summaryrefslogtreecommitdiffstats
path: root/extensions/spellcheck/hunspell/src/affixmgr.cxx
diff options
context:
space:
mode:
Diffstat (limited to 'extensions/spellcheck/hunspell/src/affixmgr.cxx')
-rw-r--r--extensions/spellcheck/hunspell/src/affixmgr.cxx4875
1 files changed, 4875 insertions, 0 deletions
diff --git a/extensions/spellcheck/hunspell/src/affixmgr.cxx b/extensions/spellcheck/hunspell/src/affixmgr.cxx
new file mode 100644
index 0000000000..adb750dba1
--- /dev/null
+++ b/extensions/spellcheck/hunspell/src/affixmgr.cxx
@@ -0,0 +1,4875 @@
+/* ***** BEGIN LICENSE BLOCK *****
+ * Version: MPL 1.1/GPL 2.0/LGPL 2.1
+ *
+ * Copyright (C) 2002-2022 Németh László
+ *
+ * The contents of this file are subject to the Mozilla Public License Version
+ * 1.1 (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.mozilla.org/MPL/
+ *
+ * Software distributed under the License is distributed on an "AS IS" basis,
+ * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
+ * for the specific language governing rights and limitations under the
+ * License.
+ *
+ * Hunspell is based on MySpell which is Copyright (C) 2002 Kevin Hendricks.
+ *
+ * Contributor(s): David Einstein, Davide Prina, Giuseppe Modugno,
+ * Gianluca Turconi, Simon Brouwer, Noll János, Bíró Árpád,
+ * Goldman Eleonóra, Sarlós Tamás, Bencsáth Boldizsár, Halácsy Péter,
+ * Dvornik László, Gefferth András, Nagy Viktor, Varga Dániel, Chris Halls,
+ * Rene Engelhard, Bram Moolenaar, Dafydd Jones, Harri Pitkänen
+ *
+ * Alternatively, the contents of this file may be used under the terms of
+ * either the GNU General Public License Version 2 or later (the "GPL"), or
+ * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
+ * in which case the provisions of the GPL or the LGPL are applicable instead
+ * of those above. If you wish to allow use of your version of this file only
+ * under the terms of either the GPL or the LGPL, and not to allow others to
+ * use your version of this file under the terms of the MPL, indicate your
+ * decision by deleting the provisions above and replace them with the notice
+ * and other provisions required by the GPL or the LGPL. If you do not delete
+ * the provisions above, a recipient may use your version of this file under
+ * the terms of any one of the MPL, the GPL or the LGPL.
+ *
+ * ***** END LICENSE BLOCK ***** */
+/*
+ * Copyright 2002 Kevin B. Hendricks, Stratford, Ontario, Canada
+ * And Contributors. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * 3. All modifications to the source code must be clearly marked as
+ * such. Binary redistributions based on modified source code
+ * must be clearly marked as modified versions in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY KEVIN B. HENDRICKS AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
+ * KEVIN B. HENDRICKS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <ctype.h>
+#include <time.h>
+
+#include <algorithm>
+#include <limits>
+#include <string>
+#include <vector>
+
+#include "affixmgr.hxx"
+#include "affentry.hxx"
+#include "langnum.hxx"
+
+#include "csutil.hxx"
+
+AffixMgr::AffixMgr(const char* affpath,
+ const std::vector<HashMgr*>& ptr,
+ const char* key)
+ : alldic(ptr)
+ , pHMgr(ptr[0]) {
+
+ // register hash manager and load affix data from aff file
+ csconv = NULL;
+ utf8 = 0;
+ complexprefixes = 0;
+ parsedmaptable = false;
+ parsedbreaktable = false;
+ iconvtable = NULL;
+ oconvtable = NULL;
+ // allow simplified compound forms (see 3rd field of CHECKCOMPOUNDPATTERN)
+ simplifiedcpd = 0;
+ parsedcheckcpd = false;
+ parseddefcpd = false;
+ phone = NULL;
+ compoundflag = FLAG_NULL; // permits word in compound forms
+ compoundbegin = FLAG_NULL; // may be first word in compound forms
+ compoundmiddle = FLAG_NULL; // may be middle word in compound forms
+ compoundend = FLAG_NULL; // may be last word in compound forms
+ compoundroot = FLAG_NULL; // compound word signing flag
+ compoundpermitflag = FLAG_NULL; // compound permitting flag for suffixed word
+ compoundforbidflag = FLAG_NULL; // compound fordidden flag for suffixed word
+ compoundmoresuffixes = 0; // allow more suffixes within compound words
+ checkcompounddup = 0; // forbid double words in compounds
+ checkcompoundrep = 0; // forbid bad compounds (may be non-compound word with
+ // a REP substitution)
+ checkcompoundcase =
+ 0; // forbid upper and lowercase combinations at word bounds
+ checkcompoundtriple = 0; // forbid compounds with triple letters
+ simplifiedtriple = 0; // allow simplified triple letters in compounds
+ // (Schiff+fahrt -> Schiffahrt)
+ forbiddenword = FORBIDDENWORD; // forbidden word signing flag
+ nosuggest = FLAG_NULL; // don't suggest words signed with NOSUGGEST flag
+ nongramsuggest = FLAG_NULL;
+ langnum = 0; // language code (see http://l10n.openoffice.org/languages.html)
+ needaffix = FLAG_NULL; // forbidden root, allowed only with suffixes
+ cpdwordmax = -1; // default: unlimited wordcount in compound words
+ cpdmin = -1; // undefined
+ cpdmaxsyllable = 0; // default: unlimited syllablecount in compound words
+ pfxappnd = NULL; // previous prefix for counting syllables of the prefix BUG
+ sfxappnd = NULL; // previous suffix for counting syllables of the suffix BUG
+ sfxextra = 0; // modifier for syllable count of sfxappnd BUG
+ checknum = 0; // checking numbers, and word with numbers
+ havecontclass = 0; // flags of possible continuing classes (double affix)
+ // LEMMA_PRESENT: not put root into the morphological output. Lemma presents
+ // in morhological description in dictionary file. It's often combined with
+ // PSEUDOROOT.
+ lemma_present = FLAG_NULL;
+ circumfix = FLAG_NULL;
+ onlyincompound = FLAG_NULL;
+ maxngramsugs = -1; // undefined
+ maxdiff = -1; // undefined
+ onlymaxdiff = 0;
+ maxcpdsugs = -1; // undefined
+ nosplitsugs = 0;
+ sugswithdots = 0;
+ keepcase = 0;
+ forceucase = 0;
+ warn = 0;
+ forbidwarn = 0;
+ checksharps = 0;
+ substandard = FLAG_NULL;
+ fullstrip = 0;
+
+ sfx = NULL;
+ pfx = NULL;
+
+ for (int i = 0; i < SETSIZE; i++) {
+ pStart[i] = NULL;
+ sStart[i] = NULL;
+ pFlag[i] = NULL;
+ sFlag[i] = NULL;
+ }
+
+ for (int j = 0; j < CONTSIZE; j++) {
+ contclasses[j] = 0;
+ }
+
+ if (parse_file(affpath, key)) {
+ HUNSPELL_WARNING(stderr, "Failure loading aff file %s\n", affpath);
+ }
+
+ if (cpdmin == -1)
+ cpdmin = MINCPDLEN;
+}
+
+AffixMgr::~AffixMgr() {
+ // pass through linked prefix entries and clean up
+ for (int i = 0; i < SETSIZE; i++) {
+ pFlag[i] = NULL;
+ PfxEntry* ptr = pStart[i];
+ PfxEntry* nptr = NULL;
+ while (ptr) {
+ nptr = ptr->getNext();
+ delete (ptr);
+ ptr = nptr;
+ nptr = NULL;
+ }
+ }
+
+ // pass through linked suffix entries and clean up
+ for (int j = 0; j < SETSIZE; j++) {
+ sFlag[j] = NULL;
+ SfxEntry* ptr = sStart[j];
+ SfxEntry* nptr = NULL;
+ while (ptr) {
+ nptr = ptr->getNext();
+ delete (ptr);
+ ptr = nptr;
+ nptr = NULL;
+ }
+ sStart[j] = NULL;
+ }
+
+ delete iconvtable;
+ delete oconvtable;
+ delete phone;
+
+ FREE_FLAG(compoundflag);
+ FREE_FLAG(compoundbegin);
+ FREE_FLAG(compoundmiddle);
+ FREE_FLAG(compoundend);
+ FREE_FLAG(compoundpermitflag);
+ FREE_FLAG(compoundforbidflag);
+ FREE_FLAG(compoundroot);
+ FREE_FLAG(forbiddenword);
+ FREE_FLAG(nosuggest);
+ FREE_FLAG(nongramsuggest);
+ FREE_FLAG(needaffix);
+ FREE_FLAG(lemma_present);
+ FREE_FLAG(circumfix);
+ FREE_FLAG(onlyincompound);
+
+ cpdwordmax = 0;
+ pHMgr = NULL;
+ cpdmin = 0;
+ cpdmaxsyllable = 0;
+ free_utf_tbl();
+ checknum = 0;
+#ifdef MOZILLA_CLIENT
+ delete[] csconv;
+#endif
+}
+
+void AffixMgr::finishFileMgr(FileMgr* afflst) {
+ delete afflst;
+
+ // convert affix trees to sorted list
+ process_pfx_tree_to_list();
+ process_sfx_tree_to_list();
+}
+
+// read in aff file and build up prefix and suffix entry objects
+int AffixMgr::parse_file(const char* affpath, const char* key) {
+
+ // checking flag duplication
+ char dupflags[CONTSIZE];
+ char dupflags_ini = 1;
+
+ // first line indicator for removing byte order mark
+ int firstline = 1;
+
+ // open the affix file
+ FileMgr* afflst = new FileMgr(affpath, key);
+ if (!afflst) {
+ HUNSPELL_WARNING(
+ stderr, "error: could not open affix description file %s\n", affpath);
+ return 1;
+ }
+
+ // step one is to parse the affix file building up the internal
+ // affix data structures
+
+ // read in each line ignoring any that do not
+ // start with a known line type indicator
+ std::string line;
+ while (afflst->getline(line)) {
+ mychomp(line);
+
+ /* remove byte order mark */
+ if (firstline) {
+ firstline = 0;
+ // Affix file begins with byte order mark: possible incompatibility with
+ // old Hunspell versions
+ if (line.compare(0, 3, "\xEF\xBB\xBF", 3) == 0) {
+ line.erase(0, 3);
+ }
+ }
+
+ /* parse in the keyboard string */
+ if (line.compare(0, 3, "KEY", 3) == 0) {
+ if (!parse_string(line, keystring, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the try string */
+ if (line.compare(0, 3, "TRY", 3) == 0) {
+ if (!parse_string(line, trystring, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the name of the character set used by the .dict and .aff */
+ if (line.compare(0, 3, "SET", 3) == 0) {
+ if (!parse_string(line, encoding, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ if (encoding == "UTF-8") {
+ utf8 = 1;
+#ifndef OPENOFFICEORG
+#ifndef MOZILLA_CLIENT
+ initialize_utf_tbl();
+#endif
+#endif
+ }
+ }
+
+ /* parse COMPLEXPREFIXES for agglutinative languages with right-to-left
+ * writing system */
+ if (line.compare(0, 15, "COMPLEXPREFIXES", 15) == 0)
+ complexprefixes = 1;
+
+ /* parse in the flag used by the controlled compound words */
+ if (line.compare(0, 12, "COMPOUNDFLAG", 12) == 0) {
+ if (!parse_flag(line, &compoundflag, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by compound words */
+ if (line.compare(0, 13, "COMPOUNDBEGIN", 13) == 0) {
+ if (complexprefixes) {
+ if (!parse_flag(line, &compoundend, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ } else {
+ if (!parse_flag(line, &compoundbegin, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+ }
+
+ /* parse in the flag used by compound words */
+ if (line.compare(0, 14, "COMPOUNDMIDDLE", 14) == 0) {
+ if (!parse_flag(line, &compoundmiddle, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by compound words */
+ if (line.compare(0, 11, "COMPOUNDEND", 11) == 0) {
+ if (complexprefixes) {
+ if (!parse_flag(line, &compoundbegin, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ } else {
+ if (!parse_flag(line, &compoundend, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+ }
+
+ /* parse in the data used by compound_check() method */
+ if (line.compare(0, 15, "COMPOUNDWORDMAX", 15) == 0) {
+ if (!parse_num(line, &cpdwordmax, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag sign compounds in dictionary */
+ if (line.compare(0, 12, "COMPOUNDROOT", 12) == 0) {
+ if (!parse_flag(line, &compoundroot, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by compound_check() method */
+ if (line.compare(0, 18, "COMPOUNDPERMITFLAG", 18) == 0) {
+ if (!parse_flag(line, &compoundpermitflag, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by compound_check() method */
+ if (line.compare(0, 18, "COMPOUNDFORBIDFLAG", 18) == 0) {
+ if (!parse_flag(line, &compoundforbidflag, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 20, "COMPOUNDMORESUFFIXES", 20) == 0) {
+ compoundmoresuffixes = 1;
+ }
+
+ if (line.compare(0, 16, "CHECKCOMPOUNDDUP", 16) == 0) {
+ checkcompounddup = 1;
+ }
+
+ if (line.compare(0, 16, "CHECKCOMPOUNDREP", 16) == 0) {
+ checkcompoundrep = 1;
+ }
+
+ if (line.compare(0, 19, "CHECKCOMPOUNDTRIPLE", 19) == 0) {
+ checkcompoundtriple = 1;
+ }
+
+ if (line.compare(0, 16, "SIMPLIFIEDTRIPLE", 16) == 0) {
+ simplifiedtriple = 1;
+ }
+
+ if (line.compare(0, 17, "CHECKCOMPOUNDCASE", 17) == 0) {
+ checkcompoundcase = 1;
+ }
+
+ if (line.compare(0, 9, "NOSUGGEST", 9) == 0) {
+ if (!parse_flag(line, &nosuggest, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 14, "NONGRAMSUGGEST", 14) == 0) {
+ if (!parse_flag(line, &nongramsuggest, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by forbidden words */
+ if (line.compare(0, 13, "FORBIDDENWORD", 13) == 0) {
+ if (!parse_flag(line, &forbiddenword, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by forbidden words (is deprecated) */
+ if (line.compare(0, 13, "LEMMA_PRESENT", 13) == 0) {
+ if (!parse_flag(line, &lemma_present, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by circumfixes */
+ if (line.compare(0, 9, "CIRCUMFIX", 9) == 0) {
+ if (!parse_flag(line, &circumfix, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by fogemorphemes */
+ if (line.compare(0, 14, "ONLYINCOMPOUND", 14) == 0) {
+ if (!parse_flag(line, &onlyincompound, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by `needaffixs' (is deprecated) */
+ if (line.compare(0, 10, "PSEUDOROOT", 10) == 0) {
+ if (!parse_flag(line, &needaffix, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by `needaffixs' */
+ if (line.compare(0, 9, "NEEDAFFIX", 9) == 0) {
+ if (!parse_flag(line, &needaffix, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the minimal length for words in compounds */
+ if (line.compare(0, 11, "COMPOUNDMIN", 11) == 0) {
+ if (!parse_num(line, &cpdmin, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ if (cpdmin < 1)
+ cpdmin = 1;
+ }
+
+ /* parse in the max. words and syllables in compounds */
+ if (line.compare(0, 16, "COMPOUNDSYLLABLE", 16) == 0) {
+ if (!parse_cpdsyllable(line, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by compound_check() method */
+ if (line.compare(0, 11, "SYLLABLENUM", 11) == 0) {
+ if (!parse_string(line, cpdsyllablenum, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by the controlled compound words */
+ if (line.compare(0, 8, "CHECKNUM", 8) == 0) {
+ checknum = 1;
+ }
+
+ /* parse in the extra word characters */
+ if (line.compare(0, 9, "WORDCHARS", 9) == 0) {
+ if (!parse_array(line, wordchars, wordchars_utf16,
+ utf8, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the ignored characters (for example, Arabic optional diacretics
+ * charachters */
+ if (line.compare(0, 6, "IGNORE", 6) == 0) {
+ if (!parse_array(line, ignorechars, ignorechars_utf16,
+ utf8, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the input conversion table */
+ if (line.compare(0, 5, "ICONV", 5) == 0) {
+ if (!parse_convtable(line, afflst, &iconvtable, "ICONV")) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the output conversion table */
+ if (line.compare(0, 5, "OCONV", 5) == 0) {
+ if (!parse_convtable(line, afflst, &oconvtable, "OCONV")) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the phonetic translation table */
+ if (line.compare(0, 5, "PHONE", 5) == 0) {
+ if (!parse_phonetable(line, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the checkcompoundpattern table */
+ if (line.compare(0, 20, "CHECKCOMPOUNDPATTERN", 20) == 0) {
+ if (!parse_checkcpdtable(line, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the defcompound table */
+ if (line.compare(0, 12, "COMPOUNDRULE", 12) == 0) {
+ if (!parse_defcpdtable(line, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the related character map table */
+ if (line.compare(0, 3, "MAP", 3) == 0) {
+ if (!parse_maptable(line, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the word breakpoints table */
+ if (line.compare(0, 5, "BREAK", 5) == 0) {
+ if (!parse_breaktable(line, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the language for language specific codes */
+ if (line.compare(0, 4, "LANG", 4) == 0) {
+ if (!parse_string(line, lang, afflst->getlinenum())) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ langnum = get_lang_num(lang);
+ }
+
+ if (line.compare(0, 7, "VERSION", 7) == 0) {
+ size_t startpos = line.find_first_not_of(" \t", 7);
+ if (startpos != std::string::npos) {
+ version = line.substr(startpos);
+ }
+ }
+
+ if (line.compare(0, 12, "MAXNGRAMSUGS", 12) == 0) {
+ if (!parse_num(line, &maxngramsugs, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 11, "ONLYMAXDIFF", 11) == 0)
+ onlymaxdiff = 1;
+
+ if (line.compare(0, 7, "MAXDIFF", 7) == 0) {
+ if (!parse_num(line, &maxdiff, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 10, "MAXCPDSUGS", 10) == 0) {
+ if (!parse_num(line, &maxcpdsugs, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 11, "NOSPLITSUGS", 11) == 0) {
+ nosplitsugs = 1;
+ }
+
+ if (line.compare(0, 9, "FULLSTRIP", 9) == 0) {
+ fullstrip = 1;
+ }
+
+ if (line.compare(0, 12, "SUGSWITHDOTS", 12) == 0) {
+ sugswithdots = 1;
+ }
+
+ /* parse in the flag used by forbidden words */
+ if (line.compare(0, 8, "KEEPCASE", 8) == 0) {
+ if (!parse_flag(line, &keepcase, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by `forceucase' */
+ if (line.compare(0, 10, "FORCEUCASE", 10) == 0) {
+ if (!parse_flag(line, &forceucase, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ /* parse in the flag used by `warn' */
+ if (line.compare(0, 4, "WARN", 4) == 0) {
+ if (!parse_flag(line, &warn, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 10, "FORBIDWARN", 10) == 0) {
+ forbidwarn = 1;
+ }
+
+ /* parse in the flag used by the affix generator */
+ if (line.compare(0, 11, "SUBSTANDARD", 11) == 0) {
+ if (!parse_flag(line, &substandard, afflst)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+
+ if (line.compare(0, 11, "CHECKSHARPS", 11) == 0) {
+ checksharps = 1;
+ }
+
+ /* parse this affix: P - prefix, S - suffix */
+ // affix type
+ char ft = ' ';
+ if (line.compare(0, 3, "PFX", 3) == 0)
+ ft = complexprefixes ? 'S' : 'P';
+ if (line.compare(0, 3, "SFX", 3) == 0)
+ ft = complexprefixes ? 'P' : 'S';
+ if (ft != ' ') {
+ if (dupflags_ini) {
+ memset(dupflags, 0, sizeof(dupflags));
+ dupflags_ini = 0;
+ }
+ if (!parse_affix(line, ft, afflst, dupflags)) {
+ finishFileMgr(afflst);
+ return 1;
+ }
+ }
+ }
+
+ finishFileMgr(afflst);
+ // affix trees are sorted now
+
+ // now we can speed up performance greatly taking advantage of the
+ // relationship between the affixes and the idea of "subsets".
