots:
**
** iSlot = (rowid / nHashPg) % nSlot
**
** Each tombstone hash page begins with an 8 byte header:
**
** 1-byte: Key-size (the size in bytes of each slot). Either 4 or 8.
** 1-byte: rowid-0-tombstone flag. This flag is only valid on the
** first tombstone hash page for each segment (iPg=0). If set,
** the hash table contains rowid 0. If clear, it does not.
** Rowid 0 is handled specially.
** 2-bytes: unused.
** 4-bytes: Big-endian integer containing number of entries on page.
**
** Following this are nSlot 4 or 8 byte slots (depending on the key-size
** in the first byte of the page header). The number of slots may be
** determined based on the size of the page record and the key-size:
**
** nSlot = (nByte - 8) / key-size
*/
/*
** Rowids for the averages and structure records in the %_data table.
*/
#define FTS5_AVERAGES_ROWID 1 /* Rowid used for the averages record */
#define FTS5_STRUCTURE_ROWID 10 /* The structure record */
/*
** Macros determining the rowids used by segment leaves and dlidx leaves
** and nodes. All nodes and leaves are stored in the %_data table with large
** positive rowids.
**
** Each segment has a unique non-zero 16-bit id.
**
** The rowid for each segment leaf is found by passing the segment id and
** the leaf page number to the FTS5_SEGMENT_ROWID macro. Leaves are numbered
** sequentially starting from 1.
*/
#define FTS5_DATA_ID_B 16 /* Max seg id number 65535 */
#define FTS5_DATA_DLI_B 1 /* Doclist-index flag (1 bit) */
#define FTS5_DATA_HEIGHT_B 5 /* Max dlidx tree height of 32 */
#define FTS5_DATA_PAGE_B 31 /* Max page number of 2147483648 */
#define fts5_dri(segid, dlidx, height, pgno) ( \
((i64)(segid) << (FTS5_DATA_PAGE_B+FTS5_DATA_HEIGHT_B+FTS5_DATA_DLI_B)) + \
((i64)(dlidx) << (FTS5_DATA_PAGE_B + FTS5_DATA_HEIGHT_B)) + \
((i64)(height) << (FTS5_DATA_PAGE_B)) + \
((i64)(pgno)) \
)
#define FTS5_SEGMENT_ROWID(segid, pgno) fts5_dri(segid, 0, 0, pgno)
#define FTS5_DLIDX_ROWID(segid, height, pgno) fts5_dri(segid, 1, height, pgno)
#define FTS5_TOMBSTONE_ROWID(segid,ipg) fts5_dri(segid+(1<<16), 0, 0, ipg)
#ifdef SQLITE_DEBUG
static int sqlite3Fts5Corrupt() { return SQLITE_CORRUPT_VTAB; }
#endif
/*
** Each time a blob is read from the %_data table, it is padded with this
** many zero bytes. This makes it easier to decode the various record formats
** without overreading if the records are corrupt.
*/
#define FTS5_DATA_ZERO_PADDING 8
#define FTS5_DATA_PADDING 20
typedef struct Fts5Data Fts5Data;
typedef struct Fts5DlidxIter Fts5DlidxIter;
typedef struct Fts5DlidxLvl Fts5DlidxLvl;
typedef struct Fts5DlidxWriter Fts5DlidxWriter;
typedef struct Fts5Iter Fts5Iter;
typedef struct Fts5PageWriter Fts5PageWriter;
typedef struct Fts5SegIter Fts5SegIter;
typedef struct Fts5DoclistIter Fts5DoclistIter;
typedef struct Fts5SegWriter Fts5SegWriter;
typedef struct Fts5Structure Fts5Structure;
typedef struct Fts5StructureLevel Fts5StructureLevel;
typedef struct Fts5StructureSegment Fts5StructureSegment;
typedef struct Fts5TokenDataIter Fts5TokenDataIter;
typedef struct Fts5TokenDataMap Fts5TokenDataMap;
typedef struct Fts5TombstoneArray Fts5TombstoneArray;
struct Fts5Data {
u8 *p; /* Pointer to buffer containing record */
int nn; /* Size of record in bytes */
int szLeaf; /* Size of leaf without page-index */
};
/*
** One object per %_data table.
**
** nContentlessDelete:
** The number of contentless delete operations since the most recent
** call to fts5IndexFlush() or fts5IndexDiscardData(). This is tracked
** so that extra auto-merge work can be done by fts5IndexFlush() to
** account for the delete operations.
*/
struct Fts5Index {
Fts5Config *pConfig; /* Virtual table configuration */
char *zDataTbl; /* Name of %_data table */
int nWorkUnit; /* Leaf pages in a "unit" of work */
/*
** Variables related to the accumulation of tokens and doclists within the
** in-memory hash tables before they are flushed to disk.
*/
Fts5Hash *pHash; /* Hash table for in-memory data */
int nPendingData; /* Current bytes of pending data */
i64 iWriteRowid; /* Rowid for current doc being written */
int bDelete; /* Current write is a delete */
int nContentlessDelete; /* Number of contentless delete ops */
int nPendingRow; /* Number of INSERT in hash table */
/* Error state. */
int rc; /* Current error code */
int flushRc;
/* State used by the fts5DataXXX() functions. */
sqlite3_blob *pReader; /* RO incr-blob open on %_data table */
sqlite3_stmt *pWriter; /* "INSERT ... %_data VALUES(?,?)" */
sqlite3_stmt *pDeleter; /* "DELETE FROM %_data ... id>=? AND id<=?" */
sqlite3_stmt *pIdxWriter; /* "INSERT ... %_idx VALUES(?,?,?,?)" */
sqlite3_stmt *pIdxDeleter; /* "DELETE FROM %_idx WHERE segid=?" */
sqlite3_stmt *pIdxSelect;
sqlite3_stmt *pIdxNextSelect;
int nRead; /* Total number of blocks read */
sqlite3_stmt *pDeleteFromIdx;
sqlite3_stmt *pDataVersion;
i64 iStructVersion; /* data_version when pStruct read */
Fts5Structure *pStruct; /* Current db structure (or NULL) */
};
struct Fts5DoclistIter {
u8 *aEof; /* Pointer to 1 byte past end of doclist */
/* Output variables. aPoslist==0 at EOF */
i64 iRowid;
u8 *aPoslist;
int nPoslist;
int nSize;
};
/*
** The contents of the "structure" record for each index are represented
** using an Fts5Structure record in memory. Which uses instances of the
** other Fts5StructureXXX types as components.
**
** nOriginCntr:
** This value is set to non-zero for structure records created for
** contentlessdelete=1 tables only. In that case it represents the
** origin value to apply to the next top-level segment created.
*/
struct Fts5StructureSegment {
int iSegid; /* Segment id */
int pgnoFirst; /* First leaf page number in segment */
int pgnoLast; /* Last leaf page number in segment */
/* contentlessdelete=1 tables only: */
u64 iOrigin1;
u64 iOrigin2;
int nPgTombstone; /* Number of tombstone hash table pages */
u64 nEntryTombstone; /* Number of tombstone entries that "count" */
u64 nEntry; /* Number of rows in this segment */
};
struct Fts5StructureLevel {
int nMerge; /* Number of segments in incr-merge */
int nSeg; /* Total number of segments on level */
Fts5StructureSegment *aSeg; /* Array of segments. aSeg[0] is oldest. */
};
struct Fts5Structure {
int nRef; /* Object reference count */
u64 nWriteCounter; /* Total leaves written to level 0 */
u64 nOriginCntr; /* Origin value for next top-level segment */
int nSegment; /* Total segments in this structure */
int nLevel; /* Number of levels in this index */
Fts5StructureLevel aLevel[1]; /* Array of nLevel level objects */
};
/*
** An object of type Fts5SegWriter is used to write to segments.
*/
struct Fts5PageWriter {
int pgno; /* Page number for this page */
int iPrevPgidx; /* Previous value written into pgidx */
Fts5Buffer buf; /* Buffer containing leaf data */
Fts5Buffer pgidx; /* Buffer containing page-index */
Fts5Buffer term; /* Buffer containing previous term on page */
};
struct Fts5DlidxWriter {
int pgno; /* Page number for this page */
int bPrevValid; /* True if iPrev is valid */
i64 iPrev; /* Previous rowid value written to page */
Fts5Buffer buf; /* Buffer containing page data */
};
struct Fts5SegWriter {
int iSegid; /* Segid to write to */
Fts5PageWriter writer; /* PageWriter object */
i64 iPrevRowid; /* Previous rowid written to current leaf */
u8 bFirstRowidInDoclist; /* True if next rowid is first in doclist */
u8 bFirstRowidInPage; /* True if next rowid is first in page */
/* TODO1: Can use (writer.pgidx.n==0) instead of bFirstTermInPage */
u8 bFirstTermInPage; /* True if next term will be first in leaf */
int nLeafWritten; /* Number of leaf pages written */
int nEmpty; /* Number of contiguous term-less nodes */
int nDlidx; /* Allocated size of aDlidx[] array */
Fts5DlidxWriter *aDlidx; /* Array of Fts5DlidxWriter objects */
/* Values to insert into the %_idx table */
Fts5Buffer btterm; /* Next term to insert into %_idx table */
int iBtPage; /* Page number corresponding to btterm */
};
typedef struct Fts5CResult Fts5CResult;
struct Fts5CResult {
u16 iFirst; /* aSeg[] index of firstest iterator */
u8 bTermEq; /* True if the terms are equal */
};
/*
** Object for iterating through a single segment, visiting each term/rowid
** pair in the segment.
**
** pSeg:
** The segment to iterate through.
**
** iLeafPgno:
** Current leaf page number within segment.
**
** iLeafOffset:
** Byte offset within the current leaf that is the first byte of the
** position list data (one byte passed the position-list size field).
**
** pLeaf:
** Buffer containing current leaf page data. Set to NULL at EOF.
**
** iTermLeafPgno, iTermLeafOffset:
** Leaf page number containing the last term read from the segment. And
** the offset immediately following the term data.
**
** flags:
** Mask of FTS5_SEGITER_XXX values. Interpreted as follows:
**
** FTS5_SEGITER_ONETERM:
** If set, set the iterator to point to EOF after the current doclist
** has been exhausted. Do not proceed to the next term in the segment.
**
** FTS5_SEGITER_REVERSE:
** This flag is only ever set if FTS5_SEGITER_ONETERM is also set. If
** it is set, iterate through rowid in descending order instead of the
** default ascending order.
**
** iRowidOffset/nRowidOffset/aRowidOffset:
** These are used if the FTS5_SEGITER_REVERSE flag is set.
**
** For each rowid on the page corresponding to the current term, the
** corresponding aRowidOffset[] entry is set to the byte offset of the
** start of the "position-list-size" field within the page.
**
** iTermIdx:
** Index of current term on iTermLeafPgno.
**
** apTombstone/nTombstone:
** These are used for contentless_delete=1 tables only. When the cursor
** is first allocated, the apTombstone[] array is allocated so that it
** is large enough for all tombstones hash pages associated with the
** segment. The pages themselves are loaded lazily from the database as
** they are required.
*/
struct Fts5SegIter {
Fts5StructureSegment *pSeg; /* Segment to iterate through */
int flags; /* Mask of configuration flags */
int iLeafPgno; /* Current leaf page number */
Fts5Data *pLeaf; /* Current leaf data */
Fts5Data *pNextLeaf; /* Leaf page (iLeafPgno+1) */
i64 iLeafOffset; /* Byte offset within current leaf */
Fts5TombstoneArray *pTombArray; /* Array of tombstone pages */
/* Next method */
void (*xNext)(Fts5Index*, Fts5SegIter*, int*);
/* The page and offset from which the current term was read. The offset
** is the offset of the first rowid in the current doclist. */
int iTermLeafPgno;
int iTermLeafOffset;
int iPgidxOff; /* Next offset in pgidx */
int iEndofDoclist;
/* The following are only used if the FTS5_SEGITER_REVERSE flag is set. */
int iRowidOffset; /* Current entry in aRowidOffset[] */
int nRowidOffset; /* Allocated size of aRowidOffset[] array */
int *aRowidOffset; /* Array of offset to rowid fields */
Fts5DlidxIter *pDlidx; /* If there is a doclist-index */
/* Variables populated based on current entry. */
Fts5Buffer term; /* Current term */
i64 iRowid; /* Current rowid */
int nPos; /* Number of bytes in current position list */
u8 bDel; /* True if the delete flag is set */
};
/*
** Array of tombstone pages. Reference counted.
*/
struct Fts5TombstoneArray {
int nRef; /* Number of pointers to this object */
int nTombstone;
Fts5Data *apTombstone[1]; /* Array of tombstone pages */
};
/*
** Argument is a pointer to an Fts5Data structure that contains a
** leaf page.
*/
#define ASSERT_SZLEAF_OK(x) assert( \
(x)->szLeaf==(x)->nn || (x)->szLeaf==fts5GetU16(&(x)->p[2]) \
)
#define FTS5_SEGITER_ONETERM 0x01
#define FTS5_SEGITER_REVERSE 0x02
/*
** Argument is a pointer to an Fts5Data structure that contains a leaf
** page. This macro evaluates to true if the leaf contains no terms, or
** false if it contains at least one term.
*/
#define fts5LeafIsTermless(x) ((x)->szLeaf >= (x)->nn)
#define fts5LeafTermOff(x, i) (fts5GetU16(&(x)->p[(x)->szLeaf + (i)*2]))
#define fts5LeafFirstRowidOff(x) (fts5GetU16((x)->p))
/*
** Object for iterating through the merged results of one or more segments,
** visiting each term/rowid pair in the merged data.
**
** nSeg is always a power of two greater than or equal to the number of
** segments that this object is merging data from. Both the aSeg[] and
** aFirst[] arrays are sized at nSeg entries. The aSeg[] array is padded
** with zeroed objects - these are handled as if they were iterators opened
** on empty segments.
**
** The results of comparing segments aSeg[N] and aSeg[N+1], where N is an
** even number, is stored in aFirst[(nSeg+N)/2]. The "result" of the
** comparison in this context is the index of the iterator that currently
** points to the smaller term/rowid combination. Iterators at EOF are
** considered to be greater than all other iterators.
**
** aFirst[1] contains the index in aSeg[] of the iterator that points to
** the smallest key overall. aFirst[0] is unused.
**
** poslist:
** Used by sqlite3Fts5IterPoslist() when the poslist needs to be buffered.
** There is no way to tell if this is populated or not.
**
** pColset:
** If not NULL, points to an object containing a set of column indices.
** Only matches that occur in one of these columns will be returned.
** The Fts5Iter does not own the Fts5Colset object, and so it is not
** freed when the iterator is closed - it is owned by the upper layer.
*/
struct Fts5Iter {
Fts5IndexIter base; /* Base class containing output vars */
Fts5TokenDataIter *pTokenDataIter;
Fts5Index *pIndex; /* Index that owns this iterator */
Fts5Buffer poslist; /* Buffer containing current poslist */
Fts5Colset *pColset; /* Restrict matches to these columns */
/* Invoked to set output variables. */
void (*xSetOutputs)(Fts5Iter*, Fts5SegIter*);
int nSeg; /* Size of aSeg[] array */
int bRev; /* True to iterate in reverse order */
u8 bSkipEmpty; /* True to skip deleted entries */
i64 iSwitchRowid; /* Firstest rowid of other than aFirst[1] */
Fts5CResult *aFirst; /* Current merge state (see above) */
Fts5SegIter aSeg[1]; /* Array of segment iterators */
};
/*
** An instance of the following type is used to iterate through the contents
** of a doclist-index record.
**
** pData:
** Record containing the doclist-index data.
**
** bEof:
** Set to true once iterator has reached EOF.
**
** iOff:
** Set to the current offset within record pData.
*/
struct Fts5DlidxLvl {
Fts5Data *pData; /* Data for current page of this level */
int iOff; /* Current offset into pData */
int bEof; /* At EOF already */
int iFirstOff; /* Used by reverse iterators */
/* Output variables */
int iLeafPgno; /* Page number of current leaf page */
i64 iRowid; /* First rowid on leaf iLeafPgno */
};
struct Fts5DlidxIter {
int nLvl;
int iSegid;
Fts5DlidxLvl aLvl[1];
};
static void fts5PutU16(u8 *aOut, u16 iVal){
aOut[0] = (iVal>>8);
aOut[1] = (iVal&0xFF);
}
static u16 fts5GetU16(const u8 *aIn){
return ((u16)aIn[0] << 8) + aIn[1];
}
/*
** The only argument points to a buffer at least 8 bytes in size. This
** function interprets the first 8 bytes of the buffer as a 64-bit big-endian
** unsigned integer and returns the result.
*/
static u64 fts5GetU64(u8 *a){
return ((u64)a[0] << 56)
+ ((u64)a[1] << 48)
+ ((u64)a[2] << 40)
+ ((u64)a[3] << 32)
+ ((u64)a[4] << 24)
+ ((u64)a[5] << 16)
+ ((u64)a[6] << 8)
+ ((u64)a[7] << 0);
}
/*
** The only argument points to a buffer at least 4 bytes in size. This
** function interprets the first 4 bytes of the buffer as a 32-bit big-endian
** unsigned integer and returns the result.
*/
static u32 fts5GetU32(const u8 *a){
return ((u32)a[0] << 24)
+ ((u32)a[1] << 16)
+ ((u32)a[2] << 8)
+ ((u32)a[3] << 0);
}
/*
** Write iVal, formated as a 64-bit big-endian unsigned integer, to the
** buffer indicated by the first argument.
*/
static void fts5PutU64(u8 *a, u64 iVal){
a[0] = ((iVal >> 56) & 0xFF);
a[1] = ((iVal >> 48) & 0xFF);
a[2] = ((iVal >> 40) & 0xFF);
a[3] = ((iVal >> 32) & 0xFF);
a[4] = ((iVal >> 24) & 0xFF);
a[5] = ((iVal >> 16) & 0xFF);
a[6] = ((iVal >> 8) & 0xFF);
a[7] = ((iVal >> 0) & 0xFF);
}
/*
** Write iVal, formated as a 32-bit big-endian unsigned integer, to the
** buffer indicated by the first argument.
*/
static void fts5PutU32(u8 *a, u32 iVal){
a[0] = ((iVal >> 24) & 0xFF);
a[1] = ((iVal >> 16) & 0xFF);
a[2] = ((iVal >> 8) & 0xFF);
a[3] = ((iVal >> 0) & 0xFF);
}
/*
** Allocate and return a buffer at least nByte bytes in size.
**
** If an OOM error is encountered, return NULL and set the error code in
** the Fts5Index handle passed as the first argument.
*/
static void *fts5IdxMalloc(Fts5Index *p, sqlite3_int64 nByte){
return sqlite3Fts5MallocZero(&p->rc, nByte);
}
/*
** Compare the contents of the pLeft buffer with the pRight/nRight blob.
**
** Return -ve if pLeft is smaller than pRight, 0 if they are equal or
** +ve if pRight is smaller than pLeft. In other words:
**
** res = *pLeft - *pRight
*/
#ifdef SQLITE_DEBUG
static int fts5BufferCompareBlob(
Fts5Buffer *pLeft, /* Left hand side of comparison */
const u8 *pRight, int nRight /* Right hand side of comparison */
){
int nCmp = MIN(pLeft->n, nRight);
int res = memcmp(pLeft->p, pRight, nCmp);
return (res==0 ? (pLeft->n - nRight) : res);
}
#endif
/*
** Compare the contents of the two buffers using memcmp(). If one buffer
** is a prefix of the other, it is considered the lesser.
**
** Return -ve if pLeft is smaller than pRight, 0 if they are equal or
** +ve if pRight is smaller than pLeft. In other words:
**
** res = *pLeft - *pRight
*/
static int fts5BufferCompare(Fts5Buffer *pLeft, Fts5Buffer *pRight){
int nCmp, res;
nCmp = MIN(pLeft->n, pRight->n);
assert( nCmp<=0 || pLeft->p!=0 );
assert( nCmp<=0 || pRight->p!=0 );
res = fts5Memcmp(pLeft->p, pRight->p, nCmp);
return (res==0 ? (pLeft->n - pRight->n) : res);
}
static int fts5LeafFirstTermOff(Fts5Data *pLeaf){
int ret;
fts5GetVarint32(&pLeaf->p[pLeaf->szLeaf], ret);
return ret;
}
/*
** Close the read-only blob handle, if it is open.
*/
static void sqlite3Fts5IndexCloseReader(Fts5Index *p){
if( p->pReader ){
sqlite3_blob *pReader = p->pReader;
p->pReader = 0;
sqlite3_blob_close(pReader);
}
}
/*
** Retrieve a record from the %_data table.
**
** If an error occurs, NULL is returned and an error left in the
** Fts5Index object.
*/
static Fts5Data *fts5DataRead(Fts5Index *p, i64 iRowid){
Fts5Data *pRet = 0;
if( p->rc==SQLITE_OK ){
int rc = SQLITE_OK;
if( p->pReader ){
/* This call may return SQLITE_ABORT if there has been a savepoint
** rollback since it was last used. In this case a new blob handle
** is required. */
sqlite3_blob *pBlob = p->pReader;
p->pReader = 0;
rc = sqlite3_blob_reopen(pBlob, iRowid);
assert( p->pReader==0 );
p->pReader = pBlob;
if( rc!=SQLITE_OK ){
sqlite3Fts5IndexCloseReader(p);
}
if( rc==SQLITE_ABORT ) rc = SQLITE_OK;
}
/* If the blob handle is not open at this point, open it and seek
** to the requested entry. */
if( p->pReader==0 && rc==SQLITE_OK ){
Fts5Config *pConfig = p->pConfig;
rc = sqlite3_blob_open(pConfig->db,
pConfig->zDb, p->zDataTbl, "block", iRowid, 0, &p->pReader
);
}
/* If either of the sqlite3_blob_open() or sqlite3_blob_reopen() calls
** above returned SQLITE_ERROR, return SQLITE_CORRUPT_VTAB instead.
** All the reasons those functions might return SQLITE_ERROR - missing
** table, missing row, non-blob/text in block column - indicate
** backing store corruption. */
if( rc==SQLITE_ERROR ) rc = FTS5_CORRUPT;
if( rc==SQLITE_OK ){
u8 *aOut = 0; /* Read blob data into this buffer */
int nByte = sqlite3_blob_bytes(p->pReader);
sqlite3_int64 nAlloc = sizeof(Fts5Data) + nByte + FTS5_DATA_PADDING;
pRet = (Fts5Data*)sqlite3_malloc64(nAlloc);
if( pRet ){
pRet->nn = nByte;
aOut = pRet->p = (u8*)&pRet[1];
}else{
rc = SQLITE_NOMEM;
}
if( rc==SQLITE_OK ){
rc = sqlite3_blob_read(p->pReader, aOut, nByte, 0);
}
if( rc!=SQLITE_OK ){
sqlite3_free(pRet);
pRet = 0;
}else{
/* TODO1: Fix this */
pRet->p[nByte] = 0x00;
pRet->p[nByte+1] = 0x00;
pRet->szLeaf = fts5GetU16(&pRet->p[2]);
}
}
p->rc = rc;
p->nRead++;
}
assert( (pRet==0)==(p->rc!=SQLITE_OK) );
return pRet;
}
/*
** Release a reference to data record returned by an earlier call to
** fts5DataRead().
*/
static void fts5DataRelease(Fts5Data *pData){
sqlite3_free(pData);
}
static Fts5Data *fts5LeafRead(Fts5Index *p, i64 iRowid){
Fts5Data *pRet = fts5DataRead(p, iRowid);
if( pRet ){
if( pRet->nn<4 || pRet->szLeaf>pRet->nn ){
p->rc = FTS5_CORRUPT;
fts5DataRelease(pRet);
pRet = 0;
}
}
return pRet;
}
static int fts5IndexPrepareStmt(
Fts5Index *p,
sqlite3_stmt **ppStmt,
char *zSql
){
if( p->rc==SQLITE_OK ){
if( zSql ){
p->rc = sqlite3_prepare_v3(p->pConfig->db, zSql, -1,
SQLITE_PREPARE_PERSISTENT|SQLITE_PREPARE_NO_VTAB,
ppStmt, 0);
}else{
p->rc = SQLITE_NOMEM;
}
}
sqlite3_free(zSql);
return p->rc;
}
/*
** INSERT OR REPLACE a record into the %_data table.
*/
static void fts5DataWrite(Fts5Index *p, i64 iRowid, const u8 *pData, int nData){
if( p->rc!=SQLITE_OK ) return;
if( p->pWriter==0 ){
Fts5Config *pConfig = p->pConfig;
fts5IndexPrepareStmt(p, &p->pWriter, sqlite3_mprintf(
"REPLACE INTO '%q'.'%q_data'(id, block) VALUES(?,?)",
pConfig->zDb, pConfig->zName
));
if( p->rc ) return;
}
sqlite3_bind_int64(p->pWriter, 1, iRowid);
sqlite3_bind_blob(p->pWriter, 2, pData, nData, SQLITE_STATIC);
sqlite3_step(p->pWriter);
p->rc = sqlite3_reset(p->pWriter);
sqlite3_bind_null(p->pWriter, 2);
}
/*
** Execute the following SQL:
**
** DELETE FROM %_data WHERE id BETWEEN $iFirst AND $iLast
*/
static void fts5DataDelete(Fts5Index *p, i64 iFirst, i64 iLast){
if( p->rc!=SQLITE_OK ) return;
if( p->pDeleter==0 ){
Fts5Config *pConfig = p->pConfig;
char *zSql = sqlite3_mprintf(
"DELETE FROM '%q'.'%q_data' WHERE id>=? AND id<=?",
pConfig->zDb, pConfig->zName
);
if( fts5IndexPrepareStmt(p, &p->pDeleter, zSql) ) return;
}
sqlite3_bind_int64(p->pDeleter, 1, iFirst);
sqlite3_bind_int64(p->pDeleter, 2, iLast);
sqlite3_step(p->pDeleter);
p->rc = sqlite3_reset(p->pDeleter);
}
/*
** Remove all records associated with segment iSegid.
*/
static void fts5DataRemoveSegment(Fts5Index *p, Fts5StructureSegment *pSeg){
int iSegid = pSeg->iSegid;
i64 iFirst = FTS5_SEGMENT_ROWID(iSegid, 0);
i64 iLast = FTS5_SEGMENT_ROWID(iSegid+1, 0)-1;
fts5DataDelete(p, iFirst, iLast);
if( pSeg->nPgTombstone ){
i64 iTomb1 = FTS5_TOMBSTONE_ROWID(iSegid, 0);
i64 iTomb2 = FTS5_TOMBSTONE_ROWID(iSegid, pSeg->nPgTombstone-1);
fts5DataDelete(p, iTomb1, iTomb2);
}
if( p->pIdxDeleter==0 ){
Fts5Config *pConfig = p->pConfig;
fts5IndexPrepareStmt(p, &p->pIdxDeleter, sqlite3_mprintf(
"DELETE FROM '%q'.'%q_idx' WHERE segid=?",
pConfig->zDb, pConfig->zName
));
}
if( p->rc==SQLITE_OK ){
sqlite3_bind_int(p->pIdxDeleter, 1, iSegid);
sqlite3_step(p->pIdxDeleter);
p->rc = sqlite3_reset(p->pIdxDeleter);
}
}
/*
** Release a reference to an Fts5Structure object returned by an earlier
** call to fts5StructureRead() or fts5StructureDecode().
*/
static void fts5StructureRelease(Fts5Structure *pStruct){
if( pStruct && 0>=(--pStruct->nRef) ){
int i;
assert( pStruct->nRef==0 );
for(i=0; inLevel; i++){
sqlite3_free(pStruct->aLevel[i].aSeg);
}
sqlite3_free(pStruct);
}
}
static void fts5StructureRef(Fts5Structure *pStruct){
pStruct->nRef++;
}
static void *sqlite3Fts5StructureRef(Fts5Index *p){
fts5StructureRef(p->pStruct);
return (void*)p->pStruct;
}
static void sqlite3Fts5StructureRelease(void *p){
if( p ){
fts5StructureRelease((Fts5Structure*)p);
}
}
static int sqlite3Fts5StructureTest(Fts5Index *p, void *pStruct){
if( p->pStruct!=(Fts5Structure*)pStruct ){
return SQLITE_ABORT;
}
return SQLITE_OK;
}
/*
** Ensure that structure object (*pp) is writable.
**
** This function is a no-op if (*pRc) is not SQLITE_OK when it is called. If
** an error occurs, (*pRc) is set to an SQLite error code before returning.
*/
static void fts5StructureMakeWritable(int *pRc, Fts5Structure **pp){
Fts5Structure *p = *pp;
if( *pRc==SQLITE_OK && p->nRef>1 ){
i64 nByte = sizeof(Fts5Structure)+(p->nLevel-1)*sizeof(Fts5StructureLevel);
Fts5Structure *pNew;
pNew = (Fts5Structure*)sqlite3Fts5MallocZero(pRc, nByte);
if( pNew ){
int i;
memcpy(pNew, p, nByte);
for(i=0; inLevel; i++) pNew->aLevel[i].aSeg = 0;
for(i=0; inLevel; i++){
Fts5StructureLevel *pLvl = &pNew->aLevel[i];
nByte = sizeof(Fts5StructureSegment) * pNew->aLevel[i].nSeg;
pLvl->aSeg = (Fts5StructureSegment*)sqlite3Fts5MallocZero(pRc, nByte);
if( pLvl->aSeg==0 ){
for(i=0; inLevel; i++){
sqlite3_free(pNew->aLevel[i].aSeg);
}
sqlite3_free(pNew);
return;
}
memcpy(pLvl->aSeg, p->aLevel[i].aSeg, nByte);
}
p->nRef--;
pNew->nRef = 1;
}
*pp = pNew;
}
}
/*
** Deserialize and return the structure record currently stored in serialized
** form within buffer pData/nData.
**
** The Fts5Structure.aLevel[] and each Fts5StructureLevel.aSeg[] array
** are over-allocated by one slot. This allows the structure contents
** to be more easily edited.
**
** If an error occurs, *ppOut is set to NULL and an SQLite error code
** returned. Otherwise, *ppOut is set to point to the new object and
** SQLITE_OK returned.
*/
static int fts5StructureDecode(
const u8 *pData, /* Buffer containing serialized structure */
int nData, /* Size of buffer pData in bytes */
int *piCookie, /* Configuration cookie value */
Fts5Structure **ppOut /* OUT: Deserialized object */
){
int rc = SQLITE_OK;
int i = 0;
int iLvl;
int nLevel = 0;
int nSegment = 0;
sqlite3_int64 nByte; /* Bytes of space to allocate at pRet */
Fts5Structure *pRet = 0; /* Structure object to return */
int bStructureV2 = 0; /* True for FTS5_STRUCTURE_V2 */
u64 nOriginCntr = 0; /* Largest origin value seen so far */
/* Grab the cookie value */
if( piCookie ) *piCookie = sqlite3Fts5Get32(pData);
i = 4;
/* Check if this is a V2 structure record. Set bStructureV2 if it is. */
if( 0==memcmp(&pData[i], FTS5_STRUCTURE_V2, 4) ){
i += 4;
bStructureV2 = 1;
}
/* Read the total number of levels and segments from the start of the
** structure record. */
i += fts5GetVarint32(&pData[i], nLevel);
i += fts5GetVarint32(&pData[i], nSegment);
if( nLevel>FTS5_MAX_SEGMENT || nLevel<0
|| nSegment>FTS5_MAX_SEGMENT || nSegment<0
){
return FTS5_CORRUPT;
}
nByte = (
sizeof(Fts5Structure) + /* Main structure */
sizeof(Fts5StructureLevel) * (nLevel-1) /* aLevel[] array */
);
pRet = (Fts5Structure*)sqlite3Fts5MallocZero(&rc, nByte);
if( pRet ){
pRet->nRef = 1;
pRet->nLevel = nLevel;
pRet->nSegment = nSegment;
i += sqlite3Fts5GetVarint(&pData[i], &pRet->nWriteCounter);
for(iLvl=0; rc==SQLITE_OK && iLvlaLevel[iLvl];
int nTotal = 0;
int iSeg;
if( i>=nData ){
rc = FTS5_CORRUPT;
}else{
i += fts5GetVarint32(&pData[i], pLvl->nMerge);
i += fts5GetVarint32(&pData[i], nTotal);
if( nTotalnMerge ) rc = FTS5_CORRUPT;
pLvl->aSeg = (Fts5StructureSegment*)sqlite3Fts5MallocZero(&rc,
nTotal * sizeof(Fts5StructureSegment)
);
nSegment -= nTotal;
}
if( rc==SQLITE_OK ){
pLvl->nSeg = nTotal;
for(iSeg=0; iSegaSeg[iSeg];
if( i>=nData ){
rc = FTS5_CORRUPT;
break;
}
assert( pSeg!=0 );
i += fts5GetVarint32(&pData[i], pSeg->iSegid);
i += fts5GetVarint32(&pData[i], pSeg->pgnoFirst);
i += fts5GetVarint32(&pData[i], pSeg->pgnoLast);
if( bStructureV2 ){
i += fts5GetVarint(&pData[i], &pSeg->iOrigin1);
i += fts5GetVarint(&pData[i], &pSeg->iOrigin2);
i += fts5GetVarint32(&pData[i], pSeg->nPgTombstone);
i += fts5GetVarint(&pData[i], &pSeg->nEntryTombstone);
i += fts5GetVarint(&pData[i], &pSeg->nEntry);
nOriginCntr = MAX(nOriginCntr, pSeg->iOrigin2);
}
if( pSeg->pgnoLastpgnoFirst ){
rc = FTS5_CORRUPT;
break;
}
}
if( iLvl>0 && pLvl[-1].nMerge && nTotal==0 ) rc = FTS5_CORRUPT;
if( iLvl==nLevel-1 && pLvl->nMerge ) rc = FTS5_CORRUPT;
}
}
if( nSegment!=0 && rc==SQLITE_OK ) rc = FTS5_CORRUPT;
if( bStructureV2 ){
pRet->nOriginCntr = nOriginCntr+1;
}
if( rc!=SQLITE_OK ){
fts5StructureRelease(pRet);
pRet = 0;
}
}
*ppOut = pRet;
return rc;
}
/*
** Add a level to the Fts5Structure.aLevel[] array of structure object
** (*ppStruct).
*/
static void fts5StructureAddLevel(int *pRc, Fts5Structure **ppStruct){
fts5StructureMakeWritable(pRc, ppStruct);
assert( (ppStruct!=0 && (*ppStruct)!=0) || (*pRc)!=SQLITE_OK );
if( *pRc==SQLITE_OK ){
Fts5Structure *pStruct = *ppStruct;
int nLevel = pStruct->nLevel;
sqlite3_int64 nByte = (
sizeof(Fts5Structure) + /* Main structure */
sizeof(Fts5StructureLevel) * (nLevel+1) /* aLevel[] array */
);
pStruct = sqlite3_realloc64(pStruct, nByte);
if( pStruct ){
memset(&pStruct->aLevel[nLevel], 0, sizeof(Fts5StructureLevel));
pStruct->nLevel++;
*ppStruct = pStruct;
}else{
*pRc = SQLITE_NOMEM;
