Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * spell.c
4 : * Normalizing word with ISpell
5 : *
6 : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
7 : *
8 : * Ispell dictionary
9 : * -----------------
10 : *
11 : * Rules of dictionaries are defined in two files with .affix and .dict
12 : * extensions. They are used by spell checker programs Ispell and Hunspell.
13 : *
14 : * An .affix file declares morphological rules to get a basic form of words.
15 : * The format of an .affix file has different structure for Ispell and Hunspell
16 : * dictionaries. The Hunspell format is more complicated. But when an .affix
17 : * file is imported and compiled, it is stored in the same structure AffixNode.
18 : *
19 : * A .dict file stores a list of basic forms of words with references to
20 : * affix rules. The format of a .dict file has the same structure for Ispell
21 : * and Hunspell dictionaries.
22 : *
23 : * Compilation of a dictionary
24 : * ---------------------------
25 : *
26 : * A compiled dictionary is stored in the IspellDict structure. Compilation of
27 : * a dictionary is divided into the several steps:
28 : * - NIImportDictionary() - stores each word of a .dict file in the
29 : * temporary Spell field.
30 : * - NIImportAffixes() - stores affix rules of an .affix file in the
31 : * Affix field (not temporary) if an .affix file has the Ispell format.
32 : * -> NIImportOOAffixes() - stores affix rules if an .affix file has the
33 : * Hunspell format. The AffixData field is initialized if AF parameter
34 : * is defined.
35 : * - NISortDictionary() - builds a prefix tree (Trie) from the words list
36 : * and stores it in the Dictionary field. The words list is got from the
37 : * Spell field. The AffixData field is initialized if AF parameter is not
38 : * defined.
39 : * - NISortAffixes():
40 : * - builds a list of compound affixes from the affix list and stores it
41 : * in the CompoundAffix.
42 : * - builds prefix trees (Trie) from the affix list for prefixes and suffixes
43 : * and stores them in Suffix and Prefix fields.
44 : * The affix list is got from the Affix field.
45 : *
46 : * Memory management
47 : * -----------------
48 : *
49 : * The IspellDict structure has the Spell field which is used only in compile
50 : * time. The Spell field stores a words list. It can take a lot of memory.
51 : * Therefore when a dictionary is compiled this field is cleared by
52 : * NIFinishBuild().
53 : *
54 : * All resources which should cleared by NIFinishBuild() is initialized using
55 : * tmpalloc() and tmpalloc0().
56 : *
57 : * IDENTIFICATION
58 : * src/backend/tsearch/spell.c
59 : *
60 : *-------------------------------------------------------------------------
61 : */
62 :
63 : #include "postgres.h"
64 :
65 : #include "catalog/pg_collation.h"
66 : #include "miscadmin.h"
67 : #include "tsearch/dicts/spell.h"
68 : #include "tsearch/ts_locale.h"
69 : #include "utils/formatting.h"
70 : #include "utils/memutils.h"
71 :
72 :
73 : /*
74 : * Initialization requires a lot of memory that's not needed
75 : * after the initialization is done. During initialization,
76 : * CurrentMemoryContext is the long-lived memory context associated
77 : * with the dictionary cache entry. We keep the short-lived stuff
78 : * in the Conf->buildCxt context.
79 : */
80 : #define tmpalloc(sz) MemoryContextAlloc(Conf->buildCxt, (sz))
81 : #define tmpalloc0(sz) MemoryContextAllocZero(Conf->buildCxt, (sz))
82 :
83 : /*
84 : * Prepare for constructing an ISpell dictionary.
85 : *
86 : * The IspellDict struct is assumed to be zeroed when allocated.
87 : */
88 : void
89 87 : NIStartBuild(IspellDict *Conf)
90 : {
91 : /*
92 : * The temp context is a child of CurTransactionContext, so that it will
93 : * go away automatically on error.