+
+ // View each prefix as a potential leading subset of another and view
+ // each suffix (reversed) as a potential trailing subset of another.
+
+ // To illustrate this relationship if we know the prefix "ab" is found in the
+ // word to examine, only prefixes that "ab" is a leading subset of need be
+ // examined.
+ // Furthermore is "ab" is not present then none of the prefixes that "ab" is
+ // is a subset need be examined.
+ // The same argument goes for suffix string that are reversed.
+
+ // Then to top this off why not examine the first char of the word to quickly
+ // limit the set of prefixes to examine (i.e. the prefixes to examine must
+ // be leading supersets of the first character of the word (if they exist)
+
+ // To take advantage of this "subset" relationship, we need to add two links
+ // from entry. One to take next if the current prefix is found (call it
+ // nexteq)
+ // and one to take next if the current prefix is not found (call it nextne).
+
+ // Since we have built ordered lists, all that remains is to properly
+ // initialize
+ // the nextne and nexteq pointers that relate them
+
+ process_pfx_order();
+ process_sfx_order();
+
+ /* get encoding for CHECKCOMPOUNDCASE */
+ if (!utf8) {
+ csconv = get_current_cs(get_encoding());
+ for (int i = 0; i <= 255; i++) {
+ if ((csconv[i].cupper != csconv[i].clower) &&
+ (wordchars.find((char)i) == std::string::npos)) {
+ wordchars.push_back((char)i);
+ }
+ }
+
+ }
+
+ // default BREAK definition
+ if (!parsedbreaktable) {
+ breaktable.push_back("-");
+ breaktable.push_back("^-");
+ breaktable.push_back("-$");
+ parsedbreaktable = true;
+ }
+ return 0;
+}
+
+// we want to be able to quickly access prefix information
+// both by prefix flag, and sorted by prefix string itself
+// so we need to set up two indexes
+
+int AffixMgr::build_pfxtree(PfxEntry* pfxptr) {
+ PfxEntry* ptr;
+ PfxEntry* pptr;
+ PfxEntry* ep = pfxptr;
+
+ // get the right starting points
+ const char* key = ep->getKey();
+ const unsigned char flg = (unsigned char)(ep->getFlag() & 0x00FF);
+
+ // first index by flag which must exist
+ ptr = pFlag[flg];
+ ep->setFlgNxt(ptr);
+ pFlag[flg] = ep;
+
+ // handle the special case of null affix string
+ if (strlen(key) == 0) {
+ // always inset them at head of list at element 0
+ ptr = pStart[0];
+ ep->setNext(ptr);
+ pStart[0] = ep;
+ return 0;
+ }
+
+ // now handle the normal case
+ ep->setNextEQ(NULL);
+ ep->setNextNE(NULL);
+
+ unsigned char sp = *((const unsigned char*)key);
+ ptr = pStart[sp];
+
+ // handle the first insert
+ if (!ptr) {
+ pStart[sp] = ep;
+ return 0;
+ }
+
+ // otherwise use binary tree insertion so that a sorted
+ // list can easily be generated later
+ pptr = NULL;
+ for (;;) {
+ pptr = ptr;
+ if (strcmp(ep->getKey(), ptr->getKey()) <= 0) {
+ ptr = ptr->getNextEQ();
+ if (!ptr) {
+ pptr->setNextEQ(ep);
+ break;
+ }
+ } else {
+ ptr = ptr->getNextNE();
+ if (!ptr) {
+ pptr->setNextNE(ep);
+ break;
+ }
+ }
+ }
+ return 0;
+}
+
+// we want to be able to quickly access suffix information
+// both by suffix flag, and sorted by the reverse of the
+// suffix string itself; so we need to set up two indexes
+int AffixMgr::build_sfxtree(SfxEntry* sfxptr) {
+
+ sfxptr->initReverseWord();
+
+ SfxEntry* ptr;
+ SfxEntry* pptr;
+ SfxEntry* ep = sfxptr;
+
+ /* get the right starting point */
+ const char* key = ep->getKey();
+ const unsigned char flg = (unsigned char)(ep->getFlag() & 0x00FF);
+
+ // first index by flag which must exist
+ ptr = sFlag[flg];
+ ep->setFlgNxt(ptr);
+ sFlag[flg] = ep;
+
+ // next index by affix string
+
+ // handle the special case of null affix string
+ if (strlen(key) == 0) {
+ // always inset them at head of list at element 0
+ ptr = sStart[0];
+ ep->setNext(ptr);
+ sStart[0] = ep;
+ return 0;
+ }
+
+ // now handle the normal case
+ ep->setNextEQ(NULL);
+ ep->setNextNE(NULL);
+
+ unsigned char sp = *((const unsigned char*)key);
+ ptr = sStart[sp];
+
+ // handle the first insert
+ if (!ptr) {
+ sStart[sp] = ep;
+ return 0;
+ }
+
+ // otherwise use binary tree insertion so that a sorted
+ // list can easily be generated later
+ pptr = NULL;
+ for (;;) {
+ pptr = ptr;
+ if (strcmp(ep->getKey(), ptr->getKey()) <= 0) {
+ ptr = ptr->getNextEQ();
+ if (!ptr) {
+ pptr->setNextEQ(ep);
+ break;
+ }
+ } else {
+ ptr = ptr->getNextNE();
+ if (!ptr) {
+ pptr->setNextNE(ep);
+ break;
+ }
+ }
+ }
+ return 0;
+}
+
+// convert from binary tree to sorted list
+int AffixMgr::process_pfx_tree_to_list() {
+ for (int i = 1; i < SETSIZE; i++) {
+ pStart[i] = process_pfx_in_order(pStart[i], NULL);
+ }
+ return 0;
+}
+
+PfxEntry* AffixMgr::process_pfx_in_order(PfxEntry* ptr, PfxEntry* nptr) {
+ if (ptr) {
+ nptr = process_pfx_in_order(ptr->getNextNE(), nptr);
+ ptr->setNext(nptr);
+ nptr = process_pfx_in_order(ptr->getNextEQ(), ptr);
+ }
+ return nptr;
+}
+
+// convert from binary tree to sorted list
+int AffixMgr::process_sfx_tree_to_list() {
+ for (int i = 1; i < SETSIZE; i++) {
+ sStart[i] = process_sfx_in_order(sStart[i], NULL);
+ }
+ return 0;
+}
+
+SfxEntry* AffixMgr::process_sfx_in_order(SfxEntry* ptr, SfxEntry* nptr) {
+ if (ptr) {
+ nptr = process_sfx_in_order(ptr->getNextNE(), nptr);
+ ptr->setNext(nptr);
+ nptr = process_sfx_in_order(ptr->getNextEQ(), ptr);
+ }
+ return nptr;
+}
+
+// reinitialize the PfxEntry links NextEQ and NextNE to speed searching
+// using the idea of leading subsets this time
+int AffixMgr::process_pfx_order() {
+ PfxEntry* ptr;
+
+ // loop through each prefix list starting point
+ for (int i = 1; i < SETSIZE; i++) {
+ ptr = pStart[i];
+
+ // look through the remainder of the list
+ // and find next entry with affix that
+ // the current one is not a subset of
+ // mark that as destination for NextNE
+ // use next in list that you are a subset
+ // of as NextEQ
+
+ for (; ptr != NULL; ptr = ptr->getNext()) {
+ PfxEntry* nptr = ptr->getNext();
+ for (; nptr != NULL; nptr = nptr->getNext()) {
+ if (!isSubset(ptr->getKey(), nptr->getKey()))
+ break;
+ }
+ ptr->setNextNE(nptr);
+ ptr->setNextEQ(NULL);
+ if ((ptr->getNext()) &&
+ isSubset(ptr->getKey(), (ptr->getNext())->getKey()))
+ ptr->setNextEQ(ptr->getNext());
+ }
+
+ // now clean up by adding smart search termination strings:
+ // if you are already a superset of the previous prefix
+ // but not a subset of the next, search can end here
+ // so set NextNE properly
+
+ ptr = pStart[i];
+ for (; ptr != NULL; ptr = ptr->getNext()) {
+ PfxEntry* nptr = ptr->getNext();
+ PfxEntry* mptr = NULL;
+ for (; nptr != NULL; nptr = nptr->getNext()) {
+ if (!isSubset(ptr->getKey(), nptr->getKey()))
+ break;
+ mptr = nptr;
+ }
+ if (mptr)
+ mptr->setNextNE(NULL);
+ }
+ }
+ return 0;
+}
+
+// initialize the SfxEntry links NextEQ and NextNE to speed searching
+// using the idea of leading subsets this time
+int AffixMgr::process_sfx_order() {
+ SfxEntry* ptr;
+
+ // loop through each prefix list starting point
+ for (int i = 1; i < SETSIZE; i++) {
+ ptr = sStart[i];
+
+ // look through the remainder of the list
+ // and find next entry with affix that
+ // the current one is not a subset of
+ // mark that as destination for NextNE
+ // use next in list that you are a subset
+ // of as NextEQ
+
+ for (; ptr != NULL; ptr = ptr->getNext()) {
+ SfxEntry* nptr = ptr->getNext();
+ for (; nptr != NULL; nptr = nptr->getNext()) {
+ if (!isSubset(ptr->getKey(), nptr->getKey()))
+ break;
+ }
+ ptr->setNextNE(nptr);
+ ptr->setNextEQ(NULL);
+ if ((ptr->getNext()) &&
+ isSubset(ptr->getKey(), (ptr->getNext())->getKey()))
+ ptr->setNextEQ(ptr->getNext());
+ }
+
+ // now clean up by adding smart search termination strings:
+ // if you are already a superset of the previous suffix
+ // but not a subset of the next, search can end here
+ // so set NextNE properly
+
+ ptr = sStart[i];
+ for (; ptr != NULL; ptr = ptr->getNext()) {
+ SfxEntry* nptr = ptr->getNext();
+ SfxEntry* mptr = NULL;
+ for (; nptr != NULL; nptr = nptr->getNext()) {
+ if (!isSubset(ptr->getKey(), nptr->getKey()))
+ break;
+ mptr = nptr;
+ }
+ if (mptr)
+ mptr->setNextNE(NULL);
+ }
+ }
+ return 0;
+}
+
+// add flags to the result for dictionary debugging
+std::string& AffixMgr::debugflag(std::string& result, unsigned short flag) {
+ char* st = encode_flag(flag);
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_FLAG);
+ if (st) {
+ result.append(st);
+ free(st);
+ }
+ return result;
+}
+
+// calculate the character length of the condition
+int AffixMgr::condlen(const char* st) {
+ int l = 0;
+ bool group = false;
+ for (; *st; st++) {
+ if (*st == '[') {
+ group = true;
+ l++;
+ } else if (*st == ']')
+ group = false;
+ else if (!group && (!utf8 || (!(*st & 0x80) || ((*st & 0xc0) == 0x80))))
+ l++;
+ }
+ return l;
+}
+
+int AffixMgr::encodeit(AffEntry& entry, const char* cs) {
+ if (strcmp(cs, ".") != 0) {
+ entry.numconds = (char)condlen(cs);
+ const size_t cslen = strlen(cs);
+ const size_t short_part = std::min<size_t>(MAXCONDLEN, cslen);
+ memcpy(entry.c.conds, cs, short_part);
+ if (short_part < MAXCONDLEN) {
+ //blank out the remaining space
+ memset(entry.c.conds + short_part, 0, MAXCONDLEN - short_part);
+ } else if (cs[MAXCONDLEN]) {
+ //there is more conditions than fit in fixed space, so its
+ //a long condition
+ entry.opts |= aeLONGCOND;
+ entry.c.l.conds2 = mystrdup(cs + MAXCONDLEN_1);
+ if (!entry.c.l.conds2)
+ return 1;
+ }
+ } else {
+ entry.numconds = 0;
+ entry.c.conds[0] = '\0';
+ }
+ return 0;
+}
+
+// return 1 if s1 is a leading subset of s2 (dots are for infixes)
+inline int AffixMgr::isSubset(const char* s1, const char* s2) {
+ while (((*s1 == *s2) || (*s1 == '.')) && (*s1 != '\0')) {
+ s1++;
+ s2++;
+ }
+ return (*s1 == '\0');
+}
+
+// check word for prefixes
+struct hentry* AffixMgr::prefix_check(const char* word,
+ int len,
+ char in_compound,
+ const FLAG needflag) {
+ struct hentry* rv = NULL;
+
+ pfx = NULL;
+ pfxappnd = NULL;
+ sfxappnd = NULL;
+ sfxextra = 0;
+
+ // first handle the special case of 0 length prefixes
+ PfxEntry* pe = pStart[0];
+ while (pe) {
+ if (
+ // fogemorpheme
+ ((in_compound != IN_CPD_NOT) ||
+ !(pe->getCont() &&
+ (TESTAFF(pe->getCont(), onlyincompound, pe->getContLen())))) &&
+ // permit prefixes in compounds
+ ((in_compound != IN_CPD_END) ||
+ (pe->getCont() &&
+ (TESTAFF(pe->getCont(), compoundpermitflag, pe->getContLen()))))) {
+ // check prefix
+ rv = pe->checkword(word, len, in_compound, needflag);
+ if (rv) {
+ pfx = pe; // BUG: pfx not stateless
+ return rv;
+ }
+ }
+ pe = pe->getNext();
+ }
+
+ // now handle the general case
+ unsigned char sp = *((const unsigned char*)word);
+ PfxEntry* pptr = pStart[sp];
+
+ while (pptr) {
+ if (isSubset(pptr->getKey(), word)) {
+ if (
+ // fogemorpheme
+ ((in_compound != IN_CPD_NOT) ||
+ !(pptr->getCont() &&
+ (TESTAFF(pptr->getCont(), onlyincompound, pptr->getContLen())))) &&
+ // permit prefixes in compounds
+ ((in_compound != IN_CPD_END) ||
+ (pptr->getCont() && (TESTAFF(pptr->getCont(), compoundpermitflag,
+ pptr->getContLen()))))) {
+ // check prefix
+ rv = pptr->checkword(word, len, in_compound, needflag);
+ if (rv) {
+ pfx = pptr; // BUG: pfx not stateless
+ return rv;
+ }
+ }
+ pptr = pptr->getNextEQ();
+ } else {
+ pptr = pptr->getNextNE();
+ }
+ }
+
+ return NULL;
+}
+
+// check word for prefixes and two-level suffixes
+struct hentry* AffixMgr::prefix_check_twosfx(const char* word,
+ int len,
+ char in_compound,
+ const FLAG needflag) {
+ struct hentry* rv = NULL;
+
+ pfx = NULL;
+ sfxappnd = NULL;
+ sfxextra = 0;
+
+ // first handle the special case of 0 length prefixes
+ PfxEntry* pe = pStart[0];
+
+ while (pe) {
+ rv = pe->check_twosfx(word, len, in_compound, needflag);
+ if (rv)
+ return rv;
+ pe = pe->getNext();
+ }
+
+ // now handle the general case
+ unsigned char sp = *((const unsigned char*)word);
+ PfxEntry* pptr = pStart[sp];
+
+ while (pptr) {
+ if (isSubset(pptr->getKey(), word)) {
+ rv = pptr->check_twosfx(word, len, in_compound, needflag);
+ if (rv) {
+ pfx = pptr;
+ return rv;
+ }
+ pptr = pptr->getNextEQ();
+ } else {
+ pptr = pptr->getNextNE();
+ }
+ }
+
+ return NULL;
+}
+
+// check word for prefixes and morph
+std::string AffixMgr::prefix_check_morph(const char* word,
+ int len,
+ char in_compound,
+ const FLAG needflag) {
+
+ std::string result;
+
+ pfx = NULL;
+ sfxappnd = NULL;
+ sfxextra = 0;
+
+ // first handle the special case of 0 length prefixes
+ PfxEntry* pe = pStart[0];
+ while (pe) {
+ std::string st = pe->check_morph(word, len, in_compound, needflag);
+ if (!st.empty()) {
+ result.append(st);
+ }
+ pe = pe->getNext();
+ }
+
+ // now handle the general case
+ unsigned char sp = *((const unsigned char*)word);
+ PfxEntry* pptr = pStart[sp];
+
+ while (pptr) {
+ if (isSubset(pptr->getKey(), word)) {
+ std::string st = pptr->check_morph(word, len, in_compound, needflag);
+ if (!st.empty()) {
+ // fogemorpheme
+ if ((in_compound != IN_CPD_NOT) ||
+ !((pptr->getCont() && (TESTAFF(pptr->getCont(), onlyincompound,
+ pptr->getContLen()))))) {
+ result.append(st);
+ pfx = pptr;
+ }
+ }
+ pptr = pptr->getNextEQ();
+ } else {
+ pptr = pptr->getNextNE();
+ }
+ }
+
+ return result;
+}
+
+// check word for prefixes and morph and two-level suffixes
+std::string AffixMgr::prefix_check_twosfx_morph(const char* word,
+ int len,
+ char in_compound,
+ const FLAG needflag) {
+ std::string result;
+
+ pfx = NULL;
+ sfxappnd = NULL;
+ sfxextra = 0;
+
+ // first handle the special case of 0 length prefixes
+ PfxEntry* pe = pStart[0];
+ while (pe) {
+ std::string st = pe->check_twosfx_morph(word, len, in_compound, needflag);
+ if (!st.empty()) {
+ result.append(st);
+ }
+ pe = pe->getNext();
+ }
+
+ // now handle the general case
+ unsigned char sp = *((const unsigned char*)word);
+ PfxEntry* pptr = pStart[sp];
+
+ while (pptr) {
+ if (isSubset(pptr->getKey(), word)) {
+ std::string st = pptr->check_twosfx_morph(word, len, in_compound, needflag);
+ if (!st.empty()) {
+ result.append(st);
+ pfx = pptr;
+ }
+ pptr = pptr->getNextEQ();
+ } else {
+ pptr = pptr->getNextNE();
+ }
+ }
+
+ return result;
+}
+
+// Is word a non-compound with a REP substitution (see checkcompoundrep)?
