summaryrefslogtreecommitdiffstats
path: root/basic/source/comp/token.cxx
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 09:06:44 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 09:06:44 +0000
commited5640d8b587fbcfed7dd7967f3de04b37a76f26 (patch)
tree7a5f7c6c9d02226d7471cb3cc8fbbf631b415303 /basic/source/comp/token.cxx
parentInitial commit. (diff)
downloadlibreoffice-ed5640d8b587fbcfed7dd7967f3de04b37a76f26.tar.xz
libreoffice-ed5640d8b587fbcfed7dd7967f3de04b37a76f26.zip
Adding upstream version 4:7.4.7.upstream/4%7.4.7upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'basic/source/comp/token.cxx')
-rw-r--r--basic/source/comp/token.cxx571
1 files changed, 571 insertions, 0 deletions
diff --git a/basic/source/comp/token.cxx b/basic/source/comp/token.cxx
new file mode 100644
index 000000000..a060b3fd9
--- /dev/null
+++ b/basic/source/comp/token.cxx
@@ -0,0 +1,571 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
+/*
+ * This file is part of the LibreOffice project.
+ *
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/.
+ *
+ * This file incorporates work covered by the following license notice:
+ *
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed
+ * with this work for additional information regarding copyright
+ * ownership. The ASF licenses this file to you under the Apache
+ * License, Version 2.0 (the "License"); you may not use this file
+ * except in compliance with the License. You may obtain a copy of
+ * the License at http://www.apache.org/licenses/LICENSE-2.0 .
+ */
+
+
+#include <array>
+
+#include <basic/sberrors.hxx>
+#include <sal/macros.h>
+#include <basiccharclass.hxx>
+#include <token.hxx>
+
+namespace {
+
+struct TokenTable { SbiToken t; const char *s; };
+
+}
+
+const TokenTable aTokTable_Basic [] = {
+ { CAT, "&" },
+ { MUL, "*" },
+ { PLUS, "+" },
+ { MINUS, "-" },
+ { DIV, "/" },
+ { EOS, ":" },
+ { ASSIGN, ":=" },
+ { LT, "<" },
+ { LE, "<=" },
+ { NE, "<>" },
+ { EQ, "=" },
+ { GT, ">" },
+ { GE, ">=" },
+ { ACCESS, "Access" },
+ { ALIAS, "Alias" },
+ { AND, "And" },
+ { ANY, "Any" },
+ { APPEND, "Append" },
+ { AS, "As" },
+ { ATTRIBUTE,"Attribute" },
+ { BASE, "Base" },
+ { BINARY, "Binary" },
+ { TBOOLEAN, "Boolean" },
+ { BYREF, "ByRef", },
+ { TBYTE, "Byte", },
+ { BYVAL, "ByVal", },
+ { CALL, "Call" },
+ { CASE, "Case" },
+ { CDECL_, "Cdecl" },
+ { CLASSMODULE, "ClassModule" },
+ { CLOSE, "Close" },
+ { COMPARE, "Compare" },
+ { COMPATIBLE,"Compatible" },
+ { CONST_, "Const" },
+ { TCURRENCY,"Currency" },
+ { TDATE, "Date" },