94 : */
95 87 : Conf->buildCxt = AllocSetContextCreate(CurTransactionContext,
96 : "Ispell dictionary init context",
97 : ALLOCSET_DEFAULT_SIZES);
98 87 : }
99 :
100 : /*
101 : * Clean up when dictionary construction is complete.
102 : */
103 : void
104 71 : NIFinishBuild(IspellDict *Conf)
105 : {
106 : /* Release no-longer-needed temp memory */
107 71 : MemoryContextDelete(Conf->buildCxt);
108 : /* Just for cleanliness, zero the now-dangling pointers */
109 71 : Conf->buildCxt = NULL;
110 71 : Conf->Spell = NULL;
111 71 : Conf->firstfree = NULL;
112 71 : Conf->CompoundAffixFlags = NULL;
113 71 : }
114 :
115 :
116 : /*
117 : * "Compact" palloc: allocate without extra palloc overhead.
118 : *
119 : * Since we have no need to free the ispell data items individually, there's
120 : * not much value in the per-chunk overhead normally consumed by palloc.
121 : * Getting rid of it is helpful since ispell can allocate a lot of small nodes.
122 : *
123 : * We currently pre-zero all data allocated this way, even though some of it
124 : * doesn't need that. The cpalloc and cpalloc0 macros are just documentation
125 : * to indicate which allocations actually require zeroing.
126 : */
127 : #define COMPACT_ALLOC_CHUNK 8192 /* amount to get from palloc at once */
128 : #define COMPACT_MAX_REQ 1024 /* must be < COMPACT_ALLOC_CHUNK */
129 :
130 : static void *
131 8013 : compact_palloc0(IspellDict *Conf, size_t size)
132 : {
133 : void *result;
134 :
135 : /* Should only be called during init */
136 : Assert(Conf->buildCxt != NULL);
137 :
138 : /* No point in this for large chunks */
139 8013 : if (size > COMPACT_MAX_REQ)
140 0 : return palloc0(size);
141 :
142 : /* Keep everything maxaligned */
143 8013 : size = MAXALIGN(size);
144 :
145 : /* Need more space? */
146 8013 : if (size > Conf->avail)
147 : {
148 83 : Conf->firstfree = palloc0(COMPACT_ALLOC_CHUNK);
149 83 : Conf->avail = COMPACT_ALLOC_CHUNK;
150 : }
151 :
152 8013 : result = Conf->firstfree;
153 8013 : Conf->firstfree += size;
154 8013 : Conf->avail -= size;
155 :
156 8013 : return result;
157 : }
158 :
159 : #define cpalloc(size) compact_palloc0(Conf, size)
160 : #define cpalloc0(size) compact_palloc0(Conf, size)
161 :
162 : static char *
163 4284 : cpstrdup(IspellDict *Conf, const char *str)
164 : {
165 4284 : char *res = cpalloc(strlen(str) + 1);
166 :
167 4284 : strcpy(res, str);
168 4284 : return res;
169 : }
170 :
171 :
172 : /*
173 : * Apply str_tolower(), producing a temporary result (in the buildCxt).