+int AffixMgr::cpdrep_check(const char* word, int wl) {
+
+ if ((wl < 2) || get_reptable().empty())
+ return 0;
+
+ for (size_t i = 0; i < get_reptable().size(); ++i) {
+ // use only available mid patterns
+ if (!get_reptable()[i].outstrings[0].empty()) {
+ const char* r = word;
+ const size_t lenp = get_reptable()[i].pattern.size();
+ // search every occurence of the pattern in the word
+ while ((r = strstr(r, get_reptable()[i].pattern.c_str())) != NULL) {
+ std::string candidate(word);
+ candidate.replace(r - word, lenp, get_reptable()[i].outstrings[0]);
+ if (candidate_check(candidate.c_str(), candidate.size()))
+ return 1;
+ ++r; // search for the next letter
+ }
+ }
+ }
+
+ return 0;
+}
+
+// forbid compound words, if they are in the dictionary as a
+// word pair separated by space
+int AffixMgr::cpdwordpair_check(const char * word, int wl) {
+ if (wl > 2) {
+ std::string candidate(word);
+ for (size_t i = 1; i < candidate.size(); i++) {
+ // go to end of the UTF-8 character
+ if (utf8 && ((word[i] & 0xc0) == 0x80))
+ continue;
+ candidate.insert(i, 1, ' ');
+ if (candidate_check(candidate.c_str(), candidate.size()))
+ return 1;
+ candidate.erase(i, 1);
+ }
+ }
+
+ return 0;
+}
+
+// forbid compoundings when there are special patterns at word bound
+int AffixMgr::cpdpat_check(const char* word,
+ int pos,
+ hentry* r1,
+ hentry* r2,
+ const char /*affixed*/) {
+ for (size_t i = 0; i < checkcpdtable.size(); ++i) {
+ size_t len;
+ if (isSubset(checkcpdtable[i].pattern2.c_str(), word + pos) &&
+ (!r1 || !checkcpdtable[i].cond ||
+ (r1->astr && TESTAFF(r1->astr, checkcpdtable[i].cond, r1->alen))) &&
+ (!r2 || !checkcpdtable[i].cond2 ||
+ (r2->astr && TESTAFF(r2->astr, checkcpdtable[i].cond2, r2->alen))) &&
+ // zero length pattern => only TESTAFF
+ // zero pattern (0/flag) => unmodified stem (zero affixes allowed)
+ (checkcpdtable[i].pattern.empty() ||
+ ((checkcpdtable[i].pattern[0] == '0' && r1->blen <= pos &&
+ strncmp(word + pos - r1->blen, r1->word, r1->blen) == 0) ||
+ (checkcpdtable[i].pattern[0] != '0' &&
+ ((len = checkcpdtable[i].pattern.size()) != 0) &&
+ strncmp(word + pos - len, checkcpdtable[i].pattern.c_str(), len) == 0)))) {
+ return 1;
+ }
+ }
+ return 0;
+}
+
+// forbid compounding with neighbouring upper and lower case characters at word
+// bounds
+int AffixMgr::cpdcase_check(const char* word, int pos) {
+ if (utf8) {
+ const char* p;
+ for (p = word + pos - 1; (*p & 0xc0) == 0x80; p--)
+ ;
+ std::string pair(p);
+ std::vector<w_char> pair_u;
+ u8_u16(pair_u, pair);
+ unsigned short a = pair_u.size() > 1 ? ((pair_u[1].h << 8) + pair_u[1].l) : 0;
+ unsigned short b = !pair_u.empty() ? ((pair_u[0].h << 8) + pair_u[0].l) : 0;
+ if (((unicodetoupper(a, langnum) == a) ||
+ (unicodetoupper(b, langnum) == b)) &&
+ (a != '-') && (b != '-'))
+ return 1;
+ } else {
+ unsigned char a = *(word + pos - 1);
+ unsigned char b = *(word + pos);
+ if ((csconv[a].ccase || csconv[b].ccase) && (a != '-') && (b != '-'))
+ return 1;
+ }
+ return 0;
+}
+
+struct metachar_data {
+ signed short btpp; // metacharacter (*, ?) position for backtracking
+ signed short btwp; // word position for metacharacters
+ int btnum; // number of matched characters in metacharacter
+};
+
+// check compound patterns
+int AffixMgr::defcpd_check(hentry*** words,
+ short wnum,
+ hentry* rv,
+ hentry** def,
+ char all) {
+ int w = 0;
+
+ if (!*words) {
+ w = 1;
+ *words = def;
+ }
+
+ if (!*words) {
+ return 0;
+ }
+
+ std::vector<metachar_data> btinfo(1);
+
+ short bt = 0;
+
+ (*words)[wnum] = rv;
+
+ // has the last word COMPOUNDRULE flag?
+ if (rv->alen == 0) {
+ (*words)[wnum] = NULL;
+ if (w)
+ *words = NULL;
+ return 0;
+ }
+ int ok = 0;
+ for (size_t i = 0; i < defcpdtable.size(); ++i) {
+ for (size_t j = 0; j < defcpdtable[i].size(); ++j) {
+ if (defcpdtable[i][j] != '*' && defcpdtable[i][j] != '?' &&
+ TESTAFF(rv->astr, defcpdtable[i][j], rv->alen)) {
+ ok = 1;
+ break;
+ }
+ }
+ }
+ if (ok == 0) {
+ (*words)[wnum] = NULL;
+ if (w)
+ *words = NULL;
+ return 0;
+ }
+
+ for (size_t i = 0; i < defcpdtable.size(); ++i) {
+ size_t pp = 0; // pattern position
+ signed short wp = 0; // "words" position
+ int ok2;
+ ok = 1;
+ ok2 = 1;
+ do {
+ while ((pp < defcpdtable[i].size()) && (wp <= wnum)) {
+ if (((pp + 1) < defcpdtable[i].size()) &&
+ ((defcpdtable[i][pp + 1] == '*') ||
+ (defcpdtable[i][pp + 1] == '?'))) {
+ int wend = (defcpdtable[i][pp + 1] == '?') ? wp : wnum;
+ ok2 = 1;
+ pp += 2;
+ btinfo[bt].btpp = pp;
+ btinfo[bt].btwp = wp;
+ while (wp <= wend) {
+ if (!(*words)[wp]->alen ||
+ !TESTAFF((*words)[wp]->astr, defcpdtable[i][pp - 2],
+ (*words)[wp]->alen)) {
+ ok2 = 0;
+ break;
+ }
+ wp++;
+ }
+ if (wp <= wnum)
+ ok2 = 0;
+ btinfo[bt].btnum = wp - btinfo[bt].btwp;
+ if (btinfo[bt].btnum > 0) {
+ ++bt;
+ btinfo.resize(bt+1);
+ }
+ if (ok2)
+ break;
+ } else {
+ ok2 = 1;
+ if (!(*words)[wp] || !(*words)[wp]->alen ||
+ !TESTAFF((*words)[wp]->astr, defcpdtable[i][pp],
+ (*words)[wp]->alen)) {
+ ok = 0;
+ break;
+ }
+ pp++;
+ wp++;
+ if ((defcpdtable[i].size() == pp) && !(wp > wnum))
+ ok = 0;
+ }
+ }
+ if (ok && ok2) {
+ size_t r = pp;
+ while ((defcpdtable[i].size() > r) && ((r + 1) < defcpdtable[i].size()) &&
+ ((defcpdtable[i][r + 1] == '*') ||
+ (defcpdtable[i][r + 1] == '?')))
+ r += 2;
+ if (defcpdtable[i].size() <= r)
+ return 1;
+ }
+ // backtrack
+ if (bt)
+ do {
+ ok = 1;
+ btinfo[bt - 1].btnum--;
+ pp = btinfo[bt - 1].btpp;
+ wp = btinfo[bt - 1].btwp + (signed short)btinfo[bt - 1].btnum;
+ } while ((btinfo[bt - 1].btnum < 0) && --bt);
+ } while (bt);
+
+ if (ok && ok2 && (!all || (defcpdtable[i].size() <= pp)))
+ return 1;
+
+ // check zero ending
+ while (ok && ok2 && (defcpdtable[i].size() > pp) &&
+ ((pp + 1) < defcpdtable[i].size()) &&
+ ((defcpdtable[i][pp + 1] == '*') ||
+ (defcpdtable[i][pp + 1] == '?')))
+ pp += 2;
+ if (ok && ok2 && (defcpdtable[i].size() <= pp))
+ return 1;
+ }
+ (*words)[wnum] = NULL;
+ if (w)
+ *words = NULL;
+ return 0;
+}
+
+inline int AffixMgr::candidate_check(const char* word, int len) {
+
+ struct hentry* rv = lookup(word);
+ if (rv)
+ return 1;
+
+ // rv = prefix_check(word,len,1);
+ // if (rv) return 1;
+
+ rv = affix_check(word, len);
+ if (rv)
+ return 1;
+ return 0;
+}
+
+// calculate number of syllable for compound-checking
+short AffixMgr::get_syllable(const std::string& word) {
+ if (cpdmaxsyllable == 0)
+ return 0;
+
+ short num = 0;
+
+ if (!utf8) {
+ for (size_t i = 0; i < word.size(); ++i) {
+ if (std::binary_search(cpdvowels.begin(), cpdvowels.end(),
+ word[i])) {
+ ++num;
+ }
+ }
+ } else if (!cpdvowels_utf16.empty()) {
+ std::vector<w_char> w;
+ u8_u16(w, word);
+ for (size_t i = 0; i < w.size(); ++i) {
+ if (std::binary_search(cpdvowels_utf16.begin(),
+ cpdvowels_utf16.end(),
+ w[i])) {
+ ++num;
+ }
+ }
+ }
+
+ return num;
+}
+
+void AffixMgr::setcminmax(int* cmin, int* cmax, const char* word, int len) {
+ if (utf8) {
+ int i;
+ for (*cmin = 0, i = 0; (i < cpdmin) && *cmin < len; i++) {
+ for ((*cmin)++; *cmin < len && (word[*cmin] & 0xc0) == 0x80; (*cmin)++)
+ ;
+ }
+ for (*cmax = len, i = 0; (i < (cpdmin - 1)) && *cmax >= 0; i++) {
+ for ((*cmax)--; *cmax >= 0 && (word[*cmax] & 0xc0) == 0x80; (*cmax)--)
+ ;
+ }
+ } else {
+ *cmin = cpdmin;
+ *cmax = len - cpdmin + 1;
+ }
+}
+
+// check if compound word is correctly spelled
+// hu_mov_rule = spec. Hungarian rule (XXX)
+struct hentry* AffixMgr::compound_check(const std::string& word,
+ short wordnum,
+ short numsyllable,
+ short maxwordnum,
+ short wnum,
+ hentry** words = NULL,
+ hentry** rwords = NULL,
+ char hu_mov_rule = 0,
+ char is_sug = 0,
+ int* info = NULL) {
+ int i;
+ short oldnumsyllable, oldnumsyllable2, oldwordnum, oldwordnum2;
+ struct hentry* rv = NULL;
+ struct hentry* rv_first;
+ std::string st;
+ char ch = '\0';
+ int cmin;
+ int cmax;
+ int striple = 0;
+ size_t scpd = 0;
+ int soldi = 0;
+ int oldcmin = 0;
+ int oldcmax = 0;
+ int oldlen = 0;
+ int checkedstriple = 0;
+ char affixed = 0;
+ hentry** oldwords = words;
+ size_t len = word.size();
+
+ int checked_prefix;
+
+ // add a time limit to handle possible
+ // combinatorical explosion of the overlapping words
+
+ HUNSPELL_THREAD_LOCAL clock_t timelimit;
+
+ if (wordnum == 0) {
+ // get the start time, seeing as we're reusing this set to 0
+ // to flag timeout, use clock() + 1 to avoid start clock()
+ // of 0 as being a timeout
+ timelimit = clock() + 1;
+ }
+ else if (timelimit != 0 && (clock() > timelimit + TIMELIMIT)) {
+ timelimit = 0;
+ }
+
+ setcminmax(&cmin, &cmax, word.c_str(), len);
+
+ st.assign(word);
+
+ for (i = cmin; i < cmax; i++) {
+ // go to end of the UTF-8 character
+ if (utf8) {
+ for (; (st[i] & 0xc0) == 0x80; i++)
+ ;
+ if (i >= cmax)
+ return NULL;
+ }
+
+ words = oldwords;
+ int onlycpdrule = (words) ? 1 : 0;
+
+ do { // onlycpdrule loop
+
+ oldnumsyllable = numsyllable;
+ oldwordnum = wordnum;
+ checked_prefix = 0;
+
+ do { // simplified checkcompoundpattern loop
+
+ if (timelimit == 0)
+ return 0;
+
+ if (scpd > 0) {
+ for (; scpd <= checkcpdtable.size() &&
+ (checkcpdtable[scpd - 1].pattern3.empty() ||
+ strncmp(word.c_str() + i, checkcpdtable[scpd - 1].pattern3.c_str(),
+ checkcpdtable[scpd - 1].pattern3.size()) != 0);
+ scpd++)
+ ;
+
+ if (scpd > checkcpdtable.size())
+ break; // break simplified checkcompoundpattern loop
+ st.replace(i, std::string::npos, checkcpdtable[scpd - 1].pattern);
+ soldi = i;
+ i += checkcpdtable[scpd - 1].pattern.size();
+ st.replace(i, std::string::npos, checkcpdtable[scpd - 1].pattern2);
+ st.replace(i + checkcpdtable[scpd - 1].pattern2.size(), std::string::npos,
+ word.substr(soldi + checkcpdtable[scpd - 1].pattern3.size()));
+
+ oldlen = len;
+ len += checkcpdtable[scpd - 1].pattern.size() +
+ checkcpdtable[scpd - 1].pattern2.size() -
+ checkcpdtable[scpd - 1].pattern3.size();
+ oldcmin = cmin;
+ oldcmax = cmax;
+ setcminmax(&cmin, &cmax, st.c_str(), len);
+
+ cmax = len - cpdmin + 1;
+ }
+
+ ch = st[i];
+ st[i] = '\0';
+
+ sfx = NULL;
+ pfx = NULL;
+
+ // FIRST WORD
+
+ affixed = 1;
+ rv = lookup(st.c_str()); // perhaps without prefix
+
+ // forbid dictionary stems with COMPOUNDFORBIDFLAG in
+ // compound words, overriding the effect of COMPOUNDPERMITFLAG
+ if ((rv) && compoundforbidflag &&
+ TESTAFF(rv->astr, compoundforbidflag, rv->alen) && !hu_mov_rule)
+ continue;
+
+ // search homonym with compound flag
+ while ((rv) && !hu_mov_rule &&
+ ((needaffix && TESTAFF(rv->astr, needaffix, rv->alen)) ||
+ !((compoundflag && !words && !onlycpdrule &&
+ TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ (compoundbegin && !wordnum && !onlycpdrule &&
+ TESTAFF(rv->astr, compoundbegin, rv->alen)) ||
+ (compoundmiddle && wordnum && !words && !onlycpdrule &&
+ TESTAFF(rv->astr, compoundmiddle, rv->alen)) ||
+ (!defcpdtable.empty() && onlycpdrule &&
+ ((!words && !wordnum &&
+ defcpd_check(&words, wnum, rv, rwords, 0)) ||
+ (words &&
+ defcpd_check(&words, wnum, rv, rwords, 0))))) ||
+ (scpd != 0 && checkcpdtable[scpd - 1].cond != FLAG_NULL &&
+ !TESTAFF(rv->astr, checkcpdtable[scpd - 1].cond, rv->alen)))) {
+ rv = rv->next_homonym;
+ }
+
+ if (rv)
+ affixed = 0;
+
+ if (!rv) {
+ if (onlycpdrule)
+ break;
+ if (compoundflag &&
+ !(rv = prefix_check(st.c_str(), i,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN,
+ compoundflag))) {
+ if (((rv = suffix_check(
+ st.c_str(), i, 0, NULL, FLAG_NULL, compoundflag,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN)) ||
+ (compoundmoresuffixes &&
+ (rv = suffix_check_twosfx(st.c_str(), i, 0, NULL, compoundflag)))) &&
+ !hu_mov_rule && sfx->getCont() &&
+ ((compoundforbidflag &&
+ TESTAFF(sfx->getCont(), compoundforbidflag,
+ sfx->getContLen())) ||
+ (compoundend &&
+ TESTAFF(sfx->getCont(), compoundend, sfx->getContLen())))) {
+ rv = NULL;
+ }
+ }
+
+ if (rv ||
+ (((wordnum == 0) && compoundbegin &&
+ ((rv = suffix_check(
+ st.c_str(), i, 0, NULL, FLAG_NULL, compoundbegin,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN)) ||
+ (compoundmoresuffixes &&
+ (rv = suffix_check_twosfx(
+ st.c_str(), i, 0, NULL,
+ compoundbegin))) || // twofold suffixes + compound
+ (rv = prefix_check(st.c_str(), i,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN,
+ compoundbegin)))) ||
+ ((wordnum > 0) && compoundmiddle &&
+ ((rv = suffix_check(
+ st.c_str(), i, 0, NULL, FLAG_NULL, compoundmiddle,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN)) ||
+ (compoundmoresuffixes &&
+ (rv = suffix_check_twosfx(
+ st.c_str(), i, 0, NULL,
+ compoundmiddle))) || // twofold suffixes + compound
+ (rv = prefix_check(st.c_str(), i,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN,
+ compoundmiddle))))))
+ checked_prefix = 1;
+ // else check forbiddenwords and needaffix
+ } else if (rv->astr && (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, needaffix, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen) ||
+ (is_sug && nosuggest &&
+ TESTAFF(rv->astr, nosuggest, rv->alen)))) {
+ st[i] = ch;
+ // continue;
+ break;
+ }
+
+ // check non_compound flag in suffix and prefix
+ if ((rv) && !hu_mov_rule &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundforbidflag, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundforbidflag,
+ sfx->getContLen())))) {
+ rv = NULL;
+ }
+
+ // check compoundend flag in suffix and prefix
+ if ((rv) && !checked_prefix && compoundend && !hu_mov_rule &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundend, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundend, sfx->getContLen())))) {
+ rv = NULL;
+ }
+
+ // check compoundmiddle flag in suffix and prefix
+ if ((rv) && !checked_prefix && (wordnum == 0) && compoundmiddle &&
+ !hu_mov_rule &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundmiddle, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundmiddle, sfx->getContLen())))) {
+ rv = NULL;
+ }
+
+ // check forbiddenwords
+ if ((rv) && (rv->astr) &&
+ (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen) ||
+ (is_sug && nosuggest && TESTAFF(rv->astr, nosuggest, rv->alen)))) {
+ return NULL;
+ }
+
+ // increment word number, if the second root has a compoundroot flag
+ if ((rv) && compoundroot &&
+ (TESTAFF(rv->astr, compoundroot, rv->alen))) {
+ wordnum++;
+ }
+
+ // first word is acceptable in compound words?