+ { DECLARE, "Declare" },
+ { DEFBOOL, "DefBool" },
+ { DEFCUR, "DefCur" },
+ { DEFDATE, "DefDate" },
+ { DEFDBL, "DefDbl" },
+ { DEFERR, "DefErr" },
+ { DEFINT, "DefInt" },
+ { DEFLNG, "DefLng" },
+ { DEFOBJ, "DefObj" },
+ { DEFSNG, "DefSng" },
+ { DEFSTR, "DefStr" },
+ { DEFVAR, "DefVar" },
+ { DIM, "Dim" },
+ { DO, "Do" },
+ { TDOUBLE, "Double" },
+ { EACH, "Each" },
+ { ELSE, "Else" },
+ { ELSEIF, "ElseIf" },
+ { END, "End" },
+ { ENDENUM, "End Enum" },
+ { ENDFUNC, "End Function" },
+ { ENDIF, "End If" },
+ { ENDPROPERTY, "End Property" },
+ { ENDSELECT,"End Select" },
+ { ENDSUB, "End Sub" },
+ { ENDTYPE, "End Type" },
+ { ENDIF, "EndIf" },
+ { ENUM, "Enum" },
+ { EQV, "Eqv" },
+ { ERASE, "Erase" },
+ { ERROR_, "Error" },
+ { EXIT, "Exit" },
+ { BASIC_EXPLICIT, "Explicit" },
+ { FOR, "For" },
+ { FUNCTION, "Function" },
+ { GET, "Get" },
+ { GLOBAL, "Global" },
+ { GOSUB, "GoSub" },
+ { GOTO, "GoTo" },
+ { IF, "If" },
+ { IMP, "Imp" },
+ { IMPLEMENTS, "Implements" },
+ { IN_, "In" },
+ { INPUT, "Input" }, // also INPUT #
+ { TINTEGER, "Integer" },
+ { IS, "Is" },
+ { LET, "Let" },
+ { LIB, "Lib" },
+ { LIKE, "Like" },
+ { LINE, "Line" },
+ { LINEINPUT,"Line Input" },
+ { LOCAL, "Local" },
+ { LOCK, "Lock" },
+ { TLONG, "Long" },
+ { LOOP, "Loop" },
+ { LPRINT, "LPrint" },
+ { LSET, "LSet" }, // JSM
+ { MOD, "Mod" },
+ { NAME, "Name" },
+ { NEW, "New" },
+ { NEXT, "Next" },
+ { NOT, "Not" },
+ { TOBJECT, "Object" },
+ { ON, "On" },
+ { OPEN, "Open" },
+ { OPTION, "Option" },
+ { OPTIONAL_, "Optional" },
+ { OR, "Or" },
+ { OUTPUT, "Output" },
+ { PARAMARRAY, "ParamArray" },
+ { PRESERVE, "Preserve" },
+ { PRINT, "Print" },
+ { PRIVATE, "Private" },
+ { PROPERTY, "Property" },
+ { PTRSAFE, "PtrSafe" },
+ { PUBLIC, "Public" },
+ { RANDOM, "Random" },
+ { READ, "Read" },
+ { REDIM, "ReDim" },
+ { REM, "Rem" },
+ { RESUME, "Resume" },
+ { RETURN, "Return" },
+ { RSET, "RSet" }, // JSM
+ { SELECT, "Select" },
+ { SET, "Set" },
+ { SHARED, "Shared" },
+ { TSINGLE, "Single" },
+ { STATIC, "Static" },
+ { STEP, "Step" },
+ { STOP, "Stop" },
+ { TSTRING, "String" },
+ { SUB, "Sub" },
+ { STOP, "System" },
+ { TEXT, "Text" },
+ { THEN, "Then" },
+ { TO, "To", },
+ { TYPE, "Type" },
+ { TYPEOF, "TypeOf" },
+ { UNTIL, "Until" },
+ { TVARIANT, "Variant" },
+ { VBASUPPORT, "VbaSupport" },
+ { WEND, "Wend" },
+ { WHILE, "While" },
+ { WITH, "With" },
+ { WITHEVENTS, "WithEvents" },
+ { WRITE, "Write" }, // also WRITE #
+ { XOR, "Xor" },
+};
+
+namespace {
+
+// #i109076
+class TokenLabelInfo
+{