}
}
}
/*
** Extend level iLvl so that there is room for at least nExtra more
** segments.
*/
static void fts5StructureExtendLevel(
int *pRc,
Fts5Structure *pStruct,
int iLvl,
int nExtra,
int bInsert
){
if( *pRc==SQLITE_OK ){
Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl];
Fts5StructureSegment *aNew;
sqlite3_int64 nByte;
nByte = (pLvl->nSeg + nExtra) * sizeof(Fts5StructureSegment);
aNew = sqlite3_realloc64(pLvl->aSeg, nByte);
if( aNew ){
if( bInsert==0 ){
memset(&aNew[pLvl->nSeg], 0, sizeof(Fts5StructureSegment) * nExtra);
}else{
int nMove = pLvl->nSeg * sizeof(Fts5StructureSegment);
memmove(&aNew[nExtra], aNew, nMove);
memset(aNew, 0, sizeof(Fts5StructureSegment) * nExtra);
}
pLvl->aSeg = aNew;
}else{
*pRc = SQLITE_NOMEM;
}
}
}
static Fts5Structure *fts5StructureReadUncached(Fts5Index *p){
Fts5Structure *pRet = 0;
Fts5Config *pConfig = p->pConfig;
int iCookie; /* Configuration cookie */
Fts5Data *pData;
pData = fts5DataRead(p, FTS5_STRUCTURE_ROWID);
if( p->rc==SQLITE_OK ){
/* TODO: Do we need this if the leaf-index is appended? Probably... */
memset(&pData->p[pData->nn], 0, FTS5_DATA_PADDING);
p->rc = fts5StructureDecode(pData->p, pData->nn, &iCookie, &pRet);
if( p->rc==SQLITE_OK && (pConfig->pgsz==0 || pConfig->iCookie!=iCookie) ){
p->rc = sqlite3Fts5ConfigLoad(pConfig, iCookie);
}
fts5DataRelease(pData);
if( p->rc!=SQLITE_OK ){
fts5StructureRelease(pRet);
pRet = 0;
}
}
return pRet;
}
static i64 fts5IndexDataVersion(Fts5Index *p){
i64 iVersion = 0;
if( p->rc==SQLITE_OK ){
if( p->pDataVersion==0 ){
p->rc = fts5IndexPrepareStmt(p, &p->pDataVersion,
sqlite3_mprintf("PRAGMA %Q.data_version", p->pConfig->zDb)
);
if( p->rc ) return 0;
}
if( SQLITE_ROW==sqlite3_step(p->pDataVersion) ){
iVersion = sqlite3_column_int64(p->pDataVersion, 0);
}
p->rc = sqlite3_reset(p->pDataVersion);
}
return iVersion;
}
/*
** Read, deserialize and return the structure record.
**
** The Fts5Structure.aLevel[] and each Fts5StructureLevel.aSeg[] array
** are over-allocated as described for function fts5StructureDecode()
** above.
**
** If an error occurs, NULL is returned and an error code left in the
** Fts5Index handle. If an error has already occurred when this function
** is called, it is a no-op.
*/
static Fts5Structure *fts5StructureRead(Fts5Index *p){
if( p->pStruct==0 ){
p->iStructVersion = fts5IndexDataVersion(p);
if( p->rc==SQLITE_OK ){
p->pStruct = fts5StructureReadUncached(p);
}
}
#if 0
else{
Fts5Structure *pTest = fts5StructureReadUncached(p);
if( pTest ){
int i, j;
assert_nc( p->pStruct->nSegment==pTest->nSegment );
assert_nc( p->pStruct->nLevel==pTest->nLevel );
for(i=0; inLevel; i++){
assert_nc( p->pStruct->aLevel[i].nMerge==pTest->aLevel[i].nMerge );
assert_nc( p->pStruct->aLevel[i].nSeg==pTest->aLevel[i].nSeg );
for(j=0; jaLevel[i].nSeg; j++){
Fts5StructureSegment *p1 = &pTest->aLevel[i].aSeg[j];
Fts5StructureSegment *p2 = &p->pStruct->aLevel[i].aSeg[j];
assert_nc( p1->iSegid==p2->iSegid );
assert_nc( p1->pgnoFirst==p2->pgnoFirst );
assert_nc( p1->pgnoLast==p2->pgnoLast );
}
}
fts5StructureRelease(pTest);
}
}
#endif
if( p->rc!=SQLITE_OK ) return 0;
assert( p->iStructVersion!=0 );
assert( p->pStruct!=0 );
fts5StructureRef(p->pStruct);
return p->pStruct;
}
static void fts5StructureInvalidate(Fts5Index *p){
if( p->pStruct ){
fts5StructureRelease(p->pStruct);
p->pStruct = 0;
}
}
/*
** Return the total number of segments in index structure pStruct. This
** function is only ever used as part of assert() conditions.
*/
#ifdef SQLITE_DEBUG
static int fts5StructureCountSegments(Fts5Structure *pStruct){
int nSegment = 0; /* Total number of segments */
if( pStruct ){
int iLvl; /* Used to iterate through levels */
for(iLvl=0; iLvlnLevel; iLvl++){
nSegment += pStruct->aLevel[iLvl].nSeg;
}
}
return nSegment;
}
#endif
#define fts5BufferSafeAppendBlob(pBuf, pBlob, nBlob) { \
assert( (pBuf)->nSpace>=((pBuf)->n+nBlob) ); \
memcpy(&(pBuf)->p[(pBuf)->n], pBlob, nBlob); \
(pBuf)->n += nBlob; \
}
#define fts5BufferSafeAppendVarint(pBuf, iVal) { \
(pBuf)->n += sqlite3Fts5PutVarint(&(pBuf)->p[(pBuf)->n], (iVal)); \
assert( (pBuf)->nSpace>=(pBuf)->n ); \
}
/*
** Serialize and store the "structure" record.
**
** If an error occurs, leave an error code in the Fts5Index object. If an
** error has already occurred, this function is a no-op.
*/
static void fts5StructureWrite(Fts5Index *p, Fts5Structure *pStruct){
if( p->rc==SQLITE_OK ){
Fts5Buffer buf; /* Buffer to serialize record into */
int iLvl; /* Used to iterate through levels */
int iCookie; /* Cookie value to store */
int nHdr = (pStruct->nOriginCntr>0 ? (4+4+9+9+9) : (4+9+9));
assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) );
memset(&buf, 0, sizeof(Fts5Buffer));
/* Append the current configuration cookie */
iCookie = p->pConfig->iCookie;
if( iCookie<0 ) iCookie = 0;
if( 0==sqlite3Fts5BufferSize(&p->rc, &buf, nHdr) ){
sqlite3Fts5Put32(buf.p, iCookie);
buf.n = 4;
if( pStruct->nOriginCntr>0 ){
fts5BufferSafeAppendBlob(&buf, FTS5_STRUCTURE_V2, 4);
}
fts5BufferSafeAppendVarint(&buf, pStruct->nLevel);
fts5BufferSafeAppendVarint(&buf, pStruct->nSegment);
fts5BufferSafeAppendVarint(&buf, (i64)pStruct->nWriteCounter);
}
for(iLvl=0; iLvlnLevel; iLvl++){
int iSeg; /* Used to iterate through segments */
Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl];
fts5BufferAppendVarint(&p->rc, &buf, pLvl->nMerge);
fts5BufferAppendVarint(&p->rc, &buf, pLvl->nSeg);
assert( pLvl->nMerge<=pLvl->nSeg );
for(iSeg=0; iSegnSeg; iSeg++){
Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg];
fts5BufferAppendVarint(&p->rc, &buf, pSeg->iSegid);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->pgnoFirst);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->pgnoLast);
if( pStruct->nOriginCntr>0 ){
fts5BufferAppendVarint(&p->rc, &buf, pSeg->iOrigin1);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->iOrigin2);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->nPgTombstone);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->nEntryTombstone);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->nEntry);
}
}
}
fts5DataWrite(p, FTS5_STRUCTURE_ROWID, buf.p, buf.n);
fts5BufferFree(&buf);
}
}
#if 0
static void fts5DebugStructure(int*,Fts5Buffer*,Fts5Structure*);
static void fts5PrintStructure(const char *zCaption, Fts5Structure *pStruct){
int rc = SQLITE_OK;
Fts5Buffer buf;
memset(&buf, 0, sizeof(buf));
fts5DebugStructure(&rc, &buf, pStruct);
fprintf(stdout, "%s: %s\n", zCaption, buf.p);
fflush(stdout);
fts5BufferFree(&buf);
}
#else
# define fts5PrintStructure(x,y)
#endif
static int fts5SegmentSize(Fts5StructureSegment *pSeg){
return 1 + pSeg->pgnoLast - pSeg->pgnoFirst;
}
/*
** Return a copy of index structure pStruct. Except, promote as many
** segments as possible to level iPromote. If an OOM occurs, NULL is
** returned.
*/
static void fts5StructurePromoteTo(
Fts5Index *p,
int iPromote,
int szPromote,
Fts5Structure *pStruct
){
int il, is;
Fts5StructureLevel *pOut = &pStruct->aLevel[iPromote];
if( pOut->nMerge==0 ){
for(il=iPromote+1; ilnLevel; il++){
Fts5StructureLevel *pLvl = &pStruct->aLevel[il];
if( pLvl->nMerge ) return;
for(is=pLvl->nSeg-1; is>=0; is--){
int sz = fts5SegmentSize(&pLvl->aSeg[is]);
if( sz>szPromote ) return;
fts5StructureExtendLevel(&p->rc, pStruct, iPromote, 1, 1);
if( p->rc ) return;
memcpy(pOut->aSeg, &pLvl->aSeg[is], sizeof(Fts5StructureSegment));
pOut->nSeg++;
pLvl->nSeg--;
}
}
}
}
/*
** A new segment has just been written to level iLvl of index structure
** pStruct. This function determines if any segments should be promoted
** as a result. Segments are promoted in two scenarios:
**
** a) If the segment just written is smaller than one or more segments
** within the previous populated level, it is promoted to the previous
** populated level.
**
** b) If the segment just written is larger than the newest segment on
** the next populated level, then that segment, and any other adjacent
** segments that are also smaller than the one just written, are
** promoted.
**
** If one or more segments are promoted, the structure object is updated
** to reflect this.
*/
static void fts5StructurePromote(
Fts5Index *p, /* FTS5 backend object */
int iLvl, /* Index level just updated */
Fts5Structure *pStruct /* Index structure */
){
if( p->rc==SQLITE_OK ){
int iTst;
int iPromote = -1;
int szPromote = 0; /* Promote anything this size or smaller */
Fts5StructureSegment *pSeg; /* Segment just written */
int szSeg; /* Size of segment just written */
int nSeg = pStruct->aLevel[iLvl].nSeg;
if( nSeg==0 ) return;
pSeg = &pStruct->aLevel[iLvl].aSeg[pStruct->aLevel[iLvl].nSeg-1];
szSeg = (1 + pSeg->pgnoLast - pSeg->pgnoFirst);
/* Check for condition (a) */
for(iTst=iLvl-1; iTst>=0 && pStruct->aLevel[iTst].nSeg==0; iTst--);
if( iTst>=0 ){
int i;
int szMax = 0;
Fts5StructureLevel *pTst = &pStruct->aLevel[iTst];
assert( pTst->nMerge==0 );
for(i=0; inSeg; i++){
int sz = pTst->aSeg[i].pgnoLast - pTst->aSeg[i].pgnoFirst + 1;
if( sz>szMax ) szMax = sz;
}
if( szMax>=szSeg ){
/* Condition (a) is true. Promote the newest segment on level
** iLvl to level iTst. */
iPromote = iTst;
szPromote = szMax;
}
}
/* If condition (a) is not met, assume (b) is true. StructurePromoteTo()
** is a no-op if it is not. */
if( iPromote<0 ){
iPromote = iLvl;
szPromote = szSeg;
}
fts5StructurePromoteTo(p, iPromote, szPromote, pStruct);
}
}
/*
** Advance the iterator passed as the only argument. If the end of the
** doclist-index page is reached, return non-zero.
*/
static int fts5DlidxLvlNext(Fts5DlidxLvl *pLvl){
Fts5Data *pData = pLvl->pData;
if( pLvl->iOff==0 ){
assert( pLvl->bEof==0 );
pLvl->iOff = 1;
pLvl->iOff += fts5GetVarint32(&pData->p[1], pLvl->iLeafPgno);
pLvl->iOff += fts5GetVarint(&pData->p[pLvl->iOff], (u64*)&pLvl->iRowid);
pLvl->iFirstOff = pLvl->iOff;
}else{
int iOff;
for(iOff=pLvl->iOff; iOffnn; iOff++){
if( pData->p[iOff] ) break;
}
if( iOffnn ){
u64 iVal;
pLvl->iLeafPgno += (iOff - pLvl->iOff) + 1;
iOff += fts5GetVarint(&pData->p[iOff], &iVal);
pLvl->iRowid += iVal;
pLvl->iOff = iOff;
}else{
pLvl->bEof = 1;
}
}
return pLvl->bEof;
}
/*
** Advance the iterator passed as the only argument.
*/
static int fts5DlidxIterNextR(Fts5Index *p, Fts5DlidxIter *pIter, int iLvl){
Fts5DlidxLvl *pLvl = &pIter->aLvl[iLvl];
assert( iLvlnLvl );
if( fts5DlidxLvlNext(pLvl) ){
if( (iLvl+1) < pIter->nLvl ){
fts5DlidxIterNextR(p, pIter, iLvl+1);
if( pLvl[1].bEof==0 ){
fts5DataRelease(pLvl->pData);
memset(pLvl, 0, sizeof(Fts5DlidxLvl));
pLvl->pData = fts5DataRead(p,
FTS5_DLIDX_ROWID(pIter->iSegid, iLvl, pLvl[1].iLeafPgno)
);
if( pLvl->pData ) fts5DlidxLvlNext(pLvl);
}
}
}
return pIter->aLvl[0].bEof;
}
static int fts5DlidxIterNext(Fts5Index *p, Fts5DlidxIter *pIter){
return fts5DlidxIterNextR(p, pIter, 0);
}
/*
** The iterator passed as the first argument has the following fields set
** as follows. This function sets up the rest of the iterator so that it
** points to the first rowid in the doclist-index.
**
** pData:
** pointer to doclist-index record,
**
** When this function is called pIter->iLeafPgno is the page number the
** doclist is associated with (the one featuring the term).
*/
static int fts5DlidxIterFirst(Fts5DlidxIter *pIter){
int i;
for(i=0; inLvl; i++){
fts5DlidxLvlNext(&pIter->aLvl[i]);
}
return pIter->aLvl[0].bEof;
}
static int fts5DlidxIterEof(Fts5Index *p, Fts5DlidxIter *pIter){
return p->rc!=SQLITE_OK || pIter->aLvl[0].bEof;
}
static void fts5DlidxIterLast(Fts5Index *p, Fts5DlidxIter *pIter){
int i;
/* Advance each level to the last entry on the last page */
for(i=pIter->nLvl-1; p->rc==SQLITE_OK && i>=0; i--){
Fts5DlidxLvl *pLvl = &pIter->aLvl[i];
while( fts5DlidxLvlNext(pLvl)==0 );
pLvl->bEof = 0;
if( i>0 ){
Fts5DlidxLvl *pChild = &pLvl[-1];
fts5DataRelease(pChild->pData);
memset(pChild, 0, sizeof(Fts5DlidxLvl));
pChild->pData = fts5DataRead(p,
FTS5_DLIDX_ROWID(pIter->iSegid, i-1, pLvl->iLeafPgno)
);
}
}
}
/*
** Move the iterator passed as the only argument to the previous entry.
*/
static int fts5DlidxLvlPrev(Fts5DlidxLvl *pLvl){
int iOff = pLvl->iOff;
assert( pLvl->bEof==0 );
if( iOff<=pLvl->iFirstOff ){
pLvl->bEof = 1;
}else{
u8 *a = pLvl->pData->p;
pLvl->iOff = 0;
fts5DlidxLvlNext(pLvl);
while( 1 ){
int nZero = 0;
int ii = pLvl->iOff;
u64 delta = 0;
while( a[ii]==0 ){
nZero++;
ii++;
}
ii += sqlite3Fts5GetVarint(&a[ii], &delta);
if( ii>=iOff ) break;
pLvl->iLeafPgno += nZero+1;
pLvl->iRowid += delta;
pLvl->iOff = ii;
}
}
return pLvl->bEof;
}
static int fts5DlidxIterPrevR(Fts5Index *p, Fts5DlidxIter *pIter, int iLvl){
Fts5DlidxLvl *pLvl = &pIter->aLvl[iLvl];
assert( iLvlnLvl );
if( fts5DlidxLvlPrev(pLvl) ){
if( (iLvl+1) < pIter->nLvl ){
fts5DlidxIterPrevR(p, pIter, iLvl+1);
if( pLvl[1].bEof==0 ){
fts5DataRelease(pLvl->pData);
memset(pLvl, 0, sizeof(Fts5DlidxLvl));
pLvl->pData = fts5DataRead(p,
FTS5_DLIDX_ROWID(pIter->iSegid, iLvl, pLvl[1].iLeafPgno)
);
if( pLvl->pData ){
while( fts5DlidxLvlNext(pLvl)==0 );
pLvl->bEof = 0;
}
}
}
}
return pIter->aLvl[0].bEof;
}
static int fts5DlidxIterPrev(Fts5Index *p, Fts5DlidxIter *pIter){
return fts5DlidxIterPrevR(p, pIter, 0);
}
/*
** Free a doclist-index iterator object allocated by fts5DlidxIterInit().
*/
static void fts5DlidxIterFree(Fts5DlidxIter *pIter){
if( pIter ){
int i;
for(i=0; inLvl; i++){
fts5DataRelease(pIter->aLvl[i].pData);
}
sqlite3_free(pIter);
}
}
static Fts5DlidxIter *fts5DlidxIterInit(
Fts5Index *p, /* Fts5 Backend to iterate within */
int bRev, /* True for ORDER BY ASC */
int iSegid, /* Segment id */
int iLeafPg /* Leaf page number to load dlidx for */
){
Fts5DlidxIter *pIter = 0;
int i;
int bDone = 0;
for(i=0; p->rc==SQLITE_OK && bDone==0; i++){
sqlite3_int64 nByte = sizeof(Fts5DlidxIter) + i * sizeof(Fts5DlidxLvl);
Fts5DlidxIter *pNew;
pNew = (Fts5DlidxIter*)sqlite3_realloc64(pIter, nByte);
if( pNew==0 ){
p->rc = SQLITE_NOMEM;
}else{
i64 iRowid = FTS5_DLIDX_ROWID(iSegid, i, iLeafPg);
Fts5DlidxLvl *pLvl = &pNew->aLvl[i];
pIter = pNew;
memset(pLvl, 0, sizeof(Fts5DlidxLvl));
pLvl->pData = fts5DataRead(p, iRowid);
if( pLvl->pData && (pLvl->pData->p[0] & 0x0001)==0 ){
bDone = 1;
}
pIter->nLvl = i+1;
}
}
if( p->rc==SQLITE_OK ){
pIter->iSegid = iSegid;
if( bRev==0 ){
fts5DlidxIterFirst(pIter);
}else{
fts5DlidxIterLast(p, pIter);
}
}
if( p->rc!=SQLITE_OK ){
fts5DlidxIterFree(pIter);
pIter = 0;
}
return pIter;
}
static i64 fts5DlidxIterRowid(Fts5DlidxIter *pIter){
return pIter->aLvl[0].iRowid;
}
static int fts5DlidxIterPgno(Fts5DlidxIter *pIter){
return pIter->aLvl[0].iLeafPgno;
}
/*
** Load the next leaf page into the segment iterator.
*/
static void fts5SegIterNextPage(
Fts5Index *p, /* FTS5 backend object */
Fts5SegIter *pIter /* Iterator to advance to next page */
){
Fts5Data *pLeaf;
Fts5StructureSegment *pSeg = pIter->pSeg;
fts5DataRelease(pIter->pLeaf);
pIter->iLeafPgno++;
if( pIter->pNextLeaf ){
pIter->pLeaf = pIter->pNextLeaf;
pIter->pNextLeaf = 0;
}else if( pIter->iLeafPgno<=pSeg->pgnoLast ){
pIter->pLeaf = fts5LeafRead(p,
FTS5_SEGMENT_ROWID(pSeg->iSegid, pIter->iLeafPgno)
);
}else{
pIter->pLeaf = 0;
}
pLeaf = pIter->pLeaf;
if( pLeaf ){
pIter->iPgidxOff = pLeaf->szLeaf;
if( fts5LeafIsTermless(pLeaf) ){
pIter->iEndofDoclist = pLeaf->nn+1;
}else{
pIter->iPgidxOff += fts5GetVarint32(&pLeaf->p[pIter->iPgidxOff],
pIter->iEndofDoclist
);
}
}
}
/*
** Argument p points to a buffer containing a varint to be interpreted as a
** position list size field. Read the varint and return the number of bytes
** read. Before returning, set *pnSz to the number of bytes in the position
** list, and *pbDel to true if the delete flag is set, or false otherwise.
*/
static int fts5GetPoslistSize(const u8 *p, int *pnSz, int *pbDel){
int nSz;
int n = 0;
fts5FastGetVarint32(p, n, nSz);
assert_nc( nSz>=0 );
*pnSz = nSz/2;
*pbDel = nSz & 0x0001;
return n;
}
/*
** Fts5SegIter.iLeafOffset currently points to the first byte of a
** position-list size field. Read the value of the field and store it
** in the following variables:
**
** Fts5SegIter.nPos
** Fts5SegIter.bDel
**
** Leave Fts5SegIter.iLeafOffset pointing to the first byte of the
** position list content (if any).
*/
static void fts5SegIterLoadNPos(Fts5Index *p, Fts5SegIter *pIter){
if( p->rc==SQLITE_OK ){
int iOff = pIter->iLeafOffset; /* Offset to read at */
ASSERT_SZLEAF_OK(pIter->pLeaf);
if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){
int iEod = MIN(pIter->iEndofDoclist, pIter->pLeaf->szLeaf);
pIter->bDel = 0;
pIter->nPos = 1;
if( iOffpLeaf->p[iOff]==0 ){
pIter->bDel = 1;
iOff++;
if( iOffpLeaf->p[iOff]==0 ){
pIter->nPos = 1;
iOff++;
}else{
pIter->nPos = 0;
}
}
}else{
int nSz;
fts5FastGetVarint32(pIter->pLeaf->p, iOff, nSz);
pIter->bDel = (nSz & 0x0001);
pIter->nPos = nSz>>1;
assert_nc( pIter->nPos>=0 );
}
pIter->iLeafOffset = iOff;
}
}
static void fts5SegIterLoadRowid(Fts5Index *p, Fts5SegIter *pIter){
u8 *a = pIter->pLeaf->p; /* Buffer to read data from */
i64 iOff = pIter->iLeafOffset;
ASSERT_SZLEAF_OK(pIter->pLeaf);
while( iOff>=pIter->pLeaf->szLeaf ){
fts5SegIterNextPage(p, pIter);
if( pIter->pLeaf==0 ){
if( p->rc==SQLITE_OK ) p->rc = FTS5_CORRUPT;
return;
}
iOff = 4;
a = pIter->pLeaf->p;
}
iOff += sqlite3Fts5GetVarint(&a[iOff], (u64*)&pIter->iRowid);
pIter->iLeafOffset = iOff;
}
/*
** Fts5SegIter.iLeafOffset currently points to the first byte of the
** "nSuffix" field of a term. Function parameter nKeep contains the value
** of the "nPrefix" field (if there was one - it is passed 0 if this is
** the first term in the segment).
**
** This function populates:
**
** Fts5SegIter.term
** Fts5SegIter.rowid
**
** accordingly and leaves (Fts5SegIter.iLeafOffset) set to the content of
** the first position list. The position list belonging to document
** (Fts5SegIter.iRowid).
*/
static void fts5SegIterLoadTerm(Fts5Index *p, Fts5SegIter *pIter, int nKeep){
u8 *a = pIter->pLeaf->p; /* Buffer to read data from */
i64 iOff = pIter->iLeafOffset; /* Offset to read at */
int nNew; /* Bytes of new data */
iOff += fts5GetVarint32(&a[iOff], nNew);
if( iOff+nNew>pIter->pLeaf->szLeaf || nKeep>pIter->term.n || nNew==0 ){
p->rc = FTS5_CORRUPT;
return;
}
pIter->term.n = nKeep;
fts5BufferAppendBlob(&p->rc, &pIter->term, nNew, &a[iOff]);
assert( pIter->term.n<=pIter->term.nSpace );
iOff += nNew;
pIter->iTermLeafOffset = iOff;
pIter->iTermLeafPgno = pIter->iLeafPgno;
pIter->iLeafOffset = iOff;
if( pIter->iPgidxOff>=pIter->pLeaf->nn ){
pIter->iEndofDoclist = pIter->pLeaf->nn+1;
}else{
int nExtra;
pIter->iPgidxOff += fts5GetVarint32(&a[pIter->iPgidxOff], nExtra);
pIter->iEndofDoclist += nExtra;
}
fts5SegIterLoadRowid(p, pIter);
}
static void fts5SegIterNext(Fts5Index*, Fts5SegIter*, int*);
static void fts5SegIterNext_Reverse(Fts5Index*, Fts5SegIter*, int*);
static void fts5SegIterNext_None(Fts5Index*, Fts5SegIter*, int*);
static void fts5SegIterSetNext(Fts5Index *p, Fts5SegIter *pIter){
if( pIter->flags & FTS5_SEGITER_REVERSE ){
pIter->xNext = fts5SegIterNext_Reverse;
}else if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){
pIter->xNext = fts5SegIterNext_None;
}else{
pIter->xNext = fts5SegIterNext;
}
}
/*
** Allocate a tombstone hash page array object (pIter->pTombArray) for
** the iterator passed as the second argument. If an OOM error occurs,
** leave an error in the Fts5Index object.
*/
static void fts5SegIterAllocTombstone(Fts5Index *p, Fts5SegIter *pIter){
const int nTomb = pIter->pSeg->nPgTombstone;
if( nTomb>0 ){
int nByte = nTomb * sizeof(Fts5Data*) + sizeof(Fts5TombstoneArray);
Fts5TombstoneArray *pNew;
pNew = (Fts5TombstoneArray*)sqlite3Fts5MallocZero(&p->rc, nByte);
if( pNew ){
pNew->nTombstone = nTomb;
pNew->nRef = 1;
pIter->pTombArray = pNew;
}
}
}
/*
** Initialize the iterator object pIter to iterate through the entries in
** segment pSeg. The iterator is left pointing to the first entry when
** this function returns.
**
** If an error occurs, Fts5Index.rc is set to an appropriate error code. If
** an error has already occurred when this function is called, it is a no-op.
*/
static void fts5SegIterInit(
Fts5Index *p, /* FTS index object */
Fts5StructureSegment *pSeg, /* Description of segment */
Fts5SegIter *pIter /* Object to populate */
){
if( pSeg->pgnoFirst==0 ){
/* This happens if the segment is being used as an input to an incremental
** merge and all data has already been "trimmed". See function
** fts5TrimSegments() for details. In this case leave the iterator empty.
** The caller will see the (pIter->pLeaf==0) and assume the iterator is
** at EOF already. */
assert( pIter->pLeaf==0 );
return;
}
if( p->rc==SQLITE_OK ){
memset(pIter, 0, sizeof(*pIter));
fts5SegIterSetNext(p, pIter);
pIter->pSeg = pSeg;
pIter->iLeafPgno = pSeg->pgnoFirst-1;
do {
fts5SegIterNextPage(p, pIter);
}while( p->rc==SQLITE_OK && pIter->pLeaf && pIter->pLeaf->nn==4 );
}
if( p->rc==SQLITE_OK && pIter->pLeaf ){
pIter->iLeafOffset = 4;
assert( pIter->pLeaf!=0 );
assert_nc( pIter->pLeaf->nn>4 );
assert_nc( fts5LeafFirstTermOff(pIter->pLeaf)==4 );
pIter->iPgidxOff = pIter->pLeaf->szLeaf+1;
fts5SegIterLoadTerm(p, pIter, 0);
fts5SegIterLoadNPos(p, pIter);
fts5SegIterAllocTombstone(p, pIter);
}
}
/*
** This function is only ever called on iterators created by calls to
** Fts5IndexQuery() with the FTS5INDEX_QUERY_DESC flag set.
**
** The iterator is in an unusual state when this function is called: the
** Fts5SegIter.iLeafOffset variable is set to the offset of the start of
** the position-list size field for the first relevant rowid on the page.
** Fts5SegIter.rowid is set, but nPos and bDel are not.
**
** This function advances the iterator so that it points to the last
** relevant rowid on the page and, if necessary, initializes the
** aRowidOffset[] and iRowidOffset variables. At this point the iterator
** is in its regular state - Fts5SegIter.iLeafOffset points to the first
** byte of the position list content associated with said rowid.
*/
static void fts5SegIterReverseInitPage(Fts5Index *p, Fts5SegIter *pIter){
int eDetail = p->pConfig->eDetail;
int n = pIter->pLeaf->szLeaf;
int i = pIter->iLeafOffset;
u8 *a = pIter->pLeaf->p;
int iRowidOffset = 0;
if( n>pIter->iEndofDoclist ){
n = pIter->iEndofDoclist;
}
ASSERT_SZLEAF_OK(pIter->pLeaf);
while( 1 ){
u64 iDelta = 0;
if( eDetail==FTS5_DETAIL_NONE ){
/* todo */
if( i=n ) break;
i += fts5GetVarint(&a[i], &iDelta);
pIter->iRowid += iDelta;
/* If necessary, grow the pIter->aRowidOffset[] array. */
if( iRowidOffset>=pIter->nRowidOffset ){
int nNew = pIter->nRowidOffset + 8;
int *aNew = (int*)sqlite3_realloc64(pIter->aRowidOffset,nNew*sizeof(int));
if( aNew==0 ){
p->rc = SQLITE_NOMEM;
break;
}
pIter->aRowidOffset = aNew;
pIter->nRowidOffset = nNew;
}
pIter->aRowidOffset[iRowidOffset++] = pIter->iLeafOffset;
pIter->iLeafOffset = i;
}
pIter->iRowidOffset = iRowidOffset;
fts5SegIterLoadNPos(p, pIter);
}
/*
**
*/
static void fts5SegIterReverseNewPage(Fts5Index *p, Fts5SegIter *pIter){
assert( pIter->flags & FTS5_SEGITER_REVERSE );
assert( pIter->flags & FTS5_SEGITER_ONETERM );
fts5DataRelease(pIter->pLeaf);
pIter->pLeaf = 0;
while( p->rc==SQLITE_OK && pIter->iLeafPgno>pIter->iTermLeafPgno ){
Fts5Data *pNew;
pIter->iLeafPgno--;
pNew = fts5DataRead(p, FTS5_SEGMENT_ROWID(
pIter->pSeg->iSegid, pIter->iLeafPgno
));
if( pNew ){
/* iTermLeafOffset may be equal to szLeaf if the term is the last
** thing on the page - i.e. the first rowid is on the following page.
** In this case leave pIter->pLeaf==0, this iterator is at EOF. */
if( pIter->iLeafPgno==pIter->iTermLeafPgno ){
assert( pIter->pLeaf==0 );
if( pIter->iTermLeafOffsetszLeaf ){
pIter->pLeaf = pNew;
pIter->iLeafOffset = pIter->iTermLeafOffset;
}
}else{
int iRowidOff;
iRowidOff = fts5LeafFirstRowidOff(pNew);
if( iRowidOff ){
if( iRowidOff>=pNew->szLeaf ){
p->rc = FTS5_CORRUPT;
}else{
pIter->pLeaf = pNew;
pIter->iLeafOffset = iRowidOff;
}
}
}
if( pIter->pLeaf ){
u8 *a = &pIter->pLeaf->p[pIter->iLeafOffset];
pIter->iLeafOffset += fts5GetVarint(a, (u64*)&pIter->iRowid);
break;
}else{
fts5DataRelease(pNew);
}
}
}
if( pIter->pLeaf ){
pIter->iEndofDoclist = pIter->pLeaf->nn+1;
fts5SegIterReverseInitPage(p, pIter);
}
}
/*
** Return true if the iterator passed as the second argument currently
** points to a delete marker. A delete marker is an entry with a 0 byte
** position-list.
*/
static int fts5MultiIterIsEmpty(Fts5Index *p, Fts5Iter *pIter){
Fts5SegIter *pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst];
return (p->rc==SQLITE_OK && pSeg->pLeaf && pSeg->nPos==0);
}
/*
** Advance iterator pIter to the next entry.
**
** This version of fts5SegIterNext() is only used by reverse iterators.
*/
static void fts5SegIterNext_Reverse(
Fts5Index *p, /* FTS5 backend object */
Fts5SegIter *pIter, /* Iterator to advance */
int *pbUnused /* Unused */
){
assert( pIter->flags & FTS5_SEGITER_REVERSE );
assert( pIter->pNextLeaf==0 );
UNUSED_PARAM(pbUnused);
if( pIter->iRowidOffset>0 ){
u8 *a = pIter->pLeaf->p;
int iOff;
u64 iDelta;
pIter->iRowidOffset--;
pIter->iLeafOffset = pIter->aRowidOffset[pIter->iRowidOffset];
fts5SegIterLoadNPos(p, pIter);
iOff = pIter->iLeafOffset;
if( p->pConfig->eDetail!=FTS5_DETAIL_NONE ){
iOff += pIter->nPos;
}
fts5GetVarint(&a[iOff], &iDelta);
pIter->iRowid -= iDelta;
}else{
fts5SegIterReverseNewPage(p, pIter);
}
}
/*
** Advance iterator pIter to the next entry.
**
** This version of fts5SegIterNext() is only used if detail=none and the
** iterator is not a reverse direction iterator.
*/
static void fts5SegIterNext_None(
Fts5Index *p, /* FTS5 backend object */
Fts5SegIter *pIter, /* Iterator to advance */
int *pbNewTerm /* OUT: Set for new term */
){
int iOff;
assert( p->rc==SQLITE_OK );
assert( (pIter->flags & FTS5_SEGITER_REVERSE)==0 );
assert( p->pConfig->eDetail==FTS5_DETAIL_NONE );
ASSERT_SZLEAF_OK(pIter->pLeaf);
iOff = pIter->iLeafOffset;
/* Next entry is on the next page */
while( pIter->pSeg && iOff>=pIter->pLeaf->szLeaf ){
fts5SegIterNextPage(p, pIter);
if( p->rc || pIter->pLeaf==0 ) return;
pIter->iRowid = 0;
iOff = 4;
}
if( iOffiEndofDoclist ){
/* Next entry is on the current page */
i64 iDelta;
iOff += sqlite3Fts5GetVarint(&pIter->pLeaf->p[iOff], (u64*)&iDelta);
pIter->iLeafOffset = iOff;
pIter->iRowid += iDelta;
}else if( (pIter->flags & FTS5_SEGITER_ONETERM)==0 ){
if( pIter->pSeg ){
int nKeep = 0;
if( iOff!=fts5LeafFirstTermOff(pIter->pLeaf) ){
iOff += fts5GetVarint32(&pIter->pLeaf->p[iOff], nKeep);
}
pIter->iLeafOffset = iOff;
fts5SegIterLoadTerm(p, pIter, nKeep);
}else{
const u8 *pList = 0;
const char *zTerm = 0;
int nTerm = 0;
int nList;
sqlite3Fts5HashScanNext(p->pHash);
sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &nTerm, &pList, &nList);
if( pList==0 ) goto next_none_eof;
pIter->pLeaf->p = (u8*)pList;
pIter->pLeaf->nn = nList;
pIter->pLeaf->szLeaf = nList;
pIter->iEndofDoclist = nList;
sqlite3Fts5BufferSet(&p->rc,&pIter->term, nTerm, (u8*)zTerm);
pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid);
}
if( pbNewTerm ) *pbNewTerm = 1;
}else{
goto next_none_eof;
}
fts5SegIterLoadNPos(p, pIter);
return;
next_none_eof:
fts5DataRelease(pIter->pLeaf);
pIter->pLeaf = 0;