+ if (((rv) &&
+ (checked_prefix || (words && words[wnum]) ||
+ (compoundflag && TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ ((oldwordnum == 0) && compoundbegin &&
+ TESTAFF(rv->astr, compoundbegin, rv->alen)) ||
+ ((oldwordnum > 0) && compoundmiddle &&
+ TESTAFF(rv->astr, compoundmiddle, rv->alen))
+
+ // LANG_hu section: spec. Hungarian rule
+ || ((langnum == LANG_hu) && hu_mov_rule &&
+ (TESTAFF(
+ rv->astr, 'F',
+ rv->alen) || // XXX hardwired Hungarian dictionary codes
+ TESTAFF(rv->astr, 'G', rv->alen) ||
+ TESTAFF(rv->astr, 'H', rv->alen)))
+ // END of LANG_hu section
+ ) &&
+ (
+ // test CHECKCOMPOUNDPATTERN conditions
+ scpd == 0 || checkcpdtable[scpd - 1].cond == FLAG_NULL ||
+ TESTAFF(rv->astr, checkcpdtable[scpd - 1].cond, rv->alen)) &&
+ !((checkcompoundtriple && scpd == 0 &&
+ !words && // test triple letters
+ (word[i - 1] == word[i]) &&
+ (((i > 1) && (word[i - 1] == word[i - 2])) ||
+ ((word[i - 1] == word[i + 1])) // may be word[i+1] == '\0'
+ )) ||
+ (checkcompoundcase && scpd == 0 && !words &&
+ cpdcase_check(word.c_str(), i))))
+ // LANG_hu section: spec. Hungarian rule
+ || ((!rv) && (langnum == LANG_hu) && hu_mov_rule &&
+ (rv = affix_check(st.c_str(), i)) &&
+ (sfx && sfx->getCont() &&
+ ( // XXX hardwired Hungarian dic. codes
+ TESTAFF(sfx->getCont(), (unsigned short)'x',
+ sfx->getContLen()) ||
+ TESTAFF(
+ sfx->getCont(), (unsigned short)'%',
+ sfx->getContLen()))))) { // first word is ok condition
+
+ // LANG_hu section: spec. Hungarian rule
+ if (langnum == LANG_hu) {
+ // calculate syllable number of the word
+ numsyllable += get_syllable(st.substr(0, i));
+ // + 1 word, if syllable number of the prefix > 1 (hungarian
+ // convention)
+ if (pfx && (get_syllable(pfx->getKey()) > 1))
+ wordnum++;
+ }
+ // END of LANG_hu section
+
+ // NEXT WORD(S)
+ rv_first = rv;
+ st[i] = ch;
+
+ do { // striple loop
+
+ // check simplifiedtriple
+ if (simplifiedtriple) {
+ if (striple) {
+ checkedstriple = 1;
+ i--; // check "fahrt" instead of "ahrt" in "Schiffahrt"
+ } else if (i > 2 && word[i - 1] == word[i - 2])
+ striple = 1;
+ }
+
+ rv = lookup(st.c_str() + i); // perhaps without prefix
+
+ // search homonym with compound flag
+ while ((rv) &&
+ ((needaffix && TESTAFF(rv->astr, needaffix, rv->alen)) ||
+ !((compoundflag && !words &&
+ TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ (compoundend && !words &&
+ TESTAFF(rv->astr, compoundend, rv->alen)) ||
+ (!defcpdtable.empty() && words &&
+ defcpd_check(&words, wnum + 1, rv, NULL, 1))) ||
+ (scpd != 0 && checkcpdtable[scpd - 1].cond2 != FLAG_NULL &&
+ !TESTAFF(rv->astr, checkcpdtable[scpd - 1].cond2,
+ rv->alen)))) {
+ rv = rv->next_homonym;
+ }
+
+ // check FORCEUCASE
+ if (rv && forceucase && (rv) &&
+ (TESTAFF(rv->astr, forceucase, rv->alen)) &&
+ !(info && *info & SPELL_ORIGCAP))
+ rv = NULL;
+
+ if (rv && words && words[wnum + 1])
+ return rv_first;
+
+ oldnumsyllable2 = numsyllable;
+ oldwordnum2 = wordnum;
+
+ // LANG_hu section: spec. Hungarian rule, XXX hardwired dictionary
+ // code
+ if ((rv) && (langnum == LANG_hu) &&
+ (TESTAFF(rv->astr, 'I', rv->alen)) &&
+ !(TESTAFF(rv->astr, 'J', rv->alen))) {
+ numsyllable--;
+ }
+ // END of LANG_hu section
+
+ // increment word number, if the second root has a compoundroot flag
+ if ((rv) && (compoundroot) &&
+ (TESTAFF(rv->astr, compoundroot, rv->alen))) {
+ wordnum++;
+ }
+
+ // check forbiddenwords
+ if ((rv) && (rv->astr) &&
+ (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen) ||
+ (is_sug && nosuggest &&
+ TESTAFF(rv->astr, nosuggest, rv->alen))))
+ return NULL;
+
+ // second word is acceptable, as a root?
+ // hungarian conventions: compounding is acceptable,
+ // when compound forms consist of 2 words, or if more,
+ // then the syllable number of root words must be 6, or lesser.
+
+ if ((rv) &&
+ ((compoundflag && TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ (compoundend && TESTAFF(rv->astr, compoundend, rv->alen))) &&
+ (((cpdwordmax == -1) || (wordnum + 1 < cpdwordmax)) ||
+ ((cpdmaxsyllable != 0) &&
+ (numsyllable + get_syllable(std::string(HENTRY_WORD(rv), rv->blen)) <=
+ cpdmaxsyllable))) &&
+ (
+ // test CHECKCOMPOUNDPATTERN
+ checkcpdtable.empty() || scpd != 0 ||
+ !cpdpat_check(word.c_str(), i, rv_first, rv, 0)) &&
+ ((!checkcompounddup || (rv != rv_first)))
+ // test CHECKCOMPOUNDPATTERN conditions
+ &&
+ (scpd == 0 || checkcpdtable[scpd - 1].cond2 == FLAG_NULL ||
+ TESTAFF(rv->astr, checkcpdtable[scpd - 1].cond2, rv->alen))) {
+ // forbid compound word, if it is a non-compound word with typical
+ // fault
+ if ((checkcompoundrep && cpdrep_check(word.c_str(), len)) ||
+ cpdwordpair_check(word.c_str(), len))
+ return NULL;
+ return rv_first;
+ }
+
+ numsyllable = oldnumsyllable2;
+ wordnum = oldwordnum2;
+
+ // perhaps second word has prefix or/and suffix
+ sfx = NULL;
+ sfxflag = FLAG_NULL;
+ rv = (compoundflag && !onlycpdrule)
+ ? affix_check((word.c_str() + i), strlen(word.c_str() + i), compoundflag,
+ IN_CPD_END)
+ : NULL;
+ if (!rv && compoundend && !onlycpdrule) {
+ sfx = NULL;
+ pfx = NULL;
+ rv = affix_check((word.c_str() + i), strlen(word.c_str() + i), compoundend,
+ IN_CPD_END);
+ }
+
+ if (!rv && !defcpdtable.empty() && words) {
+ rv = affix_check((word.c_str() + i), strlen(word.c_str() + i), 0, IN_CPD_END);
+ if (rv && defcpd_check(&words, wnum + 1, rv, NULL, 1))
+ return rv_first;
+ rv = NULL;
+ }
+
+ // test CHECKCOMPOUNDPATTERN conditions (allowed forms)
+ if (rv &&
+ !(scpd == 0 || checkcpdtable[scpd - 1].cond2 == FLAG_NULL ||
+ TESTAFF(rv->astr, checkcpdtable[scpd - 1].cond2, rv->alen)))
+ rv = NULL;
+
+ // test CHECKCOMPOUNDPATTERN conditions (forbidden compounds)
+ if (rv && !checkcpdtable.empty() && scpd == 0 &&
+ cpdpat_check(word.c_str(), i, rv_first, rv, affixed))
+ rv = NULL;
+
+ // check non_compound flag in suffix and prefix
+ if ((rv) && ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundforbidflag,
+ pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundforbidflag,
+ sfx->getContLen())))) {
+ rv = NULL;
+ }
+
+ // check FORCEUCASE
+ if (rv && forceucase && (rv) &&
+ (TESTAFF(rv->astr, forceucase, rv->alen)) &&
+ !(info && *info & SPELL_ORIGCAP))
+ rv = NULL;
+
+ // check forbiddenwords
+ if ((rv) && (rv->astr) &&
+ (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen) ||
+ (is_sug && nosuggest &&
+ TESTAFF(rv->astr, nosuggest, rv->alen))))
+ return NULL;
+
+ // pfxappnd = prefix of word+i, or NULL
+ // calculate syllable number of prefix.
+ // hungarian convention: when syllable number of prefix is more,
+ // than 1, the prefix+word counts as two words.
+
+ if (langnum == LANG_hu) {
+ // calculate syllable number of the word
+ numsyllable += get_syllable(word.c_str() + i);
+
+ // - affix syllable num.
+ // XXX only second suffix (inflections, not derivations)
+ if (sfxappnd) {
+ std::string tmp(sfxappnd);
+ reverseword(tmp);
+ numsyllable -= short(get_syllable(tmp) + sfxextra);
+ } else {
+ numsyllable -= short(sfxextra);
+ }
+
+ // + 1 word, if syllable number of the prefix > 1 (hungarian
+ // convention)
+ if (pfx && (get_syllable(pfx->getKey()) > 1))
+ wordnum++;
+
+ // increment syllable num, if last word has a SYLLABLENUM flag
+ // and the suffix is beginning `s'
+
+ if (!cpdsyllablenum.empty()) {
+ switch (sfxflag) {
+ case 'c': {
+ numsyllable += 2;
+ break;
+ }
+ case 'J': {
+ numsyllable += 1;
+ break;
+ }
+ case 'I': {
+ if (rv && TESTAFF(rv->astr, 'J', rv->alen))
+ numsyllable += 1;
+ break;
+ }
+ }
+ }
+ }
+
+ // increment word number, if the second word has a compoundroot flag
+ if ((rv) && (compoundroot) &&
+ (TESTAFF(rv->astr, compoundroot, rv->alen))) {
+ wordnum++;
+ }
+ // second word is acceptable, as a word with prefix or/and suffix?
+ // hungarian conventions: compounding is acceptable,
+ // when compound forms consist 2 word, otherwise
+ // the syllable number of root words is 6, or lesser.
+ if ((rv) &&
+ (((cpdwordmax == -1) || (wordnum + 1 < cpdwordmax)) ||
+ ((cpdmaxsyllable != 0) && (numsyllable <= cpdmaxsyllable))) &&
+ ((!checkcompounddup || (rv != rv_first)))) {
+ // forbid compound word, if it is a non-compound word with typical
+ // fault
+ if ((checkcompoundrep && cpdrep_check(word.c_str(), len)) ||
+ cpdwordpair_check(word.c_str(), len))
+ return NULL;
+ return rv_first;
+ }
+
+ numsyllable = oldnumsyllable2;
+ wordnum = oldwordnum2;
+
+ // perhaps second word is a compound word (recursive call)
+ if (wordnum + 2 < maxwordnum) {
+ rv = compound_check(st.substr(i), wordnum + 1,
+ numsyllable, maxwordnum, wnum + 1, words, rwords, 0,
+ is_sug, info);
+
+ if (rv && !checkcpdtable.empty() &&
+ ((scpd == 0 &&
+ cpdpat_check(word.c_str(), i, rv_first, rv, affixed)) ||
+ (scpd != 0 &&
+ !cpdpat_check(word.c_str(), i, rv_first, rv, affixed))))
+ rv = NULL;
+ } else {
+ rv = NULL;
+ }
+ if (rv) {
+ // forbid compound word, if it is a non-compound word with typical
+ // fault, or a dictionary word pair
+
+ if (cpdwordpair_check(word.c_str(), len))
+ return NULL;
+
+ if (checkcompoundrep || forbiddenword) {
+
+ if (checkcompoundrep && cpdrep_check(word.c_str(), len))
+ return NULL;
+
+ // check first part
+ if (strncmp(rv->word, word.c_str() + i, rv->blen) == 0) {
+ char r = st[i + rv->blen];
+ st[i + rv->blen] = '\0';
+
+ if ((checkcompoundrep && cpdrep_check(st.c_str(), i + rv->blen)) ||
+ cpdwordpair_check(st.c_str(), i + rv->blen)) {
+ st[ + i + rv->blen] = r;
+ continue;
+ }
+
+ if (forbiddenword) {
+ struct hentry* rv2 = lookup(word.c_str());
+ if (!rv2)
+ rv2 = affix_check(word.c_str(), len);
+ if (rv2 && rv2->astr &&
+ TESTAFF(rv2->astr, forbiddenword, rv2->alen) &&
+ (strncmp(rv2->word, st.c_str(), i + rv->blen) == 0)) {
+ return NULL;
+ }
+ }
+ st[i + rv->blen] = r;
+ }
+ }
+ return rv_first;
+ }
+ } while (striple && !checkedstriple); // end of striple loop
+
+ if (checkedstriple) {
+ i++;
+ checkedstriple = 0;
+ striple = 0;
+ }
+
+ } // first word is ok condition
+
+ if (soldi != 0) {
+ i = soldi;
+ soldi = 0;
+ len = oldlen;
+ cmin = oldcmin;
+ cmax = oldcmax;
+ }
+ scpd++;
+
+ } while (!onlycpdrule && simplifiedcpd &&
+ scpd <= checkcpdtable.size()); // end of simplifiedcpd loop
+
+ scpd = 0;
+ wordnum = oldwordnum;
+ numsyllable = oldnumsyllable;
+
+ if (soldi != 0) {
+ i = soldi;
+ st.assign(word); // XXX add more optim.
+ soldi = 0;
+ } else
+ st[i] = ch;
+
+ } while (!defcpdtable.empty() && oldwordnum == 0 &&
+ onlycpdrule++ < 1); // end of onlycpd loop
+ }
+
+ return NULL;
+}
+
+// check if compound word is correctly spelled
+// hu_mov_rule = spec. Hungarian rule (XXX)
+int AffixMgr::compound_check_morph(const char* word,
+ int len,
+ short wordnum,
+ short numsyllable,
+ short maxwordnum,
+ short wnum,
+ hentry** words,
+ hentry** rwords,
+ char hu_mov_rule,
+ std::string& result,
+ const std::string* partresult) {
+ int i;
+ short oldnumsyllable, oldnumsyllable2, oldwordnum, oldwordnum2;
+ int ok = 0;
+
+ struct hentry* rv = NULL;
+ struct hentry* rv_first;
+ std::string st;
+ char ch;
+
+ int checked_prefix;
+ std::string presult;
+
+ int cmin;
+ int cmax;
+
+ char affixed = 0;
+ hentry** oldwords = words;
+
+ // add a time limit to handle possible
+ // combinatorical explosion of the overlapping words
+
+ HUNSPELL_THREAD_LOCAL clock_t timelimit;
+
+ if (wordnum == 0) {
+ // get the start time, seeing as we're reusing this set to 0
+ // to flag timeout, use clock() + 1 to avoid start clock()
+ // of 0 as being a timeout
+ timelimit = clock() + 1;
+ }
+ else if (timelimit != 0 && (clock() > timelimit + TIMELIMIT)) {
+ timelimit = 0;
+ }
+
+ setcminmax(&cmin, &cmax, word, len);
+
+ st.assign(word);
+
+ for (i = cmin; i < cmax; i++) {
+ // go to end of the UTF-8 character
+ if (utf8) {
+ for (; (st[i] & 0xc0) == 0x80; i++)
+ ;
+ if (i >= cmax)
+ return 0;
+ }
+
+ words = oldwords;
+ int onlycpdrule = (words) ? 1 : 0;
+
+ do { // onlycpdrule loop
+
+ if (timelimit == 0)
+ return 0;
+
+ oldnumsyllable = numsyllable;
+ oldwordnum = wordnum;
+ checked_prefix = 0;
+
+ ch = st[i];
+ st[i] = '\0';
+ sfx = NULL;
+
+ // FIRST WORD
+
+ affixed = 1;
+
+ presult.clear();
+ if (partresult)
+ presult.append(*partresult);
+
+ rv = lookup(st.c_str()); // perhaps without prefix
+
+ // forbid dictionary stems with COMPOUNDFORBIDFLAG in
+ // compound words, overriding the effect of COMPOUNDPERMITFLAG
+ if ((rv) && compoundforbidflag &&
+ TESTAFF(rv->astr, compoundforbidflag, rv->alen) && !hu_mov_rule)
+ continue;
+
+ // search homonym with compound flag
+ while ((rv) && !hu_mov_rule &&
+ ((needaffix && TESTAFF(rv->astr, needaffix, rv->alen)) ||
+ !((compoundflag && !words && !onlycpdrule &&
+ TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ (compoundbegin && !wordnum && !onlycpdrule &&
+ TESTAFF(rv->astr, compoundbegin, rv->alen)) ||
+ (compoundmiddle && wordnum && !words && !onlycpdrule &&
+ TESTAFF(rv->astr, compoundmiddle, rv->alen)) ||
+ (!defcpdtable.empty() && onlycpdrule &&
+ ((!words && !wordnum &&
+ defcpd_check(&words, wnum, rv, rwords, 0)) ||
+ (words &&
+ defcpd_check(&words, wnum, rv, rwords, 0))))))) {
+ rv = rv->next_homonym;
+ }
+
+ if (timelimit == 0)
+ return 0;
+
+ if (rv)
+ affixed = 0;
+
+ if (rv) {
+ presult.push_back(MSEP_FLD);
+ presult.append(MORPH_PART);
+ presult.append(st.c_str());
+ if (!HENTRY_FIND(rv, MORPH_STEM)) {
+ presult.push_back(MSEP_FLD);
+ presult.append(MORPH_STEM);
+ presult.append(st.c_str());
+ }
+ if (HENTRY_DATA(rv)) {
+ presult.push_back(MSEP_FLD);
+ presult.append(HENTRY_DATA2(rv));
+ }
+ }
+
+ if (!rv) {
+ if (compoundflag &&
+ !(rv =
+ prefix_check(st.c_str(), i, hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN,
+ compoundflag))) {
+ if (((rv = suffix_check(st.c_str(), i, 0, NULL, FLAG_NULL,
+ compoundflag,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN)) ||
+ (compoundmoresuffixes &&
+ (rv = suffix_check_twosfx(st.c_str(), i, 0, NULL, compoundflag)))) &&
+ !hu_mov_rule && sfx->getCont() &&
+ ((compoundforbidflag &&
+ TESTAFF(sfx->getCont(), compoundforbidflag,
+ sfx->getContLen())) ||
+ (compoundend &&
+ TESTAFF(sfx->getCont(), compoundend, sfx->getContLen())))) {
+ rv = NULL;
+ }
+ }
+
+ if (rv ||
+ (((wordnum == 0) && compoundbegin &&
+ ((rv = suffix_check(st.c_str(), i, 0, NULL, FLAG_NULL,
+ compoundbegin,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN)) ||
+ (compoundmoresuffixes &&
+ (rv = suffix_check_twosfx(
+ st.c_str(), i, 0, NULL,
+ compoundbegin))) || // twofold suffix+compound
+ (rv = prefix_check(st.c_str(), i,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN,
+ compoundbegin)))) ||
+ ((wordnum > 0) && compoundmiddle &&
+ ((rv = suffix_check(st.c_str(), i, 0, NULL, FLAG_NULL,
+ compoundmiddle,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN)) ||
+ (compoundmoresuffixes &&
+ (rv = suffix_check_twosfx(
+ st.c_str(), i, 0, NULL,
+ compoundmiddle))) || // twofold suffix+compound
+ (rv = prefix_check(st.c_str(), i,
+ hu_mov_rule ? IN_CPD_OTHER : IN_CPD_BEGIN,
+ compoundmiddle)))))) {
+ std::string p;
+ if (compoundflag)
+ p = affix_check_morph(st.c_str(), i, compoundflag);
+ if (p.empty()) {
+ if ((wordnum == 0) && compoundbegin) {
+ p = affix_check_morph(st.c_str(), i, compoundbegin);
+ } else if ((wordnum > 0) && compoundmiddle) {
+ p = affix_check_morph(st.c_str(), i, compoundmiddle);
+ }
+ }
+ if (!p.empty()) {
+ presult.push_back(MSEP_FLD);
+ presult.append(MORPH_PART);
+ presult.append(st.c_str());
+ line_uniq_app(p, MSEP_REC);
+ presult.append(p);
+ }
+ checked_prefix = 1;
+ }
+ // else check forbiddenwords
+ } else if (rv->astr && (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen) ||
+ TESTAFF(rv->astr, needaffix, rv->alen))) {
+ st[i] = ch;
+ continue;
+ }
+
+ // check non_compound flag in suffix and prefix
+ if ((rv) && !hu_mov_rule &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundforbidflag, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundforbidflag, sfx->getContLen())))) {
+ continue;
+ }
+
+ // check compoundend flag in suffix and prefix
+ if ((rv) && !checked_prefix && compoundend && !hu_mov_rule &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundend, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundend, sfx->getContLen())))) {
+ continue;
+ }
+
+ // check compoundmiddle flag in suffix and prefix
+ if ((rv) && !checked_prefix && (wordnum == 0) && compoundmiddle &&
+ !hu_mov_rule &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundmiddle, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundmiddle, sfx->getContLen())))) {
+ rv = NULL;
+ }
+
+ // check forbiddenwords
+ if ((rv) && (rv->astr) && (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen)))
+ continue;
+
+ // increment word number, if the second root has a compoundroot flag
+ if ((rv) && (compoundroot) &&
+ (TESTAFF(rv->astr, compoundroot, rv->alen))) {
+ wordnum++;
+ }
+
+ // first word is acceptable in compound words?