174 : */
175 : static char *
176 3741 : lowerstr_ctx(IspellDict *Conf, const char *src)
177 : {
178 : MemoryContext saveCtx;
179 : char *dst;
180 :
181 3741 : saveCtx = MemoryContextSwitchTo(Conf->buildCxt);
182 3741 : dst = str_tolower(src, strlen(src), DEFAULT_COLLATION_OID);
183 3741 : MemoryContextSwitchTo(saveCtx);
184 :
185 3741 : return dst;
186 : }
187 :
188 : #define MAX_NORM 1024
189 : #define MAXNORMLEN 256
190 :
191 : #define STRNCMP(s,p) strncmp( (s), (p), strlen(p) )
192 : #define GETWCHAR(W,L,N,T) ( ((const uint8*)(W))[ ((T)==FF_PREFIX) ? (N) : ( (L) - 1 - (N) ) ] )
193 : #define GETCHAR(A,N,T) GETWCHAR( (A)->repl, (A)->replen, N, T )
194 :
195 : static const char *VoidString = "";
196 :
197 : static int
198 1866 : cmpspell(const void *s1, const void *s2)
199 : {
200 1866 : return strcmp((*(SPELL *const *) s1)->word, (*(SPELL *const *) s2)->word);
201 : }
202 :
203 : static int
204 1456 : cmpspellaffix(const void *s1, const void *s2)
205 : {
206 2912 : return strcmp((*(SPELL *const *) s1)->p.flag,
207 1456 : (*(SPELL *const *) s2)->p.flag);
208 : }
209 :
210 : static int
211 2533 : cmpcmdflag(const void *f1, const void *f2)
212 : {
213 2533 : const CompoundAffixFlag *fv1 = f1;
214 2533 : const CompoundAffixFlag *fv2 = f2;
215 :
216 : Assert(fv1->flagMode == fv2->flagMode);
217 :
218 2533 : if (fv1->flagMode == FM_NUM)
219 : {
220 489 : if (fv1->flag.i == fv2->flag.i)
221 74 : return 0;
222 :
223 415 : return (fv1->flag.i > fv2->flag.i) ? 1 : -1;
224 : }
225 :
226 2044 : return strcmp(fv1->flag.s, fv2->flag.s);
227 : }
228 :
229 : static char *
230 755 : findchar(char *str, int c)
231 : {
232 5564 : while (*str)
233 : {
234 5481 : if (t_iseq(str, c))
235 672 : return str;
236 4809 : str += pg_mblen_cstr(str);
237 : }
238 :
239 83 : return NULL;
240 : }
241 :
242 : static char *
243 27 : findchar2(char *str, int c1, int c2)
244 : {
245 567 : while (*str)
246 : {
247 567 : if (t_iseq(str, c1) || t_iseq(str, c2))
248 27 : return str;
249 540 : str += pg_mblen_cstr(str);
250 : }
251 :
252 0 : return NULL;
253 : }
254 :
255 :
256 : /* backward string compare for suffix tree operations */
257 : static int
258 745 : strbcmp(const unsigned char *s1, const unsigned char *s2)
259 : {
260 745 : int l1 = strlen((const char *) s1) - 1,
261 745 : l2 = strlen((const char *) s2) - 1;
262 :
263 997 : while (l1 >= 0 && l2 >= 0)
264 : {
265 780 : if (s1[l1] < s2[l2])
266 169 : return -1;
267 611 : if (s1[l1] > s2[l2])
268 359 : return 1;
269 252 : l1--;
270 252 : l2--;
271 : }
272 217 : if (l1 < l2)
273 58 : return -1;
274 159 : if (l1 > l2)
275 133 : return 1;
276 :
277 26 : return 0;
278 : }
279 :
280 : static int
281 26 : strbncmp(const unsigned char *s1, const unsigned char *s2, size_t count)
282 : {
283 26 : int l1 = strlen((const char *) s1) - 1,
284 26 : l2 = strlen((const char *) s2) - 1,
285 26 : l = count;
286 :
287 39 : while (l1 >= 0 && l2 >= 0 && l > 0)
288 : {
289 26 : if (s1[l1] < s2[l2])
290 13 : return -1;
291 13 : if (s1[l1] > s2[l2])
292 0 : return 1;
293 13 : l1--;
294 13 : l2--;
295 13 : l--;
296 : }
297 13 : if (l == 0)
298 13 : return 0;
299 0 : if (l1 < l2)
300 0 : return -1;
301 0 : if (l1 > l2)
302 0 : return 1;
303 0 : return 0;
304 : }
305 :
306 : /*
307 : * Compares affixes.
308 : * First compares the type of an affix. Prefixes should go before affixes.
309 : * If types are equal then compares replaceable string.