}
/*
** Advance iterator pIter to the next entry.
**
** If an error occurs, Fts5Index.rc is set to an appropriate error code. It
** is not considered an error if the iterator reaches EOF. If an error has
** already occurred when this function is called, it is a no-op.
*/
static void fts5SegIterNext(
Fts5Index *p, /* FTS5 backend object */
Fts5SegIter *pIter, /* Iterator to advance */
int *pbNewTerm /* OUT: Set for new term */
){
Fts5Data *pLeaf = pIter->pLeaf;
int iOff;
int bNewTerm = 0;
int nKeep = 0;
u8 *a;
int n;
assert( pbNewTerm==0 || *pbNewTerm==0 );
assert( p->pConfig->eDetail!=FTS5_DETAIL_NONE );
/* Search for the end of the position list within the current page. */
a = pLeaf->p;
n = pLeaf->szLeaf;
ASSERT_SZLEAF_OK(pLeaf);
iOff = pIter->iLeafOffset + pIter->nPos;
if( iOffiEndofDoclist );
if( iOff>=pIter->iEndofDoclist ){
bNewTerm = 1;
if( iOff!=fts5LeafFirstTermOff(pLeaf) ){
iOff += fts5GetVarint32(&a[iOff], nKeep);
}
}else{
u64 iDelta;
iOff += sqlite3Fts5GetVarint(&a[iOff], &iDelta);
pIter->iRowid += iDelta;
assert_nc( iDelta>0 );
}
pIter->iLeafOffset = iOff;
}else if( pIter->pSeg==0 ){
const u8 *pList = 0;
const char *zTerm = 0;
int nTerm = 0;
int nList = 0;
assert( (pIter->flags & FTS5_SEGITER_ONETERM) || pbNewTerm );
if( 0==(pIter->flags & FTS5_SEGITER_ONETERM) ){
sqlite3Fts5HashScanNext(p->pHash);
sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &nTerm, &pList, &nList);
}
if( pList==0 ){
fts5DataRelease(pIter->pLeaf);
pIter->pLeaf = 0;
}else{
pIter->pLeaf->p = (u8*)pList;
pIter->pLeaf->nn = nList;
pIter->pLeaf->szLeaf = nList;
pIter->iEndofDoclist = nList+1;
sqlite3Fts5BufferSet(&p->rc, &pIter->term, nTerm, (u8*)zTerm);
pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid);
*pbNewTerm = 1;
}
}else{
iOff = 0;
/* Next entry is not on the current page */
while( iOff==0 ){
fts5SegIterNextPage(p, pIter);
pLeaf = pIter->pLeaf;
if( pLeaf==0 ) break;
ASSERT_SZLEAF_OK(pLeaf);
if( (iOff = fts5LeafFirstRowidOff(pLeaf)) && iOffszLeaf ){
iOff += sqlite3Fts5GetVarint(&pLeaf->p[iOff], (u64*)&pIter->iRowid);
pIter->iLeafOffset = iOff;
if( pLeaf->nn>pLeaf->szLeaf ){
pIter->iPgidxOff = pLeaf->szLeaf + fts5GetVarint32(
&pLeaf->p[pLeaf->szLeaf], pIter->iEndofDoclist
);
}
}
else if( pLeaf->nn>pLeaf->szLeaf ){
pIter->iPgidxOff = pLeaf->szLeaf + fts5GetVarint32(
&pLeaf->p[pLeaf->szLeaf], iOff
);
pIter->iLeafOffset = iOff;
pIter->iEndofDoclist = iOff;
bNewTerm = 1;
}
assert_nc( iOffszLeaf );
if( iOff>pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
return;
}
}
}
/* Check if the iterator is now at EOF. If so, return early. */
if( pIter->pLeaf ){
if( bNewTerm ){
if( pIter->flags & FTS5_SEGITER_ONETERM ){
fts5DataRelease(pIter->pLeaf);
pIter->pLeaf = 0;
}else{
fts5SegIterLoadTerm(p, pIter, nKeep);
fts5SegIterLoadNPos(p, pIter);
if( pbNewTerm ) *pbNewTerm = 1;
}
}else{
/* The following could be done by calling fts5SegIterLoadNPos(). But
** this block is particularly performance critical, so equivalent
** code is inlined. */
int nSz;
assert_nc( pIter->iLeafOffset<=pIter->pLeaf->nn );
fts5FastGetVarint32(pIter->pLeaf->p, pIter->iLeafOffset, nSz);
pIter->bDel = (nSz & 0x0001);
pIter->nPos = nSz>>1;
assert_nc( pIter->nPos>=0 );
}
}
}
#define SWAPVAL(T, a, b) { T tmp; tmp=a; a=b; b=tmp; }
#define fts5IndexSkipVarint(a, iOff) { \
int iEnd = iOff+9; \
while( (a[iOff++] & 0x80) && iOffpDlidx;
Fts5Data *pLast = 0;
int pgnoLast = 0;
if( pDlidx && p->pConfig->iVersion==FTS5_CURRENT_VERSION ){
int iSegid = pIter->pSeg->iSegid;
pgnoLast = fts5DlidxIterPgno(pDlidx);
pLast = fts5LeafRead(p, FTS5_SEGMENT_ROWID(iSegid, pgnoLast));
}else{
Fts5Data *pLeaf = pIter->pLeaf; /* Current leaf data */
/* Currently, Fts5SegIter.iLeafOffset points to the first byte of
** position-list content for the current rowid. Back it up so that it
** points to the start of the position-list size field. */
int iPoslist;
if( pIter->iTermLeafPgno==pIter->iLeafPgno ){
iPoslist = pIter->iTermLeafOffset;
}else{
iPoslist = 4;
}
fts5IndexSkipVarint(pLeaf->p, iPoslist);
pIter->iLeafOffset = iPoslist;
/* If this condition is true then the largest rowid for the current
** term may not be stored on the current page. So search forward to
** see where said rowid really is. */
if( pIter->iEndofDoclist>=pLeaf->szLeaf ){
int pgno;
Fts5StructureSegment *pSeg = pIter->pSeg;
/* The last rowid in the doclist may not be on the current page. Search
** forward to find the page containing the last rowid. */
for(pgno=pIter->iLeafPgno+1; !p->rc && pgno<=pSeg->pgnoLast; pgno++){
i64 iAbs = FTS5_SEGMENT_ROWID(pSeg->iSegid, pgno);
Fts5Data *pNew = fts5LeafRead(p, iAbs);
if( pNew ){
int iRowid, bTermless;
iRowid = fts5LeafFirstRowidOff(pNew);
bTermless = fts5LeafIsTermless(pNew);
if( iRowid ){
SWAPVAL(Fts5Data*, pNew, pLast);
pgnoLast = pgno;
}
fts5DataRelease(pNew);
if( bTermless==0 ) break;
}
}
}
}
/* If pLast is NULL at this point, then the last rowid for this doclist
** lies on the page currently indicated by the iterator. In this case
** pIter->iLeafOffset is already set to point to the position-list size
** field associated with the first relevant rowid on the page.
**
** Or, if pLast is non-NULL, then it is the page that contains the last
** rowid. In this case configure the iterator so that it points to the
** first rowid on this page.
*/
if( pLast ){
int iOff;
fts5DataRelease(pIter->pLeaf);
pIter->pLeaf = pLast;
pIter->iLeafPgno = pgnoLast;
iOff = fts5LeafFirstRowidOff(pLast);
if( iOff>pLast->szLeaf ){
p->rc = FTS5_CORRUPT;
return;
}
iOff += fts5GetVarint(&pLast->p[iOff], (u64*)&pIter->iRowid);
pIter->iLeafOffset = iOff;
if( fts5LeafIsTermless(pLast) ){
pIter->iEndofDoclist = pLast->nn+1;
}else{
pIter->iEndofDoclist = fts5LeafFirstTermOff(pLast);
}
}
fts5SegIterReverseInitPage(p, pIter);
}
/*
** Iterator pIter currently points to the first rowid of a doclist.
** There is a doclist-index associated with the final term on the current
** page. If the current term is the last term on the page, load the
** doclist-index from disk and initialize an iterator at (pIter->pDlidx).
*/
static void fts5SegIterLoadDlidx(Fts5Index *p, Fts5SegIter *pIter){
int iSeg = pIter->pSeg->iSegid;
int bRev = (pIter->flags & FTS5_SEGITER_REVERSE);
Fts5Data *pLeaf = pIter->pLeaf; /* Current leaf data */
assert( pIter->flags & FTS5_SEGITER_ONETERM );
assert( pIter->pDlidx==0 );
/* Check if the current doclist ends on this page. If it does, return
** early without loading the doclist-index (as it belongs to a different
** term. */
if( pIter->iTermLeafPgno==pIter->iLeafPgno
&& pIter->iEndofDoclistszLeaf
){
return;
}
pIter->pDlidx = fts5DlidxIterInit(p, bRev, iSeg, pIter->iTermLeafPgno);
}
/*
** The iterator object passed as the second argument currently contains
** no valid values except for the Fts5SegIter.pLeaf member variable. This
** function searches the leaf page for a term matching (pTerm/nTerm).
**
** If the specified term is found on the page, then the iterator is left
** pointing to it. If argument bGe is zero and the term is not found,
** the iterator is left pointing at EOF.
**
** If bGe is non-zero and the specified term is not found, then the
** iterator is left pointing to the smallest term in the segment that
** is larger than the specified term, even if this term is not on the
** current page.
*/
static void fts5LeafSeek(
Fts5Index *p, /* Leave any error code here */
int bGe, /* True for a >= search */
Fts5SegIter *pIter, /* Iterator to seek */
const u8 *pTerm, int nTerm /* Term to search for */
){
u32 iOff;
const u8 *a = pIter->pLeaf->p;
u32 n = (u32)pIter->pLeaf->nn;
u32 nMatch = 0;
u32 nKeep = 0;
u32 nNew = 0;
u32 iTermOff;
u32 iPgidx; /* Current offset in pgidx */
int bEndOfPage = 0;
assert( p->rc==SQLITE_OK );
iPgidx = (u32)pIter->pLeaf->szLeaf;
iPgidx += fts5GetVarint32(&a[iPgidx], iTermOff);
iOff = iTermOff;
if( iOff>n ){
p->rc = FTS5_CORRUPT;
return;
}
while( 1 ){
/* Figure out how many new bytes are in this term */
fts5FastGetVarint32(a, iOff, nNew);
if( nKeep=nMatch );
if( nKeep==nMatch ){
u32 nCmp;
u32 i;
nCmp = (u32)MIN(nNew, nTerm-nMatch);
for(i=0; ipTerm[nMatch] ){
goto search_failed;
}
}
if( iPgidx>=n ){
bEndOfPage = 1;
break;
}
iPgidx += fts5GetVarint32(&a[iPgidx], nKeep);
iTermOff += nKeep;
iOff = iTermOff;
if( iOff>=n ){
p->rc = FTS5_CORRUPT;
return;
}
/* Read the nKeep field of the next term. */
fts5FastGetVarint32(a, iOff, nKeep);
}
search_failed:
if( bGe==0 ){
fts5DataRelease(pIter->pLeaf);
pIter->pLeaf = 0;
return;
}else if( bEndOfPage ){
do {
fts5SegIterNextPage(p, pIter);
if( pIter->pLeaf==0 ) return;
a = pIter->pLeaf->p;
if( fts5LeafIsTermless(pIter->pLeaf)==0 ){
iPgidx = (u32)pIter->pLeaf->szLeaf;
iPgidx += fts5GetVarint32(&pIter->pLeaf->p[iPgidx], iOff);
if( iOff<4 || (i64)iOff>=pIter->pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
return;
}else{
nKeep = 0;
iTermOff = iOff;
n = (u32)pIter->pLeaf->nn;
iOff += fts5GetVarint32(&a[iOff], nNew);
break;
}
}
}while( 1 );
}
search_success:
if( (i64)iOff+nNew>n || nNew<1 ){
p->rc = FTS5_CORRUPT;
return;
}
pIter->iLeafOffset = iOff + nNew;
pIter->iTermLeafOffset = pIter->iLeafOffset;
pIter->iTermLeafPgno = pIter->iLeafPgno;
fts5BufferSet(&p->rc, &pIter->term, nKeep, pTerm);
fts5BufferAppendBlob(&p->rc, &pIter->term, nNew, &a[iOff]);
if( iPgidx>=n ){
pIter->iEndofDoclist = pIter->pLeaf->nn+1;
}else{
int nExtra;
iPgidx += fts5GetVarint32(&a[iPgidx], nExtra);
pIter->iEndofDoclist = iTermOff + nExtra;
}
pIter->iPgidxOff = iPgidx;
fts5SegIterLoadRowid(p, pIter);
fts5SegIterLoadNPos(p, pIter);
}
static sqlite3_stmt *fts5IdxSelectStmt(Fts5Index *p){
if( p->pIdxSelect==0 ){
Fts5Config *pConfig = p->pConfig;
fts5IndexPrepareStmt(p, &p->pIdxSelect, sqlite3_mprintf(
"SELECT pgno FROM '%q'.'%q_idx' WHERE "
"segid=? AND term<=? ORDER BY term DESC LIMIT 1",
pConfig->zDb, pConfig->zName
));
}
return p->pIdxSelect;
}
/*
** Initialize the object pIter to point to term pTerm/nTerm within segment
** pSeg. If there is no such term in the index, the iterator is set to EOF.
**
** If an error occurs, Fts5Index.rc is set to an appropriate error code. If
** an error has already occurred when this function is called, it is a no-op.
*/
static void fts5SegIterSeekInit(
Fts5Index *p, /* FTS5 backend */
const u8 *pTerm, int nTerm, /* Term to seek to */
int flags, /* Mask of FTS5INDEX_XXX flags */
Fts5StructureSegment *pSeg, /* Description of segment */
Fts5SegIter *pIter /* Object to populate */
){
int iPg = 1;
int bGe = (flags & FTS5INDEX_QUERY_SCAN);
int bDlidx = 0; /* True if there is a doclist-index */
sqlite3_stmt *pIdxSelect = 0;
assert( bGe==0 || (flags & FTS5INDEX_QUERY_DESC)==0 );
assert( pTerm && nTerm );
memset(pIter, 0, sizeof(*pIter));
pIter->pSeg = pSeg;
/* This block sets stack variable iPg to the leaf page number that may
** contain term (pTerm/nTerm), if it is present in the segment. */
pIdxSelect = fts5IdxSelectStmt(p);
if( p->rc ) return;
sqlite3_bind_int(pIdxSelect, 1, pSeg->iSegid);
sqlite3_bind_blob(pIdxSelect, 2, pTerm, nTerm, SQLITE_STATIC);
if( SQLITE_ROW==sqlite3_step(pIdxSelect) ){
i64 val = sqlite3_column_int(pIdxSelect, 0);
iPg = (int)(val>>1);
bDlidx = (val & 0x0001);
}
p->rc = sqlite3_reset(pIdxSelect);
sqlite3_bind_null(pIdxSelect, 2);
if( iPgpgnoFirst ){
iPg = pSeg->pgnoFirst;
bDlidx = 0;
}
pIter->iLeafPgno = iPg - 1;
fts5SegIterNextPage(p, pIter);
if( pIter->pLeaf ){
fts5LeafSeek(p, bGe, pIter, pTerm, nTerm);
}
if( p->rc==SQLITE_OK && (bGe==0 || (flags & FTS5INDEX_QUERY_SCANONETERM)) ){
pIter->flags |= FTS5_SEGITER_ONETERM;
if( pIter->pLeaf ){
if( flags & FTS5INDEX_QUERY_DESC ){
pIter->flags |= FTS5_SEGITER_REVERSE;
}
if( bDlidx ){
fts5SegIterLoadDlidx(p, pIter);
}
if( flags & FTS5INDEX_QUERY_DESC ){
fts5SegIterReverse(p, pIter);
}
}
}
fts5SegIterSetNext(p, pIter);
if( 0==(flags & FTS5INDEX_QUERY_SCANONETERM) ){
fts5SegIterAllocTombstone(p, pIter);
}
/* Either:
**
** 1) an error has occurred, or
** 2) the iterator points to EOF, or
** 3) the iterator points to an entry with term (pTerm/nTerm), or
** 4) the FTS5INDEX_QUERY_SCAN flag was set and the iterator points
** to an entry with a term greater than or equal to (pTerm/nTerm).
*/
assert_nc( p->rc!=SQLITE_OK /* 1 */
|| pIter->pLeaf==0 /* 2 */
|| fts5BufferCompareBlob(&pIter->term, pTerm, nTerm)==0 /* 3 */
|| (bGe && fts5BufferCompareBlob(&pIter->term, pTerm, nTerm)>0) /* 4 */
);
}
/*
** SQL used by fts5SegIterNextInit() to find the page to open.
*/
static sqlite3_stmt *fts5IdxNextStmt(Fts5Index *p){
if( p->pIdxNextSelect==0 ){
Fts5Config *pConfig = p->pConfig;
fts5IndexPrepareStmt(p, &p->pIdxNextSelect, sqlite3_mprintf(
"SELECT pgno FROM '%q'.'%q_idx' WHERE "
"segid=? AND term>? ORDER BY term ASC LIMIT 1",
pConfig->zDb, pConfig->zName
));
}
return p->pIdxNextSelect;
}
/*
** This is similar to fts5SegIterSeekInit(), except that it initializes
** the segment iterator to point to the first term following the page
** with pToken/nToken on it.
*/
static void fts5SegIterNextInit(
Fts5Index *p,
const char *pTerm, int nTerm,
Fts5StructureSegment *pSeg, /* Description of segment */
Fts5SegIter *pIter /* Object to populate */
){
int iPg = -1; /* Page of segment to open */
int bDlidx = 0;
sqlite3_stmt *pSel = 0; /* SELECT to find iPg */
pSel = fts5IdxNextStmt(p);
if( pSel ){
assert( p->rc==SQLITE_OK );
sqlite3_bind_int(pSel, 1, pSeg->iSegid);
sqlite3_bind_blob(pSel, 2, pTerm, nTerm, SQLITE_STATIC);
if( sqlite3_step(pSel)==SQLITE_ROW ){
i64 val = sqlite3_column_int64(pSel, 0);
iPg = (int)(val>>1);
bDlidx = (val & 0x0001);
}
p->rc = sqlite3_reset(pSel);
sqlite3_bind_null(pSel, 2);
if( p->rc ) return;
}
memset(pIter, 0, sizeof(*pIter));
pIter->pSeg = pSeg;
pIter->flags |= FTS5_SEGITER_ONETERM;
if( iPg>=0 ){
pIter->iLeafPgno = iPg - 1;
fts5SegIterNextPage(p, pIter);
fts5SegIterSetNext(p, pIter);
}
if( pIter->pLeaf ){
const u8 *a = pIter->pLeaf->p;
int iTermOff = 0;
pIter->iPgidxOff = pIter->pLeaf->szLeaf;
pIter->iPgidxOff += fts5GetVarint32(&a[pIter->iPgidxOff], iTermOff);
pIter->iLeafOffset = iTermOff;
fts5SegIterLoadTerm(p, pIter, 0);
fts5SegIterLoadNPos(p, pIter);
if( bDlidx ) fts5SegIterLoadDlidx(p, pIter);
assert( p->rc!=SQLITE_OK ||
fts5BufferCompareBlob(&pIter->term, (const u8*)pTerm, nTerm)>0
);
}
}
/*
** Initialize the object pIter to point to term pTerm/nTerm within the
** in-memory hash table. If there is no such term in the hash-table, the
** iterator is set to EOF.
**
** If an error occurs, Fts5Index.rc is set to an appropriate error code. If
** an error has already occurred when this function is called, it is a no-op.
*/
static void fts5SegIterHashInit(
Fts5Index *p, /* FTS5 backend */
const u8 *pTerm, int nTerm, /* Term to seek to */
int flags, /* Mask of FTS5INDEX_XXX flags */
Fts5SegIter *pIter /* Object to populate */
){
int nList = 0;
const u8 *z = 0;
int n = 0;
Fts5Data *pLeaf = 0;
assert( p->pHash );
assert( p->rc==SQLITE_OK );
if( pTerm==0 || (flags & FTS5INDEX_QUERY_SCAN) ){
const u8 *pList = 0;
p->rc = sqlite3Fts5HashScanInit(p->pHash, (const char*)pTerm, nTerm);
sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &n, &pList, &nList);
if( pList ){
pLeaf = fts5IdxMalloc(p, sizeof(Fts5Data));
if( pLeaf ){
pLeaf->p = (u8*)pList;
}
}
/* The call to sqlite3Fts5HashScanInit() causes the hash table to
** fill the size field of all existing position lists. This means they
** can no longer be appended to. Since the only scenario in which they
** can be appended to is if the previous operation on this table was
** a DELETE, by clearing the Fts5Index.bDelete flag we can avoid this
** possibility altogether. */
p->bDelete = 0;
}else{
p->rc = sqlite3Fts5HashQuery(p->pHash, sizeof(Fts5Data),
(const char*)pTerm, nTerm, (void**)&pLeaf, &nList
);
if( pLeaf ){
pLeaf->p = (u8*)&pLeaf[1];
}
z = pTerm;
n = nTerm;
pIter->flags |= FTS5_SEGITER_ONETERM;
}
if( pLeaf ){
sqlite3Fts5BufferSet(&p->rc, &pIter->term, n, z);
pLeaf->nn = pLeaf->szLeaf = nList;
pIter->pLeaf = pLeaf;
pIter->iLeafOffset = fts5GetVarint(pLeaf->p, (u64*)&pIter->iRowid);
pIter->iEndofDoclist = pLeaf->nn;
if( flags & FTS5INDEX_QUERY_DESC ){
pIter->flags |= FTS5_SEGITER_REVERSE;
fts5SegIterReverseInitPage(p, pIter);
}else{
fts5SegIterLoadNPos(p, pIter);
}
}
fts5SegIterSetNext(p, pIter);
}
/*
** Array ap[] contains n elements. Release each of these elements using
** fts5DataRelease(). Then free the array itself using sqlite3_free().
*/
static void fts5IndexFreeArray(Fts5Data **ap, int n){
if( ap ){
int ii;
for(ii=0; iinRef--;
if( p->nRef<=0 ){
int ii;
for(ii=0; iinTombstone; ii++){
fts5DataRelease(p->apTombstone[ii]);
}
sqlite3_free(p);
}
}
}
/*
** Zero the iterator passed as the only argument.
*/
static void fts5SegIterClear(Fts5SegIter *pIter){
fts5BufferFree(&pIter->term);
fts5DataRelease(pIter->pLeaf);
fts5DataRelease(pIter->pNextLeaf);
fts5TombstoneArrayDelete(pIter->pTombArray);
fts5DlidxIterFree(pIter->pDlidx);
sqlite3_free(pIter->aRowidOffset);
memset(pIter, 0, sizeof(Fts5SegIter));
}
#ifdef SQLITE_DEBUG
/*
** This function is used as part of the big assert() procedure implemented by
** fts5AssertMultiIterSetup(). It ensures that the result currently stored
** in *pRes is the correct result of comparing the current positions of the
** two iterators.
*/
static void fts5AssertComparisonResult(
Fts5Iter *pIter,
Fts5SegIter *p1,
Fts5SegIter *p2,
Fts5CResult *pRes
){
int i1 = p1 - pIter->aSeg;
int i2 = p2 - pIter->aSeg;
if( p1->pLeaf || p2->pLeaf ){
if( p1->pLeaf==0 ){
assert( pRes->iFirst==i2 );
}else if( p2->pLeaf==0 ){
assert( pRes->iFirst==i1 );
}else{
int nMin = MIN(p1->term.n, p2->term.n);
int res = fts5Memcmp(p1->term.p, p2->term.p, nMin);
if( res==0 ) res = p1->term.n - p2->term.n;
if( res==0 ){
assert( pRes->bTermEq==1 );
assert( p1->iRowid!=p2->iRowid );
res = ((p1->iRowid > p2->iRowid)==pIter->bRev) ? -1 : 1;
}else{
assert( pRes->bTermEq==0 );
}
if( res<0 ){
assert( pRes->iFirst==i1 );
}else{
assert( pRes->iFirst==i2 );
}
}
}
}
/*
** This function is a no-op unless SQLITE_DEBUG is defined when this module
** is compiled. In that case, this function is essentially an assert()
** statement used to verify that the contents of the pIter->aFirst[] array
** are correct.
*/
static void fts5AssertMultiIterSetup(Fts5Index *p, Fts5Iter *pIter){
if( p->rc==SQLITE_OK ){
Fts5SegIter *pFirst = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
int i;
assert( (pFirst->pLeaf==0)==pIter->base.bEof );
/* Check that pIter->iSwitchRowid is set correctly. */
for(i=0; inSeg; i++){
Fts5SegIter *p1 = &pIter->aSeg[i];
assert( p1==pFirst
|| p1->pLeaf==0
|| fts5BufferCompare(&pFirst->term, &p1->term)
|| p1->iRowid==pIter->iSwitchRowid
|| (p1->iRowidiSwitchRowid)==pIter->bRev
);
}
for(i=0; inSeg; i+=2){
Fts5SegIter *p1 = &pIter->aSeg[i];
Fts5SegIter *p2 = &pIter->aSeg[i+1];
Fts5CResult *pRes = &pIter->aFirst[(pIter->nSeg + i) / 2];
fts5AssertComparisonResult(pIter, p1, p2, pRes);
}
for(i=1; i<(pIter->nSeg / 2); i+=2){
Fts5SegIter *p1 = &pIter->aSeg[ pIter->aFirst[i*2].iFirst ];
Fts5SegIter *p2 = &pIter->aSeg[ pIter->aFirst[i*2+1].iFirst ];
Fts5CResult *pRes = &pIter->aFirst[i];
fts5AssertComparisonResult(pIter, p1, p2, pRes);
}
}
}
#else
# define fts5AssertMultiIterSetup(x,y)
#endif
/*
** Do the comparison necessary to populate pIter->aFirst[iOut].
**
** If the returned value is non-zero, then it is the index of an entry
** in the pIter->aSeg[] array that is (a) not at EOF, and (b) pointing
** to a key that is a duplicate of another, higher priority,
** segment-iterator in the pSeg->aSeg[] array.
*/
static int fts5MultiIterDoCompare(Fts5Iter *pIter, int iOut){
int i1; /* Index of left-hand Fts5SegIter */
int i2; /* Index of right-hand Fts5SegIter */
int iRes;
Fts5SegIter *p1; /* Left-hand Fts5SegIter */
Fts5SegIter *p2; /* Right-hand Fts5SegIter */
Fts5CResult *pRes = &pIter->aFirst[iOut];
assert( iOutnSeg && iOut>0 );
assert( pIter->bRev==0 || pIter->bRev==1 );
if( iOut>=(pIter->nSeg/2) ){
i1 = (iOut - pIter->nSeg/2) * 2;
i2 = i1 + 1;
}else{
i1 = pIter->aFirst[iOut*2].iFirst;
i2 = pIter->aFirst[iOut*2+1].iFirst;
}
p1 = &pIter->aSeg[i1];
p2 = &pIter->aSeg[i2];
pRes->bTermEq = 0;
if( p1->pLeaf==0 ){ /* If p1 is at EOF */
iRes = i2;
}else if( p2->pLeaf==0 ){ /* If p2 is at EOF */
iRes = i1;
}else{
int res = fts5BufferCompare(&p1->term, &p2->term);
if( res==0 ){
assert_nc( i2>i1 );
assert_nc( i2!=0 );
pRes->bTermEq = 1;
if( p1->iRowid==p2->iRowid ){
return i2;
}
res = ((p1->iRowid > p2->iRowid)==pIter->bRev) ? -1 : +1;
}
assert( res!=0 );
if( res<0 ){
iRes = i1;
}else{
iRes = i2;
}
}
pRes->iFirst = (u16)iRes;
return 0;
}
/*
** Move the seg-iter so that it points to the first rowid on page iLeafPgno.
** It is an error if leaf iLeafPgno does not exist. Unless the db is
** a 'secure-delete' db, if it contains no rowids then this is also an error.
*/
static void fts5SegIterGotoPage(
Fts5Index *p, /* FTS5 backend object */
Fts5SegIter *pIter, /* Iterator to advance */
int iLeafPgno
){
assert( iLeafPgno>pIter->iLeafPgno );
if( iLeafPgno>pIter->pSeg->pgnoLast ){
p->rc = FTS5_CORRUPT;
}else{
fts5DataRelease(pIter->pNextLeaf);
pIter->pNextLeaf = 0;
pIter->iLeafPgno = iLeafPgno-1;
while( p->rc==SQLITE_OK ){
int iOff;
fts5SegIterNextPage(p, pIter);
if( pIter->pLeaf==0 ) break;
iOff = fts5LeafFirstRowidOff(pIter->pLeaf);
if( iOff>0 ){
u8 *a = pIter->pLeaf->p;
int n = pIter->pLeaf->szLeaf;
if( iOff<4 || iOff>=n ){
p->rc = FTS5_CORRUPT;
}else{
iOff += fts5GetVarint(&a[iOff], (u64*)&pIter->iRowid);
pIter->iLeafOffset = iOff;
fts5SegIterLoadNPos(p, pIter);
}
break;
}
}
}
}
/*
** Advance the iterator passed as the second argument until it is at or
** past rowid iFrom. Regardless of the value of iFrom, the iterator is
** always advanced at least once.
*/
static void fts5SegIterNextFrom(
Fts5Index *p, /* FTS5 backend object */
Fts5SegIter *pIter, /* Iterator to advance */
i64 iMatch /* Advance iterator at least this far */
){
int bRev = (pIter->flags & FTS5_SEGITER_REVERSE);
Fts5DlidxIter *pDlidx = pIter->pDlidx;
int iLeafPgno = pIter->iLeafPgno;
int bMove = 1;
assert( pIter->flags & FTS5_SEGITER_ONETERM );
assert( pIter->pDlidx );
assert( pIter->pLeaf );
if( bRev==0 ){
while( !fts5DlidxIterEof(p, pDlidx) && iMatch>fts5DlidxIterRowid(pDlidx) ){
iLeafPgno = fts5DlidxIterPgno(pDlidx);
fts5DlidxIterNext(p, pDlidx);
}
assert_nc( iLeafPgno>=pIter->iLeafPgno || p->rc );
if( iLeafPgno>pIter->iLeafPgno ){
fts5SegIterGotoPage(p, pIter, iLeafPgno);
bMove = 0;
}
}else{
assert( pIter->pNextLeaf==0 );
assert( iMatchiRowid );
while( !fts5DlidxIterEof(p, pDlidx) && iMatchiLeafPgno );
if( iLeafPgnoiLeafPgno ){
pIter->iLeafPgno = iLeafPgno+1;
fts5SegIterReverseNewPage(p, pIter);
bMove = 0;
}
}
do{
if( bMove && p->rc==SQLITE_OK ) pIter->xNext(p, pIter, 0);
if( pIter->pLeaf==0 ) break;
if( bRev==0 && pIter->iRowid>=iMatch ) break;
if( bRev!=0 && pIter->iRowid<=iMatch ) break;
bMove = 1;
}while( p->rc==SQLITE_OK );
}
/*
** Free the iterator object passed as the second argument.
*/
static void fts5MultiIterFree(Fts5Iter *pIter){
if( pIter ){
int i;
for(i=0; inSeg; i++){
fts5SegIterClear(&pIter->aSeg[i]);
}
fts5BufferFree(&pIter->poslist);
sqlite3_free(pIter);
}
}
static void fts5MultiIterAdvanced(
Fts5Index *p, /* FTS5 backend to iterate within */
Fts5Iter *pIter, /* Iterator to update aFirst[] array for */
int iChanged, /* Index of sub-iterator just advanced */
int iMinset /* Minimum entry in aFirst[] to set */
){
int i;
for(i=(pIter->nSeg+iChanged)/2; i>=iMinset && p->rc==SQLITE_OK; i=i/2){
int iEq;
if( (iEq = fts5MultiIterDoCompare(pIter, i)) ){
Fts5SegIter *pSeg = &pIter->aSeg[iEq];
assert( p->rc==SQLITE_OK );
pSeg->xNext(p, pSeg, 0);
i = pIter->nSeg + iEq;
}
}
}
/*
** Sub-iterator iChanged of iterator pIter has just been advanced. It still
** points to the same term though - just a different rowid. This function
** attempts to update the contents of the pIter->aFirst[] accordingly.
** If it does so successfully, 0 is returned. Otherwise 1.
**
** If non-zero is returned, the caller should call fts5MultiIterAdvanced()
** on the iterator instead. That function does the same as this one, except
** that it deals with more complicated cases as well.
*/
static int fts5MultiIterAdvanceRowid(
Fts5Iter *pIter, /* Iterator to update aFirst[] array for */
int iChanged, /* Index of sub-iterator just advanced */
Fts5SegIter **ppFirst
){
Fts5SegIter *pNew = &pIter->aSeg[iChanged];
if( pNew->iRowid==pIter->iSwitchRowid
|| (pNew->iRowidiSwitchRowid)==pIter->bRev
){
int i;
Fts5SegIter *pOther = &pIter->aSeg[iChanged ^ 0x0001];
pIter->iSwitchRowid = pIter->bRev ? SMALLEST_INT64 : LARGEST_INT64;
for(i=(pIter->nSeg+iChanged)/2; 1; i=i/2){
Fts5CResult *pRes = &pIter->aFirst[i];
assert( pNew->pLeaf );
assert( pRes->bTermEq==0 || pOther->pLeaf );
if( pRes->bTermEq ){
if( pNew->iRowid==pOther->iRowid ){
return 1;
}else if( (pOther->iRowid>pNew->iRowid)==pIter->bRev ){
pIter->iSwitchRowid = pOther->iRowid;
pNew = pOther;
}else if( (pOther->iRowid>pIter->iSwitchRowid)==pIter->bRev ){
pIter->iSwitchRowid = pOther->iRowid;
}
}
pRes->iFirst = (u16)(pNew - pIter->aSeg);
if( i==1 ) break;
pOther = &pIter->aSeg[ pIter->aFirst[i ^ 0x0001].iFirst ];
}
}
*ppFirst = pNew;
return 0;
}
/*
** Set the pIter->bEof variable based on the state of the sub-iterators.
*/
static void fts5MultiIterSetEof(Fts5Iter *pIter){
Fts5SegIter *pSeg = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
pIter->base.bEof = pSeg->pLeaf==0;
pIter->iSwitchRowid = pSeg->iRowid;
}
/*
** The argument to this macro must be an Fts5Data structure containing a
** tombstone hash page. This macro returns the key-size of the hash-page.
*/
#define TOMBSTONE_KEYSIZE(pPg) (pPg->p[0]==4 ? 4 : 8)
#define TOMBSTONE_NSLOT(pPg) \
((pPg->nn > 16) ? ((pPg->nn-8) / TOMBSTONE_KEYSIZE(pPg)) : 1)
/*
** Query a single tombstone hash table for rowid iRowid. Return true if
** it is found or false otherwise. The tombstone hash table is one of
** nHashTable tables.
*/
static int fts5IndexTombstoneQuery(
Fts5Data *pHash, /* Hash table page to query */
int nHashTable, /* Number of pages attached to segment */
u64 iRowid /* Rowid to query hash for */
){
const int szKey = TOMBSTONE_KEYSIZE(pHash);
const int nSlot = TOMBSTONE_NSLOT(pHash);
int iSlot = (iRowid / nHashTable) % nSlot;
int nCollide = nSlot;
if( iRowid==0 ){
return pHash->p[1];
}else if( szKey==4 ){
u32 *aSlot = (u32*)&pHash->p[8];
while( aSlot[iSlot] ){
if( fts5GetU32((u8*)&aSlot[iSlot])==iRowid ) return 1;
if( nCollide--==0 ) break;
iSlot = (iSlot+1)%nSlot;
}
}else{
u64 *aSlot = (u64*)&pHash->p[8];
while( aSlot[iSlot] ){
if( fts5GetU64((u8*)&aSlot[iSlot])==iRowid ) return 1;
if( nCollide--==0 ) break;
iSlot = (iSlot+1)%nSlot;
}
}
return 0;
}
/*
** Return true if the iterator passed as the only argument points
** to an segment entry for which there is a tombstone. Return false
** if there is no tombstone or if the iterator is already at EOF.
*/
static int fts5MultiIterIsDeleted(Fts5Iter *pIter){
int iFirst = pIter->aFirst[1].iFirst;
Fts5SegIter *pSeg = &pIter->aSeg[iFirst];
Fts5TombstoneArray *pArray = pSeg->pTombArray;
if( pSeg->pLeaf && pArray ){
/* Figure out which page the rowid might be present on. */
int iPg = ((u64)pSeg->iRowid) % pArray->nTombstone;
assert( iPg>=0 );
/* If tombstone hash page iPg has not yet been loaded from the
** database, load it now. */
if( pArray->apTombstone[iPg]==0 ){
pArray->apTombstone[iPg] = fts5DataRead(pIter->pIndex,
FTS5_TOMBSTONE_ROWID(pSeg->pSeg->iSegid, iPg)
);
if( pArray->apTombstone[iPg]==0 ) return 0;
}
return fts5IndexTombstoneQuery(
pArray->apTombstone[iPg],
pArray->nTombstone,
pSeg->iRowid
);
}
return 0;