+ if (((rv) &&
+ (checked_prefix || (words && words[wnum]) ||
+ (compoundflag && TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ ((oldwordnum == 0) && compoundbegin &&
+ TESTAFF(rv->astr, compoundbegin, rv->alen)) ||
+ ((oldwordnum > 0) && compoundmiddle &&
+ TESTAFF(rv->astr, compoundmiddle, rv->alen))
+ // LANG_hu section: spec. Hungarian rule
+ || ((langnum == LANG_hu) && // hu_mov_rule
+ hu_mov_rule && (TESTAFF(rv->astr, 'F', rv->alen) ||
+ TESTAFF(rv->astr, 'G', rv->alen) ||
+ TESTAFF(rv->astr, 'H', rv->alen)))
+ // END of LANG_hu section
+ ) &&
+ !((checkcompoundtriple && !words && // test triple letters
+ (word[i - 1] == word[i]) &&
+ (((i > 1) && (word[i - 1] == word[i - 2])) ||
+ ((word[i - 1] == word[i + 1])) // may be word[i+1] == '\0'
+ )) ||
+ (
+ // test CHECKCOMPOUNDPATTERN
+ !checkcpdtable.empty() && !words &&
+ cpdpat_check(word, i, rv, NULL, affixed)) ||
+ (checkcompoundcase && !words && cpdcase_check(word, i))))
+ // LANG_hu section: spec. Hungarian rule
+ ||
+ ((!rv) && (langnum == LANG_hu) && hu_mov_rule &&
+ (rv = affix_check(st.c_str(), i)) &&
+ (sfx && sfx->getCont() &&
+ (TESTAFF(sfx->getCont(), (unsigned short)'x', sfx->getContLen()) ||
+ TESTAFF(sfx->getCont(), (unsigned short)'%', sfx->getContLen()))))
+ // END of LANG_hu section
+ ) {
+ // LANG_hu section: spec. Hungarian rule
+ if (langnum == LANG_hu) {
+ // calculate syllable number of the word
+ numsyllable += get_syllable(st.substr(0, i));
+
+ // + 1 word, if syllable number of the prefix > 1 (hungarian
+ // convention)
+ if (pfx && (get_syllable(pfx->getKey()) > 1))
+ wordnum++;
+ }
+ // END of LANG_hu section
+
+ // NEXT WORD(S)
+ rv_first = rv;
+ rv = lookup((word + i)); // perhaps without prefix
+
+ // search homonym with compound flag
+ while ((rv) && ((needaffix && TESTAFF(rv->astr, needaffix, rv->alen)) ||
+ !((compoundflag && !words &&
+ TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ (compoundend && !words &&
+ TESTAFF(rv->astr, compoundend, rv->alen)) ||
+ (!defcpdtable.empty() && words &&
+ defcpd_check(&words, wnum + 1, rv, NULL, 1))))) {
+ rv = rv->next_homonym;
+ }
+
+ if (rv && words && words[wnum + 1]) {
+ result.append(presult);
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_PART);
+ result.append(word + i);
+ if (complexprefixes && HENTRY_DATA(rv))
+ result.append(HENTRY_DATA2(rv));
+ if (!HENTRY_FIND(rv, MORPH_STEM)) {
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_STEM);
+ result.append(HENTRY_WORD(rv));
+ }
+ // store the pointer of the hash entry
+ if (!complexprefixes && HENTRY_DATA(rv)) {
+ result.push_back(MSEP_FLD);
+ result.append(HENTRY_DATA2(rv));
+ }
+ result.push_back(MSEP_REC);
+ return 0;
+ }
+
+ oldnumsyllable2 = numsyllable;
+ oldwordnum2 = wordnum;
+
+ // LANG_hu section: spec. Hungarian rule
+ if ((rv) && (langnum == LANG_hu) &&
+ (TESTAFF(rv->astr, 'I', rv->alen)) &&
+ !(TESTAFF(rv->astr, 'J', rv->alen))) {
+ numsyllable--;
+ }
+ // END of LANG_hu section
+ // increment word number, if the second root has a compoundroot flag
+ if ((rv) && (compoundroot) &&
+ (TESTAFF(rv->astr, compoundroot, rv->alen))) {
+ wordnum++;
+ }
+
+ // check forbiddenwords
+ if ((rv) && (rv->astr) &&
+ (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen))) {
+ st[i] = ch;
+ continue;
+ }
+
+ // second word is acceptable, as a root?
+ // hungarian conventions: compounding is acceptable,
+ // when compound forms consist of 2 words, or if more,
+ // then the syllable number of root words must be 6, or lesser.
+ if ((rv) &&
+ ((compoundflag && TESTAFF(rv->astr, compoundflag, rv->alen)) ||
+ (compoundend && TESTAFF(rv->astr, compoundend, rv->alen))) &&
+ (((cpdwordmax == -1) || (wordnum + 1 < cpdwordmax)) ||
+ ((cpdmaxsyllable != 0) &&
+ (numsyllable + get_syllable(std::string(HENTRY_WORD(rv), rv->blen)) <=
+ cpdmaxsyllable))) &&
+ ((!checkcompounddup || (rv != rv_first)))) {
+ // bad compound word
+ result.append(presult);
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_PART);
+ result.append(word + i);
+
+ if (HENTRY_DATA(rv)) {
+ if (complexprefixes)
+ result.append(HENTRY_DATA2(rv));
+ if (!HENTRY_FIND(rv, MORPH_STEM)) {
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_STEM);
+ result.append(HENTRY_WORD(rv));
+ }
+ // store the pointer of the hash entry
+ if (!complexprefixes) {
+ result.push_back(MSEP_FLD);
+ result.append(HENTRY_DATA2(rv));
+ }
+ }
+ result.push_back(MSEP_REC);
+ ok = 1;
+ }
+
+ numsyllable = oldnumsyllable2;
+ wordnum = oldwordnum2;
+
+ // perhaps second word has prefix or/and suffix
+ sfx = NULL;
+ sfxflag = FLAG_NULL;
+
+ if (compoundflag && !onlycpdrule)
+ rv = affix_check((word + i), strlen(word + i), compoundflag);
+ else
+ rv = NULL;
+
+ if (!rv && compoundend && !onlycpdrule) {
+ sfx = NULL;
+ pfx = NULL;
+ rv = affix_check((word + i), strlen(word + i), compoundend);
+ }
+
+ if (!rv && !defcpdtable.empty() && words) {
+ rv = affix_check((word + i), strlen(word + i), 0, IN_CPD_END);
+ if (rv && words && defcpd_check(&words, wnum + 1, rv, NULL, 1)) {
+ std::string m;
+ if (compoundflag)
+ m = affix_check_morph((word + i), strlen(word + i), compoundflag);
+ if (m.empty() && compoundend) {
+ m = affix_check_morph((word + i), strlen(word + i), compoundend);
+ }
+ result.append(presult);
+ if (!m.empty()) {
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_PART);
+ result.append(word + i);
+ line_uniq_app(m, MSEP_REC);
+ result.append(m);
+ }
+ result.push_back(MSEP_REC);
+ ok = 1;
+ }
+ }
+
+ // check non_compound flag in suffix and prefix
+ if ((rv) &&
+ ((pfx && pfx->getCont() &&
+ TESTAFF(pfx->getCont(), compoundforbidflag, pfx->getContLen())) ||
+ (sfx && sfx->getCont() &&
+ TESTAFF(sfx->getCont(), compoundforbidflag,
+ sfx->getContLen())))) {
+ rv = NULL;
+ }
+
+ // check forbiddenwords
+ if ((rv) && (rv->astr) &&
+ (TESTAFF(rv->astr, forbiddenword, rv->alen) ||
+ TESTAFF(rv->astr, ONLYUPCASEFLAG, rv->alen)) &&
+ (!TESTAFF(rv->astr, needaffix, rv->alen))) {
+ st[i] = ch;
+ continue;
+ }
+
+ if (langnum == LANG_hu) {
+ // calculate syllable number of the word
+ numsyllable += get_syllable(word + i);
+
+ // - affix syllable num.
+ // XXX only second suffix (inflections, not derivations)
+ if (sfxappnd) {
+ std::string tmp(sfxappnd);
+ reverseword(tmp);
+ numsyllable -= short(get_syllable(tmp) + sfxextra);
+ } else {
+ numsyllable -= short(sfxextra);
+ }
+
+ // + 1 word, if syllable number of the prefix > 1 (hungarian
+ // convention)
+ if (pfx && (get_syllable(pfx->getKey()) > 1))
+ wordnum++;
+
+ // increment syllable num, if last word has a SYLLABLENUM flag
+ // and the suffix is beginning `s'
+
+ if (!cpdsyllablenum.empty()) {
+ switch (sfxflag) {
+ case 'c': {
+ numsyllable += 2;
+ break;
+ }
+ case 'J': {
+ numsyllable += 1;
+ break;
+ }
+ case 'I': {
+ if (rv && TESTAFF(rv->astr, 'J', rv->alen))
+ numsyllable += 1;
+ break;
+ }
+ }
+ }
+ }
+
+ // increment word number, if the second word has a compoundroot flag
+ if ((rv) && (compoundroot) &&
+ (TESTAFF(rv->astr, compoundroot, rv->alen))) {
+ wordnum++;
+ }
+ // second word is acceptable, as a word with prefix or/and suffix?
+ // hungarian conventions: compounding is acceptable,
+ // when compound forms consist 2 word, otherwise
+ // the syllable number of root words is 6, or lesser.
+ if ((rv) &&
+ (((cpdwordmax == -1) || (wordnum + 1 < cpdwordmax)) ||
+ ((cpdmaxsyllable != 0) && (numsyllable <= cpdmaxsyllable))) &&
+ ((!checkcompounddup || (rv != rv_first)))) {
+ std::string m;
+ if (compoundflag)
+ m = affix_check_morph((word + i), strlen(word + i), compoundflag);
+ if (m.empty() && compoundend) {
+ m = affix_check_morph((word + i), strlen(word + i), compoundend);
+ }
+ result.append(presult);
+ if (!m.empty()) {
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_PART);
+ result.append(word + i);
+ line_uniq_app(m, MSEP_REC);
+ result.push_back(MSEP_FLD);
+ result.append(m);
+ }
+ result.push_back(MSEP_REC);
+ ok = 1;
+ }
+
+ numsyllable = oldnumsyllable2;
+ wordnum = oldwordnum2;
+
+ // perhaps second word is a compound word (recursive call)
+ if ((wordnum + 2 < maxwordnum) && (ok == 0)) {
+ compound_check_morph((word + i), strlen(word + i), wordnum + 1,
+ numsyllable, maxwordnum, wnum + 1, words, rwords, 0,
+ result, &presult);
+ } else {
+ rv = NULL;
+ }
+ }
+ st[i] = ch;
+ wordnum = oldwordnum;
+ numsyllable = oldnumsyllable;
+
+ } while (!defcpdtable.empty() && oldwordnum == 0 &&
+ onlycpdrule++ < 1); // end of onlycpd loop
+ }
+ return 0;
+}
+
+
+inline int AffixMgr::isRevSubset(const char* s1,
+ const char* end_of_s2,
+ int len) {
+ while ((len > 0) && (*s1 != '\0') && ((*s1 == *end_of_s2) || (*s1 == '.'))) {
+ s1++;
+ end_of_s2--;
+ len--;
+ }
+ return (*s1 == '\0');
+}
+
+// check word for suffixes
+struct hentry* AffixMgr::suffix_check(const char* word,
+ int len,
+ int sfxopts,
+ PfxEntry* ppfx,
+ const FLAG cclass,
+ const FLAG needflag,
+ char in_compound) {
+ struct hentry* rv = NULL;
+ PfxEntry* ep = ppfx;
+
+ // first handle the special case of 0 length suffixes
+ SfxEntry* se = sStart[0];
+
+ while (se) {
+ if (!cclass || se->getCont()) {
+ // suffixes are not allowed in beginning of compounds
+ if ((((in_compound != IN_CPD_BEGIN)) || // && !cclass
+ // except when signed with compoundpermitflag flag
+ (se->getCont() && compoundpermitflag &&
+ TESTAFF(se->getCont(), compoundpermitflag, se->getContLen()))) &&
+ (!circumfix ||
+ // no circumfix flag in prefix and suffix
+ ((!ppfx || !(ep->getCont()) ||
+ !TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (!se->getCont() ||
+ !(TESTAFF(se->getCont(), circumfix, se->getContLen())))) ||
+ // circumfix flag in prefix AND suffix
+ ((ppfx && (ep->getCont()) &&
+ TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (se->getCont() &&
+ (TESTAFF(se->getCont(), circumfix, se->getContLen()))))) &&
+ // fogemorpheme
+ (in_compound ||
+ !(se->getCont() &&
+ (TESTAFF(se->getCont(), onlyincompound, se->getContLen())))) &&
+ // needaffix on prefix or first suffix
+ (cclass ||
+ !(se->getCont() &&
+ TESTAFF(se->getCont(), needaffix, se->getContLen())) ||
+ (ppfx &&
+ !((ep->getCont()) &&
+ TESTAFF(ep->getCont(), needaffix, ep->getContLen()))))) {
+ rv = se->checkword(word, len, sfxopts, ppfx,
+ (FLAG)cclass, needflag,
+ (in_compound ? 0 : onlyincompound));
+ if (rv) {
+ sfx = se; // BUG: sfx not stateless
+ return rv;
+ }
+ }
+ }
+ se = se->getNext();
+ }
+
+ // now handle the general case
+ if (len == 0)
+ return NULL; // FULLSTRIP
+ unsigned char sp = *((const unsigned char*)(word + len - 1));
+ SfxEntry* sptr = sStart[sp];
+
+ while (sptr) {
+ if (isRevSubset(sptr->getKey(), word + len - 1, len)) {
+ // suffixes are not allowed in beginning of compounds
+ if ((((in_compound != IN_CPD_BEGIN)) || // && !cclass
+ // except when signed with compoundpermitflag flag
+ (sptr->getCont() && compoundpermitflag &&
+ TESTAFF(sptr->getCont(), compoundpermitflag,
+ sptr->getContLen()))) &&
+ (!circumfix ||
+ // no circumfix flag in prefix and suffix
+ ((!ppfx || !(ep->getCont()) ||
+ !TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (!sptr->getCont() ||
+ !(TESTAFF(sptr->getCont(), circumfix, sptr->getContLen())))) ||
+ // circumfix flag in prefix AND suffix
+ ((ppfx && (ep->getCont()) &&
+ TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (sptr->getCont() &&
+ (TESTAFF(sptr->getCont(), circumfix, sptr->getContLen()))))) &&
+ // fogemorpheme
+ (in_compound ||
+ !((sptr->getCont() && (TESTAFF(sptr->getCont(), onlyincompound,
+ sptr->getContLen()))))) &&
+ // needaffix on prefix or first suffix
+ (cclass ||
+ !(sptr->getCont() &&
+ TESTAFF(sptr->getCont(), needaffix, sptr->getContLen())) ||
+ (ppfx &&
+ !((ep->getCont()) &&
+ TESTAFF(ep->getCont(), needaffix, ep->getContLen())))))
+ if (in_compound != IN_CPD_END || ppfx ||
+ !(sptr->getCont() &&
+ TESTAFF(sptr->getCont(), onlyincompound, sptr->getContLen()))) {
+ rv = sptr->checkword(word, len, sfxopts, ppfx,
+ cclass, needflag,
+ (in_compound ? 0 : onlyincompound));
+ if (rv) {
+ sfx = sptr; // BUG: sfx not stateless
+ sfxflag = sptr->getFlag(); // BUG: sfxflag not stateless
+ if (!sptr->getCont())
+ sfxappnd = sptr->getKey(); // BUG: sfxappnd not stateless
+ // LANG_hu section: spec. Hungarian rule
+ else if (langnum == LANG_hu && sptr->getKeyLen() &&
+ sptr->getKey()[0] == 'i' && sptr->getKey()[1] != 'y' &&
+ sptr->getKey()[1] != 't') {
+ sfxextra = 1;
+ }
+ // END of LANG_hu section
+ return rv;
+ }
+ }
+ sptr = sptr->getNextEQ();
+ } else {
+ sptr = sptr->getNextNE();
+ }
+ }
+
+ return NULL;
+}
+
+// check word for two-level suffixes
+struct hentry* AffixMgr::suffix_check_twosfx(const char* word,
+ int len,
+ int sfxopts,
+ PfxEntry* ppfx,
+ const FLAG needflag) {
+ struct hentry* rv = NULL;
+
+ // first handle the special case of 0 length suffixes
+ SfxEntry* se = sStart[0];
+ while (se) {
+ if (contclasses[se->getFlag()]) {
+ rv = se->check_twosfx(word, len, sfxopts, ppfx, needflag);
+ if (rv)
+ return rv;
+ }
+ se = se->getNext();
+ }
+
+ // now handle the general case
+ if (len == 0)
+ return NULL; // FULLSTRIP
+ unsigned char sp = *((const unsigned char*)(word + len - 1));
+ SfxEntry* sptr = sStart[sp];
+
+ while (sptr) {
+ if (isRevSubset(sptr->getKey(), word + len - 1, len)) {
+ if (contclasses[sptr->getFlag()]) {
+ rv = sptr->check_twosfx(word, len, sfxopts, ppfx, needflag);
+ if (rv) {
+ sfxflag = sptr->getFlag(); // BUG: sfxflag not stateless
+ if (!sptr->getCont())
+ sfxappnd = sptr->getKey(); // BUG: sfxappnd not stateless
+ return rv;
+ }
+ }
+ sptr = sptr->getNextEQ();
+ } else {
+ sptr = sptr->getNextNE();
+ }
+ }
+
+ return NULL;
+}
+
+// check word for two-level suffixes and morph
+std::string AffixMgr::suffix_check_twosfx_morph(const char* word,
+ int len,
+ int sfxopts,
+ PfxEntry* ppfx,
+ const FLAG needflag) {
+ std::string result;
+ std::string result2;
+ std::string result3;
+
+ // first handle the special case of 0 length suffixes
+ SfxEntry* se = sStart[0];
+ while (se) {
+ if (contclasses[se->getFlag()]) {
+ std::string st = se->check_twosfx_morph(word, len, sfxopts, ppfx, needflag);
+ if (!st.empty()) {
+ if (ppfx) {
+ if (ppfx->getMorph()) {
+ result.append(ppfx->getMorph());
+ result.push_back(MSEP_FLD);
+ } else
+ debugflag(result, ppfx->getFlag());
+ }
+ result.append(st);
+ if (se->getMorph()) {
+ result.push_back(MSEP_FLD);
+ result.append(se->getMorph());
+ } else
+ debugflag(result, se->getFlag());
+ result.push_back(MSEP_REC);
+ }
+ }
+ se = se->getNext();
+ }
+
+ // now handle the general case
+ if (len == 0)