310 : */
311 : static int
312 1260 : cmpaffix(const void *s1, const void *s2)
313 : {
314 1260 : const AFFIX *a1 = (const AFFIX *) s1;
315 1260 : const AFFIX *a2 = (const AFFIX *) s2;
316 :
317 1260 : if (a1->type < a2->type)
318 288 : return -1;
319 972 : if (a1->type > a2->type)
320 85 : return 1;
321 887 : if (a1->type == FF_PREFIX)
322 142 : return strcmp(a1->repl, a2->repl);
323 : else
324 745 : return strbcmp((const unsigned char *) a1->repl,
325 745 : (const unsigned char *) a2->repl);
326 : }
327 :
328 : /*
329 : * Gets an affix flag from the set of affix flags (sflagset).
330 : *
331 : * Several flags can be stored in a single string. Flags can be represented by:
332 : * - 1 character (FM_CHAR). A character may be Unicode.
333 : * - 2 characters (FM_LONG). A character may be Unicode.
334 : * - numbers from 1 to 65000 (FM_NUM).
335 : *
336 : * Depending on the flagMode an affix string can have the following format:
337 : * - FM_CHAR: ABCD
338 : * Here we have 4 flags: A, B, C and D
339 : * - FM_LONG: ABCDE*
340 : * Here we have 3 flags: AB, CD and E*
341 : * - FM_NUM: 200,205,50
342 : * Here we have 3 flags: 200, 205 and 50
343 : *
344 : * Conf: current dictionary.
345 : * sflagset: the set of affix flags. Returns a reference to the start of a next
346 : * affix flag.
347 : * sflag: returns an affix flag from sflagset.
348 : */
349 : static void
350 4086 : getNextFlagFromString(IspellDict *Conf, const char **sflagset, char *sflag)
351 : {
352 : int32 s;
353 : char *next;
354 4086 : const char *sbuf = *sflagset;
355 : int maxstep;
356 : int clen;
357 4086 : bool stop = false;
358 4086 : bool met_comma = false;
359 :
360 4086 : maxstep = (Conf->flagMode == FM_LONG) ? 2 : 1;
361 :
362 5360 : while (**sflagset)
363 : {
364 5360 : switch (Conf->flagMode)
365 : {
366 4571 : case FM_LONG:
367 : case FM_CHAR:
368 4571 : clen = ts_copychar_cstr(sflag, *sflagset);
369 4571 : sflag += clen;
370 :
371 : /* Go to start of the next flag */
372 4571 : *sflagset += clen;
373 :
374 : /* Check if we get all characters of flag */
375 4571 : maxstep--;
376 4571 : stop = (maxstep == 0);
377 4571 : break;
378 789 : case FM_NUM:
379 789 : errno = 0;
380 789 : s = strtol(*sflagset, &next, 10);
381 789 : if (*sflagset == next || errno == ERANGE)
382 4 : ereport(ERROR,
383 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
384 : errmsg("invalid affix flag \"%s\"", *sflagset)));
385 785 : if (s < 0 || s > FLAGNUM_MAXSIZE)
386 0 : ereport(ERROR,
387 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
388 : errmsg("affix flag \"%s\" is out of range",
389 : *sflagset)));
390 785 : sflag += sprintf(sflag, "%0d", s);
391 :
392 : /* Go to start of the next flag */
393 785 : *sflagset = next;
394 1219 : while (**sflagset)
395 : {
396 868 : if (isdigit((unsigned char) **sflagset))
397 : {
398 434 : if (!met_comma)
399 0 : ereport(ERROR,
400 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
401 : errmsg("invalid affix flag \"%s\"",
402 : *sflagset)));
403 434 : break;
404 : }
405 434 : else if (t_iseq(*sflagset, ','))
406 : {
407 434 : if (met_comma)
408 0 : ereport(ERROR,
409 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
410 : errmsg("invalid affix flag \"%s\"",
411 : *sflagset)));
412 434 : met_comma = true;
413 : }
414 0 : else if (!isspace((unsigned char) **sflagset))
415 : {
416 0 : ereport(ERROR,
417 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
418 : errmsg("invalid character in affix flag \"%s\"",
419 : *sflagset)));
420 : }
421 :
422 434 : *sflagset += pg_mblen_cstr(*sflagset);
423 : }
424 785 : stop = true;
425 785 : break;
426 0 : default:
427 0 : elog(ERROR, "unrecognized type of Conf->flagMode: %d",
428 : Conf->flagMode);
429 : }
430 :
431 5356 : if (stop)
432 4082 : break;
433 : }
434 :
435 4082 : if (Conf->flagMode == FM_LONG && maxstep > 0)
436 0 : ereport(ERROR,
437 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
438 : errmsg("invalid affix flag \"%s\" with \"long\" flag value",
439 : sbuf)));
440 :
441 4082 : *sflag = '\0';
442 4082 : }
443 :
444 : /*
445 : * Checks if the affix set Conf->AffixData[affix] contains affixflag.