}
/*
** Move the iterator to the next entry.
**
** If an error occurs, an error code is left in Fts5Index.rc. It is not
** considered an error if the iterator reaches EOF, or if it is already at
** EOF when this function is called.
*/
static void fts5MultiIterNext(
Fts5Index *p,
Fts5Iter *pIter,
int bFrom, /* True if argument iFrom is valid */
i64 iFrom /* Advance at least as far as this */
){
int bUseFrom = bFrom;
assert( pIter->base.bEof==0 );
while( p->rc==SQLITE_OK ){
int iFirst = pIter->aFirst[1].iFirst;
int bNewTerm = 0;
Fts5SegIter *pSeg = &pIter->aSeg[iFirst];
assert( p->rc==SQLITE_OK );
if( bUseFrom && pSeg->pDlidx ){
fts5SegIterNextFrom(p, pSeg, iFrom);
}else{
pSeg->xNext(p, pSeg, &bNewTerm);
}
if( pSeg->pLeaf==0 || bNewTerm
|| fts5MultiIterAdvanceRowid(pIter, iFirst, &pSeg)
){
fts5MultiIterAdvanced(p, pIter, iFirst, 1);
fts5MultiIterSetEof(pIter);
pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst];
if( pSeg->pLeaf==0 ) return;
}
fts5AssertMultiIterSetup(p, pIter);
assert( pSeg==&pIter->aSeg[pIter->aFirst[1].iFirst] && pSeg->pLeaf );
if( (pIter->bSkipEmpty==0 || pSeg->nPos)
&& 0==fts5MultiIterIsDeleted(pIter)
){
pIter->xSetOutputs(pIter, pSeg);
return;
}
bUseFrom = 0;
}
}
static void fts5MultiIterNext2(
Fts5Index *p,
Fts5Iter *pIter,
int *pbNewTerm /* OUT: True if *might* be new term */
){
assert( pIter->bSkipEmpty );
if( p->rc==SQLITE_OK ){
*pbNewTerm = 0;
do{
int iFirst = pIter->aFirst[1].iFirst;
Fts5SegIter *pSeg = &pIter->aSeg[iFirst];
int bNewTerm = 0;
assert( p->rc==SQLITE_OK );
pSeg->xNext(p, pSeg, &bNewTerm);
if( pSeg->pLeaf==0 || bNewTerm
|| fts5MultiIterAdvanceRowid(pIter, iFirst, &pSeg)
){
fts5MultiIterAdvanced(p, pIter, iFirst, 1);
fts5MultiIterSetEof(pIter);
*pbNewTerm = 1;
}
fts5AssertMultiIterSetup(p, pIter);
}while( (fts5MultiIterIsEmpty(p, pIter) || fts5MultiIterIsDeleted(pIter))
&& (p->rc==SQLITE_OK)
);
}
}
static void fts5IterSetOutputs_Noop(Fts5Iter *pUnused1, Fts5SegIter *pUnused2){
UNUSED_PARAM2(pUnused1, pUnused2);
}
static Fts5Iter *fts5MultiIterAlloc(
Fts5Index *p, /* FTS5 backend to iterate within */
int nSeg
){
Fts5Iter *pNew;
i64 nSlot; /* Power of two >= nSeg */
for(nSlot=2; nSlotaSeg[] */
sizeof(Fts5CResult) * nSlot /* pNew->aFirst[] */
);
if( pNew ){
pNew->nSeg = nSlot;
pNew->aFirst = (Fts5CResult*)&pNew->aSeg[nSlot];
pNew->pIndex = p;
pNew->xSetOutputs = fts5IterSetOutputs_Noop;
}
return pNew;
}
static void fts5PoslistCallback(
Fts5Index *pUnused,
void *pContext,
const u8 *pChunk, int nChunk
){
UNUSED_PARAM(pUnused);
assert_nc( nChunk>=0 );
if( nChunk>0 ){
fts5BufferSafeAppendBlob((Fts5Buffer*)pContext, pChunk, nChunk);
}
}
typedef struct PoslistCallbackCtx PoslistCallbackCtx;
struct PoslistCallbackCtx {
Fts5Buffer *pBuf; /* Append to this buffer */
Fts5Colset *pColset; /* Restrict matches to this column */
int eState; /* See above */
};
typedef struct PoslistOffsetsCtx PoslistOffsetsCtx;
struct PoslistOffsetsCtx {
Fts5Buffer *pBuf; /* Append to this buffer */
Fts5Colset *pColset; /* Restrict matches to this column */
int iRead;
int iWrite;
};
/*
** TODO: Make this more efficient!
*/
static int fts5IndexColsetTest(Fts5Colset *pColset, int iCol){
int i;
for(i=0; inCol; i++){
if( pColset->aiCol[i]==iCol ) return 1;
}
return 0;
}
static void fts5PoslistOffsetsCallback(
Fts5Index *pUnused,
void *pContext,
const u8 *pChunk, int nChunk
){
PoslistOffsetsCtx *pCtx = (PoslistOffsetsCtx*)pContext;
UNUSED_PARAM(pUnused);
assert_nc( nChunk>=0 );
if( nChunk>0 ){
int i = 0;
while( iiRead - 2;
pCtx->iRead = iVal;
if( fts5IndexColsetTest(pCtx->pColset, iVal) ){
fts5BufferSafeAppendVarint(pCtx->pBuf, iVal + 2 - pCtx->iWrite);
pCtx->iWrite = iVal;
}
}
}
}
static void fts5PoslistFilterCallback(
Fts5Index *pUnused,
void *pContext,
const u8 *pChunk, int nChunk
){
PoslistCallbackCtx *pCtx = (PoslistCallbackCtx*)pContext;
UNUSED_PARAM(pUnused);
assert_nc( nChunk>=0 );
if( nChunk>0 ){
/* Search through to find the first varint with value 1. This is the
** start of the next columns hits. */
int i = 0;
int iStart = 0;
if( pCtx->eState==2 ){
int iCol;
fts5FastGetVarint32(pChunk, i, iCol);
if( fts5IndexColsetTest(pCtx->pColset, iCol) ){
pCtx->eState = 1;
fts5BufferSafeAppendVarint(pCtx->pBuf, 1);
}else{
pCtx->eState = 0;
}
}
do {
while( ieState ){
fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart);
}
if( i=nChunk ){
pCtx->eState = 2;
}else{
fts5FastGetVarint32(pChunk, i, iCol);
pCtx->eState = fts5IndexColsetTest(pCtx->pColset, iCol);
if( pCtx->eState ){
fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart);
iStart = i;
}
}
}
}while( inPos; /* Number of bytes still to come */
Fts5Data *pData = 0;
u8 *pChunk = &pSeg->pLeaf->p[pSeg->iLeafOffset];
int nChunk = MIN(nRem, pSeg->pLeaf->szLeaf - pSeg->iLeafOffset);
int pgno = pSeg->iLeafPgno;
int pgnoSave = 0;
/* This function does not work with detail=none databases. */
assert( p->pConfig->eDetail!=FTS5_DETAIL_NONE );
if( (pSeg->flags & FTS5_SEGITER_REVERSE)==0 ){
pgnoSave = pgno+1;
}
while( 1 ){
xChunk(p, pCtx, pChunk, nChunk);
nRem -= nChunk;
fts5DataRelease(pData);
if( nRem<=0 ){
break;
}else if( pSeg->pSeg==0 ){
p->rc = FTS5_CORRUPT;
return;
}else{
pgno++;
pData = fts5LeafRead(p, FTS5_SEGMENT_ROWID(pSeg->pSeg->iSegid, pgno));
if( pData==0 ) break;
pChunk = &pData->p[4];
nChunk = MIN(nRem, pData->szLeaf - 4);
if( pgno==pgnoSave ){
assert( pSeg->pNextLeaf==0 );
pSeg->pNextLeaf = pData;
pData = 0;
}
}
}
}
/*
** Iterator pIter currently points to a valid entry (not EOF). This
** function appends the position list data for the current entry to
** buffer pBuf. It does not make a copy of the position-list size
** field.
*/
static void fts5SegiterPoslist(
Fts5Index *p,
Fts5SegIter *pSeg,
Fts5Colset *pColset,
Fts5Buffer *pBuf
){
assert( pBuf!=0 );
assert( pSeg!=0 );
if( 0==fts5BufferGrow(&p->rc, pBuf, pSeg->nPos+FTS5_DATA_ZERO_PADDING) ){
assert( pBuf->p!=0 );
assert( pBuf->nSpace >= pBuf->n+pSeg->nPos+FTS5_DATA_ZERO_PADDING );
memset(&pBuf->p[pBuf->n+pSeg->nPos], 0, FTS5_DATA_ZERO_PADDING);
if( pColset==0 ){
fts5ChunkIterate(p, pSeg, (void*)pBuf, fts5PoslistCallback);
}else{
if( p->pConfig->eDetail==FTS5_DETAIL_FULL ){
PoslistCallbackCtx sCtx;
sCtx.pBuf = pBuf;
sCtx.pColset = pColset;
sCtx.eState = fts5IndexColsetTest(pColset, 0);
assert( sCtx.eState==0 || sCtx.eState==1 );
fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistFilterCallback);
}else{
PoslistOffsetsCtx sCtx;
memset(&sCtx, 0, sizeof(sCtx));
sCtx.pBuf = pBuf;
sCtx.pColset = pColset;
fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistOffsetsCallback);
}
}
}
}
/*
** Parameter pPos points to a buffer containing a position list, size nPos.
** This function filters it according to pColset (which must be non-NULL)
** and sets pIter->base.pData/nData to point to the new position list.
** If memory is required for the new position list, use buffer pIter->poslist.
** Or, if the new position list is a contiguous subset of the input, set
** pIter->base.pData/nData to point directly to it.
**
** This function is a no-op if *pRc is other than SQLITE_OK when it is
** called. If an OOM error is encountered, *pRc is set to SQLITE_NOMEM
** before returning.
*/
static void fts5IndexExtractColset(
int *pRc,
Fts5Colset *pColset, /* Colset to filter on */
const u8 *pPos, int nPos, /* Position list */
Fts5Iter *pIter
){
if( *pRc==SQLITE_OK ){
const u8 *p = pPos;
const u8 *aCopy = p;
const u8 *pEnd = &p[nPos]; /* One byte past end of position list */
int i = 0;
int iCurrent = 0;
if( pColset->nCol>1 && sqlite3Fts5BufferSize(pRc, &pIter->poslist, nPos) ){
return;
}
while( 1 ){
while( pColset->aiCol[i]nCol ){
pIter->base.pData = pIter->poslist.p;
pIter->base.nData = pIter->poslist.n;
return;
}
}
/* Advance pointer p until it points to pEnd or an 0x01 byte that is
** not part of a varint */
while( paiCol[i]==iCurrent ){
if( pColset->nCol==1 ){
pIter->base.pData = aCopy;
pIter->base.nData = p-aCopy;
return;
}
fts5BufferSafeAppendBlob(&pIter->poslist, aCopy, p-aCopy);
}
if( p>=pEnd ){
pIter->base.pData = pIter->poslist.p;
pIter->base.nData = pIter->poslist.n;
return;
}
aCopy = p++;
iCurrent = *p++;
if( iCurrent & 0x80 ){
p--;
p += fts5GetVarint32(p, iCurrent);
}
}
}
}
/*
** xSetOutputs callback used by detail=none tables.
*/
static void fts5IterSetOutputs_None(Fts5Iter *pIter, Fts5SegIter *pSeg){
assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_NONE );
pIter->base.iRowid = pSeg->iRowid;
pIter->base.nData = pSeg->nPos;
}
/*
** xSetOutputs callback used by detail=full and detail=col tables when no
** column filters are specified.
*/
static void fts5IterSetOutputs_Nocolset(Fts5Iter *pIter, Fts5SegIter *pSeg){
pIter->base.iRowid = pSeg->iRowid;
pIter->base.nData = pSeg->nPos;
assert( pIter->pIndex->pConfig->eDetail!=FTS5_DETAIL_NONE );
assert( pIter->pColset==0 );
if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){
/* All data is stored on the current page. Populate the output
** variables to point into the body of the page object. */
pIter->base.pData = &pSeg->pLeaf->p[pSeg->iLeafOffset];
}else{
/* The data is distributed over two or more pages. Copy it into the
** Fts5Iter.poslist buffer and then set the output pointer to point
** to this buffer. */
fts5BufferZero(&pIter->poslist);
fts5SegiterPoslist(pIter->pIndex, pSeg, 0, &pIter->poslist);
pIter->base.pData = pIter->poslist.p;
}
}
/*
** xSetOutputs callback used when the Fts5Colset object has nCol==0 (match
** against no columns at all).
*/
static void fts5IterSetOutputs_ZeroColset(Fts5Iter *pIter, Fts5SegIter *pSeg){
UNUSED_PARAM(pSeg);
pIter->base.nData = 0;
}
/*
** xSetOutputs callback used by detail=col when there is a column filter
** and there are 100 or more columns. Also called as a fallback from
** fts5IterSetOutputs_Col100 if the column-list spans more than one page.
*/
static void fts5IterSetOutputs_Col(Fts5Iter *pIter, Fts5SegIter *pSeg){
fts5BufferZero(&pIter->poslist);
fts5SegiterPoslist(pIter->pIndex, pSeg, pIter->pColset, &pIter->poslist);
pIter->base.iRowid = pSeg->iRowid;
pIter->base.pData = pIter->poslist.p;
pIter->base.nData = pIter->poslist.n;
}
/*
** xSetOutputs callback used when:
**
** * detail=col,
** * there is a column filter, and
** * the table contains 100 or fewer columns.
**
** The last point is to ensure all column numbers are stored as
** single-byte varints.
*/
static void fts5IterSetOutputs_Col100(Fts5Iter *pIter, Fts5SegIter *pSeg){
assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_COLUMNS );
assert( pIter->pColset );
if( pSeg->iLeafOffset+pSeg->nPos>pSeg->pLeaf->szLeaf ){
fts5IterSetOutputs_Col(pIter, pSeg);
}else{
u8 *a = (u8*)&pSeg->pLeaf->p[pSeg->iLeafOffset];
u8 *pEnd = (u8*)&a[pSeg->nPos];
int iPrev = 0;
int *aiCol = pIter->pColset->aiCol;
int *aiColEnd = &aiCol[pIter->pColset->nCol];
u8 *aOut = pIter->poslist.p;
int iPrevOut = 0;
pIter->base.iRowid = pSeg->iRowid;
while( abase.pData = pIter->poslist.p;
pIter->base.nData = aOut - pIter->poslist.p;
}
}
/*
** xSetOutputs callback used by detail=full when there is a column filter.
*/
static void fts5IterSetOutputs_Full(Fts5Iter *pIter, Fts5SegIter *pSeg){
Fts5Colset *pColset = pIter->pColset;
pIter->base.iRowid = pSeg->iRowid;
assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_FULL );
assert( pColset );
if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){
/* All data is stored on the current page. Populate the output
** variables to point into the body of the page object. */
const u8 *a = &pSeg->pLeaf->p[pSeg->iLeafOffset];
int *pRc = &pIter->pIndex->rc;
fts5BufferZero(&pIter->poslist);
fts5IndexExtractColset(pRc, pColset, a, pSeg->nPos, pIter);
}else{
/* The data is distributed over two or more pages. Copy it into the
** Fts5Iter.poslist buffer and then set the output pointer to point
** to this buffer. */
fts5BufferZero(&pIter->poslist);
fts5SegiterPoslist(pIter->pIndex, pSeg, pColset, &pIter->poslist);
pIter->base.pData = pIter->poslist.p;
pIter->base.nData = pIter->poslist.n;
}
}
static void fts5IterSetOutputCb(int *pRc, Fts5Iter *pIter){
assert( pIter!=0 || (*pRc)!=SQLITE_OK );
if( *pRc==SQLITE_OK ){
Fts5Config *pConfig = pIter->pIndex->pConfig;
if( pConfig->eDetail==FTS5_DETAIL_NONE ){
pIter->xSetOutputs = fts5IterSetOutputs_None;
}
else if( pIter->pColset==0 ){
pIter->xSetOutputs = fts5IterSetOutputs_Nocolset;
}
else if( pIter->pColset->nCol==0 ){
pIter->xSetOutputs = fts5IterSetOutputs_ZeroColset;
}
else if( pConfig->eDetail==FTS5_DETAIL_FULL ){
pIter->xSetOutputs = fts5IterSetOutputs_Full;
}
else{
assert( pConfig->eDetail==FTS5_DETAIL_COLUMNS );
if( pConfig->nCol<=100 ){
pIter->xSetOutputs = fts5IterSetOutputs_Col100;
sqlite3Fts5BufferSize(pRc, &pIter->poslist, pConfig->nCol);
}else{
pIter->xSetOutputs = fts5IterSetOutputs_Col;
}
}
}
}
/*
** All the component segment-iterators of pIter have been set up. This
** functions finishes setup for iterator pIter itself.
*/
static void fts5MultiIterFinishSetup(Fts5Index *p, Fts5Iter *pIter){
int iIter;
for(iIter=pIter->nSeg-1; iIter>0; iIter--){
int iEq;
if( (iEq = fts5MultiIterDoCompare(pIter, iIter)) ){
Fts5SegIter *pSeg = &pIter->aSeg[iEq];
if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0);
fts5MultiIterAdvanced(p, pIter, iEq, iIter);
}
}
fts5MultiIterSetEof(pIter);
fts5AssertMultiIterSetup(p, pIter);
if( (pIter->bSkipEmpty && fts5MultiIterIsEmpty(p, pIter))
|| fts5MultiIterIsDeleted(pIter)
){
fts5MultiIterNext(p, pIter, 0, 0);
}else if( pIter->base.bEof==0 ){
Fts5SegIter *pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst];
pIter->xSetOutputs(pIter, pSeg);
}
}
/*
** Allocate a new Fts5Iter object.
**
** The new object will be used to iterate through data in structure pStruct.
** If iLevel is -ve, then all data in all segments is merged. Or, if iLevel
** is zero or greater, data from the first nSegment segments on level iLevel
** is merged.
**
** The iterator initially points to the first term/rowid entry in the
** iterated data.
*/
static void fts5MultiIterNew(
Fts5Index *p, /* FTS5 backend to iterate within */
Fts5Structure *pStruct, /* Structure of specific index */
int flags, /* FTS5INDEX_QUERY_XXX flags */
Fts5Colset *pColset, /* Colset to filter on (or NULL) */
const u8 *pTerm, int nTerm, /* Term to seek to (or NULL/0) */
int iLevel, /* Level to iterate (-1 for all) */
int nSegment, /* Number of segments to merge (iLevel>=0) */
Fts5Iter **ppOut /* New object */
){
int nSeg = 0; /* Number of segment-iters in use */
int iIter = 0; /* */
int iSeg; /* Used to iterate through segments */
Fts5StructureLevel *pLvl;
Fts5Iter *pNew;
assert( (pTerm==0 && nTerm==0) || iLevel<0 );
/* Allocate space for the new multi-seg-iterator. */
if( p->rc==SQLITE_OK ){
if( iLevel<0 ){
assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) );
nSeg = pStruct->nSegment;
nSeg += (p->pHash && 0==(flags & FTS5INDEX_QUERY_SKIPHASH));
}else{
nSeg = MIN(pStruct->aLevel[iLevel].nSeg, nSegment);
}
}
*ppOut = pNew = fts5MultiIterAlloc(p, nSeg);
if( pNew==0 ){
assert( p->rc!=SQLITE_OK );
goto fts5MultiIterNew_post_check;
}
pNew->bRev = (0!=(flags & FTS5INDEX_QUERY_DESC));
pNew->bSkipEmpty = (0!=(flags & FTS5INDEX_QUERY_SKIPEMPTY));
pNew->pColset = pColset;
if( (flags & FTS5INDEX_QUERY_NOOUTPUT)==0 ){
fts5IterSetOutputCb(&p->rc, pNew);
}
/* Initialize each of the component segment iterators. */
if( p->rc==SQLITE_OK ){
if( iLevel<0 ){
Fts5StructureLevel *pEnd = &pStruct->aLevel[pStruct->nLevel];
if( p->pHash && 0==(flags & FTS5INDEX_QUERY_SKIPHASH) ){
/* Add a segment iterator for the current contents of the hash table. */
Fts5SegIter *pIter = &pNew->aSeg[iIter++];
fts5SegIterHashInit(p, pTerm, nTerm, flags, pIter);
}
for(pLvl=&pStruct->aLevel[0]; pLvlnSeg-1; iSeg>=0; iSeg--){
Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg];
Fts5SegIter *pIter = &pNew->aSeg[iIter++];
if( pTerm==0 ){
fts5SegIterInit(p, pSeg, pIter);
}else{
fts5SegIterSeekInit(p, pTerm, nTerm, flags, pSeg, pIter);
}
}
}
}else{
pLvl = &pStruct->aLevel[iLevel];
for(iSeg=nSeg-1; iSeg>=0; iSeg--){
fts5SegIterInit(p, &pLvl->aSeg[iSeg], &pNew->aSeg[iIter++]);
}
}
assert( iIter==nSeg );
}
/* If the above was successful, each component iterator now points
** to the first entry in its segment. In this case initialize the
** aFirst[] array. Or, if an error has occurred, free the iterator
** object and set the output variable to NULL. */
if( p->rc==SQLITE_OK ){
fts5MultiIterFinishSetup(p, pNew);
}else{
fts5MultiIterFree(pNew);
*ppOut = 0;
}
fts5MultiIterNew_post_check:
assert( (*ppOut)!=0 || p->rc!=SQLITE_OK );
return;
}
/*
** Create an Fts5Iter that iterates through the doclist provided
** as the second argument.
*/
static void fts5MultiIterNew2(
Fts5Index *p, /* FTS5 backend to iterate within */
Fts5Data *pData, /* Doclist to iterate through */
int bDesc, /* True for descending rowid order */
Fts5Iter **ppOut /* New object */
){
Fts5Iter *pNew;
pNew = fts5MultiIterAlloc(p, 2);
if( pNew ){
Fts5SegIter *pIter = &pNew->aSeg[1];
pIter->flags = FTS5_SEGITER_ONETERM;
if( pData->szLeaf>0 ){
pIter->pLeaf = pData;
pIter->iLeafOffset = fts5GetVarint(pData->p, (u64*)&pIter->iRowid);
pIter->iEndofDoclist = pData->nn;
pNew->aFirst[1].iFirst = 1;
if( bDesc ){
pNew->bRev = 1;
pIter->flags |= FTS5_SEGITER_REVERSE;
fts5SegIterReverseInitPage(p, pIter);
}else{
fts5SegIterLoadNPos(p, pIter);
}
pData = 0;
}else{
pNew->base.bEof = 1;
}
fts5SegIterSetNext(p, pIter);
*ppOut = pNew;
}
fts5DataRelease(pData);
}
/*
** Return true if the iterator is at EOF or if an error has occurred.
** False otherwise.
*/
static int fts5MultiIterEof(Fts5Index *p, Fts5Iter *pIter){
assert( pIter!=0 || p->rc!=SQLITE_OK );
assert( p->rc!=SQLITE_OK
|| (pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf==0)==pIter->base.bEof
);
return (p->rc || pIter->base.bEof);
}
/*
** Return the rowid of the entry that the iterator currently points
** to. If the iterator points to EOF when this function is called the
** results are undefined.
*/
static i64 fts5MultiIterRowid(Fts5Iter *pIter){
assert( pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf );
return pIter->aSeg[ pIter->aFirst[1].iFirst ].iRowid;
}
/*
** Move the iterator to the next entry at or following iMatch.
*/
static void fts5MultiIterNextFrom(
Fts5Index *p,
Fts5Iter *pIter,
i64 iMatch
){
while( 1 ){
i64 iRowid;
fts5MultiIterNext(p, pIter, 1, iMatch);
if( fts5MultiIterEof(p, pIter) ) break;
iRowid = fts5MultiIterRowid(pIter);
if( pIter->bRev==0 && iRowid>=iMatch ) break;
if( pIter->bRev!=0 && iRowid<=iMatch ) break;
}
}
/*
** Return a pointer to a buffer containing the term associated with the
** entry that the iterator currently points to.
*/
static const u8 *fts5MultiIterTerm(Fts5Iter *pIter, int *pn){
Fts5SegIter *p = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
*pn = p->term.n;
return p->term.p;
}
/*
** Allocate a new segment-id for the structure pStruct. The new segment
** id must be between 1 and 65335 inclusive, and must not be used by
** any currently existing segment. If a free segment id cannot be found,
** SQLITE_FULL is returned.
**
** If an error has already occurred, this function is a no-op. 0 is
** returned in this case.
*/
static int fts5AllocateSegid(Fts5Index *p, Fts5Structure *pStruct){
int iSegid = 0;
if( p->rc==SQLITE_OK ){
if( pStruct->nSegment>=FTS5_MAX_SEGMENT ){
p->rc = SQLITE_FULL;
}else{
/* FTS5_MAX_SEGMENT is currently defined as 2000. So the following
** array is 63 elements, or 252 bytes, in size. */
u32 aUsed[(FTS5_MAX_SEGMENT+31) / 32];
int iLvl, iSeg;
int i;
u32 mask;
memset(aUsed, 0, sizeof(aUsed));
for(iLvl=0; iLvlnLevel; iLvl++){
for(iSeg=0; iSegaLevel[iLvl].nSeg; iSeg++){
int iId = pStruct->aLevel[iLvl].aSeg[iSeg].iSegid;
if( iId<=FTS5_MAX_SEGMENT && iId>0 ){
aUsed[(iId-1) / 32] |= (u32)1 << ((iId-1) % 32);
}
}
}
for(i=0; aUsed[i]==0xFFFFFFFF; i++);
mask = aUsed[i];
for(iSegid=0; mask & ((u32)1 << iSegid); iSegid++);
iSegid += 1 + i*32;
#ifdef SQLITE_DEBUG
for(iLvl=0; iLvlnLevel; iLvl++){
for(iSeg=0; iSegaLevel[iLvl].nSeg; iSeg++){
assert_nc( iSegid!=pStruct->aLevel[iLvl].aSeg[iSeg].iSegid );
}
}
assert_nc( iSegid>0 && iSegid<=FTS5_MAX_SEGMENT );
{
sqlite3_stmt *pIdxSelect = fts5IdxSelectStmt(p);
if( p->rc==SQLITE_OK ){
u8 aBlob[2] = {0xff, 0xff};
sqlite3_bind_int(pIdxSelect, 1, iSegid);
sqlite3_bind_blob(pIdxSelect, 2, aBlob, 2, SQLITE_STATIC);
assert_nc( sqlite3_step(pIdxSelect)!=SQLITE_ROW );
p->rc = sqlite3_reset(pIdxSelect);
sqlite3_bind_null(pIdxSelect, 2);
}
}
#endif
}
}
return iSegid;
}
/*
** Discard all data currently cached in the hash-tables.
*/
static void fts5IndexDiscardData(Fts5Index *p){
assert( p->pHash || p->nPendingData==0 );
if( p->pHash ){
sqlite3Fts5HashClear(p->pHash);
p->nPendingData = 0;
p->nPendingRow = 0;
p->flushRc = SQLITE_OK;
}
p->nContentlessDelete = 0;
}
/*
** Return the size of the prefix, in bytes, that buffer
** (pNew/) shares with buffer (pOld/nOld).
**
** Buffer (pNew/) is guaranteed to be greater
** than buffer (pOld/nOld).
*/
static int fts5PrefixCompress(int nOld, const u8 *pOld, const u8 *pNew){
int i;
for(i=0; inDlidx>0 && pWriter->aDlidx[0].buf.n>0) );
for(i=0; inDlidx; i++){
Fts5DlidxWriter *pDlidx = &pWriter->aDlidx[i];
if( pDlidx->buf.n==0 ) break;
if( bFlush ){
assert( pDlidx->pgno!=0 );
fts5DataWrite(p,
FTS5_DLIDX_ROWID(pWriter->iSegid, i, pDlidx->pgno),
pDlidx->buf.p, pDlidx->buf.n
);
}
sqlite3Fts5BufferZero(&pDlidx->buf);
pDlidx->bPrevValid = 0;
}
}
/*
** Grow the pWriter->aDlidx[] array to at least nLvl elements in size.
** Any new array elements are zeroed before returning.
*/
static int fts5WriteDlidxGrow(
Fts5Index *p,
Fts5SegWriter *pWriter,
int nLvl
){
if( p->rc==SQLITE_OK && nLvl>=pWriter->nDlidx ){
Fts5DlidxWriter *aDlidx = (Fts5DlidxWriter*)sqlite3_realloc64(
pWriter->aDlidx, sizeof(Fts5DlidxWriter) * nLvl
);
if( aDlidx==0 ){
p->rc = SQLITE_NOMEM;
}else{
size_t nByte = sizeof(Fts5DlidxWriter) * (nLvl - pWriter->nDlidx);
memset(&aDlidx[pWriter->nDlidx], 0, nByte);
pWriter->aDlidx = aDlidx;
pWriter->nDlidx = nLvl;
}
}
return p->rc;
}
/*
** If the current doclist-index accumulating in pWriter->aDlidx[] is large
** enough, flush it to disk and return 1. Otherwise discard it and return
** zero.
*/
static int fts5WriteFlushDlidx(Fts5Index *p, Fts5SegWriter *pWriter){
int bFlag = 0;
/* If there were FTS5_MIN_DLIDX_SIZE or more empty leaf pages written
** to the database, also write the doclist-index to disk. */
if( pWriter->aDlidx[0].buf.n>0 && pWriter->nEmpty>=FTS5_MIN_DLIDX_SIZE ){
bFlag = 1;
}
fts5WriteDlidxClear(p, pWriter, bFlag);
pWriter->nEmpty = 0;
return bFlag;
}
/*
** This function is called whenever processing of the doclist for the
** last term on leaf page (pWriter->iBtPage) is completed.
**
** The doclist-index for that term is currently stored in-memory within the
** Fts5SegWriter.aDlidx[] array. If it is large enough, this function
** writes it out to disk. Or, if it is too small to bother with, discards
** it.
**
** Fts5SegWriter.btterm currently contains the first term on page iBtPage.
*/
static void fts5WriteFlushBtree(Fts5Index *p, Fts5SegWriter *pWriter){
int bFlag;
assert( pWriter->iBtPage || pWriter->nEmpty==0 );
if( pWriter->iBtPage==0 ) return;
bFlag = fts5WriteFlushDlidx(p, pWriter);
if( p->rc==SQLITE_OK ){
const char *z = (pWriter->btterm.n>0?(const char*)pWriter->btterm.p:"");
/* The following was already done in fts5WriteInit(): */
/* sqlite3_bind_int(p->pIdxWriter, 1, pWriter->iSegid); */
sqlite3_bind_blob(p->pIdxWriter, 2, z, pWriter->btterm.n, SQLITE_STATIC);
sqlite3_bind_int64(p->pIdxWriter, 3, bFlag + ((i64)pWriter->iBtPage<<1));
sqlite3_step(p->pIdxWriter);
p->rc = sqlite3_reset(p->pIdxWriter);
sqlite3_bind_null(p->pIdxWriter, 2);
}
pWriter->iBtPage = 0;
}
/*
** This is called once for each leaf page except the first that contains
** at least one term. Argument (nTerm/pTerm) is the split-key - a term that
** is larger than all terms written to earlier leaves, and equal to or
** smaller than the first term on the new leaf.
**
** If an error occurs, an error code is left in Fts5Index.rc. If an error
** has already occurred when this function is called, it is a no-op.
*/
static void fts5WriteBtreeTerm(
Fts5Index *p, /* FTS5 backend object */
Fts5SegWriter *pWriter, /* Writer object */
int nTerm, const u8 *pTerm /* First term on new page */
){
fts5WriteFlushBtree(p, pWriter);
if( p->rc==SQLITE_OK ){
fts5BufferSet(&p->rc, &pWriter->btterm, nTerm, pTerm);
pWriter->iBtPage = pWriter->writer.pgno;
}
}
/*
** This function is called when flushing a leaf page that contains no
** terms at all to disk.
*/
static void fts5WriteBtreeNoTerm(
Fts5Index *p, /* FTS5 backend object */
Fts5SegWriter *pWriter /* Writer object */
){
/* If there were no rowids on the leaf page either and the doclist-index
** has already been started, append an 0x00 byte to it. */
if( pWriter->bFirstRowidInPage && pWriter->aDlidx[0].buf.n>0 ){
Fts5DlidxWriter *pDlidx = &pWriter->aDlidx[0];
assert( pDlidx->bPrevValid );
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, 0);
}
/* Increment the "number of sequential leaves without a term" counter. */
pWriter->nEmpty++;
}
static i64 fts5DlidxExtractFirstRowid(Fts5Buffer *pBuf){
i64 iRowid;
int iOff;
iOff = 1 + fts5GetVarint(&pBuf->p[1], (u64*)&iRowid);
fts5GetVarint(&pBuf->p[iOff], (u64*)&iRowid);
return iRowid;
}
/*
** Rowid iRowid has just been appended to the current leaf page. It is the
** first on the page. This function appends an appropriate entry to the current
** doclist-index.
*/
static void fts5WriteDlidxAppend(
Fts5Index *p,
Fts5SegWriter *pWriter,
i64 iRowid
){
int i;
int bDone = 0;
for(i=0; p->rc==SQLITE_OK && bDone==0; i++){
i64 iVal;
Fts5DlidxWriter *pDlidx = &pWriter->aDlidx[i];
if( pDlidx->buf.n>=p->pConfig->pgsz ){
/* The current doclist-index page is full. Write it to disk and push
** a copy of iRowid (which will become the first rowid on the next
** doclist-index leaf page) up into the next level of the b-tree
** hierarchy. If the node being flushed is currently the root node,
** also push its first rowid upwards. */
pDlidx->buf.p[0] = 0x01; /* Not the root node */
fts5DataWrite(p,
FTS5_DLIDX_ROWID(pWriter->iSegid, i, pDlidx->pgno),
pDlidx->buf.p, pDlidx->buf.n
);
fts5WriteDlidxGrow(p, pWriter, i+2);
pDlidx = &pWriter->aDlidx[i];
if( p->rc==SQLITE_OK && pDlidx[1].buf.n==0 ){
i64 iFirst = fts5DlidxExtractFirstRowid(&pDlidx->buf);
/* This was the root node. Push its first rowid up to the new root. */
pDlidx[1].pgno = pDlidx->pgno;
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx[1].buf, 0);
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx[1].buf, pDlidx->pgno);
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx[1].buf, iFirst);
pDlidx[1].bPrevValid = 1;
pDlidx[1].iPrev = iFirst;
}
sqlite3Fts5BufferZero(&pDlidx->buf);
pDlidx->bPrevValid = 0;
pDlidx->pgno++;
}else{
bDone = 1;
}
if( pDlidx->bPrevValid ){
iVal = (u64)iRowid - (u64)pDlidx->iPrev;
}else{
i64 iPgno = (i==0 ? pWriter->writer.pgno : pDlidx[-1].pgno);
assert( pDlidx->buf.n==0 );
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, !bDone);
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, iPgno);
iVal = iRowid;
}
sqlite3Fts5BufferAppendVarint(&p->rc, &pDlidx->buf, iVal);
pDlidx->bPrevValid = 1;
pDlidx->iPrev = iRowid;
}
}
static void fts5WriteFlushLeaf(Fts5Index *p, Fts5SegWriter *pWriter){
static const u8 zero[] = { 0x00, 0x00, 0x00, 0x00 };
Fts5PageWriter *pPage = &pWriter->writer;
i64 iRowid;
assert( (pPage->pgidx.n==0)==(pWriter->bFirstTermInPage) );
/* Set the szLeaf header field. */
assert( 0==fts5GetU16(&pPage->buf.p[2]) );
fts5PutU16(&pPage->buf.p[2], (u16)pPage->buf.n);
if( pWriter->bFirstTermInPage ){
/* No term was written to this page. */
assert( pPage->pgidx.n==0 );
fts5WriteBtreeNoTerm(p, pWriter);
}else{
/* Append the pgidx to the page buffer. Set the szLeaf header field. */
fts5BufferAppendBlob(&p->rc, &pPage->buf, pPage->pgidx.n, pPage->pgidx.p);
}
/* Write the page out to disk */
iRowid = FTS5_SEGMENT_ROWID(pWriter->iSegid, pPage->pgno);
fts5DataWrite(p, iRowid, pPage->buf.p, pPage->buf.n);
/* Initialize the next page. */
fts5BufferZero(&pPage->buf);
fts5BufferZero(&pPage->pgidx);
fts5BufferAppendBlob(&p->rc, &pPage->buf, 4, zero);
pPage->iPrevPgidx = 0;
pPage->pgno++;
/* Increase the leaves written counter */
pWriter->nLeafWritten++;
/* The new leaf holds no terms or rowids */
pWriter->bFirstTermInPage = 1;
pWriter->bFirstRowidInPage = 1;