+ std::array<bool,VBASUPPORT+1> m_pTokenCanBeLabelTab;
+
+public:
+ TokenLabelInfo();
+
+ bool canTokenBeLabel( SbiToken eTok )
+ { return m_pTokenCanBeLabelTab[eTok]; }
+};
+
+}
+
+// #i109076
+TokenLabelInfo::TokenLabelInfo()
+{
+ m_pTokenCanBeLabelTab.fill(false);
+
+ // Token accepted as label by VBA
+ static const SbiToken eLabelToken[] = { ACCESS, ALIAS, APPEND, BASE, BINARY, CLASSMODULE,
+ COMPARE, COMPATIBLE, DEFERR, ERROR_, BASIC_EXPLICIT, LIB, LINE, LPRINT, NAME,
+ TOBJECT, OUTPUT, PROPERTY, RANDOM, READ, STEP, STOP, TEXT, VBASUPPORT };
+ for( SbiToken eTok : eLabelToken )
+ {
+ m_pTokenCanBeLabelTab[eTok] = true;
+ }
+}
+
+
+SbiTokenizer::SbiTokenizer( const OUString& rSrc, StarBASIC* pb )
+ : SbiScanner(rSrc, pb)
+ , eCurTok(NIL)
+ , ePush(NIL)
+ , nPLine(0)
+ , nPCol1(0)
+ , nPCol2(0)
+ , bEof(false)
+ , bEos(true)
+ , bAs(false)
+ , bErrorIsSymbol(true)
+{
+}
+
+void SbiTokenizer::Push( SbiToken t )
+{
+ if( ePush != NIL )
+ Error( ERRCODE_BASIC_INTERNAL_ERROR, "PUSH" );
+ else ePush = t;
+}
+
+void SbiTokenizer::Error( ErrCode code, const OUString &aMsg )
+{
+ aError = aMsg;
+ Error( code );
+}
+
+void SbiTokenizer::Error( ErrCode code, SbiToken tok )
+{
+ aError = Symbol( tok );
+ Error( code );
+}
+
+// reading in the next token without absorbing it
+
+SbiToken SbiTokenizer::Peek()
+{
+ if( ePush == NIL )
+ {
+ sal_Int32 nOldLine = nLine;
+ sal_Int32 nOldCol1 = nCol1;
+ sal_Int32 nOldCol2 = nCol2;
+ ePush = Next();
+ nPLine = nLine; nLine = nOldLine;
+ nPCol1 = nCol1; nCol1 = nOldCol1;
+ nPCol2 = nCol2; nCol2 = nOldCol2;
+ }
+ eCurTok = ePush;
+ return eCurTok;
+}
+
+// For decompilation. Numbers and symbols return an empty string.
+
+const OUString& SbiTokenizer::Symbol( SbiToken t )
+{
+ // character token?
+ if( t < FIRSTKWD )
+ {
+ aSym = OUString(sal::static_int_cast<sal_Unicode>(t));
+ return aSym;
+ }
+ switch( t )
+ {
+ case NEG :
+ aSym = "-";
+ return aSym;
+ case EOS :
+ aSym = ":/CRLF";
+ return aSym;
+ case EOLN :
+ aSym = "CRLF";
+ return aSym;
+ default:
+ break;
+ }
+ for( auto& rTok : aTokTable_Basic )
+ {
+ if( rTok.t == t )
+ {
+ aSym = OStringToOUString(rTok.s, RTL_TEXTENCODING_ASCII_US);
+ return aSym;
+ }
+ }
+ const sal_Unicode *p = aSym.getStr();
+ if (*p <= ' ')
+ {
+ aSym = "???";
+ }
+ return aSym;
+}
+
+// Reading in the next token and put it down.
+// Tokens that don't appear in the token table
+// are directly returned as a character.
+// Some words are treated in a special way.
+
+SbiToken SbiTokenizer::Next()
+{
+ if (bEof)
+ {
+ return EOLN;
+ }
+ // have read in one already?