446 : * Conf->AffixData[affix] does not contain affixflag if this flag is not used
447 : * actually by the .dict file.
448 : *
449 : * Conf: current dictionary.
450 : * affix: index of the Conf->AffixData array.
451 : * affixflag: the affix flag.
452 : *
453 : * Returns true if the string Conf->AffixData[affix] contains affixflag,
454 : * otherwise returns false.
455 : */
456 : static bool
457 1650 : IsAffixFlagInUse(IspellDict *Conf, int affix, const char *affixflag)
458 : {
459 : const char *flagcur;
460 : char flag[BUFSIZ];
461 :
462 1650 : if (*affixflag == 0)
463 530 : return true;
464 :
465 : Assert(affix < Conf->nAffixData);
466 :
467 1120 : flagcur = Conf->AffixData[affix];
468 :
469 3194 : while (*flagcur)
470 : {
471 2466 : getNextFlagFromString(Conf, &flagcur, flag);
472 : /* Compare first affix flag in flagcur with affixflag */
473 2466 : if (strcmp(flag, affixflag) == 0)
474 392 : return true;
475 : }
476 :
477 : /* Could not find affixflag */
478 728 : return false;
479 : }
480 :
481 : /*
482 : * Adds the new word into the temporary array Spell.
483 : *
484 : * Conf: current dictionary.
485 : * word: new word.
486 : * flag: set of affix flags. Single flag can be get by getNextFlagFromString().
487 : */
488 : static void
489 755 : NIAddSpell(IspellDict *Conf, const char *word, const char *flag)
490 : {
491 755 : if (Conf->nspell >= Conf->mspell)
492 : {
493 83 : if (Conf->mspell)
494 : {
495 0 : Conf->mspell *= 2;
496 0 : Conf->Spell = (SPELL **) repalloc(Conf->Spell, Conf->mspell * sizeof(SPELL *));
497 : }
498 : else
499 : {
500 83 : Conf->mspell = 1024 * 20;
501 83 : Conf->Spell = (SPELL **) tmpalloc(Conf->mspell * sizeof(SPELL *));
502 : }
503 : }
504 755 : Conf->Spell[Conf->nspell] = (SPELL *) tmpalloc(SPELLHDRSZ + strlen(word) + 1);
505 755 : strcpy(Conf->Spell[Conf->nspell]->word, word);
506 1510 : Conf->Spell[Conf->nspell]->p.flag = (*flag != '\0')
507 755 : ? cpstrdup(Conf, flag) : VoidString;
508 755 : Conf->nspell++;
509 755 : }
510 :
511 : /*
512 : * Imports dictionary into the temporary array Spell.
513 : *
514 : * Note caller must already have applied get_tsearch_config_filename.
515 : *
516 : * Conf: current dictionary.
517 : * filename: path to the .dict file.