+ return std::string(); // FULLSTRIP
+ unsigned char sp = *((const unsigned char*)(word + len - 1));
+ SfxEntry* sptr = sStart[sp];
+
+ while (sptr) {
+ if (isRevSubset(sptr->getKey(), word + len - 1, len)) {
+ if (contclasses[sptr->getFlag()]) {
+ std::string st = sptr->check_twosfx_morph(word, len, sfxopts, ppfx, needflag);
+ if (!st.empty()) {
+ sfxflag = sptr->getFlag(); // BUG: sfxflag not stateless
+ if (!sptr->getCont())
+ sfxappnd = sptr->getKey(); // BUG: sfxappnd not stateless
+ result2.assign(st);
+
+ result3.clear();
+
+ if (sptr->getMorph()) {
+ result3.push_back(MSEP_FLD);
+ result3.append(sptr->getMorph());
+ } else
+ debugflag(result3, sptr->getFlag());
+ strlinecat(result2, result3);
+ result2.push_back(MSEP_REC);
+ result.append(result2);
+ }
+ }
+ sptr = sptr->getNextEQ();
+ } else {
+ sptr = sptr->getNextNE();
+ }
+ }
+
+ return result;
+}
+
+std::string AffixMgr::suffix_check_morph(const char* word,
+ int len,
+ int sfxopts,
+ PfxEntry* ppfx,
+ const FLAG cclass,
+ const FLAG needflag,
+ char in_compound) {
+ std::string result;
+
+ struct hentry* rv = NULL;
+
+ PfxEntry* ep = ppfx;
+
+ // first handle the special case of 0 length suffixes
+ SfxEntry* se = sStart[0];
+ while (se) {
+ if (!cclass || se->getCont()) {
+ // suffixes are not allowed in beginning of compounds
+ if (((((in_compound != IN_CPD_BEGIN)) || // && !cclass
+ // except when signed with compoundpermitflag flag
+ (se->getCont() && compoundpermitflag &&
+ TESTAFF(se->getCont(), compoundpermitflag, se->getContLen()))) &&
+ (!circumfix ||
+ // no circumfix flag in prefix and suffix
+ ((!ppfx || !(ep->getCont()) ||
+ !TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (!se->getCont() ||
+ !(TESTAFF(se->getCont(), circumfix, se->getContLen())))) ||
+ // circumfix flag in prefix AND suffix
+ ((ppfx && (ep->getCont()) &&
+ TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (se->getCont() &&
+ (TESTAFF(se->getCont(), circumfix, se->getContLen()))))) &&
+ // fogemorpheme
+ (in_compound ||
+ !((se->getCont() &&
+ (TESTAFF(se->getCont(), onlyincompound, se->getContLen()))))) &&
+ // needaffix on prefix or first suffix
+ (cclass ||
+ !(se->getCont() &&
+ TESTAFF(se->getCont(), needaffix, se->getContLen())) ||
+ (ppfx &&
+ !((ep->getCont()) &&
+ TESTAFF(ep->getCont(), needaffix, ep->getContLen()))))))
+ rv = se->checkword(word, len, sfxopts, ppfx, cclass,
+ needflag, FLAG_NULL);
+ while (rv) {
+ if (ppfx) {
+ if (ppfx->getMorph()) {
+ result.append(ppfx->getMorph());
+ result.push_back(MSEP_FLD);
+ } else
+ debugflag(result, ppfx->getFlag());
+ }
+ if (complexprefixes && HENTRY_DATA(rv))
+ result.append(HENTRY_DATA2(rv));
+ if (!HENTRY_FIND(rv, MORPH_STEM)) {
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_STEM);
+ result.append(HENTRY_WORD(rv));
+ }
+
+ if (!complexprefixes && HENTRY_DATA(rv)) {
+ result.push_back(MSEP_FLD);
+ result.append(HENTRY_DATA2(rv));
+ }
+ if (se->getMorph()) {
+ result.push_back(MSEP_FLD);
+ result.append(se->getMorph());
+ } else
+ debugflag(result, se->getFlag());
+ result.push_back(MSEP_REC);
+ rv = se->get_next_homonym(rv, sfxopts, ppfx, cclass, needflag);
+ }
+ }
+ se = se->getNext();
+ }
+
+ // now handle the general case
+ if (len == 0)
+ return std::string(); // FULLSTRIP
+ unsigned char sp = *((const unsigned char*)(word + len - 1));
+ SfxEntry* sptr = sStart[sp];
+
+ while (sptr) {
+ if (isRevSubset(sptr->getKey(), word + len - 1, len)) {
+ // suffixes are not allowed in beginning of compounds
+ if (((((in_compound != IN_CPD_BEGIN)) || // && !cclass
+ // except when signed with compoundpermitflag flag
+ (sptr->getCont() && compoundpermitflag &&
+ TESTAFF(sptr->getCont(), compoundpermitflag,
+ sptr->getContLen()))) &&
+ (!circumfix ||
+ // no circumfix flag in prefix and suffix
+ ((!ppfx || !(ep->getCont()) ||
+ !TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (!sptr->getCont() ||
+ !(TESTAFF(sptr->getCont(), circumfix, sptr->getContLen())))) ||
+ // circumfix flag in prefix AND suffix
+ ((ppfx && (ep->getCont()) &&
+ TESTAFF(ep->getCont(), circumfix, ep->getContLen())) &&
+ (sptr->getCont() &&
+ (TESTAFF(sptr->getCont(), circumfix, sptr->getContLen()))))) &&
+ // fogemorpheme
+ (in_compound ||
+ !((sptr->getCont() && (TESTAFF(sptr->getCont(), onlyincompound,
+ sptr->getContLen()))))) &&
+ // needaffix on first suffix
+ (cclass ||
+ !(sptr->getCont() &&
+ TESTAFF(sptr->getCont(), needaffix, sptr->getContLen())))))
+ rv = sptr->checkword(word, len, sfxopts, ppfx, cclass,
+ needflag, FLAG_NULL);
+ while (rv) {
+ if (ppfx) {
+ if (ppfx->getMorph()) {
+ result.append(ppfx->getMorph());
+ result.push_back(MSEP_FLD);
+ } else
+ debugflag(result, ppfx->getFlag());
+ }
+ if (complexprefixes && HENTRY_DATA(rv))
+ result.append(HENTRY_DATA2(rv));
+ if (!HENTRY_FIND(rv, MORPH_STEM)) {
+ result.push_back(MSEP_FLD);
+ result.append(MORPH_STEM);
+ result.append(HENTRY_WORD(rv));
+ }
+
+ if (!complexprefixes && HENTRY_DATA(rv)) {
+ result.push_back(MSEP_FLD);
+ result.append(HENTRY_DATA2(rv));
+ }
+
+ if (sptr->getMorph()) {
+ result.push_back(MSEP_FLD);
+ result.append(sptr->getMorph());
+ } else
+ debugflag(result, sptr->getFlag());
+ result.push_back(MSEP_REC);
+ rv = sptr->get_next_homonym(rv, sfxopts, ppfx, cclass, needflag);
+ }
+ sptr = sptr->getNextEQ();
+ } else {
+ sptr = sptr->getNextNE();
+ }
+ }
+
+ return result;
+}
+
+// check if word with affixes is correctly spelled
+struct hentry* AffixMgr::affix_check(const char* word,
+ int len,
+ const FLAG needflag,
+ char in_compound) {
+
+ // check all prefixes (also crossed with suffixes if allowed)
+ struct hentry* rv = prefix_check(word, len, in_compound, needflag);
+ if (rv)
+ return rv;
+
+ // if still not found check all suffixes
+ rv = suffix_check(word, len, 0, NULL, FLAG_NULL, needflag, in_compound);
+
+ if (havecontclass) {
+ sfx = NULL;
+ pfx = NULL;
+
+ if (rv)
+ return rv;
+ // if still not found check all two-level suffixes
+ rv = suffix_check_twosfx(word, len, 0, NULL, needflag);
+
+ if (rv)
+ return rv;
+ // if still not found check all two-level suffixes
+ rv = prefix_check_twosfx(word, len, IN_CPD_NOT, needflag);
+ }
+
+ return rv;
+}
+
+// check if word with affixes is correctly spelled
+std::string AffixMgr::affix_check_morph(const char* word,
+ int len,
+ const FLAG needflag,
+ char in_compound) {
+ std::string result;
+
+ // check all prefixes (also crossed with suffixes if allowed)
+ std::string st = prefix_check_morph(word, len, in_compound);
+ if (!st.empty()) {
+ result.append(st);
+ }
+
+ // if still not found check all suffixes
+ st = suffix_check_morph(word, len, 0, NULL, '\0', needflag, in_compound);
+ if (!st.empty()) {
+ result.append(st);
+ }
+
+ if (havecontclass) {
+ sfx = NULL;
+ pfx = NULL;
+ // if still not found check all two-level suffixes
+ st = suffix_check_twosfx_morph(word, len, 0, NULL, needflag);
+ if (!st.empty()) {
+ result.append(st);
+ }
+
+ // if still not found check all two-level suffixes
+ st = prefix_check_twosfx_morph(word, len, IN_CPD_NOT, needflag);
+ if (!st.empty()) {
+ result.append(st);
+ }
+ }
+
+ return result;
+}
+
+// morphcmp(): compare MORPH_DERI_SFX, MORPH_INFL_SFX and MORPH_TERM_SFX fields
+// in the first line of the inputs
+// return 0, if inputs equal
+// return 1, if inputs may equal with a secondary suffix
+// otherwise return -1
+static int morphcmp(const char* s, const char* t) {
+ int se = 0;
+ int te = 0;
+ const char* sl;
+ const char* tl;
+ const char* olds;
+ const char* oldt;
+ if (!s || !t)
+ return 1;
+ olds = s;
+ sl = strchr(s, '\n');
+ s = strstr(s, MORPH_DERI_SFX);
+ if (!s || (sl && sl < s))
+ s = strstr(olds, MORPH_INFL_SFX);
+ if (!s || (sl && sl < s)) {
+ s = strstr(olds, MORPH_TERM_SFX);
+ olds = NULL;
+ }
+ oldt = t;
+ tl = strchr(t, '\n');
+ t = strstr(t, MORPH_DERI_SFX);
+ if (!t || (tl && tl < t))
+ t = strstr(oldt, MORPH_INFL_SFX);
+ if (!t || (tl && tl < t)) {
+ t = strstr(oldt, MORPH_TERM_SFX);
+ oldt = NULL;
+ }
+ while (s && t && (!sl || sl > s) && (!tl || tl > t)) {
+ s += MORPH_TAG_LEN;
+ t += MORPH_TAG_LEN;
+ se = 0;
+ te = 0;
+ while ((*s == *t) && !se && !te) {
+ s++;
+ t++;
+ switch (*s) {
+ case ' ':
+ case '\n':
+ case '\t':
+ case '\0':
+ se = 1;
+ }
+ switch (*t) {
+ case ' ':
+ case '\n':
+ case '\t':
+ case '\0':
+ te = 1;
+ }
+ }
+ if (!se || !te) {
+ // not terminal suffix difference
+ if (olds)
+ return -1;
+ return 1;
+ }
+ olds = s;
+ s = strstr(s, MORPH_DERI_SFX);
+ if (!s || (sl && sl < s))
+ s = strstr(olds, MORPH_INFL_SFX);
+ if (!s || (sl && sl < s)) {
+ s = strstr(olds, MORPH_TERM_SFX);
+ olds = NULL;
+ }
+ oldt = t;
+ t = strstr(t, MORPH_DERI_SFX);
+ if (!t || (tl && tl < t))
+ t = strstr(oldt, MORPH_INFL_SFX);
+ if (!t || (tl && tl < t)) {
+ t = strstr(oldt, MORPH_TERM_SFX);
+ oldt = NULL;
+ }
+ }
+ if (!s && !t && se && te)
+ return 0;
+ return 1;
+}
+
+std::string AffixMgr::morphgen(const char* ts,
+ int wl,
+ const unsigned short* ap,
+ unsigned short al,
+ const char* morph,
+ const char* targetmorph,
+ int level) {
+ // handle suffixes
+ if (!morph)
+ return std::string();
+
+ // check substandard flag
+ if (TESTAFF(ap, substandard, al))
+ return std::string();
+
+ if (morphcmp(morph, targetmorph) == 0)
+ return ts;
+
+ size_t stemmorphcatpos;
+ std::string mymorph;
+
+ // use input suffix fields, if exist
+ if (strstr(morph, MORPH_INFL_SFX) || strstr(morph, MORPH_DERI_SFX)) {
+ mymorph.assign(morph);
+ mymorph.push_back(MSEP_FLD);
+ stemmorphcatpos = mymorph.size();
+ } else {
+ stemmorphcatpos = std::string::npos;
+ }
+
+ for (int i = 0; i < al; i++) {
+ const unsigned char c = (unsigned char)(ap[i] & 0x00FF);
+ SfxEntry* sptr = sFlag[c];
+ while (sptr) {
+ if (sptr->getFlag() == ap[i] && sptr->getMorph() &&
+ ((sptr->getContLen() == 0) ||
+ // don't generate forms with substandard affixes
+ !TESTAFF(sptr->getCont(), substandard, sptr->getContLen()))) {
+ const char* stemmorph;
+ if (stemmorphcatpos != std::string::npos) {
+ mymorph.replace(stemmorphcatpos, std::string::npos, sptr->getMorph());
+ stemmorph = mymorph.c_str();
+ } else {
+ stemmorph = sptr->getMorph();
+ }
+
+ int cmp = morphcmp(stemmorph, targetmorph);
+
+ if (cmp == 0) {
+ std::string newword = sptr->add(ts, wl);
+ if (!newword.empty()) {
+ hentry* check = pHMgr->lookup(newword.c_str()); // XXX extra dic
+ if (!check || !check->astr ||
+ !(TESTAFF(check->astr, forbiddenword, check->alen) ||
+ TESTAFF(check->astr, ONLYUPCASEFLAG, check->alen))) {
+ return newword;
+ }
+ }
+ }
+
+ // recursive call for secondary suffixes
+ if ((level == 0) && (cmp == 1) && (sptr->getContLen() > 0) &&
+ !TESTAFF(sptr->getCont(), substandard, sptr->getContLen())) {
+ std::string newword = sptr->add(ts, wl);
+ if (!newword.empty()) {
+ std::string newword2 =
+ morphgen(newword.c_str(), newword.size(), sptr->getCont(),
+ sptr->getContLen(), stemmorph, targetmorph, 1);
+
+ if (!newword2.empty()) {
+ return newword2;
+ }
+ }
+ }
+ }
+ sptr = sptr->getFlgNxt();
+ }
+ }
+ return std::string();
+}
+
+int AffixMgr::expand_rootword(struct guessword* wlst,
+ int maxn,
+ const char* ts,
+ int wl,
+ const unsigned short* ap,
+ unsigned short al,
+ const char* bad,
+ int badl,
+ const char* phon) {
+ int nh = 0;
+ // first add root word to list
+ if ((nh < maxn) &&
+ !(al && ((needaffix && TESTAFF(ap, needaffix, al)) ||
+ (onlyincompound && TESTAFF(ap, onlyincompound, al))))) {
+ wlst[nh].word = mystrdup(ts);
+ if (!wlst[nh].word)
+ return 0;
+ wlst[nh].allow = false;
+ wlst[nh].orig = NULL;
+ nh++;
+ // add special phonetic version
+ if (phon && (nh < maxn)) {
+ wlst[nh].word = mystrdup(phon);
+ if (!wlst[nh].word)
+ return nh - 1;
+ wlst[nh].allow = false;
+ wlst[nh].orig = mystrdup(ts);
+ if (!wlst[nh].orig)
+ return nh - 1;
+ nh++;
+ }
+ }
+
+ // handle suffixes
+ for (int i = 0; i < al; i++) {
+ const unsigned char c = (unsigned char)(ap[i] & 0x00FF);
+ SfxEntry* sptr = sFlag[c];
+ while (sptr) {
+ if ((sptr->getFlag() == ap[i]) &&
+ (!sptr->getKeyLen() ||
+ ((badl > sptr->getKeyLen()) &&
+ (strcmp(sptr->getAffix(), bad + badl - sptr->getKeyLen()) == 0))) &&
+ // check needaffix flag
+ !(sptr->getCont() &&
+ ((needaffix &&
+ TESTAFF(sptr->getCont(), needaffix, sptr->getContLen())) ||
+ (circumfix &&
+ TESTAFF(sptr->getCont(), circumfix, sptr->getContLen())) ||
+ (onlyincompound &&
+ TESTAFF(sptr->getCont(), onlyincompound, sptr->getContLen()))))) {
+ std::string newword = sptr->add(ts, wl);
+ if (!newword.empty()) {
+ if (nh < maxn) {
+ wlst[nh].word = mystrdup(newword.c_str());
+ wlst[nh].allow = sptr->allowCross();
+ wlst[nh].orig = NULL;
+ nh++;
+ // add special phonetic version
+ if (phon && (nh < maxn)) {
+ std::string prefix(phon);
+ std::string key(sptr->getKey());
+ reverseword(key);
+ prefix.append(key);
+ wlst[nh].word = mystrdup(prefix.c_str());
+ if (!wlst[nh].word)
+ return nh - 1;
+ wlst[nh].allow = false;
+ wlst[nh].orig = mystrdup(newword.c_str());
+ if (!wlst[nh].orig)
+ return nh - 1;
+ nh++;
+ }
+ }
+ }
+ }
+ sptr = sptr->getFlgNxt();
+ }
+ }
+
+ int n = nh;
+
+ // handle cross products of prefixes and suffixes
+ for (int j = 1; j < n; j++)
+ if (wlst[j].allow) {
+ for (int k = 0; k < al; k++) {
+ const unsigned char c = (unsigned char)(ap[k] & 0x00FF);
+ PfxEntry* cptr = pFlag[c];
+ while (cptr) {
+ if ((cptr->getFlag() == ap[k]) && cptr->allowCross() &&
+ (!cptr->getKeyLen() ||
+ ((badl > cptr->getKeyLen()) &&
+ (strncmp(cptr->getKey(), bad, cptr->getKeyLen()) == 0)))) {
+ int l1 = strlen(wlst[j].word);
+ std::string newword = cptr->add(wlst[j].word, l1);
+ if (!newword.empty()) {
+ if (nh < maxn) {
+ wlst[nh].word = mystrdup(newword.c_str());
+ wlst[nh].allow = cptr->allowCross();
+ wlst[nh].orig = NULL;
+ nh++;
+ }
+ }
+ }
+ cptr = cptr->getFlgNxt();
+ }
+ }
+ }
+
+ // now handle pure prefixes
+ for (int m = 0; m < al; m++) {
+ const unsigned char c = (unsigned char)(ap[m] & 0x00FF);
+ PfxEntry* ptr = pFlag[c];
+ while (ptr) {
+ if ((ptr->getFlag() == ap[m]) &&
+ (!ptr->getKeyLen() ||
+ ((badl > ptr->getKeyLen()) &&
+ (strncmp(ptr->getKey(), bad, ptr->getKeyLen()) == 0))) &&
+ // check needaffix flag
+ !(ptr->getCont() &&
+ ((needaffix &&
+ TESTAFF(ptr->getCont(), needaffix, ptr->getContLen())) ||
+ (circumfix &&
+ TESTAFF(ptr->getCont(), circumfix, ptr->getContLen())) ||
+ (onlyincompound &&
+ TESTAFF(ptr->getCont(), onlyincompound, ptr->getContLen()))))) {
+ std::string newword = ptr->add(ts, wl);
+ if (!newword.empty()) {
+ if (nh < maxn) {
+ wlst[nh].word = mystrdup(newword.c_str());
+ wlst[nh].allow = ptr->allowCross();
+ wlst[nh].orig = NULL;