}
/*
** Append term pTerm/nTerm to the segment being written by the writer passed
** as the second argument.
**
** If an error occurs, set the Fts5Index.rc error code. If an error has
** already occurred, this function is a no-op.
*/
static void fts5WriteAppendTerm(
Fts5Index *p,
Fts5SegWriter *pWriter,
int nTerm, const u8 *pTerm
){
int nPrefix; /* Bytes of prefix compression for term */
Fts5PageWriter *pPage = &pWriter->writer;
Fts5Buffer *pPgidx = &pWriter->writer.pgidx;
int nMin = MIN(pPage->term.n, nTerm);
assert( p->rc==SQLITE_OK );
assert( pPage->buf.n>=4 );
assert( pPage->buf.n>4 || pWriter->bFirstTermInPage );
/* If the current leaf page is full, flush it to disk. */
if( (pPage->buf.n + pPgidx->n + nTerm + 2)>=p->pConfig->pgsz ){
if( pPage->buf.n>4 ){
fts5WriteFlushLeaf(p, pWriter);
if( p->rc!=SQLITE_OK ) return;
}
fts5BufferGrow(&p->rc, &pPage->buf, nTerm+FTS5_DATA_PADDING);
}
/* TODO1: Updating pgidx here. */
pPgidx->n += sqlite3Fts5PutVarint(
&pPgidx->p[pPgidx->n], pPage->buf.n - pPage->iPrevPgidx
);
pPage->iPrevPgidx = pPage->buf.n;
#if 0
fts5PutU16(&pPgidx->p[pPgidx->n], pPage->buf.n);
pPgidx->n += 2;
#endif
if( pWriter->bFirstTermInPage ){
nPrefix = 0;
if( pPage->pgno!=1 ){
/* This is the first term on a leaf that is not the leftmost leaf in
** the segment b-tree. In this case it is necessary to add a term to
** the b-tree hierarchy that is (a) larger than the largest term
** already written to the segment and (b) smaller than or equal to
** this term. In other words, a prefix of (pTerm/nTerm) that is one
** byte longer than the longest prefix (pTerm/nTerm) shares with the
** previous term.
**
** Usually, the previous term is available in pPage->term. The exception
** is if this is the first term written in an incremental-merge step.
** In this case the previous term is not available, so just write a
** copy of (pTerm/nTerm) into the parent node. This is slightly
** inefficient, but still correct. */
int n = nTerm;
if( pPage->term.n ){
n = 1 + fts5PrefixCompress(nMin, pPage->term.p, pTerm);
}
fts5WriteBtreeTerm(p, pWriter, n, pTerm);
if( p->rc!=SQLITE_OK ) return;
pPage = &pWriter->writer;
}
}else{
nPrefix = fts5PrefixCompress(nMin, pPage->term.p, pTerm);
fts5BufferAppendVarint(&p->rc, &pPage->buf, nPrefix);
}
/* Append the number of bytes of new data, then the term data itself
** to the page. */
fts5BufferAppendVarint(&p->rc, &pPage->buf, nTerm - nPrefix);
fts5BufferAppendBlob(&p->rc, &pPage->buf, nTerm - nPrefix, &pTerm[nPrefix]);
/* Update the Fts5PageWriter.term field. */
fts5BufferSet(&p->rc, &pPage->term, nTerm, pTerm);
pWriter->bFirstTermInPage = 0;
pWriter->bFirstRowidInPage = 0;
pWriter->bFirstRowidInDoclist = 1;
assert( p->rc || (pWriter->nDlidx>0 && pWriter->aDlidx[0].buf.n==0) );
pWriter->aDlidx[0].pgno = pPage->pgno;
}
/*
** Append a rowid and position-list size field to the writers output.
*/
static void fts5WriteAppendRowid(
Fts5Index *p,
Fts5SegWriter *pWriter,
i64 iRowid
){
if( p->rc==SQLITE_OK ){
Fts5PageWriter *pPage = &pWriter->writer;
if( (pPage->buf.n + pPage->pgidx.n)>=p->pConfig->pgsz ){
fts5WriteFlushLeaf(p, pWriter);
}
/* If this is to be the first rowid written to the page, set the
** rowid-pointer in the page-header. Also append a value to the dlidx
** buffer, in case a doclist-index is required. */
if( pWriter->bFirstRowidInPage ){
fts5PutU16(pPage->buf.p, (u16)pPage->buf.n);
fts5WriteDlidxAppend(p, pWriter, iRowid);
}
/* Write the rowid. */
if( pWriter->bFirstRowidInDoclist || pWriter->bFirstRowidInPage ){
fts5BufferAppendVarint(&p->rc, &pPage->buf, iRowid);
}else{
assert_nc( p->rc || iRowid>pWriter->iPrevRowid );
fts5BufferAppendVarint(&p->rc, &pPage->buf,
(u64)iRowid - (u64)pWriter->iPrevRowid
);
}
pWriter->iPrevRowid = iRowid;
pWriter->bFirstRowidInDoclist = 0;
pWriter->bFirstRowidInPage = 0;
}
}
static void fts5WriteAppendPoslistData(
Fts5Index *p,
Fts5SegWriter *pWriter,
const u8 *aData,
int nData
){
Fts5PageWriter *pPage = &pWriter->writer;
const u8 *a = aData;
int n = nData;
assert( p->pConfig->pgsz>0 || p->rc!=SQLITE_OK );
while( p->rc==SQLITE_OK
&& (pPage->buf.n + pPage->pgidx.n + n)>=p->pConfig->pgsz
){
int nReq = p->pConfig->pgsz - pPage->buf.n - pPage->pgidx.n;
int nCopy = 0;
while( nCopyrc, &pPage->buf, nCopy, a);
a += nCopy;
n -= nCopy;
fts5WriteFlushLeaf(p, pWriter);
}
if( n>0 ){
fts5BufferAppendBlob(&p->rc, &pPage->buf, n, a);
}
}
/*
** Flush any data cached by the writer object to the database. Free any
** allocations associated with the writer.
*/
static void fts5WriteFinish(
Fts5Index *p,
Fts5SegWriter *pWriter, /* Writer object */
int *pnLeaf /* OUT: Number of leaf pages in b-tree */
){
int i;
Fts5PageWriter *pLeaf = &pWriter->writer;
if( p->rc==SQLITE_OK ){
assert( pLeaf->pgno>=1 );
if( pLeaf->buf.n>4 ){
fts5WriteFlushLeaf(p, pWriter);
}
*pnLeaf = pLeaf->pgno-1;
if( pLeaf->pgno>1 ){
fts5WriteFlushBtree(p, pWriter);
}
}
fts5BufferFree(&pLeaf->term);
fts5BufferFree(&pLeaf->buf);
fts5BufferFree(&pLeaf->pgidx);
fts5BufferFree(&pWriter->btterm);
for(i=0; inDlidx; i++){
sqlite3Fts5BufferFree(&pWriter->aDlidx[i].buf);
}
sqlite3_free(pWriter->aDlidx);
}
static void fts5WriteInit(
Fts5Index *p,
Fts5SegWriter *pWriter,
int iSegid
){
const int nBuffer = p->pConfig->pgsz + FTS5_DATA_PADDING;
memset(pWriter, 0, sizeof(Fts5SegWriter));
pWriter->iSegid = iSegid;
fts5WriteDlidxGrow(p, pWriter, 1);
pWriter->writer.pgno = 1;
pWriter->bFirstTermInPage = 1;
pWriter->iBtPage = 1;
assert( pWriter->writer.buf.n==0 );
assert( pWriter->writer.pgidx.n==0 );
/* Grow the two buffers to pgsz + padding bytes in size. */
sqlite3Fts5BufferSize(&p->rc, &pWriter->writer.pgidx, nBuffer);
sqlite3Fts5BufferSize(&p->rc, &pWriter->writer.buf, nBuffer);
if( p->pIdxWriter==0 ){
Fts5Config *pConfig = p->pConfig;
fts5IndexPrepareStmt(p, &p->pIdxWriter, sqlite3_mprintf(
"INSERT INTO '%q'.'%q_idx'(segid,term,pgno) VALUES(?,?,?)",
pConfig->zDb, pConfig->zName
));
}
if( p->rc==SQLITE_OK ){
/* Initialize the 4-byte leaf-page header to 0x00. */
memset(pWriter->writer.buf.p, 0, 4);
pWriter->writer.buf.n = 4;
/* Bind the current output segment id to the index-writer. This is an
** optimization over binding the same value over and over as rows are
** inserted into %_idx by the current writer. */
sqlite3_bind_int(p->pIdxWriter, 1, pWriter->iSegid);
}
}
/*
** Iterator pIter was used to iterate through the input segments of on an
** incremental merge operation. This function is called if the incremental
** merge step has finished but the input has not been completely exhausted.
*/
static void fts5TrimSegments(Fts5Index *p, Fts5Iter *pIter){
int i;
Fts5Buffer buf;
memset(&buf, 0, sizeof(Fts5Buffer));
for(i=0; inSeg && p->rc==SQLITE_OK; i++){
Fts5SegIter *pSeg = &pIter->aSeg[i];
if( pSeg->pSeg==0 ){
/* no-op */
}else if( pSeg->pLeaf==0 ){
/* All keys from this input segment have been transfered to the output.
** Set both the first and last page-numbers to 0 to indicate that the
** segment is now empty. */
pSeg->pSeg->pgnoLast = 0;
pSeg->pSeg->pgnoFirst = 0;
}else{
int iOff = pSeg->iTermLeafOffset; /* Offset on new first leaf page */
i64 iLeafRowid;
Fts5Data *pData;
int iId = pSeg->pSeg->iSegid;
u8 aHdr[4] = {0x00, 0x00, 0x00, 0x00};
iLeafRowid = FTS5_SEGMENT_ROWID(iId, pSeg->iTermLeafPgno);
pData = fts5LeafRead(p, iLeafRowid);
if( pData ){
if( iOff>pData->szLeaf ){
/* This can occur if the pages that the segments occupy overlap - if
** a single page has been assigned to more than one segment. In
** this case a prior iteration of this loop may have corrupted the
** segment currently being trimmed. */
p->rc = FTS5_CORRUPT;
}else{
fts5BufferZero(&buf);
fts5BufferGrow(&p->rc, &buf, pData->nn);
fts5BufferAppendBlob(&p->rc, &buf, sizeof(aHdr), aHdr);
fts5BufferAppendVarint(&p->rc, &buf, pSeg->term.n);
fts5BufferAppendBlob(&p->rc, &buf, pSeg->term.n, pSeg->term.p);
fts5BufferAppendBlob(&p->rc, &buf,pData->szLeaf-iOff,&pData->p[iOff]);
if( p->rc==SQLITE_OK ){
/* Set the szLeaf field */
fts5PutU16(&buf.p[2], (u16)buf.n);
}
/* Set up the new page-index array */
fts5BufferAppendVarint(&p->rc, &buf, 4);
if( pSeg->iLeafPgno==pSeg->iTermLeafPgno
&& pSeg->iEndofDoclistszLeaf
&& pSeg->iPgidxOff<=pData->nn
){
int nDiff = pData->szLeaf - pSeg->iEndofDoclist;
fts5BufferAppendVarint(&p->rc, &buf, buf.n - 1 - nDiff - 4);
fts5BufferAppendBlob(&p->rc, &buf,
pData->nn - pSeg->iPgidxOff, &pData->p[pSeg->iPgidxOff]
);
}
pSeg->pSeg->pgnoFirst = pSeg->iTermLeafPgno;
fts5DataDelete(p, FTS5_SEGMENT_ROWID(iId, 1), iLeafRowid);
fts5DataWrite(p, iLeafRowid, buf.p, buf.n);
}
fts5DataRelease(pData);
}
}
}
fts5BufferFree(&buf);
}
static void fts5MergeChunkCallback(
Fts5Index *p,
void *pCtx,
const u8 *pChunk, int nChunk
){
Fts5SegWriter *pWriter = (Fts5SegWriter*)pCtx;
fts5WriteAppendPoslistData(p, pWriter, pChunk, nChunk);
}
/*
**
*/
static void fts5IndexMergeLevel(
Fts5Index *p, /* FTS5 backend object */
Fts5Structure **ppStruct, /* IN/OUT: Stucture of index */
int iLvl, /* Level to read input from */
int *pnRem /* Write up to this many output leaves */
){
Fts5Structure *pStruct = *ppStruct;
Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl];
Fts5StructureLevel *pLvlOut;
Fts5Iter *pIter = 0; /* Iterator to read input data */
int nRem = pnRem ? *pnRem : 0; /* Output leaf pages left to write */
int nInput; /* Number of input segments */
Fts5SegWriter writer; /* Writer object */
Fts5StructureSegment *pSeg; /* Output segment */
Fts5Buffer term;
int bOldest; /* True if the output segment is the oldest */
int eDetail = p->pConfig->eDetail;
const int flags = FTS5INDEX_QUERY_NOOUTPUT;
int bTermWritten = 0; /* True if current term already output */
assert( iLvlnLevel );
assert( pLvl->nMerge<=pLvl->nSeg );
memset(&writer, 0, sizeof(Fts5SegWriter));
memset(&term, 0, sizeof(Fts5Buffer));
if( pLvl->nMerge ){
pLvlOut = &pStruct->aLevel[iLvl+1];
assert( pLvlOut->nSeg>0 );
nInput = pLvl->nMerge;
pSeg = &pLvlOut->aSeg[pLvlOut->nSeg-1];
fts5WriteInit(p, &writer, pSeg->iSegid);
writer.writer.pgno = pSeg->pgnoLast+1;
writer.iBtPage = 0;
}else{
int iSegid = fts5AllocateSegid(p, pStruct);
/* Extend the Fts5Structure object as required to ensure the output
** segment exists. */
if( iLvl==pStruct->nLevel-1 ){
fts5StructureAddLevel(&p->rc, ppStruct);
pStruct = *ppStruct;
}
fts5StructureExtendLevel(&p->rc, pStruct, iLvl+1, 1, 0);
if( p->rc ) return;
pLvl = &pStruct->aLevel[iLvl];
pLvlOut = &pStruct->aLevel[iLvl+1];
fts5WriteInit(p, &writer, iSegid);
/* Add the new segment to the output level */
pSeg = &pLvlOut->aSeg[pLvlOut->nSeg];
pLvlOut->nSeg++;
pSeg->pgnoFirst = 1;
pSeg->iSegid = iSegid;
pStruct->nSegment++;
/* Read input from all segments in the input level */
nInput = pLvl->nSeg;
/* Set the range of origins that will go into the output segment. */
if( pStruct->nOriginCntr>0 ){
pSeg->iOrigin1 = pLvl->aSeg[0].iOrigin1;
pSeg->iOrigin2 = pLvl->aSeg[pLvl->nSeg-1].iOrigin2;
}
}
bOldest = (pLvlOut->nSeg==1 && pStruct->nLevel==iLvl+2);
assert( iLvl>=0 );
for(fts5MultiIterNew(p, pStruct, flags, 0, 0, 0, iLvl, nInput, &pIter);
fts5MultiIterEof(p, pIter)==0;
fts5MultiIterNext(p, pIter, 0, 0)
){
Fts5SegIter *pSegIter = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
int nPos; /* position-list size field value */
int nTerm;
const u8 *pTerm;
pTerm = fts5MultiIterTerm(pIter, &nTerm);
if( nTerm!=term.n || fts5Memcmp(pTerm, term.p, nTerm) ){
if( pnRem && writer.nLeafWritten>nRem ){
break;
}
fts5BufferSet(&p->rc, &term, nTerm, pTerm);
bTermWritten =0;
}
/* Check for key annihilation. */
if( pSegIter->nPos==0 && (bOldest || pSegIter->bDel==0) ) continue;
if( p->rc==SQLITE_OK && bTermWritten==0 ){
/* This is a new term. Append a term to the output segment. */
fts5WriteAppendTerm(p, &writer, nTerm, pTerm);
bTermWritten = 1;
}
/* Append the rowid to the output */
/* WRITEPOSLISTSIZE */
fts5WriteAppendRowid(p, &writer, fts5MultiIterRowid(pIter));
if( eDetail==FTS5_DETAIL_NONE ){
if( pSegIter->bDel ){
fts5BufferAppendVarint(&p->rc, &writer.writer.buf, 0);
if( pSegIter->nPos>0 ){
fts5BufferAppendVarint(&p->rc, &writer.writer.buf, 0);
}
}
}else{
/* Append the position-list data to the output */
nPos = pSegIter->nPos*2 + pSegIter->bDel;
fts5BufferAppendVarint(&p->rc, &writer.writer.buf, nPos);
fts5ChunkIterate(p, pSegIter, (void*)&writer, fts5MergeChunkCallback);
}
}
/* Flush the last leaf page to disk. Set the output segment b-tree height
** and last leaf page number at the same time. */
fts5WriteFinish(p, &writer, &pSeg->pgnoLast);
assert( pIter!=0 || p->rc!=SQLITE_OK );
if( fts5MultiIterEof(p, pIter) ){
int i;
/* Remove the redundant segments from the %_data table */
assert( pSeg->nEntry==0 );
for(i=0; iaSeg[i];
pSeg->nEntry += (pOld->nEntry - pOld->nEntryTombstone);
fts5DataRemoveSegment(p, pOld);
}
/* Remove the redundant segments from the input level */
if( pLvl->nSeg!=nInput ){
int nMove = (pLvl->nSeg - nInput) * sizeof(Fts5StructureSegment);
memmove(pLvl->aSeg, &pLvl->aSeg[nInput], nMove);
}
pStruct->nSegment -= nInput;
pLvl->nSeg -= nInput;
pLvl->nMerge = 0;
if( pSeg->pgnoLast==0 ){
pLvlOut->nSeg--;
pStruct->nSegment--;
}
}else{
assert( pSeg->pgnoLast>0 );
fts5TrimSegments(p, pIter);
pLvl->nMerge = nInput;
}
fts5MultiIterFree(pIter);
fts5BufferFree(&term);
if( pnRem ) *pnRem -= writer.nLeafWritten;
}
/*
** If this is not a contentless_delete=1 table, or if the 'deletemerge'
** configuration option is set to 0, then this function always returns -1.
** Otherwise, it searches the structure object passed as the second argument
** for a level suitable for merging due to having a large number of
** tombstones in the tombstone hash. If one is found, its index is returned.
** Otherwise, if there is no suitable level, -1.
*/
static int fts5IndexFindDeleteMerge(Fts5Index *p, Fts5Structure *pStruct){
Fts5Config *pConfig = p->pConfig;
int iRet = -1;
if( pConfig->bContentlessDelete && pConfig->nDeleteMerge>0 ){
int ii;
int nBest = 0;
for(ii=0; iinLevel; ii++){
Fts5StructureLevel *pLvl = &pStruct->aLevel[ii];
i64 nEntry = 0;
i64 nTomb = 0;
int iSeg;
for(iSeg=0; iSegnSeg; iSeg++){
nEntry += pLvl->aSeg[iSeg].nEntry;
nTomb += pLvl->aSeg[iSeg].nEntryTombstone;
}
assert_nc( nEntry>0 || pLvl->nSeg==0 );
if( nEntry>0 ){
int nPercent = (nTomb * 100) / nEntry;
if( nPercent>=pConfig->nDeleteMerge && nPercent>nBest ){
iRet = ii;
nBest = nPercent;
}
}
}
}
return iRet;
}
/*
** Do up to nPg pages of automerge work on the index.
**
** Return true if any changes were actually made, or false otherwise.
*/
static int fts5IndexMerge(
Fts5Index *p, /* FTS5 backend object */
Fts5Structure **ppStruct, /* IN/OUT: Current structure of index */
int nPg, /* Pages of work to do */
int nMin /* Minimum number of segments to merge */
){
int nRem = nPg;
int bRet = 0;
Fts5Structure *pStruct = *ppStruct;
while( nRem>0 && p->rc==SQLITE_OK ){
int iLvl; /* To iterate through levels */
int iBestLvl = 0; /* Level offering the most input segments */
int nBest = 0; /* Number of input segments on best level */
/* Set iBestLvl to the level to read input segments from. Or to -1 if
** there is no level suitable to merge segments from. */
assert( pStruct->nLevel>0 );
for(iLvl=0; iLvlnLevel; iLvl++){
Fts5StructureLevel *pLvl = &pStruct->aLevel[iLvl];
if( pLvl->nMerge ){
if( pLvl->nMerge>nBest ){
iBestLvl = iLvl;
nBest = nMin;
}
break;
}
if( pLvl->nSeg>nBest ){
nBest = pLvl->nSeg;
iBestLvl = iLvl;
}
}
if( nBestrc==SQLITE_OK && pStruct->aLevel[iBestLvl].nMerge==0 ){
fts5StructurePromote(p, iBestLvl+1, pStruct);
}
if( nMin==1 ) nMin = 2;
}
*ppStruct = pStruct;
return bRet;
}
/*
** A total of nLeaf leaf pages of data has just been flushed to a level-0
** segment. This function updates the write-counter accordingly and, if
** necessary, performs incremental merge work.
**
** If an error occurs, set the Fts5Index.rc error code. If an error has
** already occurred, this function is a no-op.
*/
static void fts5IndexAutomerge(
Fts5Index *p, /* FTS5 backend object */
Fts5Structure **ppStruct, /* IN/OUT: Current structure of index */
int nLeaf /* Number of output leaves just written */
){
if( p->rc==SQLITE_OK && p->pConfig->nAutomerge>0 && ALWAYS((*ppStruct)!=0) ){
Fts5Structure *pStruct = *ppStruct;
u64 nWrite; /* Initial value of write-counter */
int nWork; /* Number of work-quanta to perform */
int nRem; /* Number of leaf pages left to write */
/* Update the write-counter. While doing so, set nWork. */
nWrite = pStruct->nWriteCounter;
nWork = (int)(((nWrite + nLeaf) / p->nWorkUnit) - (nWrite / p->nWorkUnit));
pStruct->nWriteCounter += nLeaf;
nRem = (int)(p->nWorkUnit * nWork * pStruct->nLevel);
fts5IndexMerge(p, ppStruct, nRem, p->pConfig->nAutomerge);
}
}
static void fts5IndexCrisismerge(
Fts5Index *p, /* FTS5 backend object */
Fts5Structure **ppStruct /* IN/OUT: Current structure of index */
){
const int nCrisis = p->pConfig->nCrisisMerge;
Fts5Structure *pStruct = *ppStruct;
if( pStruct && pStruct->nLevel>0 ){
int iLvl = 0;
while( p->rc==SQLITE_OK && pStruct->aLevel[iLvl].nSeg>=nCrisis ){
fts5IndexMergeLevel(p, &pStruct, iLvl, 0);
assert( p->rc!=SQLITE_OK || pStruct->nLevel>(iLvl+1) );
fts5StructurePromote(p, iLvl+1, pStruct);
iLvl++;
}
*ppStruct = pStruct;
}
}
static int fts5IndexReturn(Fts5Index *p){
int rc = p->rc;
p->rc = SQLITE_OK;
return rc;
}
typedef struct Fts5FlushCtx Fts5FlushCtx;
struct Fts5FlushCtx {
Fts5Index *pIdx;
Fts5SegWriter writer;
};
/*
** Buffer aBuf[] contains a list of varints, all small enough to fit
** in a 32-bit integer. Return the size of the largest prefix of this
** list nMax bytes or less in size.
*/
static int fts5PoslistPrefix(const u8 *aBuf, int nMax){
int ret;
u32 dummy;
ret = fts5GetVarint32(aBuf, dummy);
if( ret nMax ) break;
ret += i;
}
}
return ret;
}
/*
** Execute the SQL statement:
**
** DELETE FROM %_idx WHERE (segid, (pgno/2)) = ($iSegid, $iPgno);
**
** This is used when a secure-delete operation removes the last term
** from a segment leaf page. In that case the %_idx entry is removed
** too. This is done to ensure that if all instances of a token are
** removed from an fts5 database in secure-delete mode, no trace of
** the token itself remains in the database.
*/
static void fts5SecureDeleteIdxEntry(
Fts5Index *p, /* FTS5 backend object */
int iSegid, /* Id of segment to delete entry for */
int iPgno /* Page number within segment */
){
if( iPgno!=1 ){
assert( p->pConfig->iVersion==FTS5_CURRENT_VERSION_SECUREDELETE );
if( p->pDeleteFromIdx==0 ){
fts5IndexPrepareStmt(p, &p->pDeleteFromIdx, sqlite3_mprintf(
"DELETE FROM '%q'.'%q_idx' WHERE (segid, (pgno/2)) = (?1, ?2)",
p->pConfig->zDb, p->pConfig->zName
));
}
if( p->rc==SQLITE_OK ){
sqlite3_bind_int(p->pDeleteFromIdx, 1, iSegid);
sqlite3_bind_int(p->pDeleteFromIdx, 2, iPgno);
sqlite3_step(p->pDeleteFromIdx);
p->rc = sqlite3_reset(p->pDeleteFromIdx);
}
}
}
/*
** This is called when a secure-delete operation removes a position-list
** that overflows onto segment page iPgno of segment pSeg. This function
** rewrites node iPgno, and possibly one or more of its right-hand peers,
** to remove this portion of the position list.
**
** Output variable (*pbLastInDoclist) is set to true if the position-list
** removed is followed by a new term or the end-of-segment, or false if
** it is followed by another rowid/position list.
*/
static void fts5SecureDeleteOverflow(
Fts5Index *p,
Fts5StructureSegment *pSeg,
int iPgno,
int *pbLastInDoclist
){
const int bDetailNone = (p->pConfig->eDetail==FTS5_DETAIL_NONE);
int pgno;
Fts5Data *pLeaf = 0;
assert( iPgno!=1 );
*pbLastInDoclist = 1;
for(pgno=iPgno; p->rc==SQLITE_OK && pgno<=pSeg->pgnoLast; pgno++){
i64 iRowid = FTS5_SEGMENT_ROWID(pSeg->iSegid, pgno);
int iNext = 0;
u8 *aPg = 0;
pLeaf = fts5DataRead(p, iRowid);
if( pLeaf==0 ) break;
aPg = pLeaf->p;
iNext = fts5GetU16(&aPg[0]);
if( iNext!=0 ){
*pbLastInDoclist = 0;
}
if( iNext==0 && pLeaf->szLeaf!=pLeaf->nn ){
fts5GetVarint32(&aPg[pLeaf->szLeaf], iNext);
}
if( iNext==0 ){
/* The page contains no terms or rowids. Replace it with an empty
** page and move on to the right-hand peer. */
const u8 aEmpty[] = {0x00, 0x00, 0x00, 0x04};
assert_nc( bDetailNone==0 || pLeaf->nn==4 );
if( bDetailNone==0 ) fts5DataWrite(p, iRowid, aEmpty, sizeof(aEmpty));
fts5DataRelease(pLeaf);
pLeaf = 0;
}else if( bDetailNone ){
break;
}else if( iNext>=pLeaf->szLeaf || pLeaf->nnszLeaf || iNext<4 ){
p->rc = FTS5_CORRUPT;
break;
}else{
int nShift = iNext - 4;
int nPg;
int nIdx = 0;
u8 *aIdx = 0;
/* Unless the current page footer is 0 bytes in size (in which case
** the new page footer will be as well), allocate and populate a
** buffer containing the new page footer. Set stack variables aIdx
** and nIdx accordingly. */
if( pLeaf->nn>pLeaf->szLeaf ){
int iFirst = 0;
int i1 = pLeaf->szLeaf;
int i2 = 0;
i1 += fts5GetVarint32(&aPg[i1], iFirst);
if( iFirstrc = FTS5_CORRUPT;
break;
}
aIdx = sqlite3Fts5MallocZero(&p->rc, (pLeaf->nn-pLeaf->szLeaf)+2);
if( aIdx==0 ) break;
i2 = sqlite3Fts5PutVarint(aIdx, iFirst-nShift);
if( i1nn ){
memcpy(&aIdx[i2], &aPg[i1], pLeaf->nn-i1);
i2 += (pLeaf->nn-i1);
}
nIdx = i2;
}
/* Modify the contents of buffer aPg[]. Set nPg to the new size
** in bytes. The new page is always smaller than the old. */
nPg = pLeaf->szLeaf - nShift;
memmove(&aPg[4], &aPg[4+nShift], nPg-4);
fts5PutU16(&aPg[2], nPg);
if( fts5GetU16(&aPg[0]) ) fts5PutU16(&aPg[0], 4);
if( nIdx>0 ){
memcpy(&aPg[nPg], aIdx, nIdx);
nPg += nIdx;
}
sqlite3_free(aIdx);
/* Write the new page to disk and exit the loop */
assert( nPg>4 || fts5GetU16(aPg)==0 );
fts5DataWrite(p, iRowid, aPg, nPg);
break;
}
}
fts5DataRelease(pLeaf);
}
/*
** Completely remove the entry that pSeg currently points to from
** the database.
*/
static void fts5DoSecureDelete(
Fts5Index *p,
Fts5SegIter *pSeg
){
const int bDetailNone = (p->pConfig->eDetail==FTS5_DETAIL_NONE);
int iSegid = pSeg->pSeg->iSegid;
u8 *aPg = pSeg->pLeaf->p;
int nPg = pSeg->pLeaf->nn;
int iPgIdx = pSeg->pLeaf->szLeaf;
u64 iDelta = 0;
int iNextOff = 0;
int iOff = 0;
int nIdx = 0;
u8 *aIdx = 0;
int bLastInDoclist = 0;
int iIdx = 0;
int iStart = 0;
int iDelKeyOff = 0; /* Offset of deleted key, if any */
nIdx = nPg-iPgIdx;
aIdx = sqlite3Fts5MallocZero(&p->rc, nIdx+16);
if( p->rc ) return;
memcpy(aIdx, &aPg[iPgIdx], nIdx);
/* At this point segment iterator pSeg points to the entry
** this function should remove from the b-tree segment.
**
** In detail=full or detail=column mode, pSeg->iLeafOffset is the
** offset of the first byte in the position-list for the entry to
** remove. Immediately before this comes two varints that will also
** need to be removed:
**
** + the rowid or delta rowid value for the entry, and
** + the size of the position list in bytes.
**
** Or, in detail=none mode, there is a single varint prior to
** pSeg->iLeafOffset - the rowid or delta rowid value.
**
** This block sets the following variables:
**
** iStart:
** The offset of the first byte of the rowid or delta-rowid
** value for the doclist entry being removed.
**
** iDelta:
** The value of the rowid or delta-rowid value for the doclist
** entry being removed.
**
** iNextOff:
** The offset of the next entry following the position list
** for the one being removed. If the position list for this
** entry overflows onto the next leaf page, this value will be
** greater than pLeaf->szLeaf.
*/
{
int iSOP; /* Start-Of-Position-list */
if( pSeg->iLeafPgno==pSeg->iTermLeafPgno ){
iStart = pSeg->iTermLeafOffset;
}else{
iStart = fts5GetU16(&aPg[0]);
}
iSOP = iStart + fts5GetVarint(&aPg[iStart], &iDelta);
assert_nc( iSOP<=pSeg->iLeafOffset );
if( bDetailNone ){
while( iSOPiLeafOffset ){
if( aPg[iSOP]==0x00 ) iSOP++;
if( aPg[iSOP]==0x00 ) iSOP++;
iStart = iSOP;
iSOP = iStart + fts5GetVarint(&aPg[iStart], &iDelta);
}
iNextOff = iSOP;
if( iNextOffiEndofDoclist && aPg[iNextOff]==0x00 ) iNextOff++;
if( iNextOffiEndofDoclist && aPg[iNextOff]==0x00 ) iNextOff++;
}else{
int nPos = 0;
iSOP += fts5GetVarint32(&aPg[iSOP], nPos);
while( iSOPiLeafOffset ){
iStart = iSOP + (nPos/2);
iSOP = iStart + fts5GetVarint(&aPg[iStart], &iDelta);
iSOP += fts5GetVarint32(&aPg[iSOP], nPos);
}
assert_nc( iSOP==pSeg->iLeafOffset );
iNextOff = pSeg->iLeafOffset + pSeg->nPos;
}
}
iOff = iStart;
/* If the position-list for the entry being removed flows over past
** the end of this page, delete the portion of the position-list on the
** next page and beyond.
**
** Set variable bLastInDoclist to true if this entry happens
** to be the last rowid in the doclist for its term. */
if( iNextOff>=iPgIdx ){
int pgno = pSeg->iLeafPgno+1;
fts5SecureDeleteOverflow(p, pSeg->pSeg, pgno, &bLastInDoclist);
iNextOff = iPgIdx;
}
if( pSeg->bDel==0 ){
if( iNextOff!=iPgIdx ){
/* Loop through the page-footer. If iNextOff (offset of the
** entry following the one we are removing) is equal to the
** offset of a key on this page, then the entry is the last
** in its doclist. */
int iKeyOff = 0;
for(iIdx=0; iIdxbDel ){
iOff += sqlite3Fts5PutVarint(&aPg[iOff], iDelta);
aPg[iOff++] = 0x01;
}else if( bLastInDoclist==0 ){
if( iNextOff!=iPgIdx ){
u64 iNextDelta = 0;
iNextOff += fts5GetVarint(&aPg[iNextOff], &iNextDelta);
iOff += sqlite3Fts5PutVarint(&aPg[iOff], iDelta + iNextDelta);
}
}else if(
pSeg->iLeafPgno==pSeg->iTermLeafPgno
&& iStart==pSeg->iTermLeafOffset
){
/* The entry being removed was the only position list in its
** doclist. Therefore the term needs to be removed as well. */
int iKey = 0;
int iKeyOff = 0;
/* Set iKeyOff to the offset of the term that will be removed - the
** last offset in the footer that is not greater than iStart. */
for(iIdx=0; iIdx(u32)iStart ) break;
iKeyOff += iVal;
}
assert_nc( iKey>=1 );
/* Set iDelKeyOff to the value of the footer entry to remove from
** the page. */
iDelKeyOff = iOff = iKeyOff;
if( iNextOff!=iPgIdx ){
/* This is the only position-list associated with the term, and there
** is another term following it on this page. So the subsequent term
** needs to be moved to replace the term associated with the entry
** being removed. */
int nPrefix = 0;
int nSuffix = 0;
int nPrefix2 = 0;
int nSuffix2 = 0;
iDelKeyOff = iNextOff;
iNextOff += fts5GetVarint32(&aPg[iNextOff], nPrefix2);
iNextOff += fts5GetVarint32(&aPg[iNextOff], nSuffix2);
if( iKey!=1 ){
iKeyOff += fts5GetVarint32(&aPg[iKeyOff], nPrefix);
}
iKeyOff += fts5GetVarint32(&aPg[iKeyOff], nSuffix);
nPrefix = MIN(nPrefix, nPrefix2);
nSuffix = (nPrefix2 + nSuffix2) - nPrefix;
if( (iKeyOff+nSuffix)>iPgIdx || (iNextOff+nSuffix2)>iPgIdx ){
p->rc = FTS5_CORRUPT;
}else{
if( iKey!=1 ){
iOff += sqlite3Fts5PutVarint(&aPg[iOff], nPrefix);
}
iOff += sqlite3Fts5PutVarint(&aPg[iOff], nSuffix);
if( nPrefix2>pSeg->term.n ){
p->rc = FTS5_CORRUPT;
}else if( nPrefix2>nPrefix ){
memcpy(&aPg[iOff], &pSeg->term.p[nPrefix], nPrefix2-nPrefix);
iOff += (nPrefix2-nPrefix);
}
memmove(&aPg[iOff], &aPg[iNextOff], nSuffix2);
iOff += nSuffix2;
iNextOff += nSuffix2;
}
}
}else if( iStart==4 ){
int iPgno;
assert_nc( pSeg->iLeafPgno>pSeg->iTermLeafPgno );
/* The entry being removed may be the only position list in
** its doclist. */
for(iPgno=pSeg->iLeafPgno-1; iPgno>pSeg->iTermLeafPgno; iPgno-- ){
Fts5Data *pPg = fts5DataRead(p, FTS5_SEGMENT_ROWID(iSegid, iPgno));
int bEmpty = (pPg && pPg->nn==4);
fts5DataRelease(pPg);
if( bEmpty==0 ) break;
}
if( iPgno==pSeg->iTermLeafPgno ){
i64 iId = FTS5_SEGMENT_ROWID(iSegid, pSeg->iTermLeafPgno);
Fts5Data *pTerm = fts5DataRead(p, iId);
if( pTerm && pTerm->szLeaf==pSeg->iTermLeafOffset ){
u8 *aTermIdx = &pTerm->p[pTerm->szLeaf];
int nTermIdx = pTerm->nn - pTerm->szLeaf;
int iTermIdx = 0;
int iTermOff = 0;
while( 1 ){
u32 iVal = 0;
int nByte = fts5GetVarint32(&aTermIdx[iTermIdx], iVal);
iTermOff += iVal;
if( (iTermIdx+nByte)>=nTermIdx ) break;
iTermIdx += nByte;
}
nTermIdx = iTermIdx;
memmove(&pTerm->p[iTermOff], &pTerm->p[pTerm->szLeaf], nTermIdx);
fts5PutU16(&pTerm->p[2], iTermOff);
fts5DataWrite(p, iId, pTerm->p, iTermOff+nTermIdx);
if( nTermIdx==0 ){
fts5SecureDeleteIdxEntry(p, iSegid, pSeg->iTermLeafPgno);
}
}
fts5DataRelease(pTerm);