+ if( ePush != NIL )
+ {
+ eCurTok = ePush;
+ ePush = NIL;
+ nLine = nPLine;
+ nCol1 = nPCol1;
+ nCol2 = nPCol2;
+ bEos = IsEoln( eCurTok );
+ return eCurTok;
+ }
+ const TokenTable *tp;
+
+ if( !NextSym() )
+ {
+ bEof = bEos = true;
+ eCurTok = EOLN;
+ return eCurTok;
+ }
+
+ if( aSym.startsWith("\n") )
+ {
+ bEos = true;
+ eCurTok = EOLN;
+ return eCurTok;
+ }
+ bEos = false;
+
+ if( bNumber )
+ {
+ eCurTok = NUMBER;
+ return eCurTok;
+ }
+ else if( ( eScanType == SbxDATE || eScanType == SbxSTRING ) && !bSymbol )
+ {
+ eCurTok = FIXSTRING;
+ return eCurTok;
+ }
+ else if( aSym.isEmpty() )
+ {
+ //something went wrong
+ bEof = bEos = true;
+ eCurTok = EOLN;
+ return eCurTok;
+ }
+ // Special cases of characters that are between "Z" and "a". ICompare()
+ // evaluates the position of these characters in different ways.
+ else if( aSym[0] == '^' )
+ {
+ eCurTok = EXPON;
+ return eCurTok;
+ }
+ else if( aSym[0] == '\\' )
+ {
+ eCurTok = IDIV;
+ return eCurTok;
+ }
+ else
+ {
+ if( eScanType != SbxVARIANT )
+ {
+ eCurTok = SYMBOL;
+ return eCurTok;
+ }
+ // valid token?
+ short lb = 0;
+ short ub = std::size(aTokTable_Basic)-1;
+ short delta;
+ do
+ {
+ delta = (ub - lb) >> 1;
+ tp = &aTokTable_Basic[ lb + delta ];
+ sal_Int32 res = aSym.compareToIgnoreAsciiCaseAscii( tp->s );
+
+ if( res == 0 )
+ {
+ goto special;
+ }
+ if( res < 0 )
+ {
+ if ((ub - lb) == 2)
+ {
+ ub = lb;
+ }
+ else
+ {
+ ub = ub - delta;
+ }
+ }
+ else
+ {
+ if ((ub -lb) == 2)
+ {
+ lb = ub;
+ }
+ else
+ {
+ lb = lb + delta;
+ }
+ }
+ }
+ while( delta );
+ // Symbol? if not >= token
+ sal_Unicode ch = aSym[0];
+ if( !BasicCharClass::isAlpha( ch, bCompatible ) && !bSymbol )
+ {
+ eCurTok = static_cast<SbiToken>(ch & 0x00FF);
+ return eCurTok;
+ }
+ eCurTok = SYMBOL;
+ return eCurTok;
+ }
+special:
+ // #i92642
+ bool bStartOfLine = (eCurTok == NIL || eCurTok == REM || eCurTok == EOLN ||
+ eCurTok == THEN || eCurTok == ELSE); // single line If
+ if( !bStartOfLine && (tp->t == NAME || tp->t == LINE) )
+ {
+ eCurTok = SYMBOL;
+ return eCurTok;
+ }
+ else if( tp->t == TEXT )
+ {
+ eCurTok = SYMBOL;
+ return eCurTok;
+ }
+ // maybe we can expand this for other statements that have parameters
+ // that are keywords ( and those keywords are only used within such
+ // statements )
+ // what's happening here is that if we come across 'append' ( and we are
+ // not in the middle of parsing a special statement ( like 'Open')
+ // we just treat keyword 'append' as a normal 'SYMBOL'.