+ nh++;
+ }
+ }
+ }
+ ptr = ptr->getFlgNxt();
+ }
+ }
+
+ return nh;
+}
+
+// return replacing table
+const std::vector<replentry>& AffixMgr::get_reptable() const {
+ return pHMgr->get_reptable();
+}
+
+// return iconv table
+RepList* AffixMgr::get_iconvtable() const {
+ if (!iconvtable)
+ return NULL;
+ return iconvtable;
+}
+
+// return oconv table
+RepList* AffixMgr::get_oconvtable() const {
+ if (!oconvtable)
+ return NULL;
+ return oconvtable;
+}
+
+// return replacing table
+struct phonetable* AffixMgr::get_phonetable() const {
+ if (!phone)
+ return NULL;
+ return phone;
+}
+
+// return character map table
+const std::vector<mapentry>& AffixMgr::get_maptable() const {
+ return maptable;
+}
+
+// return character map table
+const std::vector<std::string>& AffixMgr::get_breaktable() const {
+ return breaktable;
+}
+
+// return text encoding of dictionary
+const std::string& AffixMgr::get_encoding() {
+ if (encoding.empty())
+ encoding = SPELL_ENCODING;
+ return encoding;
+}
+
+// return text encoding of dictionary
+int AffixMgr::get_langnum() const {
+ return langnum;
+}
+
+// return double prefix option
+int AffixMgr::get_complexprefixes() const {
+ return complexprefixes;
+}
+
+// return FULLSTRIP option
+int AffixMgr::get_fullstrip() const {
+ return fullstrip;
+}
+
+FLAG AffixMgr::get_keepcase() const {
+ return keepcase;
+}
+
+FLAG AffixMgr::get_forceucase() const {
+ return forceucase;
+}
+
+FLAG AffixMgr::get_warn() const {
+ return warn;
+}
+
+int AffixMgr::get_forbidwarn() const {
+ return forbidwarn;
+}
+
+int AffixMgr::get_checksharps() const {
+ return checksharps;
+}
+
+char* AffixMgr::encode_flag(unsigned short aflag) const {
+ return pHMgr->encode_flag(aflag);
+}
+
+// return the preferred ignore string for suggestions
+const char* AffixMgr::get_ignore() const {
+ if (ignorechars.empty())
+ return NULL;
+ return ignorechars.c_str();
+}
+
+// return the preferred ignore string for suggestions
+const std::vector<w_char>& AffixMgr::get_ignore_utf16() const {
+ return ignorechars_utf16;
+}
+
+// return the keyboard string for suggestions
+char* AffixMgr::get_key_string() {
+ if (keystring.empty())
+ keystring = SPELL_KEYSTRING;
+ return mystrdup(keystring.c_str());
+}
+
+// return the preferred try string for suggestions
+char* AffixMgr::get_try_string() const {
+ if (trystring.empty())
+ return NULL;
+ return mystrdup(trystring.c_str());
+}
+
+// return the preferred try string for suggestions
+const std::string& AffixMgr::get_wordchars() const {
+ return wordchars;
+}
+
+const std::vector<w_char>& AffixMgr::get_wordchars_utf16() const {
+ return wordchars_utf16;
+}
+
+// is there compounding?
+int AffixMgr::get_compound() const {
+ return compoundflag || compoundbegin || !defcpdtable.empty();
+}
+
+// return the compound words control flag
+FLAG AffixMgr::get_compoundflag() const {
+ return compoundflag;
+}
+
+// return the forbidden words control flag
+FLAG AffixMgr::get_forbiddenword() const {
+ return forbiddenword;
+}
+
+// return the forbidden words control flag
+FLAG AffixMgr::get_nosuggest() const {
+ return nosuggest;
+}
+
+// return the forbidden words control flag
+FLAG AffixMgr::get_nongramsuggest() const {
+ return nongramsuggest;
+}
+
+// return the substandard root/affix control flag
+FLAG AffixMgr::get_substandard() const {
+ return substandard;
+}
+
+// return the forbidden words flag modify flag
+FLAG AffixMgr::get_needaffix() const {
+ return needaffix;
+}
+
+// return the onlyincompound flag
+FLAG AffixMgr::get_onlyincompound() const {
+ return onlyincompound;
+}
+
+// return the value of suffix
+const std::string& AffixMgr::get_version() const {
+ return version;
+}
+
+// utility method to look up root words in hash table
+struct hentry* AffixMgr::lookup(const char* word) {
+ struct hentry* he = NULL;
+ for (size_t i = 0; i < alldic.size() && !he; ++i) {
+ he = alldic[i]->lookup(word);
+ }
+ return he;
+}
+
+// return the value of suffix
+int AffixMgr::have_contclass() const {
+ return havecontclass;
+}
+
+// return utf8
+int AffixMgr::get_utf8() const {
+ return utf8;
+}
+
+int AffixMgr::get_maxngramsugs(void) const {
+ return maxngramsugs;
+}
+
+int AffixMgr::get_maxcpdsugs(void) const {
+ return maxcpdsugs;
+}
+
+int AffixMgr::get_maxdiff(void) const {
+ return maxdiff;
+}
+
+int AffixMgr::get_onlymaxdiff(void) const {
+ return onlymaxdiff;
+}
+
+// return nosplitsugs
+int AffixMgr::get_nosplitsugs(void) const {
+ return nosplitsugs;
+}
+
+// return sugswithdots
+int AffixMgr::get_sugswithdots(void) const {
+ return sugswithdots;
+}
+
+/* parse flag */
+bool AffixMgr::parse_flag(const std::string& line, unsigned short* out, FileMgr* af) {
+ if (*out != FLAG_NULL && !(*out >= DEFAULTFLAGS)) {
+ HUNSPELL_WARNING(
+ stderr,
+ "error: line %d: multiple definitions of an affix file parameter\n",
+ af->getlinenum());
+ return false;
+ }
+ std::string s;
+ if (!parse_string(line, s, af->getlinenum()))
+ return false;
+ *out = pHMgr->decode_flag(s.c_str());
+ return true;
+}
+
+/* parse num */
+bool AffixMgr::parse_num(const std::string& line, int* out, FileMgr* af) {
+ if (*out != -1) {
+ HUNSPELL_WARNING(
+ stderr,
+ "error: line %d: multiple definitions of an affix file parameter\n",
+ af->getlinenum());
+ return false;
+ }
+ std::string s;
+ if (!parse_string(line, s, af->getlinenum()))
+ return false;
+ *out = atoi(s.c_str());
+ return true;
+}
+
+/* parse in the max syllablecount of compound words and */
+bool AffixMgr::parse_cpdsyllable(const std::string& line, FileMgr* af) {
+ int i = 0;
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ cpdmaxsyllable = atoi(std::string(start_piece, iter).c_str());
+ np++;
+ break;
+ }
+ case 2: {
+ if (!utf8) {
+ cpdvowels.assign(start_piece, iter);
+ std::sort(cpdvowels.begin(), cpdvowels.end());
+ } else {
+ std::string piece(start_piece, iter);
+ u8_u16(cpdvowels_utf16, piece);
+ std::sort(cpdvowels_utf16.begin(), cpdvowels_utf16.end());
+ }
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np < 2) {
+ HUNSPELL_WARNING(stderr,
+ "error: line %d: missing compoundsyllable information\n",
+ af->getlinenum());
+ return false;
+ }
+ if (np == 2)
+ cpdvowels = "AEIOUaeiou";
+ return true;
+}
+
+bool AffixMgr::parse_convtable(const std::string& line,
+ FileMgr* af,
+ RepList** rl,
+ const std::string& keyword) {
+ if (*rl) {
+ HUNSPELL_WARNING(stderr, "error: line %d: multiple table definitions\n",
+ af->getlinenum());
+ return false;
+ }
+ int i = 0;
+ int np = 0;
+ int numrl = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ numrl = atoi(std::string(start_piece, iter).c_str());
+ if (numrl < 1) {
+ HUNSPELL_WARNING(stderr, "error: line %d: incorrect entry number\n",
+ af->getlinenum());
+ return false;
+ }
+ *rl = new RepList(numrl);
+ if (!*rl)
+ return false;
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np != 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ return false;
+ }
+
+ /* now parse the num lines to read in the remainder of the table */
+ for (int j = 0; j < numrl; j++) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+ i = 0;
+ std::string pattern;
+ std::string pattern2;
+ iter = nl.begin();
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ {
+ switch (i) {
+ case 0: {
+ if (nl.compare(start_piece - nl.begin(), keyword.size(), keyword, 0, keyword.size()) != 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ delete *rl;
+ *rl = NULL;
+ return false;
+ }
+ break;
+ }
+ case 1: {
+ pattern.assign(start_piece, iter);
+ break;
+ }
+ case 2: {
+ pattern2.assign(start_piece, iter);
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ }
+ start_piece = mystrsep(nl, iter);
+ }
+ if (pattern.empty() || pattern2.empty()) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ return false;
+ }
+ (*rl)->add(pattern, pattern2);
+ }
+ return true;
+}
+
+/* parse in the typical fault correcting table */
+bool AffixMgr::parse_phonetable(const std::string& line, FileMgr* af) {
+ if (phone) {
+ HUNSPELL_WARNING(stderr, "error: line %d: multiple table definitions\n",
+ af->getlinenum());
+ return false;
+ }
+ int num = -1;
+ int i = 0;
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ num = atoi(std::string(start_piece, iter).c_str());
+ if (num < 1) {
+ HUNSPELL_WARNING(stderr, "error: line %d: bad entry number\n",
+ af->getlinenum());
+ return false;
+ }
+ phone = new phonetable;
+ phone->utf8 = (char)utf8;
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np != 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ return false;
+ }
+
+ /* now parse the phone->num lines to read in the remainder of the table */
+ for (int j = 0; j < num; ++j) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+ i = 0;
+ const size_t old_size = phone->rules.size();
+ iter = nl.begin();
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ {
+ switch (i) {
+ case 0: {
+ if (nl.compare(start_piece - nl.begin(), 5, "PHONE", 5) != 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ return false;
+ }
+ break;
+ }
+ case 1: {
+ phone->rules.push_back(std::string(start_piece, iter));
+ break;
+ }
+ case 2: {
+ phone->rules.push_back(std::string(start_piece, iter));
+ mystrrep(phone->rules.back(), "_", "");
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ }
+ start_piece = mystrsep(nl, iter);
+ }
+ if (phone->rules.size() != old_size + 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ phone->rules.clear();
+ return false;
+ }
+ }
+ phone->rules.push_back("");
+ phone->rules.push_back("");
+ init_phonet_hash(*phone);
+ return true;
+}
+
+/* parse in the checkcompoundpattern table */
+bool AffixMgr::parse_checkcpdtable(const std::string& line, FileMgr* af) {
+ if (parsedcheckcpd) {
+ HUNSPELL_WARNING(stderr, "error: line %d: multiple table definitions\n",
+ af->getlinenum());
+ return false;
+ }
+ parsedcheckcpd = true;
+ int numcheckcpd = -1;
+ int i = 0;
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ numcheckcpd = atoi(std::string(start_piece, iter).c_str());
+ if (numcheckcpd < 1) {
+ HUNSPELL_WARNING(stderr, "error: line %d: bad entry number\n",
+ af->getlinenum());
+ return false;
+ }
+ checkcpdtable.reserve(numcheckcpd);
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np != 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ return false;
+ }
+
+ /* now parse the numcheckcpd lines to read in the remainder of the table */
+ for (int j = 0; j < numcheckcpd; ++j) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+ i = 0;
+ checkcpdtable.push_back(patentry());
+ iter = nl.begin();
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ switch (i) {
+ case 0: {
+ if (nl.compare(start_piece - nl.begin(), 20, "CHECKCOMPOUNDPATTERN", 20) != 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ return false;
+ }
+ break;
+ }
+ case 1: {
+ checkcpdtable.back().pattern.assign(start_piece, iter);
+ size_t slash_pos = checkcpdtable.back().pattern.find('/');
+ if (slash_pos != std::string::npos) {
+ std::string chunk(checkcpdtable.back().pattern, slash_pos + 1);
+ checkcpdtable.back().pattern.resize(slash_pos);
+ checkcpdtable.back().cond = pHMgr->decode_flag(chunk.c_str());
+ }
+ break;
+ }
+ case 2: {
+ checkcpdtable.back().pattern2.assign(start_piece, iter);
+ size_t slash_pos = checkcpdtable.back().pattern2.find('/');
+ if (slash_pos != std::string::npos) {
+ std::string chunk(checkcpdtable.back().pattern2, slash_pos + 1);
+ checkcpdtable.back().pattern2.resize(slash_pos);
+ checkcpdtable.back().cond2 = pHMgr->decode_flag(chunk.c_str());
+ }
+ break;
+ }
+ case 3: {
+ checkcpdtable.back().pattern3.assign(start_piece, iter);
+ simplifiedcpd = 1;
+ break;
+ }
+ default:
+ break;
+ }
+ i++;
+ start_piece = mystrsep(nl, iter);
+ }
+ }
+ return true;
+}
+
+/* parse in the compound rule table */
+bool AffixMgr::parse_defcpdtable(const std::string& line, FileMgr* af) {
+ if (parseddefcpd) {
+ HUNSPELL_WARNING(stderr, "error: line %d: multiple table definitions\n",
+ af->getlinenum());
+ return false;
+ }
+ parseddefcpd = true;
+ int numdefcpd = -1;
+ int i = 0;
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ numdefcpd = atoi(std::string(start_piece, iter).c_str());
+ if (numdefcpd < 1) {
+ HUNSPELL_WARNING(stderr, "error: line %d: bad entry number\n",
+ af->getlinenum());
+ return false;
+ }
+ defcpdtable.reserve(numdefcpd);
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np != 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ return false;
+ }
+
+ /* now parse the numdefcpd lines to read in the remainder of the table */
+ for (int j = 0; j < numdefcpd; ++j) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+ i = 0;
+ defcpdtable.push_back(flagentry());
+ iter = nl.begin();
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ switch (i) {
+ case 0: {
+ if (nl.compare(start_piece - nl.begin(), 12, "COMPOUNDRULE", 12) != 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ numdefcpd = 0;
+ return false;
+ }
+ break;
+ }
+ case 1: { // handle parenthesized flags
+ if (std::find(start_piece, iter, '(') != iter) {
+ for (std::string::const_iterator k = start_piece; k != iter; ++k) {
+ std::string::const_iterator chb = k;
+ std::string::const_iterator che = k + 1;
+ if (*k == '(') {
+ std::string::const_iterator parpos = std::find(k, iter, ')');
+ if (parpos != iter) {
+ chb = k + 1;
+ che = parpos;
+ k = parpos;
+ }
+ }
+
+ if (*chb == '*' || *chb == '?') {
+ defcpdtable.back().push_back((FLAG)*chb);
+ } else {
+ pHMgr->decode_flags(defcpdtable.back(), std::string(chb, che), af);
+ }
+ }
+ } else {
+ pHMgr->decode_flags(defcpdtable.back(), std::string(start_piece, iter), af);
+ }
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(nl, iter);
+ }
+ if (defcpdtable.back().empty()) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ return false;
+ }
+ }
+ return true;
+}
+
+/* parse in the character map table */
+bool AffixMgr::parse_maptable(const std::string& line, FileMgr* af) {
+ if (parsedmaptable) {
+ HUNSPELL_WARNING(stderr, "error: line %d: multiple table definitions\n",
+ af->getlinenum());
+ return false;
+ }
+ parsedmaptable = true;
+ int nummap = -1;
+ int i = 0;
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ nummap = atoi(std::string(start_piece, iter).c_str());
+ if (nummap < 1) {
+ HUNSPELL_WARNING(stderr, "error: line %d: bad entry number\n",
+ af->getlinenum());
+ return false;
+ }
+ maptable.reserve(nummap);
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np != 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ return false;
+ }
+