}
}
/* Assuming no error has occurred, this block does final edits to the
** leaf page before writing it back to disk. Input variables are:
**
** nPg: Total initial size of leaf page.
** iPgIdx: Initial offset of page footer.
**
** iOff: Offset to move data to
** iNextOff: Offset to move data from
*/
if( p->rc==SQLITE_OK ){
const int nMove = nPg - iNextOff; /* Number of bytes to move */
int nShift = iNextOff - iOff; /* Distance to move them */
int iPrevKeyOut = 0;
int iKeyIn = 0;
memmove(&aPg[iOff], &aPg[iNextOff], nMove);
iPgIdx -= nShift;
nPg = iPgIdx;
fts5PutU16(&aPg[2], iPgIdx);
for(iIdx=0; iIdxiOff ? nShift : 0));
nPg += sqlite3Fts5PutVarint(&aPg[nPg], iKeyOut - iPrevKeyOut);
iPrevKeyOut = iKeyOut;
}
}
if( iPgIdx==nPg && nIdx>0 && pSeg->iLeafPgno!=1 ){
fts5SecureDeleteIdxEntry(p, iSegid, pSeg->iLeafPgno);
}
assert_nc( nPg>4 || fts5GetU16(aPg)==0 );
fts5DataWrite(p, FTS5_SEGMENT_ROWID(iSegid,pSeg->iLeafPgno), aPg, nPg);
}
sqlite3_free(aIdx);
}
/*
** This is called as part of flushing a delete to disk in 'secure-delete'
** mode. It edits the segments within the database described by argument
** pStruct to remove the entries for term zTerm, rowid iRowid.
*/
static void fts5FlushSecureDelete(
Fts5Index *p,
Fts5Structure *pStruct,
const char *zTerm,
int nTerm,
i64 iRowid
){
const int f = FTS5INDEX_QUERY_SKIPHASH;
Fts5Iter *pIter = 0; /* Used to find term instance */
fts5MultiIterNew(p, pStruct, f, 0, (const u8*)zTerm, nTerm, -1, 0, &pIter);
if( fts5MultiIterEof(p, pIter)==0 ){
i64 iThis = fts5MultiIterRowid(pIter);
if( iThisrc==SQLITE_OK
&& fts5MultiIterEof(p, pIter)==0
&& iRowid==fts5MultiIterRowid(pIter)
){
Fts5SegIter *pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst];
fts5DoSecureDelete(p, pSeg);
}
}
fts5MultiIterFree(pIter);
}
/*
** Flush the contents of in-memory hash table iHash to a new level-0
** segment on disk. Also update the corresponding structure record.
**
** If an error occurs, set the Fts5Index.rc error code. If an error has
** already occurred, this function is a no-op.
*/
static void fts5FlushOneHash(Fts5Index *p){
Fts5Hash *pHash = p->pHash;
Fts5Structure *pStruct;
int iSegid;
int pgnoLast = 0; /* Last leaf page number in segment */
/* Obtain a reference to the index structure and allocate a new segment-id
** for the new level-0 segment. */
pStruct = fts5StructureRead(p);
fts5StructureInvalidate(p);
if( sqlite3Fts5HashIsEmpty(pHash)==0 ){
iSegid = fts5AllocateSegid(p, pStruct);
if( iSegid ){
const int pgsz = p->pConfig->pgsz;
int eDetail = p->pConfig->eDetail;
int bSecureDelete = p->pConfig->bSecureDelete;
Fts5StructureSegment *pSeg; /* New segment within pStruct */
Fts5Buffer *pBuf; /* Buffer in which to assemble leaf page */
Fts5Buffer *pPgidx; /* Buffer in which to assemble pgidx */
Fts5SegWriter writer;
fts5WriteInit(p, &writer, iSegid);
pBuf = &writer.writer.buf;
pPgidx = &writer.writer.pgidx;
/* fts5WriteInit() should have initialized the buffers to (most likely)
** the maximum space required. */
assert( p->rc || pBuf->nSpace>=(pgsz + FTS5_DATA_PADDING) );
assert( p->rc || pPgidx->nSpace>=(pgsz + FTS5_DATA_PADDING) );
/* Begin scanning through hash table entries. This loop runs once for each
** term/doclist currently stored within the hash table. */
if( p->rc==SQLITE_OK ){
p->rc = sqlite3Fts5HashScanInit(pHash, 0, 0);
}
while( p->rc==SQLITE_OK && 0==sqlite3Fts5HashScanEof(pHash) ){
const char *zTerm; /* Buffer containing term */
int nTerm; /* Size of zTerm in bytes */
const u8 *pDoclist; /* Pointer to doclist for this term */
int nDoclist; /* Size of doclist in bytes */
/* Get the term and doclist for this entry. */
sqlite3Fts5HashScanEntry(pHash, &zTerm, &nTerm, &pDoclist, &nDoclist);
if( bSecureDelete==0 ){
fts5WriteAppendTerm(p, &writer, nTerm, (const u8*)zTerm);
if( p->rc!=SQLITE_OK ) break;
assert( writer.bFirstRowidInPage==0 );
}
if( !bSecureDelete && pgsz>=(pBuf->n + pPgidx->n + nDoclist + 1) ){
/* The entire doclist will fit on the current leaf. */
fts5BufferSafeAppendBlob(pBuf, pDoclist, nDoclist);
}else{
int bTermWritten = !bSecureDelete;
i64 iRowid = 0;
i64 iPrev = 0;
int iOff = 0;
/* The entire doclist will not fit on this leaf. The following
** loop iterates through the poslists that make up the current
** doclist. */
while( p->rc==SQLITE_OK && iOffrc!=SQLITE_OK || pDoclist[iOff]==0x01 ){
iOff++;
continue;
}
}
}
if( p->rc==SQLITE_OK && bTermWritten==0 ){
fts5WriteAppendTerm(p, &writer, nTerm, (const u8*)zTerm);
bTermWritten = 1;
assert( p->rc!=SQLITE_OK || writer.bFirstRowidInPage==0 );
}
if( writer.bFirstRowidInPage ){
fts5PutU16(&pBuf->p[0], (u16)pBuf->n); /* first rowid on page */
pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], iRowid);
writer.bFirstRowidInPage = 0;
fts5WriteDlidxAppend(p, &writer, iRowid);
}else{
u64 iRowidDelta = (u64)iRowid - (u64)iPrev;
pBuf->n += sqlite3Fts5PutVarint(&pBuf->p[pBuf->n], iRowidDelta);
}
if( p->rc!=SQLITE_OK ) break;
assert( pBuf->n<=pBuf->nSpace );
iPrev = iRowid;
if( eDetail==FTS5_DETAIL_NONE ){
if( iOffp[pBuf->n++] = 0;
iOff++;
if( iOffp[pBuf->n++] = 0;
iOff++;
}
}
if( (pBuf->n + pPgidx->n)>=pgsz ){
fts5WriteFlushLeaf(p, &writer);
}
}else{
int bDel = 0;
int nPos = 0;
int nCopy = fts5GetPoslistSize(&pDoclist[iOff], &nPos, &bDel);
if( bDel && bSecureDelete ){
fts5BufferAppendVarint(&p->rc, pBuf, nPos*2);
iOff += nCopy;
nCopy = nPos;
}else{
nCopy += nPos;
}
if( (pBuf->n + pPgidx->n + nCopy) <= pgsz ){
/* The entire poslist will fit on the current leaf. So copy
** it in one go. */
fts5BufferSafeAppendBlob(pBuf, &pDoclist[iOff], nCopy);
}else{
/* The entire poslist will not fit on this leaf. So it needs
** to be broken into sections. The only qualification being
** that each varint must be stored contiguously. */
const u8 *pPoslist = &pDoclist[iOff];
int iPos = 0;
while( p->rc==SQLITE_OK ){
int nSpace = pgsz - pBuf->n - pPgidx->n;
int n = 0;
if( (nCopy - iPos)<=nSpace ){
n = nCopy - iPos;
}else{
n = fts5PoslistPrefix(&pPoslist[iPos], nSpace);
}
assert( n>0 );
fts5BufferSafeAppendBlob(pBuf, &pPoslist[iPos], n);
iPos += n;
if( (pBuf->n + pPgidx->n)>=pgsz ){
fts5WriteFlushLeaf(p, &writer);
}
if( iPos>=nCopy ) break;
}
}
iOff += nCopy;
}
}
}
/* TODO2: Doclist terminator written here. */
/* pBuf->p[pBuf->n++] = '\0'; */
assert( pBuf->n<=pBuf->nSpace );
if( p->rc==SQLITE_OK ) sqlite3Fts5HashScanNext(pHash);
}
fts5WriteFinish(p, &writer, &pgnoLast);
assert( p->rc!=SQLITE_OK || bSecureDelete || pgnoLast>0 );
if( pgnoLast>0 ){
/* Update the Fts5Structure. It is written back to the database by the
** fts5StructureRelease() call below. */
if( pStruct->nLevel==0 ){
fts5StructureAddLevel(&p->rc, &pStruct);
}
fts5StructureExtendLevel(&p->rc, pStruct, 0, 1, 0);
if( p->rc==SQLITE_OK ){
pSeg = &pStruct->aLevel[0].aSeg[ pStruct->aLevel[0].nSeg++ ];
pSeg->iSegid = iSegid;
pSeg->pgnoFirst = 1;
pSeg->pgnoLast = pgnoLast;
if( pStruct->nOriginCntr>0 ){
pSeg->iOrigin1 = pStruct->nOriginCntr;
pSeg->iOrigin2 = pStruct->nOriginCntr;
pSeg->nEntry = p->nPendingRow;
pStruct->nOriginCntr++;
}
pStruct->nSegment++;
}
fts5StructurePromote(p, 0, pStruct);
}
}
}
fts5IndexAutomerge(p, &pStruct, pgnoLast + p->nContentlessDelete);
fts5IndexCrisismerge(p, &pStruct);
fts5StructureWrite(p, pStruct);
fts5StructureRelease(pStruct);
}
/*
** Flush any data stored in the in-memory hash tables to the database.
*/
static void fts5IndexFlush(Fts5Index *p){
/* Unless it is empty, flush the hash table to disk */
if( p->flushRc ){
p->rc = p->flushRc;
return;
}
if( p->nPendingData || p->nContentlessDelete ){
assert( p->pHash );
fts5FlushOneHash(p);
if( p->rc==SQLITE_OK ){
sqlite3Fts5HashClear(p->pHash);
p->nPendingData = 0;
p->nPendingRow = 0;
p->nContentlessDelete = 0;
}else if( p->nPendingData || p->nContentlessDelete ){
p->flushRc = p->rc;
}
}
}
static Fts5Structure *fts5IndexOptimizeStruct(
Fts5Index *p,
Fts5Structure *pStruct
){
Fts5Structure *pNew = 0;
sqlite3_int64 nByte = sizeof(Fts5Structure);
int nSeg = pStruct->nSegment;
int i;
/* Figure out if this structure requires optimization. A structure does
** not require optimization if either:
**
** 1. it consists of fewer than two segments, or
** 2. all segments are on the same level, or
** 3. all segments except one are currently inputs to a merge operation.
**
** In the first case, if there are no tombstone hash pages, return NULL. In
** the second, increment the ref-count on *pStruct and return a copy of the
** pointer to it.
*/
if( nSeg==0 ) return 0;
for(i=0; inLevel; i++){
int nThis = pStruct->aLevel[i].nSeg;
int nMerge = pStruct->aLevel[i].nMerge;
if( nThis>0 && (nThis==nSeg || (nThis==nSeg-1 && nMerge==nThis)) ){
if( nSeg==1 && nThis==1 && pStruct->aLevel[i].aSeg[0].nPgTombstone==0 ){
return 0;
}
fts5StructureRef(pStruct);
return pStruct;
}
assert( pStruct->aLevel[i].nMerge<=nThis );
}
nByte += (pStruct->nLevel+1) * sizeof(Fts5StructureLevel);
pNew = (Fts5Structure*)sqlite3Fts5MallocZero(&p->rc, nByte);
if( pNew ){
Fts5StructureLevel *pLvl;
nByte = nSeg * sizeof(Fts5StructureSegment);
pNew->nLevel = MIN(pStruct->nLevel+1, FTS5_MAX_LEVEL);
pNew->nRef = 1;
pNew->nWriteCounter = pStruct->nWriteCounter;
pNew->nOriginCntr = pStruct->nOriginCntr;
pLvl = &pNew->aLevel[pNew->nLevel-1];
pLvl->aSeg = (Fts5StructureSegment*)sqlite3Fts5MallocZero(&p->rc, nByte);
if( pLvl->aSeg ){
int iLvl, iSeg;
int iSegOut = 0;
/* Iterate through all segments, from oldest to newest. Add them to
** the new Fts5Level object so that pLvl->aSeg[0] is the oldest
** segment in the data structure. */
for(iLvl=pStruct->nLevel-1; iLvl>=0; iLvl--){
for(iSeg=0; iSegaLevel[iLvl].nSeg; iSeg++){
pLvl->aSeg[iSegOut] = pStruct->aLevel[iLvl].aSeg[iSeg];
iSegOut++;
}
}
pNew->nSegment = pLvl->nSeg = nSeg;
}else{
sqlite3_free(pNew);
pNew = 0;
}
}
return pNew;
}
static int sqlite3Fts5IndexOptimize(Fts5Index *p){
Fts5Structure *pStruct;
Fts5Structure *pNew = 0;
assert( p->rc==SQLITE_OK );
fts5IndexFlush(p);
assert( p->rc!=SQLITE_OK || p->nContentlessDelete==0 );
pStruct = fts5StructureRead(p);
assert( p->rc!=SQLITE_OK || pStruct!=0 );
fts5StructureInvalidate(p);
if( pStruct ){
pNew = fts5IndexOptimizeStruct(p, pStruct);
}
fts5StructureRelease(pStruct);
assert( pNew==0 || pNew->nSegment>0 );
if( pNew ){
int iLvl;
for(iLvl=0; pNew->aLevel[iLvl].nSeg==0; iLvl++){}
while( p->rc==SQLITE_OK && pNew->aLevel[iLvl].nSeg>0 ){
int nRem = FTS5_OPT_WORK_UNIT;
fts5IndexMergeLevel(p, &pNew, iLvl, &nRem);
}
fts5StructureWrite(p, pNew);
fts5StructureRelease(pNew);
}
return fts5IndexReturn(p);
}
/*
** This is called to implement the special "VALUES('merge', $nMerge)"
** INSERT command.
*/
static int sqlite3Fts5IndexMerge(Fts5Index *p, int nMerge){
Fts5Structure *pStruct = 0;
fts5IndexFlush(p);
pStruct = fts5StructureRead(p);
if( pStruct ){
int nMin = p->pConfig->nUsermerge;
fts5StructureInvalidate(p);
if( nMerge<0 ){
Fts5Structure *pNew = fts5IndexOptimizeStruct(p, pStruct);
fts5StructureRelease(pStruct);
pStruct = pNew;
nMin = 1;
nMerge = nMerge*-1;
}
if( pStruct && pStruct->nLevel ){
if( fts5IndexMerge(p, &pStruct, nMerge, nMin) ){
fts5StructureWrite(p, pStruct);
}
}
fts5StructureRelease(pStruct);
}
return fts5IndexReturn(p);
}
static void fts5AppendRowid(
Fts5Index *p,
u64 iDelta,
Fts5Iter *pUnused,
Fts5Buffer *pBuf
){
UNUSED_PARAM(pUnused);
fts5BufferAppendVarint(&p->rc, pBuf, iDelta);
}
static void fts5AppendPoslist(
Fts5Index *p,
u64 iDelta,
Fts5Iter *pMulti,
Fts5Buffer *pBuf
){
int nData = pMulti->base.nData;
int nByte = nData + 9 + 9 + FTS5_DATA_ZERO_PADDING;
assert( nData>0 );
if( p->rc==SQLITE_OK && 0==fts5BufferGrow(&p->rc, pBuf, nByte) ){
fts5BufferSafeAppendVarint(pBuf, iDelta);
fts5BufferSafeAppendVarint(pBuf, nData*2);
fts5BufferSafeAppendBlob(pBuf, pMulti->base.pData, nData);
memset(&pBuf->p[pBuf->n], 0, FTS5_DATA_ZERO_PADDING);
}
}
static void fts5DoclistIterNext(Fts5DoclistIter *pIter){
u8 *p = pIter->aPoslist + pIter->nSize + pIter->nPoslist;
assert( pIter->aPoslist || (p==0 && pIter->aPoslist==0) );
if( p>=pIter->aEof ){
pIter->aPoslist = 0;
}else{
i64 iDelta;
p += fts5GetVarint(p, (u64*)&iDelta);
pIter->iRowid += iDelta;
/* Read position list size */
if( p[0] & 0x80 ){
int nPos;
pIter->nSize = fts5GetVarint32(p, nPos);
pIter->nPoslist = (nPos>>1);
}else{
pIter->nPoslist = ((int)(p[0])) >> 1;
pIter->nSize = 1;
}
pIter->aPoslist = p;
if( &pIter->aPoslist[pIter->nPoslist]>pIter->aEof ){
pIter->aPoslist = 0;
}
}
}
static void fts5DoclistIterInit(
Fts5Buffer *pBuf,
Fts5DoclistIter *pIter
){
memset(pIter, 0, sizeof(*pIter));
if( pBuf->n>0 ){
pIter->aPoslist = pBuf->p;
pIter->aEof = &pBuf->p[pBuf->n];
fts5DoclistIterNext(pIter);
}
}
#if 0
/*
** Append a doclist to buffer pBuf.
**
** This function assumes that space within the buffer has already been
** allocated.
*/
static void fts5MergeAppendDocid(
Fts5Buffer *pBuf, /* Buffer to write to */
i64 *piLastRowid, /* IN/OUT: Previous rowid written (if any) */
i64 iRowid /* Rowid to append */
){
assert( pBuf->n!=0 || (*piLastRowid)==0 );
fts5BufferSafeAppendVarint(pBuf, iRowid - *piLastRowid);
*piLastRowid = iRowid;
}
#endif
#define fts5MergeAppendDocid(pBuf, iLastRowid, iRowid) { \
assert( (pBuf)->n!=0 || (iLastRowid)==0 ); \
fts5BufferSafeAppendVarint((pBuf), (u64)(iRowid) - (u64)(iLastRowid)); \
(iLastRowid) = (iRowid); \
}
/*
** Swap the contents of buffer *p1 with that of *p2.
*/
static void fts5BufferSwap(Fts5Buffer *p1, Fts5Buffer *p2){
Fts5Buffer tmp = *p1;
*p1 = *p2;
*p2 = tmp;
}
static void fts5NextRowid(Fts5Buffer *pBuf, int *piOff, i64 *piRowid){
int i = *piOff;
if( i>=pBuf->n ){
*piOff = -1;
}else{
u64 iVal;
*piOff = i + sqlite3Fts5GetVarint(&pBuf->p[i], &iVal);
*piRowid += iVal;
}
}
/*
** This is the equivalent of fts5MergePrefixLists() for detail=none mode.
** In this case the buffers consist of a delta-encoded list of rowids only.
*/
static void fts5MergeRowidLists(
Fts5Index *p, /* FTS5 backend object */
Fts5Buffer *p1, /* First list to merge */
int nBuf, /* Number of entries in apBuf[] */
Fts5Buffer *aBuf /* Array of other lists to merge into p1 */
){
int i1 = 0;
int i2 = 0;
i64 iRowid1 = 0;
i64 iRowid2 = 0;
i64 iOut = 0;
Fts5Buffer *p2 = &aBuf[0];
Fts5Buffer out;
(void)nBuf;
memset(&out, 0, sizeof(out));
assert( nBuf==1 );
sqlite3Fts5BufferSize(&p->rc, &out, p1->n + p2->n);
if( p->rc ) return;
fts5NextRowid(p1, &i1, &iRowid1);
fts5NextRowid(p2, &i2, &iRowid2);
while( i1>=0 || i2>=0 ){
if( i1>=0 && (i2<0 || iRowid1iOut );
fts5BufferSafeAppendVarint(&out, iRowid1 - iOut);
iOut = iRowid1;
fts5NextRowid(p1, &i1, &iRowid1);
}else{
assert( iOut==0 || iRowid2>iOut );
fts5BufferSafeAppendVarint(&out, iRowid2 - iOut);
iOut = iRowid2;
if( i1>=0 && iRowid1==iRowid2 ){
fts5NextRowid(p1, &i1, &iRowid1);
}
fts5NextRowid(p2, &i2, &iRowid2);
}
}
fts5BufferSwap(&out, p1);
fts5BufferFree(&out);
}
typedef struct PrefixMerger PrefixMerger;
struct PrefixMerger {
Fts5DoclistIter iter; /* Doclist iterator */
i64 iPos; /* For iterating through a position list */
int iOff;
u8 *aPos;
PrefixMerger *pNext; /* Next in docid/poslist order */
};
static void fts5PrefixMergerInsertByRowid(
PrefixMerger **ppHead,
PrefixMerger *p
){
if( p->iter.aPoslist ){
PrefixMerger **pp = ppHead;
while( *pp && p->iter.iRowid>(*pp)->iter.iRowid ){
pp = &(*pp)->pNext;
}
p->pNext = *pp;
*pp = p;
}
}
static void fts5PrefixMergerInsertByPosition(
PrefixMerger **ppHead,
PrefixMerger *p
){
if( p->iPos>=0 ){
PrefixMerger **pp = ppHead;
while( *pp && p->iPos>(*pp)->iPos ){
pp = &(*pp)->pNext;
}
p->pNext = *pp;
*pp = p;
}
}
/*
** Array aBuf[] contains nBuf doclists. These are all merged in with the
** doclist in buffer p1.
*/
static void fts5MergePrefixLists(
Fts5Index *p, /* FTS5 backend object */
Fts5Buffer *p1, /* First list to merge */
int nBuf, /* Number of buffers in array aBuf[] */
Fts5Buffer *aBuf /* Other lists to merge in */
){
#define fts5PrefixMergerNextPosition(p) \
sqlite3Fts5PoslistNext64((p)->aPos,(p)->iter.nPoslist,&(p)->iOff,&(p)->iPos)
#define FTS5_MERGE_NLIST 16
PrefixMerger aMerger[FTS5_MERGE_NLIST];
PrefixMerger *pHead = 0;
int i;
int nOut = 0;
Fts5Buffer out = {0, 0, 0};
Fts5Buffer tmp = {0, 0, 0};
i64 iLastRowid = 0;
/* Initialize a doclist-iterator for each input buffer. Arrange them in
** a linked-list starting at pHead in ascending order of rowid. Avoid
** linking any iterators already at EOF into the linked list at all. */
assert( nBuf+1<=(int)(sizeof(aMerger)/sizeof(aMerger[0])) );
memset(aMerger, 0, sizeof(PrefixMerger)*(nBuf+1));
pHead = &aMerger[nBuf];
fts5DoclistIterInit(p1, &pHead->iter);
for(i=0; in + 9 + 10*nBuf;
/* The maximum size of the output is equal to the sum of the
** input sizes + 1 varint (9 bytes). The extra varint is because if the
** first rowid in one input is a large negative number, and the first in
** the other a non-negative number, the delta for the non-negative
** number will be larger on disk than the literal integer value
** was.
**
** Or, if the input position-lists are corrupt, then the output might
** include up to (nBuf+1) extra 10-byte positions created by interpreting -1
** (the value PoslistNext64() uses for EOF) as a position and appending
** it to the output. This can happen at most once for each input
** position-list, hence (nBuf+1) 10 byte paddings. */
if( sqlite3Fts5BufferSize(&p->rc, &out, nOut) ) return;
while( pHead ){
fts5MergeAppendDocid(&out, iLastRowid, pHead->iter.iRowid);
if( pHead->pNext && iLastRowid==pHead->pNext->iter.iRowid ){
/* Merge data from two or more poslists */
i64 iPrev = 0;
int nTmp = FTS5_DATA_ZERO_PADDING;
int nMerge = 0;
PrefixMerger *pSave = pHead;
PrefixMerger *pThis = 0;
int nTail = 0;
pHead = 0;
while( pSave && pSave->iter.iRowid==iLastRowid ){
PrefixMerger *pNext = pSave->pNext;
pSave->iOff = 0;
pSave->iPos = 0;
pSave->aPos = &pSave->iter.aPoslist[pSave->iter.nSize];
fts5PrefixMergerNextPosition(pSave);
nTmp += pSave->iter.nPoslist + 10;
nMerge++;
fts5PrefixMergerInsertByPosition(&pHead, pSave);
pSave = pNext;
}
if( pHead==0 || pHead->pNext==0 ){
p->rc = FTS5_CORRUPT;
break;
}
/* See the earlier comment in this function for an explanation of why
** corrupt input position lists might cause the output to consume
** at most nMerge*10 bytes of unexpected space. */
if( sqlite3Fts5BufferSize(&p->rc, &tmp, nTmp+nMerge*10) ){
break;
}
fts5BufferZero(&tmp);
pThis = pHead;
pHead = pThis->pNext;
sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, pThis->iPos);
fts5PrefixMergerNextPosition(pThis);
fts5PrefixMergerInsertByPosition(&pHead, pThis);
while( pHead->pNext ){
pThis = pHead;
if( pThis->iPos!=iPrev ){
sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, pThis->iPos);
}
fts5PrefixMergerNextPosition(pThis);
pHead = pThis->pNext;
fts5PrefixMergerInsertByPosition(&pHead, pThis);
}
if( pHead->iPos!=iPrev ){
sqlite3Fts5PoslistSafeAppend(&tmp, &iPrev, pHead->iPos);
}
nTail = pHead->iter.nPoslist - pHead->iOff;
/* WRITEPOSLISTSIZE */
assert_nc( tmp.n+nTail<=nTmp );
assert( tmp.n+nTail<=nTmp+nMerge*10 );
if( tmp.n+nTail>nTmp-FTS5_DATA_ZERO_PADDING ){
if( p->rc==SQLITE_OK ) p->rc = FTS5_CORRUPT;
break;
}
fts5BufferSafeAppendVarint(&out, (tmp.n+nTail) * 2);
fts5BufferSafeAppendBlob(&out, tmp.p, tmp.n);
if( nTail>0 ){
fts5BufferSafeAppendBlob(&out, &pHead->aPos[pHead->iOff], nTail);
}
pHead = pSave;
for(i=0; iiter.aPoslist && pX->iter.iRowid==iLastRowid ){
fts5DoclistIterNext(&pX->iter);
fts5PrefixMergerInsertByRowid(&pHead, pX);
}
}
}else{
/* Copy poslist from pHead to output */
PrefixMerger *pThis = pHead;
Fts5DoclistIter *pI = &pThis->iter;
fts5BufferSafeAppendBlob(&out, pI->aPoslist, pI->nPoslist+pI->nSize);
fts5DoclistIterNext(pI);
pHead = pThis->pNext;
fts5PrefixMergerInsertByRowid(&pHead, pThis);
}
}
fts5BufferFree(p1);
fts5BufferFree(&tmp);
memset(&out.p[out.n], 0, FTS5_DATA_ZERO_PADDING);
*p1 = out;
}
static void fts5SetupPrefixIter(
Fts5Index *p, /* Index to read from */
int bDesc, /* True for "ORDER BY rowid DESC" */
int iIdx, /* Index to scan for data */
u8 *pToken, /* Buffer containing prefix to match */
int nToken, /* Size of buffer pToken in bytes */
Fts5Colset *pColset, /* Restrict matches to these columns */
Fts5Iter **ppIter /* OUT: New iterator */
){
Fts5Structure *pStruct;
Fts5Buffer *aBuf;
int nBuf = 32;
int nMerge = 1;
void (*xMerge)(Fts5Index*, Fts5Buffer*, int, Fts5Buffer*);
void (*xAppend)(Fts5Index*, u64, Fts5Iter*, Fts5Buffer*);
if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){
xMerge = fts5MergeRowidLists;
xAppend = fts5AppendRowid;
}else{
nMerge = FTS5_MERGE_NLIST-1;
nBuf = nMerge*8; /* Sufficient to merge (16^8)==(2^32) lists */
xMerge = fts5MergePrefixLists;
xAppend = fts5AppendPoslist;
}
aBuf = (Fts5Buffer*)fts5IdxMalloc(p, sizeof(Fts5Buffer)*nBuf);
pStruct = fts5StructureRead(p);
assert( p->rc!=SQLITE_OK || (aBuf && pStruct) );
if( p->rc==SQLITE_OK ){
const int flags = FTS5INDEX_QUERY_SCAN
| FTS5INDEX_QUERY_SKIPEMPTY
| FTS5INDEX_QUERY_NOOUTPUT;
int i;
i64 iLastRowid = 0;
Fts5Iter *p1 = 0; /* Iterator used to gather data from index */
Fts5Data *pData;
Fts5Buffer doclist;
int bNewTerm = 1;
memset(&doclist, 0, sizeof(doclist));
/* If iIdx is non-zero, then it is the number of a prefix-index for
** prefixes 1 character longer than the prefix being queried for. That
** index contains all the doclists required, except for the one
** corresponding to the prefix itself. That one is extracted from the
** main term index here. */
if( iIdx!=0 ){
int dummy = 0;
const int f2 = FTS5INDEX_QUERY_SKIPEMPTY|FTS5INDEX_QUERY_NOOUTPUT;
pToken[0] = FTS5_MAIN_PREFIX;
fts5MultiIterNew(p, pStruct, f2, pColset, pToken, nToken, -1, 0, &p1);
fts5IterSetOutputCb(&p->rc, p1);
for(;
fts5MultiIterEof(p, p1)==0;
fts5MultiIterNext2(p, p1, &dummy)
){
Fts5SegIter *pSeg = &p1->aSeg[ p1->aFirst[1].iFirst ];
p1->xSetOutputs(p1, pSeg);
if( p1->base.nData ){
xAppend(p, (u64)p1->base.iRowid-(u64)iLastRowid, p1, &doclist);
iLastRowid = p1->base.iRowid;
}
}
fts5MultiIterFree(p1);
}
pToken[0] = FTS5_MAIN_PREFIX + iIdx;
fts5MultiIterNew(p, pStruct, flags, pColset, pToken, nToken, -1, 0, &p1);
fts5IterSetOutputCb(&p->rc, p1);
for( /* no-op */ ;
fts5MultiIterEof(p, p1)==0;
fts5MultiIterNext2(p, p1, &bNewTerm)
){
Fts5SegIter *pSeg = &p1->aSeg[ p1->aFirst[1].iFirst ];
int nTerm = pSeg->term.n;
const u8 *pTerm = pSeg->term.p;
p1->xSetOutputs(p1, pSeg);
assert_nc( memcmp(pToken, pTerm, MIN(nToken, nTerm))<=0 );
if( bNewTerm ){
if( nTermbase.nData==0 ) continue;
if( p1->base.iRowid<=iLastRowid && doclist.n>0 ){
for(i=0; p->rc==SQLITE_OK && doclist.n; i++){
int i1 = i*nMerge;
int iStore;
assert( i1+nMerge<=nBuf );
for(iStore=i1; iStorebase.iRowid-(u64)iLastRowid, p1, &doclist);
iLastRowid = p1->base.iRowid;
}
assert( (nBuf%nMerge)==0 );
for(i=0; irc==SQLITE_OK ){
xMerge(p, &doclist, nMerge, &aBuf[i]);
}
for(iFree=i; iFreep = (u8*)&pData[1];
pData->nn = pData->szLeaf = doclist.n;
if( doclist.n ) memcpy(pData->p, doclist.p, doclist.n);
fts5MultiIterNew2(p, pData, bDesc, ppIter);
}
fts5BufferFree(&doclist);
}
fts5StructureRelease(pStruct);
sqlite3_free(aBuf);
}
/*
** Indicate that all subsequent calls to sqlite3Fts5IndexWrite() pertain
** to the document with rowid iRowid.
*/
static int sqlite3Fts5IndexBeginWrite(Fts5Index *p, int bDelete, i64 iRowid){
assert( p->rc==SQLITE_OK );
/* Allocate the hash table if it has not already been allocated */
if( p->pHash==0 ){
p->rc = sqlite3Fts5HashNew(p->pConfig, &p->pHash, &p->nPendingData);
}
/* Flush the hash table to disk if required */
if( iRowidiWriteRowid
|| (iRowid==p->iWriteRowid && p->bDelete==0)
|| (p->nPendingData > p->pConfig->nHashSize)
){
fts5IndexFlush(p);
}
p->iWriteRowid = iRowid;
p->bDelete = bDelete;
if( bDelete==0 ){
p->nPendingRow++;
}
return fts5IndexReturn(p);
}
/*
** Commit data to disk.
*/
static int sqlite3Fts5IndexSync(Fts5Index *p){
assert( p->rc==SQLITE_OK );
fts5IndexFlush(p);
sqlite3Fts5IndexCloseReader(p);
return fts5IndexReturn(p);
}
/*
** Discard any data stored in the in-memory hash tables. Do not write it
** to the database. Additionally, assume that the contents of the %_data
** table may have changed on disk. So any in-memory caches of %_data
** records must be invalidated.
*/
static int sqlite3Fts5IndexRollback(Fts5Index *p){
sqlite3Fts5IndexCloseReader(p);
fts5IndexDiscardData(p);
fts5StructureInvalidate(p);
/* assert( p->rc==SQLITE_OK ); */
return SQLITE_OK;
}
/*
** The %_data table is completely empty when this function is called. This
** function populates it with the initial structure objects for each index,
** and the initial version of the "averages" record (a zero-byte blob).
*/
static int sqlite3Fts5IndexReinit(Fts5Index *p){
Fts5Structure s;
fts5StructureInvalidate(p);
fts5IndexDiscardData(p);
memset(&s, 0, sizeof(Fts5Structure));
if( p->pConfig->bContentlessDelete ){
s.nOriginCntr = 1;
}
fts5DataWrite(p, FTS5_AVERAGES_ROWID, (const u8*)"", 0);
fts5StructureWrite(p, &s);
return fts5IndexReturn(p);
}
/*
** Open a new Fts5Index handle. If the bCreate argument is true, create
** and initialize the underlying %_data table.
**
** If successful, set *pp to point to the new object and return SQLITE_OK.
** Otherwise, set *pp to NULL and return an SQLite error code.
*/
static int sqlite3Fts5IndexOpen(
Fts5Config *pConfig,
int bCreate,
Fts5Index **pp,
char **pzErr
){
int rc = SQLITE_OK;
Fts5Index *p; /* New object */
*pp = p = (Fts5Index*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Index));
if( rc==SQLITE_OK ){
p->pConfig = pConfig;
p->nWorkUnit = FTS5_WORK_UNIT;
p->zDataTbl = sqlite3Fts5Mprintf(&rc, "%s_data", pConfig->zName);
if( p->zDataTbl && bCreate ){
rc = sqlite3Fts5CreateTable(
pConfig, "data", "id INTEGER PRIMARY KEY, block BLOB", 0, pzErr
);
if( rc==SQLITE_OK ){
rc = sqlite3Fts5CreateTable(pConfig, "idx",
"segid, term, pgno, PRIMARY KEY(segid, term)",
1, pzErr
);
}
if( rc==SQLITE_OK ){
rc = sqlite3Fts5IndexReinit(p);
}
}
}
assert( rc!=SQLITE_OK || p->rc==SQLITE_OK );
if( rc ){
sqlite3Fts5IndexClose(p);
*pp = 0;
}
return rc;