+ // Also we accept Dim APPEND
+ else if ( ( !bInStatement || eCurTok == DIM ) && tp->t == APPEND )
+ {
+ eCurTok = SYMBOL;
+ return eCurTok;
+ }
+ // #i92642: Special LINE token handling -> SbiParser::Line()
+
+ // END IF, CASE, SUB, DEF, FUNCTION, TYPE, CLASS, WITH
+ if( tp->t == END )
+ {
+ // from 15.3.96, special treatment for END, at Peek() the current
+ // time is lost, so memorize everything and restore after
+ sal_Int32 nOldLine = nLine;
+ sal_Int32 nOldCol = nCol;
+ sal_Int32 nOldCol1 = nCol1;
+ sal_Int32 nOldCol2 = nCol2;
+ OUString aOldSym = aSym;
+ SaveLine(); // save pLine in the scanner
+
+ eCurTok = Peek();
+ switch( eCurTok )
+ {
+ case IF: Next(); eCurTok = ENDIF; break;
+ case SELECT: Next(); eCurTok = ENDSELECT; break;
+ case SUB: Next(); eCurTok = ENDSUB; break;
+ case FUNCTION: Next(); eCurTok = ENDFUNC; break;
+ case PROPERTY: Next(); eCurTok = ENDPROPERTY; break;
+ case TYPE: Next(); eCurTok = ENDTYPE; break;
+ case ENUM: Next(); eCurTok = ENDENUM; break;
+ case WITH: Next(); eCurTok = ENDWITH; break;
+ default : eCurTok = END; break;
+ }
+ nCol1 = nOldCol1;
+ if( eCurTok == END )
+ {
+ // reset everything so that token is read completely newly after END
+ ePush = NIL;
+ nLine = nOldLine;
+ nCol = nOldCol;
+ nCol2 = nOldCol2;
+ aSym = aOldSym;
+ RestoreLine();
+ }
+ return eCurTok;
+ }
+ // are data types keywords?
+ // there is ERROR(), DATA(), STRING() etc.
+ eCurTok = tp->t;
+ // AS: data types are keywords
+ if( tp->t == AS )
+ {
+ bAs = true;
+ }
+ else
+ {
+ if( bAs )
+ {
+ bAs = false;
+ }
+ else if( eCurTok >= DATATYPE1 && eCurTok <= DATATYPE2 && (bErrorIsSymbol || eCurTok != ERROR_) )
+ {
+ eCurTok = SYMBOL;
+ }
+ }
+
+ // CLASSMODULE, PROPERTY, GET, ENUM token only visible in compatible mode
+ SbiToken eTok = tp->t;
+ if( bCompatible )
+ {
+ // #129904 Suppress system
+ if( eTok == STOP && aSym.equalsIgnoreAsciiCase("system") )
+ {
+ eCurTok = SYMBOL;
+ }
+ if( eTok == GET && bStartOfLine )
+ {
+ eCurTok = SYMBOL;
+ }
+ }
+ else
+ {
+ if( eTok == CLASSMODULE ||
+ eTok == IMPLEMENTS ||
+ eTok == PARAMARRAY ||
+ eTok == ENUM ||
+ eTok == PROPERTY ||
+ eTok == GET ||
+ eTok == TYPEOF )
+ {
+ eCurTok = SYMBOL;
+ }
+ }
+
+ bEos = IsEoln( eCurTok );
+ return eCurTok;
+}
+
+bool SbiTokenizer::MayBeLabel( bool bNeedsColon )
+{
+ static TokenLabelInfo gaStaticTokenLabelInfo;
+
+ if( eCurTok == SYMBOL || gaStaticTokenLabelInfo.canTokenBeLabel( eCurTok ) )
+ {
+ return !bNeedsColon || DoesColonFollow();
+ }
+ else
+ {
+ return ( eCurTok == NUMBER
+ && eScanType == SbxINTEGER
+ && nVal >= 0 );
+ }
+}
+
+
+OUString SbiTokenizer::GetKeywordCase( const OUString& sKeyword )
+{
+ for( auto& rTok : aTokTable_Basic )
+ {
+ if( sKeyword.equalsIgnoreAsciiCaseAscii(rTok.s) )
+ return OStringToOUString(rTok.s, RTL_TEXTENCODING_ASCII_US);
+ }
+ return OUString();
+}
+
+/* vim:set shiftwidth=4 softtabstop=4 expandtab: */