+ /* now parse the nummap lines to read in the remainder of the table */
+ for (int j = 0; j < nummap; ++j) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+ i = 0;
+ maptable.push_back(mapentry());
+ iter = nl.begin();
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ switch (i) {
+ case 0: {
+ if (nl.compare(start_piece - nl.begin(), 3, "MAP", 3) != 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ nummap = 0;
+ return false;
+ }
+ break;
+ }
+ case 1: {
+ for (std::string::const_iterator k = start_piece; k != iter; ++k) {
+ std::string::const_iterator chb = k;
+ std::string::const_iterator che = k + 1;
+ if (*k == '(') {
+ std::string::const_iterator parpos = std::find(k, iter, ')');
+ if (parpos != iter) {
+ chb = k + 1;
+ che = parpos;
+ k = parpos;
+ }
+ } else {
+ if (utf8 && (*k & 0xc0) == 0xc0) {
+ ++k;
+ while (k != iter && (*k & 0xc0) == 0x80)
+ ++k;
+ che = k;
+ --k;
+ }
+ }
+ maptable.back().push_back(std::string(chb, che));
+ }
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(nl, iter);
+ }
+ if (maptable.back().empty()) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ return false;
+ }
+ }
+ return true;
+}
+
+/* parse in the word breakpoint table */
+bool AffixMgr::parse_breaktable(const std::string& line, FileMgr* af) {
+ if (parsedbreaktable) {
+ HUNSPELL_WARNING(stderr, "error: line %d: multiple table definitions\n",
+ af->getlinenum());
+ return false;
+ }
+ parsedbreaktable = true;
+ int numbreak = -1;
+ int i = 0;
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ case 0: {
+ np++;
+ break;
+ }
+ case 1: {
+ numbreak = atoi(std::string(start_piece, iter).c_str());
+ if (numbreak < 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: bad entry number\n",
+ af->getlinenum());
+ return false;
+ }
+ if (numbreak == 0)
+ return true;
+ breaktable.reserve(numbreak);
+ np++;
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ if (np != 2) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ return false;
+ }
+
+ /* now parse the numbreak lines to read in the remainder of the table */
+ for (int j = 0; j < numbreak; ++j) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+ i = 0;
+ iter = nl.begin();
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ switch (i) {
+ case 0: {
+ if (nl.compare(start_piece - nl.begin(), 5, "BREAK", 5) != 0) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ numbreak = 0;
+ return false;
+ }
+ break;
+ }
+ case 1: {
+ breaktable.push_back(std::string(start_piece, iter));
+ break;
+ }
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(nl, iter);
+ }
+ }
+
+ if (breaktable.size() != static_cast<size_t>(numbreak)) {
+ HUNSPELL_WARNING(stderr, "error: line %d: table is corrupt\n",
+ af->getlinenum());
+ return false;
+ }
+
+ return true;
+}
+
+void AffixMgr::reverse_condition(std::string& piece) {
+ if (piece.empty())
+ return;
+
+ int neg = 0;
+ for (std::string::reverse_iterator k = piece.rbegin(); k != piece.rend(); ++k) {
+ switch (*k) {
+ case '[': {
+ if (neg)
+ *(k - 1) = '[';
+ else
+ *k = ']';
+ break;
+ }
+ case ']': {
+ *k = '[';
+ if (neg)
+ *(k - 1) = '^';
+ neg = 0;
+ break;
+ }
+ case '^': {
+ if (*(k - 1) == ']')
+ neg = 1;
+ else if (neg)
+ *(k - 1) = *k;
+ break;
+ }
+ default: {
+ if (neg)
+ *(k - 1) = *k;
+ }
+ }
+ }
+}
+
+class entries_container {
+ std::vector<AffEntry*> entries;
+ AffixMgr* m_mgr;
+ char m_at;
+public:
+ entries_container(char at, AffixMgr* mgr)
+ : m_mgr(mgr)
+ , m_at(at) {
+ }
+ void release() {
+ entries.clear();
+ }
+ void initialize(int numents,
+ char opts, unsigned short aflag) {
+ entries.reserve(numents);
+
+ if (m_at == 'P') {
+ entries.push_back(new PfxEntry(m_mgr));
+ } else {
+ entries.push_back(new SfxEntry(m_mgr));
+ }
+
+ entries.back()->opts = opts;
+ entries.back()->aflag = aflag;
+ }
+
+ AffEntry* add_entry(char opts) {
+ if (m_at == 'P') {
+ entries.push_back(new PfxEntry(m_mgr));
+ } else {
+ entries.push_back(new SfxEntry(m_mgr));
+ }
+ AffEntry* ret = entries.back();
+ ret->opts = entries[0]->opts & opts;
+ return ret;
+ }
+
+ AffEntry* first_entry() {
+ return entries.empty() ? NULL : entries[0];
+ }
+
+ ~entries_container() {
+ for (size_t i = 0; i < entries.size(); ++i) {
+ delete entries[i];
+ }
+ }
+
+ std::vector<AffEntry*>::iterator begin() { return entries.begin(); }
+ std::vector<AffEntry*>::iterator end() { return entries.end(); }
+};
+
+bool AffixMgr::parse_affix(const std::string& line,
+ const char at,
+ FileMgr* af,
+ char* dupflags) {
+ int numents = 0; // number of AffEntry structures to parse
+
+ unsigned short aflag = 0; // affix char identifier
+
+ char ff = 0;
+ entries_container affentries(at, this);
+
+ int i = 0;
+
+// checking lines with bad syntax
+#ifdef DEBUG
+ int basefieldnum = 0;
+#endif
+
+ // split affix header line into pieces
+
+ int np = 0;
+ std::string::const_iterator iter = line.begin();
+ std::string::const_iterator start_piece = mystrsep(line, iter);
+ while (start_piece != line.end()) {
+ switch (i) {
+ // piece 1 - is type of affix
+ case 0: {
+ np++;
+ break;
+ }
+
+ // piece 2 - is affix char
+ case 1: {
+ np++;
+ aflag = pHMgr->decode_flag(std::string(start_piece, iter).c_str());
+ if (((at == 'S') && (dupflags[aflag] & dupSFX)) ||
+ ((at == 'P') && (dupflags[aflag] & dupPFX))) {
+ HUNSPELL_WARNING(
+ stderr,
+ "error: line %d: multiple definitions of an affix flag\n",
+ af->getlinenum());
+ }
+ dupflags[aflag] += (char)((at == 'S') ? dupSFX : dupPFX);
+ break;
+ }
+ // piece 3 - is cross product indicator
+ case 2: {
+ np++;
+ if (*start_piece == 'Y')
+ ff = aeXPRODUCT;
+ break;
+ }
+
+ // piece 4 - is number of affentries
+ case 3: {
+ np++;
+ numents = atoi(std::string(start_piece, iter).c_str());
+ if ((numents <= 0) || ((std::numeric_limits<size_t>::max() /
+ sizeof(AffEntry)) < static_cast<size_t>(numents))) {
+ char* err = pHMgr->encode_flag(aflag);
+ if (err) {
+ HUNSPELL_WARNING(stderr, "error: line %d: bad entry number\n",
+ af->getlinenum());
+ free(err);
+ }
+ return false;
+ }
+
+ char opts = ff;
+ if (utf8)
+ opts |= aeUTF8;
+ if (pHMgr->is_aliasf())
+ opts |= aeALIASF;
+ if (pHMgr->is_aliasm())
+ opts |= aeALIASM;
+ affentries.initialize(numents, opts, aflag);
+ }
+
+ default:
+ break;
+ }
+ ++i;
+ start_piece = mystrsep(line, iter);
+ }
+ // check to make sure we parsed enough pieces
+ if (np != 4) {
+ char* err = pHMgr->encode_flag(aflag);
+ if (err) {
+ HUNSPELL_WARNING(stderr, "error: line %d: missing data\n",
+ af->getlinenum());
+ free(err);
+ }
+ return false;
+ }
+
+ // now parse numents affentries for this affix
+ AffEntry* entry = affentries.first_entry();
+ for (int ent = 0; ent < numents; ++ent) {
+ std::string nl;
+ if (!af->getline(nl))
+ return false;
+ mychomp(nl);
+
+ iter = nl.begin();
+ i = 0;
+ np = 0;
+
+ // split line into pieces
+ start_piece = mystrsep(nl, iter);
+ while (start_piece != nl.end()) {
+ switch (i) {
+ // piece 1 - is type
+ case 0: {
+ np++;
+ if (ent != 0)
+ entry = affentries.add_entry((char)(aeXPRODUCT + aeUTF8 + aeALIASF + aeALIASM));
+ break;
+ }
+
+ // piece 2 - is affix char
+ case 1: {
+ np++;
+ std::string chunk(start_piece, iter);
+ if (pHMgr->decode_flag(chunk.c_str()) != aflag) {
+ char* err = pHMgr->encode_flag(aflag);
+ if (err) {
+ HUNSPELL_WARNING(stderr,
+ "error: line %d: affix %s is corrupt\n",
+ af->getlinenum(), err);
+ free(err);
+ }
+ return false;
+ }
+
+ if (ent != 0) {
+ AffEntry* start_entry = affentries.first_entry();
+ entry->aflag = start_entry->aflag;
+ }
+ break;
+ }
+
+ // piece 3 - is string to strip or 0 for null
+ case 2: {
+ np++;
+ entry->strip = std::string(start_piece, iter);
+ if (complexprefixes) {
+ if (utf8)
+ reverseword_utf(entry->strip);
+ else
+ reverseword(entry->strip);
+ }
+ if (entry->strip.compare("0") == 0) {
+ entry->strip.clear();
+ }
+ break;
+ }
+
+ // piece 4 - is affix string or 0 for null
+ case 3: {
+ entry->morphcode = NULL;
+ entry->contclass = NULL;
+ entry->contclasslen = 0;
+ np++;
+ std::string::const_iterator dash = std::find(start_piece, iter, '/');
+ if (dash != iter) {
+ entry->appnd = std::string(start_piece, dash);
+ std::string dash_str(dash + 1, iter);
+
+ if (!ignorechars.empty() && !has_no_ignored_chars(entry->appnd, ignorechars)) {
+ if (utf8) {
+ remove_ignored_chars_utf(entry->appnd, ignorechars_utf16);
+ } else {
+ remove_ignored_chars(entry->appnd, ignorechars);
+ }
+ }
+
+ if (complexprefixes) {
+ if (utf8)
+ reverseword_utf(entry->appnd);
+ else
+ reverseword(entry->appnd);
+ }
+
+ if (pHMgr->is_aliasf()) {
+ int index = atoi(dash_str.c_str());
+ entry->contclasslen = (unsigned short)pHMgr->get_aliasf(
+ index, &(entry->contclass), af);
+ if (!entry->contclasslen)
+ HUNSPELL_WARNING(stderr,
+ "error: bad affix flag alias: \"%s\"\n",
+ dash_str.c_str());
+ } else {
+ entry->contclasslen = (unsigned short)pHMgr->decode_flags(
+ &(entry->contclass), dash_str.c_str(), af);
+ std::sort(entry->contclass, entry->contclass + entry->contclasslen);
+ }
+
+ havecontclass = 1;
+ for (unsigned short _i = 0; _i < entry->contclasslen; _i++) {
+ contclasses[(entry->contclass)[_i]] = 1;
+ }
+ } else {
+ entry->appnd = std::string(start_piece, iter);
+
+ if (!ignorechars.empty() && !has_no_ignored_chars(entry->appnd, ignorechars)) {
+ if (utf8) {
+ remove_ignored_chars_utf(entry->appnd, ignorechars_utf16);
+ } else {
+ remove_ignored_chars(entry->appnd, ignorechars);
+ }
+ }
+
+ if (complexprefixes) {
+ if (utf8)
+ reverseword_utf(entry->appnd);
+ else
+ reverseword(entry->appnd);
+ }
+ }
+
+ if (entry->appnd.compare("0") == 0) {
+ entry->appnd.clear();
+ }
+ break;
+ }
+
+ // piece 5 - is the conditions descriptions
+ case 4: {
+ std::string chunk(start_piece, iter);
+ np++;
+ if (complexprefixes) {
+ if (utf8)
+ reverseword_utf(chunk);
+ else
+ reverseword(chunk);
+ reverse_condition(chunk);
+ }
+ if (!entry->strip.empty() && chunk != "." &&
+ redundant_condition(at, entry->strip.c_str(), entry->strip.size(), chunk.c_str(),
+ af->getlinenum()))
+ chunk = ".";
+ if (at == 'S') {
+ reverseword(chunk);
+ reverse_condition(chunk);
+ }
+ if (encodeit(*entry, chunk.c_str()))
+ return false;
+ break;
+ }
+
+ case 5: {
+ std::string chunk(start_piece, iter);
+ np++;
+ if (pHMgr->is_aliasm()) {
+ int index = atoi(chunk.c_str());
+ entry->morphcode = pHMgr->get_aliasm(index);
+ } else {
+ if (complexprefixes) { // XXX - fix me for morph. gen.
+ if (utf8)
+ reverseword_utf(chunk);
+ else
+ reverseword(chunk);
+ }
+ // add the remaining of the line
+ std::string::const_iterator end = nl.end();
+ if (iter != end) {
+ chunk.append(iter, end);
+ }
+ entry->morphcode = mystrdup(chunk.c_str());
+ if (!entry->morphcode)
+ return false;
+ }
+ break;
+ }
+ default:
+ break;
+ }
+ i++;
+ start_piece = mystrsep(nl, iter);
+ }
+ // check to make sure we parsed enough pieces
+ if (np < 4) {
+ char* err = pHMgr->encode_flag(aflag);
+ if (err) {
+ HUNSPELL_WARNING(stderr, "error: line %d: affix %s is corrupt\n",
+ af->getlinenum(), err);
+ free(err);
+ }
+ return false;
+ }
+
+#ifdef DEBUG
+ // detect unnecessary fields, excepting comments
+ if (basefieldnum) {
+ int fieldnum =
+ !(entry->morphcode) ? 5 : ((*(entry->morphcode) == '#') ? 5 : 6);
+ if (fieldnum != basefieldnum)
+ HUNSPELL_WARNING(stderr, "warning: line %d: bad field number\n",
+ af->getlinenum());
+ } else {
+ basefieldnum =
+ !(entry->morphcode) ? 5 : ((*(entry->morphcode) == '#') ? 5 : 6);
+ }
+#endif
+ }
+
+ // now create SfxEntry or PfxEntry objects and use links to
+ // build an ordered (sorted by affix string) list
+ std::vector<AffEntry*>::iterator start = affentries.begin();
+ std::vector<AffEntry*>::iterator end = affentries.end();
+ for (std::vector<AffEntry*>::iterator affentry = start; affentry != end; ++affentry) {
+ if (at == 'P') {
+ build_pfxtree(static_cast<PfxEntry*>(*affentry));
+ } else {
+ build_sfxtree(static_cast<SfxEntry*>(*affentry));
+ }
+ }
+
+ //contents belong to AffixMgr now
+ affentries.release();
+
+ return true;
+}
+
+int AffixMgr::redundant_condition(char ft,
+ const char* strip,
+ int stripl,
+ const char* cond,
+ int linenum) {
+ int condl = strlen(cond);
+ int i;
+ int j;
+ int neg;
+ int in;
+ if (ft == 'P') { // prefix
+ if (strncmp(strip, cond, condl) == 0)
+ return 1;
+ if (utf8) {
+ } else {
+ for (i = 0, j = 0; (i < stripl) && (j < condl); i++, j++) {
+ if (cond[j] != '[') {
+ if (cond[j] != strip[i]) {
+ HUNSPELL_WARNING(stderr,
+ "warning: line %d: incompatible stripping "
+ "characters and condition\n",
+ linenum);
+ return 0;
+ }
+ } else {
+ neg = (cond[j + 1] == '^') ? 1 : 0;
+ in = 0;
+ do {
+ j++;
+ if (strip[i] == cond[j])
+ in = 1;
+ } while ((j < (condl - 1)) && (cond[j] != ']'));
+ if (j == (condl - 1) && (cond[j] != ']')) {
+ HUNSPELL_WARNING(stderr,
+ "error: line %d: missing ] in condition:\n%s\n",
+ linenum, cond);
+ return 0;
+ }
+ if ((!neg && !in) || (neg && in)) {
+ HUNSPELL_WARNING(stderr,
+ "warning: line %d: incompatible stripping "
+ "characters and condition\n",
+ linenum);
+ return 0;
+ }
+ }
+ }
+ if (j >= condl)
+ return 1;
+ }
+ } else { // suffix
+ if ((stripl >= condl) && strcmp(strip + stripl - condl, cond) == 0)
+ return 1;
+ if (utf8) {
+ } else {
+ for (i = stripl - 1, j = condl - 1; (i >= 0) && (j >= 0); i--, j--) {
+ if (cond[j] != ']') {
+ if (cond[j] != strip[i]) {
+ HUNSPELL_WARNING(stderr,
+ "warning: line %d: incompatible stripping "
+ "characters and condition\n",
+ linenum);
+ return 0;
+ }
+ } else {
+ in = 0;
+ do {
+ j--;
+ if (strip[i] == cond[j])
+ in = 1;
+ } while ((j > 0) && (cond[j] != '['));
+ if ((j == 0) && (cond[j] != '[')) {
+ HUNSPELL_WARNING(stderr,
+ "error: line: %d: missing ] in condition:\n%s\n",
+ linenum, cond);
+ return 0;
+ }
+ neg = (cond[j + 1] == '^') ? 1 : 0;
+ if ((!neg && !in) || (neg && in)) {
+ HUNSPELL_WARNING(stderr,
+ "warning: line %d: incompatible stripping "
+ "characters and condition\n",
+ linenum);
+ return 0;
+ }
+ }
+ }
+ if (j < 0)
+ return 1;
+ }
+ }
+ return 0;
+}
+
+std::vector<std::string> AffixMgr::get_suffix_words(short unsigned* suff,
+ int len,
+ const char* root_word) {
+ std::vector<std::string> slst;
+ short unsigned* start_ptr = suff;
+ for (int j = 0; j < SETSIZE; j++) {
+ SfxEntry* ptr = sStart[j];
+ while (ptr) {
+ suff = start_ptr;
+ for (int i = 0; i < len; i++) {
+ if ((*suff) == ptr->getFlag()) {
+ std::string nw(root_word);
+ nw.append(ptr->getAffix());
+ hentry* ht = ptr->checkword(nw.c_str(), nw.size(), 0, NULL, 0, 0, 0);
+ if (ht) {
+ slst.push_back(nw);
+ }
+ }
+ suff++;
+ }
+ ptr = ptr->getNext();
+ }
+ }
+ return slst;
+}