}
/*
** Close a handle opened by an earlier call to sqlite3Fts5IndexOpen().
*/
static int sqlite3Fts5IndexClose(Fts5Index *p){
int rc = SQLITE_OK;
if( p ){
assert( p->pReader==0 );
fts5StructureInvalidate(p);
sqlite3_finalize(p->pWriter);
sqlite3_finalize(p->pDeleter);
sqlite3_finalize(p->pIdxWriter);
sqlite3_finalize(p->pIdxDeleter);
sqlite3_finalize(p->pIdxSelect);
sqlite3_finalize(p->pIdxNextSelect);
sqlite3_finalize(p->pDataVersion);
sqlite3_finalize(p->pDeleteFromIdx);
sqlite3Fts5HashFree(p->pHash);
sqlite3_free(p->zDataTbl);
sqlite3_free(p);
}
return rc;
}
/*
** Argument p points to a buffer containing utf-8 text that is n bytes in
** size. Return the number of bytes in the nChar character prefix of the
** buffer, or 0 if there are less than nChar characters in total.
*/
static int sqlite3Fts5IndexCharlenToBytelen(
const char *p,
int nByte,
int nChar
){
int n = 0;
int i;
for(i=0; i=nByte ) return 0; /* Input contains fewer than nChar chars */
if( (unsigned char)p[n++]>=0xc0 ){
if( n>=nByte ) return 0;
while( (p[n] & 0xc0)==0x80 ){
n++;
if( n>=nByte ){
if( i+1==nChar ) break;
return 0;
}
}
}
}
return n;
}
/*
** pIn is a UTF-8 encoded string, nIn bytes in size. Return the number of
** unicode characters in the string.
*/
static int fts5IndexCharlen(const char *pIn, int nIn){
int nChar = 0;
int i = 0;
while( i=0xc0 ){
while( i delete) */
int iPos, /* Position of token within column */
const char *pToken, int nToken /* Token to add or remove to or from index */
){
int i; /* Used to iterate through indexes */
int rc = SQLITE_OK; /* Return code */
Fts5Config *pConfig = p->pConfig;
assert( p->rc==SQLITE_OK );
assert( (iCol<0)==p->bDelete );
/* Add the entry to the main terms index. */
rc = sqlite3Fts5HashWrite(
p->pHash, p->iWriteRowid, iCol, iPos, FTS5_MAIN_PREFIX, pToken, nToken
);
for(i=0; inPrefix && rc==SQLITE_OK; i++){
const int nChar = pConfig->aPrefix[i];
int nByte = sqlite3Fts5IndexCharlenToBytelen(pToken, nToken, nChar);
if( nByte ){
rc = sqlite3Fts5HashWrite(p->pHash,
p->iWriteRowid, iCol, iPos, (char)(FTS5_MAIN_PREFIX+i+1), pToken,
nByte
);
}
}
return rc;
}
/*
** pToken points to a buffer of size nToken bytes containing a search
** term, including the index number at the start, used on a tokendata=1
** table. This function returns true if the term in buffer pBuf matches
** token pToken/nToken.
*/
static int fts5IsTokendataPrefix(
Fts5Buffer *pBuf,
const u8 *pToken,
int nToken
){
return (
pBuf->n>=nToken
&& 0==memcmp(pBuf->p, pToken, nToken)
&& (pBuf->n==nToken || pBuf->p[nToken]==0x00)
);
}
/*
** Ensure the segment-iterator passed as the only argument points to EOF.
*/
static void fts5SegIterSetEOF(Fts5SegIter *pSeg){
fts5DataRelease(pSeg->pLeaf);
pSeg->pLeaf = 0;
}
/*
** Usually, a tokendata=1 iterator (struct Fts5TokenDataIter) accumulates an
** array of these for each row it visits. Or, for an iterator used by an
** "ORDER BY rank" query, it accumulates an array of these for the entire
** query.
**
** Each instance in the array indicates the iterator (and therefore term)
** associated with position iPos of rowid iRowid. This is used by the
** xInstToken() API.
*/
struct Fts5TokenDataMap {
i64 iRowid; /* Row this token is located in */
i64 iPos; /* Position of token */
int iIter; /* Iterator token was read from */
};
/*
** An object used to supplement Fts5Iter for tokendata=1 iterators.
*/
struct Fts5TokenDataIter {
int nIter;
int nIterAlloc;
int nMap;
int nMapAlloc;
Fts5TokenDataMap *aMap;
Fts5PoslistReader *aPoslistReader;
int *aPoslistToIter;
Fts5Iter *apIter[1];
};
/*
** This function appends iterator pAppend to Fts5TokenDataIter pIn and
** returns the result.
*/
static Fts5TokenDataIter *fts5AppendTokendataIter(
Fts5Index *p, /* Index object (for error code) */
Fts5TokenDataIter *pIn, /* Current Fts5TokenDataIter struct */
Fts5Iter *pAppend /* Append this iterator */
){
Fts5TokenDataIter *pRet = pIn;
if( p->rc==SQLITE_OK ){
if( pIn==0 || pIn->nIter==pIn->nIterAlloc ){
int nAlloc = pIn ? pIn->nIterAlloc*2 : 16;
int nByte = nAlloc * sizeof(Fts5Iter*) + sizeof(Fts5TokenDataIter);
Fts5TokenDataIter *pNew = (Fts5TokenDataIter*)sqlite3_realloc(pIn, nByte);
if( pNew==0 ){
p->rc = SQLITE_NOMEM;
}else{
if( pIn==0 ) memset(pNew, 0, nByte);
pRet = pNew;
pNew->nIterAlloc = nAlloc;
}
}
}
if( p->rc ){
sqlite3Fts5IterClose((Fts5IndexIter*)pAppend);
}else{
pRet->apIter[pRet->nIter++] = pAppend;
}
assert( pRet==0 || pRet->nIter<=pRet->nIterAlloc );
return pRet;
}
/*
** Delete an Fts5TokenDataIter structure and its contents.
*/
static void fts5TokendataIterDelete(Fts5TokenDataIter *pSet){
if( pSet ){
int ii;
for(ii=0; iinIter; ii++){
fts5MultiIterFree(pSet->apIter[ii]);
}
sqlite3_free(pSet->aPoslistReader);
sqlite3_free(pSet->aMap);
sqlite3_free(pSet);
}
}
/*
** Append a mapping to the token-map belonging to object pT.
*/
static void fts5TokendataIterAppendMap(
Fts5Index *p,
Fts5TokenDataIter *pT,
int iIter,
i64 iRowid,
i64 iPos
){
if( p->rc==SQLITE_OK ){
if( pT->nMap==pT->nMapAlloc ){
int nNew = pT->nMapAlloc ? pT->nMapAlloc*2 : 64;
int nByte = nNew * sizeof(Fts5TokenDataMap);
Fts5TokenDataMap *aNew;
aNew = (Fts5TokenDataMap*)sqlite3_realloc(pT->aMap, nByte);
if( aNew==0 ){
p->rc = SQLITE_NOMEM;
return;
}
pT->aMap = aNew;
pT->nMapAlloc = nNew;
}
pT->aMap[pT->nMap].iRowid = iRowid;
pT->aMap[pT->nMap].iPos = iPos;
pT->aMap[pT->nMap].iIter = iIter;
pT->nMap++;
}
}
/*
** The iterator passed as the only argument must be a tokendata=1 iterator
** (pIter->pTokenDataIter!=0). This function sets the iterator output
** variables (pIter->base.*) according to the contents of the current
** row.
*/
static void fts5IterSetOutputsTokendata(Fts5Iter *pIter){
int ii;
int nHit = 0;
i64 iRowid = SMALLEST_INT64;
int iMin = 0;
Fts5TokenDataIter *pT = pIter->pTokenDataIter;
pIter->base.nData = 0;
pIter->base.pData = 0;
for(ii=0; iinIter; ii++){
Fts5Iter *p = pT->apIter[ii];
if( p->base.bEof==0 ){
if( nHit==0 || p->base.iRowidbase.iRowid;
nHit = 1;
pIter->base.pData = p->base.pData;
pIter->base.nData = p->base.nData;
iMin = ii;
}else if( p->base.iRowid==iRowid ){
nHit++;
}
}
}
if( nHit==0 ){
pIter->base.bEof = 1;
}else{
int eDetail = pIter->pIndex->pConfig->eDetail;
pIter->base.bEof = 0;
pIter->base.iRowid = iRowid;
if( nHit==1 && eDetail==FTS5_DETAIL_FULL ){
fts5TokendataIterAppendMap(pIter->pIndex, pT, iMin, iRowid, -1);
}else
if( nHit>1 && eDetail!=FTS5_DETAIL_NONE ){
int nReader = 0;
int nByte = 0;
i64 iPrev = 0;
/* Allocate array of iterators if they are not already allocated. */
if( pT->aPoslistReader==0 ){
pT->aPoslistReader = (Fts5PoslistReader*)sqlite3Fts5MallocZero(
&pIter->pIndex->rc,
pT->nIter * (sizeof(Fts5PoslistReader) + sizeof(int))
);
if( pT->aPoslistReader==0 ) return;
pT->aPoslistToIter = (int*)&pT->aPoslistReader[pT->nIter];
}
/* Populate an iterator for each poslist that will be merged */
for(ii=0; iinIter; ii++){
Fts5Iter *p = pT->apIter[ii];
if( iRowid==p->base.iRowid ){
pT->aPoslistToIter[nReader] = ii;
sqlite3Fts5PoslistReaderInit(
p->base.pData, p->base.nData, &pT->aPoslistReader[nReader++]
);
nByte += p->base.nData;
}
}
/* Ensure the output buffer is large enough */
if( fts5BufferGrow(&pIter->pIndex->rc, &pIter->poslist, nByte+nHit*10) ){
return;
}
/* Ensure the token-mapping is large enough */
if( eDetail==FTS5_DETAIL_FULL && pT->nMapAlloc<(pT->nMap + nByte) ){
int nNew = (pT->nMapAlloc + nByte) * 2;
Fts5TokenDataMap *aNew = (Fts5TokenDataMap*)sqlite3_realloc(
pT->aMap, nNew*sizeof(Fts5TokenDataMap)
);
if( aNew==0 ){
pIter->pIndex->rc = SQLITE_NOMEM;
return;
}
pT->aMap = aNew;
pT->nMapAlloc = nNew;
}
pIter->poslist.n = 0;
while( 1 ){
i64 iMinPos = LARGEST_INT64;
/* Find smallest position */
iMin = 0;
for(ii=0; iiaPoslistReader[ii];
if( pReader->bEof==0 ){
if( pReader->iPosiPos;
iMin = ii;
}
}
}
/* If all readers were at EOF, break out of the loop. */
if( iMinPos==LARGEST_INT64 ) break;
sqlite3Fts5PoslistSafeAppend(&pIter->poslist, &iPrev, iMinPos);
sqlite3Fts5PoslistReaderNext(&pT->aPoslistReader[iMin]);
if( eDetail==FTS5_DETAIL_FULL ){
pT->aMap[pT->nMap].iPos = iMinPos;
pT->aMap[pT->nMap].iIter = pT->aPoslistToIter[iMin];
pT->aMap[pT->nMap].iRowid = iRowid;
pT->nMap++;
}
}
pIter->base.pData = pIter->poslist.p;
pIter->base.nData = pIter->poslist.n;
}
}
}
/*
** The iterator passed as the only argument must be a tokendata=1 iterator
** (pIter->pTokenDataIter!=0). This function advances the iterator. If
** argument bFrom is false, then the iterator is advanced to the next
** entry. Or, if bFrom is true, it is advanced to the first entry with
** a rowid of iFrom or greater.
*/
static void fts5TokendataIterNext(Fts5Iter *pIter, int bFrom, i64 iFrom){
int ii;
Fts5TokenDataIter *pT = pIter->pTokenDataIter;
for(ii=0; iinIter; ii++){
Fts5Iter *p = pT->apIter[ii];
if( p->base.bEof==0
&& (p->base.iRowid==pIter->base.iRowid || (bFrom && p->base.iRowidpIndex, p, bFrom, iFrom);
while( bFrom && p->base.bEof==0
&& p->base.iRowidpIndex->rc==SQLITE_OK
){
fts5MultiIterNext(p->pIndex, p, 0, 0);
}
}
}
fts5IterSetOutputsTokendata(pIter);
}
/*
** If the segment-iterator passed as the first argument is at EOF, then
** set pIter->term to a copy of buffer pTerm.
*/
static void fts5TokendataSetTermIfEof(Fts5Iter *pIter, Fts5Buffer *pTerm){
if( pIter && pIter->aSeg[0].pLeaf==0 ){
fts5BufferSet(&pIter->pIndex->rc, &pIter->aSeg[0].term, pTerm->n, pTerm->p);
}
}
/*
** This function sets up an iterator to use for a non-prefix query on a
** tokendata=1 table.
*/
static Fts5Iter *fts5SetupTokendataIter(
Fts5Index *p, /* FTS index to query */
const u8 *pToken, /* Buffer containing query term */
int nToken, /* Size of buffer pToken in bytes */
Fts5Colset *pColset /* Colset to filter on */
){
Fts5Iter *pRet = 0;
Fts5TokenDataIter *pSet = 0;
Fts5Structure *pStruct = 0;
const int flags = FTS5INDEX_QUERY_SCANONETERM | FTS5INDEX_QUERY_SCAN;
Fts5Buffer bSeek = {0, 0, 0};
Fts5Buffer *pSmall = 0;
fts5IndexFlush(p);
pStruct = fts5StructureRead(p);
while( p->rc==SQLITE_OK ){
Fts5Iter *pPrev = pSet ? pSet->apIter[pSet->nIter-1] : 0;
Fts5Iter *pNew = 0;
Fts5SegIter *pNewIter = 0;
Fts5SegIter *pPrevIter = 0;
int iLvl, iSeg, ii;
pNew = fts5MultiIterAlloc(p, pStruct->nSegment);
if( pSmall ){
fts5BufferSet(&p->rc, &bSeek, pSmall->n, pSmall->p);
fts5BufferAppendBlob(&p->rc, &bSeek, 1, (const u8*)"\0");
}else{
fts5BufferSet(&p->rc, &bSeek, nToken, pToken);
}
if( p->rc ){
sqlite3Fts5IterClose((Fts5IndexIter*)pNew);
break;
}
pNewIter = &pNew->aSeg[0];
pPrevIter = (pPrev ? &pPrev->aSeg[0] : 0);
for(iLvl=0; iLvlnLevel; iLvl++){
for(iSeg=pStruct->aLevel[iLvl].nSeg-1; iSeg>=0; iSeg--){
Fts5StructureSegment *pSeg = &pStruct->aLevel[iLvl].aSeg[iSeg];
int bDone = 0;
if( pPrevIter ){
if( fts5BufferCompare(pSmall, &pPrevIter->term) ){
memcpy(pNewIter, pPrevIter, sizeof(Fts5SegIter));
memset(pPrevIter, 0, sizeof(Fts5SegIter));
bDone = 1;
}else if( pPrevIter->iEndofDoclist>pPrevIter->pLeaf->szLeaf ){
fts5SegIterNextInit(p,(const char*)bSeek.p,bSeek.n-1,pSeg,pNewIter);
bDone = 1;
}
}
if( bDone==0 ){
fts5SegIterSeekInit(p, bSeek.p, bSeek.n, flags, pSeg, pNewIter);
}
if( pPrevIter ){
if( pPrevIter->pTombArray ){
pNewIter->pTombArray = pPrevIter->pTombArray;
pNewIter->pTombArray->nRef++;
}
}else{
fts5SegIterAllocTombstone(p, pNewIter);
}
pNewIter++;
if( pPrevIter ) pPrevIter++;
if( p->rc ) break;
}
}
fts5TokendataSetTermIfEof(pPrev, pSmall);
pNew->bSkipEmpty = 1;
pNew->pColset = pColset;
fts5IterSetOutputCb(&p->rc, pNew);
/* Loop through all segments in the new iterator. Find the smallest
** term that any segment-iterator points to. Iterator pNew will be
** used for this term. Also, set any iterator that points to a term that
** does not match pToken/nToken to point to EOF */
pSmall = 0;
for(ii=0; iinSeg; ii++){
Fts5SegIter *pII = &pNew->aSeg[ii];
if( 0==fts5IsTokendataPrefix(&pII->term, pToken, nToken) ){
fts5SegIterSetEOF(pII);
}
if( pII->pLeaf && (!pSmall || fts5BufferCompare(pSmall, &pII->term)>0) ){
pSmall = &pII->term;
}
}
/* If pSmall is still NULL at this point, then the new iterator does
** not point to any terms that match the query. So delete it and break
** out of the loop - all required iterators have been collected. */
if( pSmall==0 ){
sqlite3Fts5IterClose((Fts5IndexIter*)pNew);
break;
}
/* Append this iterator to the set and continue. */
pSet = fts5AppendTokendataIter(p, pSet, pNew);
}
if( p->rc==SQLITE_OK && pSet ){
int ii;
for(ii=0; iinIter; ii++){
Fts5Iter *pIter = pSet->apIter[ii];
int iSeg;
for(iSeg=0; iSegnSeg; iSeg++){
pIter->aSeg[iSeg].flags |= FTS5_SEGITER_ONETERM;
}
fts5MultiIterFinishSetup(p, pIter);
}
}
if( p->rc==SQLITE_OK ){
pRet = fts5MultiIterAlloc(p, 0);
}
if( pRet ){
pRet->pTokenDataIter = pSet;
if( pSet ){
fts5IterSetOutputsTokendata(pRet);
}else{
pRet->base.bEof = 1;
}
}else{
fts5TokendataIterDelete(pSet);
}
fts5StructureRelease(pStruct);
fts5BufferFree(&bSeek);
return pRet;
}
/*
** Open a new iterator to iterate though all rowid that match the
** specified token or token prefix.
*/
static int sqlite3Fts5IndexQuery(
Fts5Index *p, /* FTS index to query */
const char *pToken, int nToken, /* Token (or prefix) to query for */
int flags, /* Mask of FTS5INDEX_QUERY_X flags */
Fts5Colset *pColset, /* Match these columns only */
Fts5IndexIter **ppIter /* OUT: New iterator object */
){
Fts5Config *pConfig = p->pConfig;
Fts5Iter *pRet = 0;
Fts5Buffer buf = {0, 0, 0};
/* If the QUERY_SCAN flag is set, all other flags must be clear. */
assert( (flags & FTS5INDEX_QUERY_SCAN)==0 || flags==FTS5INDEX_QUERY_SCAN );
if( sqlite3Fts5BufferSize(&p->rc, &buf, nToken+1)==0 ){
int iIdx = 0; /* Index to search */
int iPrefixIdx = 0; /* +1 prefix index */
int bTokendata = pConfig->bTokendata;
if( nToken>0 ) memcpy(&buf.p[1], pToken, nToken);
if( flags & (FTS5INDEX_QUERY_NOTOKENDATA|FTS5INDEX_QUERY_SCAN) ){
bTokendata = 0;
}
/* Figure out which index to search and set iIdx accordingly. If this
** is a prefix query for which there is no prefix index, set iIdx to
** greater than pConfig->nPrefix to indicate that the query will be
** satisfied by scanning multiple terms in the main index.
**
** If the QUERY_TEST_NOIDX flag was specified, then this must be a
** prefix-query. Instead of using a prefix-index (if one exists),
** evaluate the prefix query using the main FTS index. This is used
** for internal sanity checking by the integrity-check in debug
** mode only. */
#ifdef SQLITE_DEBUG
if( pConfig->bPrefixIndex==0 || (flags & FTS5INDEX_QUERY_TEST_NOIDX) ){
assert( flags & FTS5INDEX_QUERY_PREFIX );
iIdx = 1+pConfig->nPrefix;
}else
#endif
if( flags & FTS5INDEX_QUERY_PREFIX ){
int nChar = fts5IndexCharlen(pToken, nToken);
for(iIdx=1; iIdx<=pConfig->nPrefix; iIdx++){
int nIdxChar = pConfig->aPrefix[iIdx-1];
if( nIdxChar==nChar ) break;
if( nIdxChar==nChar+1 ) iPrefixIdx = iIdx;
}
}
if( bTokendata && iIdx==0 ){
buf.p[0] = '0';
pRet = fts5SetupTokendataIter(p, buf.p, nToken+1, pColset);
}else if( iIdx<=pConfig->nPrefix ){
/* Straight index lookup */
Fts5Structure *pStruct = fts5StructureRead(p);
buf.p[0] = (u8)(FTS5_MAIN_PREFIX + iIdx);
if( pStruct ){
fts5MultiIterNew(p, pStruct, flags | FTS5INDEX_QUERY_SKIPEMPTY,
pColset, buf.p, nToken+1, -1, 0, &pRet
);
fts5StructureRelease(pStruct);
}
}else{
/* Scan multiple terms in the main index */
int bDesc = (flags & FTS5INDEX_QUERY_DESC)!=0;
fts5SetupPrefixIter(p, bDesc, iPrefixIdx, buf.p, nToken+1, pColset,&pRet);
if( pRet==0 ){
assert( p->rc!=SQLITE_OK );
}else{
assert( pRet->pColset==0 );
fts5IterSetOutputCb(&p->rc, pRet);
if( p->rc==SQLITE_OK ){
Fts5SegIter *pSeg = &pRet->aSeg[pRet->aFirst[1].iFirst];
if( pSeg->pLeaf ) pRet->xSetOutputs(pRet, pSeg);
}
}
}
if( p->rc ){
sqlite3Fts5IterClose((Fts5IndexIter*)pRet);
pRet = 0;
sqlite3Fts5IndexCloseReader(p);
}
*ppIter = (Fts5IndexIter*)pRet;
sqlite3Fts5BufferFree(&buf);
}
return fts5IndexReturn(p);
}
/*
** Return true if the iterator passed as the only argument is at EOF.
*/
/*
** Move to the next matching rowid.
*/
static int sqlite3Fts5IterNext(Fts5IndexIter *pIndexIter){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
assert( pIter->pIndex->rc==SQLITE_OK );
if( pIter->pTokenDataIter ){
fts5TokendataIterNext(pIter, 0, 0);
}else{
fts5MultiIterNext(pIter->pIndex, pIter, 0, 0);
}
return fts5IndexReturn(pIter->pIndex);
}
/*
** Move to the next matching term/rowid. Used by the fts5vocab module.
*/
static int sqlite3Fts5IterNextScan(Fts5IndexIter *pIndexIter){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
Fts5Index *p = pIter->pIndex;
assert( pIter->pIndex->rc==SQLITE_OK );
fts5MultiIterNext(p, pIter, 0, 0);
if( p->rc==SQLITE_OK ){
Fts5SegIter *pSeg = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
if( pSeg->pLeaf && pSeg->term.p[0]!=FTS5_MAIN_PREFIX ){
fts5DataRelease(pSeg->pLeaf);
pSeg->pLeaf = 0;
pIter->base.bEof = 1;
}
}
return fts5IndexReturn(pIter->pIndex);
}
/*
** Move to the next matching rowid that occurs at or after iMatch. The
** definition of "at or after" depends on whether this iterator iterates
** in ascending or descending rowid order.
*/
static int sqlite3Fts5IterNextFrom(Fts5IndexIter *pIndexIter, i64 iMatch){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
if( pIter->pTokenDataIter ){
fts5TokendataIterNext(pIter, 1, iMatch);
}else{
fts5MultiIterNextFrom(pIter->pIndex, pIter, iMatch);
}
return fts5IndexReturn(pIter->pIndex);
}
/*
** Return the current term.
*/
static const char *sqlite3Fts5IterTerm(Fts5IndexIter *pIndexIter, int *pn){
int n;
const char *z = (const char*)fts5MultiIterTerm((Fts5Iter*)pIndexIter, &n);
assert_nc( z || n<=1 );
*pn = n-1;
return (z ? &z[1] : 0);
}
/*
** This is used by xInstToken() to access the token at offset iOff, column
** iCol of row iRowid. The token is returned via output variables *ppOut
** and *pnOut. The iterator passed as the first argument must be a tokendata=1
** iterator (pIter->pTokenDataIter!=0).
*/
static int sqlite3Fts5IterToken(
Fts5IndexIter *pIndexIter,
i64 iRowid,
int iCol,
int iOff,
const char **ppOut, int *pnOut
){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
Fts5TokenDataIter *pT = pIter->pTokenDataIter;
Fts5TokenDataMap *aMap = pT->aMap;
i64 iPos = (((i64)iCol)<<32) + iOff;
int i1 = 0;
int i2 = pT->nMap;
int iTest = 0;
while( i2>i1 ){
iTest = (i1 + i2) / 2;
if( aMap[iTest].iRowidiRowid ){
i2 = iTest;
}else{
if( aMap[iTest].iPosiPos ){
i2 = iTest;
}else{
break;
}
}
}
if( i2>i1 ){
Fts5Iter *pMap = pT->apIter[aMap[iTest].iIter];
*ppOut = (const char*)pMap->aSeg[0].term.p+1;
*pnOut = pMap->aSeg[0].term.n-1;
}
return SQLITE_OK;
}
/*
** Clear any existing entries from the token-map associated with the
** iterator passed as the only argument.
*/
static void sqlite3Fts5IndexIterClearTokendata(Fts5IndexIter *pIndexIter){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
if( pIter && pIter->pTokenDataIter ){
pIter->pTokenDataIter->nMap = 0;
}
}
/*
** Set a token-mapping for the iterator passed as the first argument. This
** is used in detail=column or detail=none mode when a token is requested
** using the xInstToken() API. In this case the caller tokenizers the
** current row and configures the token-mapping via multiple calls to this
** function.
*/
static int sqlite3Fts5IndexIterWriteTokendata(
Fts5IndexIter *pIndexIter,
const char *pToken, int nToken,
i64 iRowid, int iCol, int iOff
){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
Fts5TokenDataIter *pT = pIter->pTokenDataIter;
Fts5Index *p = pIter->pIndex;
int ii;
assert( p->pConfig->eDetail!=FTS5_DETAIL_FULL );
assert( pIter->pTokenDataIter );
for(ii=0; iinIter; ii++){
Fts5Buffer *pTerm = &pT->apIter[ii]->aSeg[0].term;
if( nToken==pTerm->n-1 && memcmp(pToken, pTerm->p+1, nToken)==0 ) break;
}
if( iinIter ){
fts5TokendataIterAppendMap(p, pT, ii, iRowid, (((i64)iCol)<<32) + iOff);
}
return fts5IndexReturn(p);
}
/*
** Close an iterator opened by an earlier call to sqlite3Fts5IndexQuery().
*/
static void sqlite3Fts5IterClose(Fts5IndexIter *pIndexIter){
if( pIndexIter ){
Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
Fts5Index *pIndex = pIter->pIndex;
fts5TokendataIterDelete(pIter->pTokenDataIter);
fts5MultiIterFree(pIter);
sqlite3Fts5IndexCloseReader(pIndex);
}
}
/*
** Read and decode the "averages" record from the database.
**
** Parameter anSize must point to an array of size nCol, where nCol is
** the number of user defined columns in the FTS table.
*/
static int sqlite3Fts5IndexGetAverages(Fts5Index *p, i64 *pnRow, i64 *anSize){
int nCol = p->pConfig->nCol;
Fts5Data *pData;
*pnRow = 0;
memset(anSize, 0, sizeof(i64) * nCol);
pData = fts5DataRead(p, FTS5_AVERAGES_ROWID);
if( p->rc==SQLITE_OK && pData->nn ){
int i = 0;
int iCol;
i += fts5GetVarint(&pData->p[i], (u64*)pnRow);
for(iCol=0; inn && iColp[i], (u64*)&anSize[iCol]);
}
}
fts5DataRelease(pData);
return fts5IndexReturn(p);
}
/*
** Replace the current "averages" record with the contents of the buffer
** supplied as the second argument.
*/
static int sqlite3Fts5IndexSetAverages(Fts5Index *p, const u8 *pData, int nData){
assert( p->rc==SQLITE_OK );
fts5DataWrite(p, FTS5_AVERAGES_ROWID, pData, nData);
return fts5IndexReturn(p);
}
/*
** Return the total number of blocks this module has read from the %_data
** table since it was created.
*/
static int sqlite3Fts5IndexReads(Fts5Index *p){
return p->nRead;
}
/*
** Set the 32-bit cookie value stored at the start of all structure
** records to the value passed as the second argument.
**
** Return SQLITE_OK if successful, or an SQLite error code if an error
** occurs.
*/
static int sqlite3Fts5IndexSetCookie(Fts5Index *p, int iNew){
int rc; /* Return code */
Fts5Config *pConfig = p->pConfig; /* Configuration object */
u8 aCookie[4]; /* Binary representation of iNew */
sqlite3_blob *pBlob = 0;
assert( p->rc==SQLITE_OK );
sqlite3Fts5Put32(aCookie, iNew);
rc = sqlite3_blob_open(pConfig->db, pConfig->zDb, p->zDataTbl,
"block", FTS5_STRUCTURE_ROWID, 1, &pBlob
);
if( rc==SQLITE_OK ){
sqlite3_blob_write(pBlob, aCookie, 4, 0);
rc = sqlite3_blob_close(pBlob);
}
return rc;
}
static int sqlite3Fts5IndexLoadConfig(Fts5Index *p){
Fts5Structure *pStruct;
pStruct = fts5StructureRead(p);
fts5StructureRelease(pStruct);
return fts5IndexReturn(p);
}
/*
** Retrieve the origin value that will be used for the segment currently
** being accumulated in the in-memory hash table when it is flushed to
** disk. If successful, SQLITE_OK is returned and (*piOrigin) set to
** the queried value. Or, if an error occurs, an error code is returned
** and the final value of (*piOrigin) is undefined.
*/
static int sqlite3Fts5IndexGetOrigin(Fts5Index *p, i64 *piOrigin){
Fts5Structure *pStruct;
pStruct = fts5StructureRead(p);
if( pStruct ){
*piOrigin = pStruct->nOriginCntr;
fts5StructureRelease(pStruct);
}
return fts5IndexReturn(p);
}
/*
** Buffer pPg contains a page of a tombstone hash table - one of nPg pages
** associated with the same segment. This function adds rowid iRowid to
** the hash table. The caller is required to guarantee that there is at
** least one free slot on the page.
**
** If parameter bForce is false and the hash table is deemed to be full
** (more than half of the slots are occupied), then non-zero is returned
** and iRowid not inserted. Or, if bForce is true or if the hash table page
** is not full, iRowid is inserted and zero returned.
*/
static int fts5IndexTombstoneAddToPage(
Fts5Data *pPg,
int bForce,
int nPg,
u64 iRowid
){
const int szKey = TOMBSTONE_KEYSIZE(pPg);
const int nSlot = TOMBSTONE_NSLOT(pPg);
const int nElem = fts5GetU32(&pPg->p[4]);
int iSlot = (iRowid / nPg) % nSlot;
int nCollide = nSlot;
if( szKey==4 && iRowid>0xFFFFFFFF ) return 2;
if( iRowid==0 ){
pPg->p[1] = 0x01;
return 0;
}
if( bForce==0 && nElem>=(nSlot/2) ){
return 1;
}
fts5PutU32(&pPg->p[4], nElem+1);
if( szKey==4 ){
u32 *aSlot = (u32*)&pPg->p[8];
while( aSlot[iSlot] ){
iSlot = (iSlot + 1) % nSlot;
if( nCollide--==0 ) return 0;
}
fts5PutU32((u8*)&aSlot[iSlot], (u32)iRowid);
}else{
u64 *aSlot = (u64*)&pPg->p[8];
while( aSlot[iSlot] ){
iSlot = (iSlot + 1) % nSlot;
if( nCollide--==0 ) return 0;
}
fts5PutU64((u8*)&aSlot[iSlot], iRowid);
}
return 0;
}
/*
** This function attempts to build a new hash containing all the keys
** currently in the tombstone hash table for segment pSeg. The new
** hash will be stored in the nOut buffers passed in array apOut[].
** All pages of the new hash use key-size szKey (4 or 8).
**
** Return 0 if the hash is successfully rebuilt into the nOut pages.
** Or non-zero if it is not (because one page became overfull). In this
** case the caller should retry with a larger nOut parameter.
**
** Parameter pData1 is page iPg1 of the hash table being rebuilt.
*/
static int fts5IndexTombstoneRehash(
Fts5Index *p,
Fts5StructureSegment *pSeg, /* Segment to rebuild hash of */
Fts5Data *pData1, /* One page of current hash - or NULL */
int iPg1, /* Which page of the current hash is pData1 */
int szKey, /* 4 or 8, the keysize */
int nOut, /* Number of output pages */
Fts5Data **apOut /* Array of output hash pages */
){
int ii;
int res = 0;
/* Initialize the headers of all the output pages */
for(ii=0; iip[0] = szKey;
fts5PutU32(&apOut[ii]->p[4], 0);
}
/* Loop through the current pages of the hash table. */
for(ii=0; res==0 && iinPgTombstone; ii++){
Fts5Data *pData = 0; /* Page ii of the current hash table */
Fts5Data *pFree = 0; /* Free this at the end of the loop */
if( iPg1==ii ){
pData = pData1;
}else{
pFree = pData = fts5DataRead(p, FTS5_TOMBSTONE_ROWID(pSeg->iSegid, ii));
}
if( pData ){
int szKeyIn = TOMBSTONE_KEYSIZE(pData);
int nSlotIn = (pData->nn - 8) / szKeyIn;
int iIn;
for(iIn=0; iInp[8];
if( aSlot[iIn] ) iVal = fts5GetU32((u8*)&aSlot[iIn]);
}else{
u64 *aSlot = (u64*)&pData->p[8];
if( aSlot[iIn] ) iVal = fts5GetU64((u8*)&aSlot[iIn]);
}
/* If iVal is not 0 at this point, insert it into the new hash table */
if( iVal ){
Fts5Data *pPg = apOut[(iVal % nOut)];
res = fts5IndexTombstoneAddToPage(pPg, 0, nOut, iVal);
if( res ) break;
}
}
/* If this is page 0 of the old hash, copy the rowid-0-flag from the
** old hash to the new. */
if( ii==0 ){
apOut[0]->p[1] = pData->p[1];
}
}
fts5DataRelease(pFree);
}
return res;
}
/*
** This is called to rebuild the hash table belonging to segment pSeg.
** If parameter pData1 is not NULL, then one page of the existing hash table
** has already been loaded - pData1, which is page iPg1. The key-size for
** the new hash table is szKey (4 or 8).
**
** If successful, the new hash table is not written to disk. Instead,
** output parameter (*pnOut) is set to the number of pages in the new
** hash table, and (*papOut) to point to an array of buffers containing
** the new page data.
**
** If an error occurs, an error code is left in the Fts5Index object and
** both output parameters set to 0 before returning.
*/
static void fts5IndexTombstoneRebuild(
Fts5Index *p,
Fts5StructureSegment *pSeg, /* Segment to rebuild hash of */
Fts5Data *pData1, /* One page of current hash - or NULL */
int iPg1, /* Which page of the current hash is pData1 */
int szKey, /* 4 or 8, the keysize */
int *pnOut, /* OUT: Number of output pages */
Fts5Data ***papOut /* OUT: Output hash pages */
){
const int MINSLOT = 32;
int nSlotPerPage = MAX(MINSLOT, (p->pConfig->pgsz - 8) / szKey);
int nSlot = 0; /* Number of slots in each output page */
int nOut = 0;
/* Figure out how many output pages (nOut) and how many slots per
** page (nSlot). There are three possibilities:
**
** 1. The hash table does not yet exist. In this case the new hash
** table will consist of a single page with MINSLOT slots.
**
** 2. The hash table exists but is currently a single page. In this
** case an attempt is made to grow the page to accommodate the new
** entry. The page is allowed to grow up to nSlotPerPage (see above)
** slots.
**
** 3. The hash table already consists of more than one page, or of
** a single page already so large that it cannot be grown. In this
** case the new hash consists of (nPg*2+1) pages of nSlotPerPage
** slots each, where nPg is the current number of pages in the
** hash table.
*/
if( pSeg->nPgTombstone==0 ){
/* Case 1. */
nOut = 1;
nSlot = MINSLOT;
}else if( pSeg->nPgTombstone==1 ){
/* Case 2. */
int nElem = (int)fts5GetU32(&pData1->p[4]);
assert( pData1 && iPg1==0 );
nOut = 1;
nSlot = MAX(nElem*4, MINSLOT);
if( nSlot>nSlotPerPage ) nOut = 0;
}
if( nOut==0 ){
/* Case 3. */
nOut = (pSeg->nPgTombstone * 2 + 1);
nSlot = nSlotPerPage;
}
/* Allocate the required array and output pages */
while( 1 ){
int res = 0;
int ii = 0;
int szPage = 0;
Fts5Data **apOut = 0;
/* Allocate space for the new hash table */
assert( nSlot>=MINSLOT );
apOut = (Fts5Data**)sqlite3Fts5MallocZero(&p->rc, sizeof(Fts5Data*) * nOut);
szPage = 8 + nSlot*szKey;
for(ii=0; iirc,
sizeof(Fts5Data)+szPage
);
if( pNew ){
pNew->nn = szPage;
pNew->p = (u8*)&pNew[1];
apOut[ii] = pNew;
}
}
/* Rebuild the hash table. */
if( p->rc==SQLITE_OK ){
res = fts5IndexTombstoneRehash(p, pSeg, pData1, iPg1, szKey, nOut, apOut);
}
if( res==0 ){
if( p->rc ){
fts5IndexFreeArray(apOut, nOut);
apOut = 0;
nOut = 0;
}
*pnOut = nOut;
*papOut = apOut;
break;
}
/* If control flows to here, it was not possible to rebuild the hash
** table. Free all buffers and then try again with more pages. */
assert( p->rc==SQLITE_OK );
fts5IndexFreeArray(apOut, nOut);
nSlot = nSlotPerPage;
nOut = nOut*2 + 1;
}
}
/*
** Add a tombstone for rowid iRowid to segment pSeg.
*/
static void fts5IndexTombstoneAdd(
Fts5Index *p,
Fts5StructureSegment *pSeg,
u64 iRowid
){
Fts5Data *pPg = 0;
int iPg = -1;
int szKey = 0;
int nHash = 0;
Fts5Data **apHash = 0;
p->nContentlessDelete++;
if( pSeg->nPgTombstone>0 ){
iPg = iRowid % pSeg->nPgTombstone;
pPg = fts5DataRead(p, FTS5_TOMBSTONE_ROWID(pSeg->iSegid,iPg));
if( pPg==0 ){
assert( p->rc!=SQLITE_OK );
return;
}
if( 0==fts5IndexTombstoneAddToPage(pPg, 0, pSeg->nPgTombstone, iRowid) ){
fts5DataWrite(p, FTS5_TOMBSTONE_ROWID(pSeg->iSegid,iPg), pPg->p, pPg->nn);
fts5DataRelease(pPg);
return;
}
}
/* Have to rebuild the hash table. First figure out the key-size (4 or 8). */
szKey = pPg ? TOMBSTONE_KEYSIZE(pPg) : 4;
if( iRowid>0xFFFFFFFF ) szKey = 8;
/* Rebuild the hash table */
fts5IndexTombstoneRebuild(p, pSeg, pPg, iPg, szKey, &nHash, &apHash);
assert( p->rc==SQLITE_OK || (nHash==0 && apHash==0) );
/* If all has succeeded, write the new rowid into one of the new hash
** table pages, then write them all out to disk. */
if( nHash ){
int ii = 0;
fts5IndexTombstoneAddToPage(apHash[iRowid % nHash], 1, nHash, iRowid);
for(ii=0; iiiSegid, ii);
fts5DataWrite(p, iTombstoneRowid, apHash[ii]->p, apHash[ii]->nn);
}
pSeg->nPgTombstone = nHash;
fts5StructureWrite(p, p->pStruct);
}
fts5DataRelease(pPg);
fts5IndexFreeArray(apHash, nHash);
}
/*
** Add iRowid to the tombstone list of the segment or segments that contain
** rows from origin iOrigin. Return SQLITE_OK if successful, or an SQLite
** error code otherwise.
*/
static int sqlite3Fts5IndexContentlessDelete(Fts5Index *p, i64 iOrigin, i64 iRowid){
Fts5Structure *pStruct;
pStruct = fts5StructureRead(p);
if( pStruct ){
int bFound = 0; /* True after pSeg->nEntryTombstone incr. */
int iLvl;
for(iLvl=pStruct->nLevel-1; iLvl>=0; iLvl--){
int iSeg;
for(iSeg=pStruct->aLevel[iLvl].nSeg-1; iSeg>=0; iSeg--){
Fts5StructureSegment *pSeg = &pStruct->aLevel[iLvl].aSeg[iSeg];
if( pSeg->iOrigin1<=(u64)iOrigin && pSeg->iOrigin2>=(u64)iOrigin ){
if( bFound==0 ){
pSeg->nEntryTombstone++;
bFound = 1;
}
fts5IndexTombstoneAdd(p, pSeg, iRowid);
}
}
}
fts5StructureRelease(pStruct);
}
return fts5IndexReturn(p);
}
/*************************************************************************
**************************************************************************
** Below this point is the implementation of the integrity-check
** functionality.
*/
/*
** Return a simple checksum value based on the arguments.
*/
static u64 sqlite3Fts5IndexEntryCksum(
i64 iRowid,
int iCol,
int iPos,
int iIdx,
const char *pTerm,
int nTerm
){
int i;
u64 ret = iRowid;
ret += (ret<<3) + iCol;
ret += (ret<<3) + iPos;
if( iIdx>=0 ) ret += (ret<<3) + (FTS5_MAIN_PREFIX + iIdx);
for(i=0; iiLeaf );
cksum1 += iRowid + ((i64)pgno<<32);
}
fts5DlidxIterFree(pDlidx);
pDlidx = 0;
for(pDlidx=fts5DlidxIterInit(p, 1, iSegid, iLeaf);
fts5DlidxIterEof(p, pDlidx)==0;
fts5DlidxIterPrev(p, pDlidx)
){
i64 iRowid = fts5DlidxIterRowid(pDlidx);
int pgno = fts5DlidxIterPgno(pDlidx);
assert( fts5DlidxIterPgno(pDlidx)>iLeaf );
cksum2 += iRowid + ((i64)pgno<<32);
}
fts5DlidxIterFree(pDlidx);
pDlidx = 0;
if( p->rc==SQLITE_OK && cksum1!=cksum2 ) p->rc = FTS5_CORRUPT;
}
static int fts5QueryCksum(
Fts5Index *p, /* Fts5 index object */
int iIdx,
const char *z, /* Index key to query for */
int n, /* Size of index key in bytes */
int flags, /* Flags for Fts5IndexQuery */
u64 *pCksum /* IN/OUT: Checksum value */
){
int eDetail = p->pConfig->eDetail;
u64 cksum = *pCksum;
Fts5IndexIter *pIter = 0;
int rc = sqlite3Fts5IndexQuery(
p, z, n, (flags | FTS5INDEX_QUERY_NOTOKENDATA), 0, &pIter
);
while( rc==SQLITE_OK && ALWAYS(pIter!=0) && 0==sqlite3Fts5IterEof(pIter) ){
i64 rowid = pIter->iRowid;
if( eDetail==FTS5_DETAIL_NONE ){
cksum ^= sqlite3Fts5IndexEntryCksum(rowid, 0, 0, iIdx, z, n);
}else{
Fts5PoslistReader sReader;
for(sqlite3Fts5PoslistReaderInit(pIter->pData, pIter->nData, &sReader);
sReader.bEof==0;
sqlite3Fts5PoslistReaderNext(&sReader)
){
int iCol = FTS5_POS2COLUMN(sReader.iPos);
int iOff = FTS5_POS2OFFSET(sReader.iPos);
cksum ^= sqlite3Fts5IndexEntryCksum(rowid, iCol, iOff, iIdx, z, n);
}
}
if( rc==SQLITE_OK ){
rc = sqlite3Fts5IterNext(pIter);
}
}
sqlite3Fts5IterClose(pIter);
*pCksum = cksum;
return rc;
}
/*
** Check if buffer z[], size n bytes, contains as series of valid utf-8
** encoded codepoints. If so, return 0. Otherwise, if the buffer does not
** contain valid utf-8, return non-zero.
*/
static int fts5TestUtf8(const char *z, int n){
int i = 0;
assert_nc( n>0 );
while( i=n || (z[i+1] & 0xC0)!=0x80 ) return 1;
i += 2;
}else
if( (z[i] & 0xF0)==0xE0 ){
if( i+2>=n || (z[i+1] & 0xC0)!=0x80 || (z[i+2] & 0xC0)!=0x80 ) return 1;
i += 3;
}else
if( (z[i] & 0xF8)==0xF0 ){
if( i+3>=n || (z[i+1] & 0xC0)!=0x80 || (z[i+2] & 0xC0)!=0x80 ) return 1;
if( (z[i+2] & 0xC0)!=0x80 ) return 1;
i += 3;
}else{
return 1;
}
}
return 0;
}
/*
** This function is also purely an internal test. It does not contribute to
** FTS functionality, or even the integrity-check, in any way.
*/
static void fts5TestTerm(
Fts5Index *p,
Fts5Buffer *pPrev, /* Previous term */
const char *z, int n, /* Possibly new term to test */
u64 expected,
u64 *pCksum
){
int rc = p->rc;
if( pPrev->n==0 ){
fts5BufferSet(&rc, pPrev, n, (const u8*)z);
}else
if( rc==SQLITE_OK && (pPrev->n!=n || memcmp(pPrev->p, z, n)) ){
u64 cksum3 = *pCksum;
const char *zTerm = (const char*)&pPrev->p[1]; /* term sans prefix-byte */
int nTerm = pPrev->n-1; /* Size of zTerm in bytes */
int iIdx = (pPrev->p[0] - FTS5_MAIN_PREFIX);
int flags = (iIdx==0 ? 0 : FTS5INDEX_QUERY_PREFIX);
u64 ck1 = 0;
u64 ck2 = 0;
/* Check that the results returned for ASC and DESC queries are
** the same. If not, call this corruption. */
rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, flags, &ck1);
if( rc==SQLITE_OK ){
int f = flags|FTS5INDEX_QUERY_DESC;
rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, f, &ck2);
}
if( rc==SQLITE_OK && ck1!=ck2 ) rc = FTS5_CORRUPT;
/* If this is a prefix query, check that the results returned if the
** the index is disabled are the same. In both ASC and DESC order.
**
** This check may only be performed if the hash table is empty. This
** is because the hash table only supports a single scan query at
** a time, and the multi-iter loop from which this function is called
** is already performing such a scan.
**
** Also only do this if buffer zTerm contains nTerm bytes of valid
** utf-8. Otherwise, the last part of the buffer contents might contain
** a non-utf-8 sequence that happens to be a prefix of a valid utf-8
** character stored in the main fts index, which will cause the
** test to fail. */
if( p->nPendingData==0 && 0==fts5TestUtf8(zTerm, nTerm) ){
if( iIdx>0 && rc==SQLITE_OK ){
int f = flags|FTS5INDEX_QUERY_TEST_NOIDX;
ck2 = 0;
rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, f, &ck2);
if( rc==SQLITE_OK && ck1!=ck2 ) rc = FTS5_CORRUPT;
}
if( iIdx>0 && rc==SQLITE_OK ){
int f = flags|FTS5INDEX_QUERY_TEST_NOIDX|FTS5INDEX_QUERY_DESC;
ck2 = 0;
rc = fts5QueryCksum(p, iIdx, zTerm, nTerm, f, &ck2);
if( rc==SQLITE_OK && ck1!=ck2 ) rc = FTS5_CORRUPT;
}
}
cksum3 ^= ck1;
fts5BufferSet(&rc, pPrev, n, (const u8*)z);
if( rc==SQLITE_OK && cksum3!=expected ){
rc = FTS5_CORRUPT;
}
*pCksum = cksum3;
}
p->rc = rc;
}
#else
# define fts5TestDlidxReverse(x,y,z)
# define fts5TestTerm(u,v,w,x,y,z)
#endif
/*
** Check that:
**
** 1) All leaves of pSeg between iFirst and iLast (inclusive) exist and
** contain zero terms.
** 2) All leaves of pSeg between iNoRowid and iLast (inclusive) exist and
** contain zero rowids.
*/
static void fts5IndexIntegrityCheckEmpty(
Fts5Index *p,
Fts5StructureSegment *pSeg, /* Segment to check internal consistency */
int iFirst,
int iNoRowid,
int iLast
){
int i;
/* Now check that the iter.nEmpty leaves following the current leaf
** (a) exist and (b) contain no terms. */
for(i=iFirst; p->rc==SQLITE_OK && i<=iLast; i++){
Fts5Data *pLeaf = fts5DataRead(p, FTS5_SEGMENT_ROWID(pSeg->iSegid, i));
if( pLeaf ){
if( !fts5LeafIsTermless(pLeaf) ) p->rc = FTS5_CORRUPT;
if( i>=iNoRowid && 0!=fts5LeafFirstRowidOff(pLeaf) ) p->rc = FTS5_CORRUPT;
}
fts5DataRelease(pLeaf);
}
}
static void fts5IntegrityCheckPgidx(Fts5Index *p, Fts5Data *pLeaf){
i64 iTermOff = 0;
int ii;
Fts5Buffer buf1 = {0,0,0};
Fts5Buffer buf2 = {0,0,0};
ii = pLeaf->szLeaf;
while( iinn && p->rc==SQLITE_OK ){
int res;
i64 iOff;
int nIncr;
ii += fts5GetVarint32(&pLeaf->p[ii], nIncr);
iTermOff += nIncr;
iOff = iTermOff;
if( iOff>=pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
}else if( iTermOff==nIncr ){
int nByte;
iOff += fts5GetVarint32(&pLeaf->p[iOff], nByte);
if( (iOff+nByte)>pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
}else{
fts5BufferSet(&p->rc, &buf1, nByte, &pLeaf->p[iOff]);
}
}else{
int nKeep, nByte;
iOff += fts5GetVarint32(&pLeaf->p[iOff], nKeep);
iOff += fts5GetVarint32(&pLeaf->p[iOff], nByte);
if( nKeep>buf1.n || (iOff+nByte)>pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
}else{
buf1.n = nKeep;
fts5BufferAppendBlob(&p->rc, &buf1, nByte, &pLeaf->p[iOff]);
}
if( p->rc==SQLITE_OK ){
res = fts5BufferCompare(&buf1, &buf2);
if( res<=0 ) p->rc = FTS5_CORRUPT;
}
}
fts5BufferSet(&p->rc, &buf2, buf1.n, buf1.p);
}
fts5BufferFree(&buf1);
fts5BufferFree(&buf2);
}
static void fts5IndexIntegrityCheckSegment(
Fts5Index *p, /* FTS5 backend object */
Fts5StructureSegment *pSeg /* Segment to check internal consistency */
){
Fts5Config *pConfig = p->pConfig;
int bSecureDelete = (pConfig->iVersion==FTS5_CURRENT_VERSION_SECUREDELETE);
sqlite3_stmt *pStmt = 0;
int rc2;
int iIdxPrevLeaf = pSeg->pgnoFirst-1;
int iDlidxPrevLeaf = pSeg->pgnoLast;
if( pSeg->pgnoFirst==0 ) return;
fts5IndexPrepareStmt(p, &pStmt, sqlite3_mprintf(
"SELECT segid, term, (pgno>>1), (pgno&1) FROM %Q.'%q_idx' WHERE segid=%d "
"ORDER BY 1, 2",
pConfig->zDb, pConfig->zName, pSeg->iSegid
));
/* Iterate through the b-tree hierarchy. */
while( p->rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){
i64 iRow; /* Rowid for this leaf */
Fts5Data *pLeaf; /* Data for this leaf */
const char *zIdxTerm = (const char*)sqlite3_column_blob(pStmt, 1);
int nIdxTerm = sqlite3_column_bytes(pStmt, 1);
int iIdxLeaf = sqlite3_column_int(pStmt, 2);
int bIdxDlidx = sqlite3_column_int(pStmt, 3);
/* If the leaf in question has already been trimmed from the segment,
** ignore this b-tree entry. Otherwise, load it into memory. */
if( iIdxLeafpgnoFirst ) continue;
iRow = FTS5_SEGMENT_ROWID(pSeg->iSegid, iIdxLeaf);
pLeaf = fts5LeafRead(p, iRow);
if( pLeaf==0 ) break;
/* Check that the leaf contains at least one term, and that it is equal
** to or larger than the split-key in zIdxTerm. Also check that if there
** is also a rowid pointer within the leaf page header, it points to a
** location before the term. */
if( pLeaf->nn<=pLeaf->szLeaf ){
if( nIdxTerm==0
&& pConfig->iVersion==FTS5_CURRENT_VERSION_SECUREDELETE
&& pLeaf->nn==pLeaf->szLeaf
&& pLeaf->nn==4
){
/* special case - the very first page in a segment keeps its %_idx
** entry even if all the terms are removed from it by secure-delete
** operations. */
}else{
p->rc = FTS5_CORRUPT;
}
}else{
int iOff; /* Offset of first term on leaf */
int iRowidOff; /* Offset of first rowid on leaf */
int nTerm; /* Size of term on leaf in bytes */
int res; /* Comparison of term and split-key */
iOff = fts5LeafFirstTermOff(pLeaf);
iRowidOff = fts5LeafFirstRowidOff(pLeaf);
if( iRowidOff>=iOff || iOff>=pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
}else{
iOff += fts5GetVarint32(&pLeaf->p[iOff], nTerm);
res = fts5Memcmp(&pLeaf->p[iOff], zIdxTerm, MIN(nTerm, nIdxTerm));
if( res==0 ) res = nTerm - nIdxTerm;
if( res<0 ) p->rc = FTS5_CORRUPT;
}
fts5IntegrityCheckPgidx(p, pLeaf);
}
fts5DataRelease(pLeaf);
if( p->rc ) break;
/* Now check that the iter.nEmpty leaves following the current leaf
** (a) exist and (b) contain no terms. */
fts5IndexIntegrityCheckEmpty(
p, pSeg, iIdxPrevLeaf+1, iDlidxPrevLeaf+1, iIdxLeaf-1
);
if( p->rc ) break;
/* If there is a doclist-index, check that it looks right. */
if( bIdxDlidx ){
Fts5DlidxIter *pDlidx = 0; /* For iterating through doclist index */
int iPrevLeaf = iIdxLeaf;
int iSegid = pSeg->iSegid;
int iPg = 0;
i64 iKey;
for(pDlidx=fts5DlidxIterInit(p, 0, iSegid, iIdxLeaf);
fts5DlidxIterEof(p, pDlidx)==0;
fts5DlidxIterNext(p, pDlidx)
){
/* Check any rowid-less pages that occur before the current leaf. */
for(iPg=iPrevLeaf+1; iPgrc = FTS5_CORRUPT;
fts5DataRelease(pLeaf);
}
}
iPrevLeaf = fts5DlidxIterPgno(pDlidx);
/* Check that the leaf page indicated by the iterator really does
** contain the rowid suggested by the same. */
iKey = FTS5_SEGMENT_ROWID(iSegid, iPrevLeaf);
pLeaf = fts5DataRead(p, iKey);
if( pLeaf ){
i64 iRowid;
int iRowidOff = fts5LeafFirstRowidOff(pLeaf);
ASSERT_SZLEAF_OK(pLeaf);
if( iRowidOff>=pLeaf->szLeaf ){
p->rc = FTS5_CORRUPT;
}else if( bSecureDelete==0 || iRowidOff>0 ){
i64 iDlRowid = fts5DlidxIterRowid(pDlidx);
fts5GetVarint(&pLeaf->p[iRowidOff], (u64*)&iRowid);
if( iRowidrc = FTS5_CORRUPT;
}
}
fts5DataRelease(pLeaf);
}
}
iDlidxPrevLeaf = iPg;
fts5DlidxIterFree(pDlidx);
fts5TestDlidxReverse(p, iSegid, iIdxLeaf);
}else{
iDlidxPrevLeaf = pSeg->pgnoLast;
/* TODO: Check there is no doclist index */
}
iIdxPrevLeaf = iIdxLeaf;
}
rc2 = sqlite3_finalize(pStmt);
if( p->rc==SQLITE_OK ) p->rc = rc2;
/* Page iter.iLeaf must now be the rightmost leaf-page in the segment */
#if 0
if( p->rc==SQLITE_OK && iter.iLeaf!=pSeg->pgnoLast ){
p->rc = FTS5_CORRUPT;
}
#endif
}
/*
** Run internal checks to ensure that the FTS index (a) is internally
** consistent and (b) contains entries for which the XOR of the checksums
** as calculated by sqlite3Fts5IndexEntryCksum() is cksum.
**
** Return SQLITE_CORRUPT if any of the internal checks fail, or if the
** checksum does not match. Return SQLITE_OK if all checks pass without
** error, or some other SQLite error code if another error (e.g. OOM)
** occurs.
*/
static int sqlite3Fts5IndexIntegrityCheck(Fts5Index *p, u64 cksum, int bUseCksum){
int eDetail = p->pConfig->eDetail;
u64 cksum2 = 0; /* Checksum based on contents of indexes */
Fts5Buffer poslist = {0,0,0}; /* Buffer used to hold a poslist */
Fts5Iter *pIter; /* Used to iterate through entire index */
Fts5Structure *pStruct; /* Index structure */
int iLvl, iSeg;
#ifdef SQLITE_DEBUG
/* Used by extra internal tests only run if NDEBUG is not defined */
u64 cksum3 = 0; /* Checksum based on contents of indexes */
Fts5Buffer term = {0,0,0}; /* Buffer used to hold most recent term */
#endif
const int flags = FTS5INDEX_QUERY_NOOUTPUT;
/* Load the FTS index structure */
pStruct = fts5StructureRead(p);
if( pStruct==0 ){
assert( p->rc!=SQLITE_OK );
return fts5IndexReturn(p);
}
/* Check that the internal nodes of each segment match the leaves */
for(iLvl=0; iLvlnLevel; iLvl++){
for(iSeg=0; iSegaLevel[iLvl].nSeg; iSeg++){
Fts5StructureSegment *pSeg = &pStruct->aLevel[iLvl].aSeg[iSeg];
fts5IndexIntegrityCheckSegment(p, pSeg);
}
}
/* The cksum argument passed to this function is a checksum calculated
** based on all expected entries in the FTS index (including prefix index
** entries). This block checks that a checksum calculated based on the
** actual contents of FTS index is identical.
**
** Two versions of the same checksum are calculated. The first (stack
** variable cksum2) based on entries extracted from the full-text index
** while doing a linear scan of each individual index in turn.
**
** As each term visited by the linear scans, a separate query for the
** same term is performed. cksum3 is calculated based on the entries
** extracted by these queries.
*/
for(fts5MultiIterNew(p, pStruct, flags, 0, 0, 0, -1, 0, &pIter);
fts5MultiIterEof(p, pIter)==0;
fts5MultiIterNext(p, pIter, 0, 0)
){
int n; /* Size of term in bytes */
i64 iPos = 0; /* Position read from poslist */
int iOff = 0; /* Offset within poslist */
i64 iRowid = fts5MultiIterRowid(pIter);
char *z = (char*)fts5MultiIterTerm(pIter, &n);
/* If this is a new term, query for it. Update cksum3 with the results. */
fts5TestTerm(p, &term, z, n, cksum2, &cksum3);
if( p->rc ) break;
if( eDetail==FTS5_DETAIL_NONE ){
if( 0==fts5MultiIterIsEmpty(p, pIter) ){
cksum2 ^= sqlite3Fts5IndexEntryCksum(iRowid, 0, 0, -1, z, n);
}
}else{
poslist.n = 0;
fts5SegiterPoslist(p, &pIter->aSeg[pIter->aFirst[1].iFirst], 0, &poslist);
fts5BufferAppendBlob(&p->rc, &poslist, 4, (const u8*)"\0\0\0\0");
while( 0==sqlite3Fts5PoslistNext64(poslist.p, poslist.n, &iOff, &iPos) ){
int iCol = FTS5_POS2COLUMN(iPos);
int iTokOff = FTS5_POS2OFFSET(iPos);
cksum2 ^= sqlite3Fts5IndexEntryCksum(iRowid, iCol, iTokOff, -1, z, n);
}
}
}
fts5TestTerm(p, &term, 0, 0, cksum2, &cksum3);
fts5MultiIterFree(pIter);
if( p->rc==SQLITE_OK && bUseCksum && cksum!=cksum2 ) p->rc = FTS5_CORRUPT;
fts5StructureRelease(pStruct);
#ifdef SQLITE_DEBUG
fts5BufferFree(&term);
#endif
fts5BufferFree(&poslist);
return fts5IndexReturn(p);
}
/*************************************************************************
**************************************************************************
** Below this point is the implementation of the fts5_decode() scalar
** function only.
*/
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
/*
** Decode a segment-data rowid from the %_data table. This function is
** the opposite of macro FTS5_SEGMENT_ROWID().
*/
static void fts5DecodeRowid(
i64 iRowid, /* Rowid from %_data table */
int *pbTombstone, /* OUT: Tombstone hash flag */
int *piSegid, /* OUT: Segment id */
int *pbDlidx, /* OUT: Dlidx flag */
int *piHeight, /* OUT: Height */
int *piPgno /* OUT: Page number */
){
*piPgno = (int)(iRowid & (((i64)1 << FTS5_DATA_PAGE_B) - 1));
iRowid >>= FTS5_DATA_PAGE_B;
*piHeight = (int)(iRowid & (((i64)1 << FTS5_DATA_HEIGHT_B) - 1));
iRowid >>= FTS5_DATA_HEIGHT_B;
*pbDlidx = (int)(iRowid & 0x0001);
iRowid >>= FTS5_DATA_DLI_B;
*piSegid = (int)(iRowid & (((i64)1 << FTS5_DATA_ID_B) - 1));
iRowid >>= FTS5_DATA_ID_B;
*pbTombstone = (int)(iRowid & 0x0001);
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
static void fts5DebugRowid(int *pRc, Fts5Buffer *pBuf, i64 iKey){
int iSegid, iHeight, iPgno, bDlidx, bTomb; /* Rowid compenents */
fts5DecodeRowid(iKey, &bTomb, &iSegid, &bDlidx, &iHeight, &iPgno);
if( iSegid==0 ){
if( iKey==FTS5_AVERAGES_ROWID ){
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "{averages} ");
}else{
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "{structure}");
}
}
else{
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "{%s%ssegid=%d h=%d pgno=%d}",
bDlidx ? "dlidx " : "",
bTomb ? "tombstone " : "",
iSegid, iHeight, iPgno
);
}
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
static void fts5DebugStructure(
int *pRc, /* IN/OUT: error code */
Fts5Buffer *pBuf,
Fts5Structure *p
){
int iLvl, iSeg; /* Iterate through levels, segments */
for(iLvl=0; iLvlnLevel; iLvl++){
Fts5StructureLevel *pLvl = &p->aLevel[iLvl];
sqlite3Fts5BufferAppendPrintf(pRc, pBuf,
" {lvl=%d nMerge=%d nSeg=%d", iLvl, pLvl->nMerge, pLvl->nSeg
);
for(iSeg=0; iSegnSeg; iSeg++){
Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg];
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " {id=%d leaves=%d..%d",
pSeg->iSegid, pSeg->pgnoFirst, pSeg->pgnoLast
);
if( pSeg->iOrigin1>0 ){
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " origin=%lld..%lld",
pSeg->iOrigin1, pSeg->iOrigin2
);
}
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "}");
}
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, "}");
}
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
/*
** This is part of the fts5_decode() debugging aid.
**
** Arguments pBlob/nBlob contain a serialized Fts5Structure object. This
** function appends a human-readable representation of the same object
** to the buffer passed as the second argument.
*/
static void fts5DecodeStructure(
int *pRc, /* IN/OUT: error code */
Fts5Buffer *pBuf,
const u8 *pBlob, int nBlob
){
int rc; /* Return code */
Fts5Structure *p = 0; /* Decoded structure object */
rc = fts5StructureDecode(pBlob, nBlob, 0, &p);
if( rc!=SQLITE_OK ){
*pRc = rc;
return;
}
fts5DebugStructure(pRc, pBuf, p);
fts5StructureRelease(p);
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
/*
** This is part of the fts5_decode() debugging aid.
**
** Arguments pBlob/nBlob contain an "averages" record. This function
** appends a human-readable representation of record to the buffer passed
** as the second argument.
*/
static void fts5DecodeAverages(
int *pRc, /* IN/OUT: error code */
Fts5Buffer *pBuf,
const u8 *pBlob, int nBlob
){
int i = 0;
const char *zSpace = "";
while( i0 ){
iOff = sqlite3Fts5GetVarint(a, (u64*)&iDocid);
sqlite3Fts5BufferAppendPrintf(pRc, pBuf, " id=%lld", iDocid);
}
while( iOffn*2 + 1);
if( *pRc==SQLITE_OK ){
for(ii=0; iin; ii++){
if( pTerm->p[ii]==0x00 ){
pBuf->p[pBuf->n++] = '\\';
pBuf->p[pBuf->n++] = '0';
}else{
pBuf->p[pBuf->n++] = pTerm->p[ii];
}
}
pBuf->p[pBuf->n] = 0x00;
}
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
/*
** The implementation of user-defined scalar function fts5_decode().
*/
static void fts5DecodeFunction(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args (always 2) */
sqlite3_value **apVal /* Function arguments */
){
i64 iRowid; /* Rowid for record being decoded */
int iSegid,iHeight,iPgno,bDlidx;/* Rowid components */
int bTomb;
const u8 *aBlob; int n; /* Record to decode */
u8 *a = 0;
Fts5Buffer s; /* Build up text to return here */
int rc = SQLITE_OK; /* Return code */
sqlite3_int64 nSpace = 0;
int eDetailNone = (sqlite3_user_data(pCtx)!=0);
assert( nArg==2 );
UNUSED_PARAM(nArg);
memset(&s, 0, sizeof(Fts5Buffer));
iRowid = sqlite3_value_int64(apVal[0]);
/* Make a copy of the second argument (a blob) in aBlob[]. The aBlob[]
** copy is followed by FTS5_DATA_ZERO_PADDING 0x00 bytes, which prevents
** buffer overreads even if the record is corrupt. */
n = sqlite3_value_bytes(apVal[1]);
aBlob = sqlite3_value_blob(apVal[1]);
nSpace = n + FTS5_DATA_ZERO_PADDING;
a = (u8*)sqlite3Fts5MallocZero(&rc, nSpace);
if( a==0 ) goto decode_out;
if( n>0 ) memcpy(a, aBlob, n);
fts5DecodeRowid(iRowid, &bTomb, &iSegid, &bDlidx, &iHeight, &iPgno);
fts5DebugRowid(&rc, &s, iRowid);
if( bDlidx ){
Fts5Data dlidx;
Fts5DlidxLvl lvl;
dlidx.p = a;
dlidx.nn = n;
memset(&lvl, 0, sizeof(Fts5DlidxLvl));
lvl.pData = &dlidx;
lvl.iLeafPgno = iPgno;
for(fts5DlidxLvlNext(&lvl); lvl.bEof==0; fts5DlidxLvlNext(&lvl)){
sqlite3Fts5BufferAppendPrintf(&rc, &s,
" %d(%lld)", lvl.iLeafPgno, lvl.iRowid
);
}
}else if( bTomb ){
u32 nElem = fts5GetU32(&a[4]);
int szKey = (aBlob[0]==4 || aBlob[0]==8) ? aBlob[0] : 8;
int nSlot = (n - 8) / szKey;
int ii;
sqlite3Fts5BufferAppendPrintf(&rc, &s, " nElem=%d", (int)nElem);
if( aBlob[1] ){
sqlite3Fts5BufferAppendPrintf(&rc, &s, " 0");
}
for(ii=0; iin ){
rc = FTS5_CORRUPT;
goto decode_out;
}
}
/* Decode the position list tail at the start of the page */
if( iRowidOff!=0 ){
iOff = iRowidOff;
}else if( iTermOff!=0 ){
iOff = iTermOff;
}else{
iOff = szLeaf;
}
if( iOff>n ){
rc = FTS5_CORRUPT;
goto decode_out;
}
fts5DecodePoslist(&rc, &s, &a[4], iOff-4);
/* Decode any more doclist data that appears on the page before the
** first term. */
nDoclist = (iTermOff ? iTermOff : szLeaf) - iOff;
if( nDoclist+iOff>n ){
rc = FTS5_CORRUPT;
goto decode_out;
}
fts5DecodeDoclist(&rc, &s, &a[iOff], nDoclist);
while( iPgidxOffszLeaf ){
rc = FTS5_CORRUPT;
break;
}
if( bFirst==0 ){
iOff += fts5GetVarint32(&a[iOff], nByte);
if( nByte>term.n ){
rc = FTS5_CORRUPT;
break;
}
term.n = nByte;
}
iOff += fts5GetVarint32(&a[iOff], nByte);
if( iOff+nByte>n ){
rc = FTS5_CORRUPT;
break;
}
fts5BufferAppendBlob(&rc, &term, nByte, &a[iOff]);
iOff += nByte;
sqlite3Fts5BufferAppendPrintf(&rc, &s, " term=");
fts5BufferAppendTerm(&rc, &s, &term);
iOff += fts5DecodeDoclist(&rc, &s, &a[iOff], iEnd-iOff);
}
fts5BufferFree(&term);
}
decode_out:
sqlite3_free(a);
if( rc==SQLITE_OK ){
sqlite3_result_text(pCtx, (const char*)s.p, s.n, SQLITE_TRANSIENT);
}else{
sqlite3_result_error_code(pCtx, rc);
}
fts5BufferFree(&s);
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
/*
** The implementation of user-defined scalar function fts5_rowid().
*/
static void fts5RowidFunction(
sqlite3_context *pCtx, /* Function call context */
int nArg, /* Number of args (always 2) */
sqlite3_value **apVal /* Function arguments */
){
const char *zArg;
if( nArg==0 ){
sqlite3_result_error(pCtx, "should be: fts5_rowid(subject, ....)", -1);
}else{
zArg = (const char*)sqlite3_value_text(apVal[0]);
if( 0==sqlite3_stricmp(zArg, "segment") ){
i64 iRowid;
int segid, pgno;
if( nArg!=3 ){
sqlite3_result_error(pCtx,
"should be: fts5_rowid('segment', segid, pgno))", -1
);
}else{
segid = sqlite3_value_int(apVal[1]);
pgno = sqlite3_value_int(apVal[2]);
iRowid = FTS5_SEGMENT_ROWID(segid, pgno);
sqlite3_result_int64(pCtx, iRowid);
}
}else{
sqlite3_result_error(pCtx,
"first arg to fts5_rowid() must be 'segment'" , -1
);
}
}
}
#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */
#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG)
typedef struct Fts5StructVtab Fts5StructVtab;
struct Fts5StructVtab {
sqlite3_vtab base;
};
typedef struct Fts5StructVcsr Fts5StructVcsr;
struct Fts5StructVcsr {
sqlite3_vtab_cursor base;
Fts5Structure *pStruct;
int iLevel;
int iSeg;
int iRowid;
};
/*
** Create a new fts5_structure() table-valued function.
*/
static int fts5structConnectMethod(
sqlite3 *db,
void *pAux,
int argc, const char *const*argv,
sqlite3_vtab **ppVtab,
char **pzErr
){
Fts5StructVtab *pNew = 0;
int rc = SQLITE_OK;
rc = sqlite3_declare_vtab(db,
"CREATE TABLE xyz("
"level, segment, merge, segid, leaf1, leaf2, loc1, loc2, "
"npgtombstone, nentrytombstone, nentry, struct HIDDEN);"
);
if( rc==SQLITE_OK ){
pNew = sqlite3Fts5MallocZero(&rc, sizeof(*pNew));
}
*ppVtab = (sqlite3_vtab*)pNew;
return rc;
}
/*
** We must have a single struct=? constraint that will be passed through
** into the xFilter method. If there is no valid stmt=? constraint,
** then return an SQLITE_CONSTRAINT error.
*/
static int fts5structBestIndexMethod(
sqlite3_vtab *tab,
sqlite3_index_info *pIdxInfo
){
int i;
int rc = SQLITE_CONSTRAINT;
struct sqlite3_index_constraint *p;
pIdxInfo->estimatedCost = (double)100;
pIdxInfo->estimatedRows = 100;
pIdxInfo->idxNum = 0;
for(i=0, p=pIdxInfo->aConstraint; i