518 : */
519 : void
520 83 : NIImportDictionary(IspellDict *Conf, const char *filename)
521 : {
522 : tsearch_readline_state trst;
523 : char *line;
524 :
525 83 : if (!tsearch_readline_begin(&trst, filename))
526 0 : ereport(ERROR,
527 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
528 : errmsg("could not open dictionary file \"%s\": %m",
529 : filename)));
530 :
531 838 : while ((line = tsearch_readline(&trst)) != NULL)
532 : {
533 : char *s,
534 : *pstr;
535 :
536 : /* Set of affix flags */
537 : const char *flag;
538 :
539 : /* Extract flag from the line */
540 755 : flag = NULL;
541 755 : if ((s = findchar(line, '/')))
542 : {
543 672 : *s++ = '\0';
544 672 : flag = s;
545 2689 : while (*s)
546 : {
547 : /* we allow only single encoded flags for faster works */
548 2689 : if (pg_mblen_cstr(s) == 1 && isprint((unsigned char) *s) && !isspace((unsigned char) *s))
549 2017 : s++;
550 : else
551 : {
552 672 : *s = '\0';
553 672 : break;
554 : }
555 : }
556 : }
557 : else
558 83 : flag = "";
559 :
560 : /* Remove trailing spaces */
561 755 : s = line;
562 5481 : while (*s)
563 : {
564 4809 : if (isspace((unsigned char) *s))
565 : {
566 83 : *s = '\0';
567 83 : break;
568 : }
569 4726 : s += pg_mblen_cstr(s);
570 : }
571 755 : pstr = lowerstr_ctx(Conf, line);
572 :
573 755 : NIAddSpell(Conf, pstr, flag);
574 755 : pfree(pstr);
575 :
576 755 : pfree(line);
577 : }
578 83 : tsearch_readline_end(&trst);
579 83 : }
580 :
581 : /*
582 : * Searches a basic form of word in the prefix tree. This word was generated
583 : * using an affix rule. This rule may not be presented in an affix set of
584 : * a basic form of word.
585 : *
586 : * For example, we have the entry in the .dict file:
587 : * meter/GMD
588 : *
589 : * The affix rule with the flag S:
590 : * SFX S y ies [^aeiou]y
591 : * is not presented here.
592 : *
593 : * The affix rule with the flag M:
594 : * SFX M 0 's .
595 : * is presented here.
596 : *
597 : * Conf: current dictionary.
598 : * word: basic form of word.
599 : * affixflag: affix flag, by which a basic form of word was generated.
600 : * flag: compound flag used to compare with StopMiddle->compoundflag.
601 : *
602 : * Returns 1 if the word was found in the prefix tree, else returns 0.
603 : */
604 : static int
605 2495 : FindWord(IspellDict *Conf, const char *word, const char *affixflag, int flag)
606 : {
607 2495 : SPNode *node = Conf->Dictionary;
608 : SPNodeData *StopLow,
609 : *StopHigh,
610 : *StopMiddle;
611 2495 : const uint8 *ptr = (const uint8 *) word;
612 :
613 2495 : flag &= FF_COMPOUNDFLAGMASK;
614 :
615 11620 : while (node && *ptr)
616 : {
617 11020 : StopLow = node->data;
618 11020 : StopHigh = node->data + node->length;
619 15765 : while (StopLow < StopHigh)
620 : {
621 14710 : StopMiddle = StopLow + ((StopHigh - StopLow) >> 1);
622 14710 : if (StopMiddle->val == *ptr)
623 : {
624 9965 : if (*(ptr + 1) == '\0' && StopMiddle->isword)
625 : {
626 955 : if (flag == 0)
627 : {
628 : /*
629 : * The word can be formed only with another word. And
630 : * in the flag parameter there is not a sign that we
631 : * search compound words.
632 : */
633 605 : if (StopMiddle->compoundflag & FF_COMPOUNDONLY)
634 0 : return 0;
635 : }
636 350 : else if ((flag & StopMiddle->compoundflag) == 0)
637 0 : return 0;
638 :
639 : /*
640 : * Check if this affix rule is presented in the affix set
641 : * with index StopMiddle->affix.
642 : */
643 955 : if (IsAffixFlagInUse(Conf, StopMiddle->affix, affixflag))
644 840 : return 1;
645 : }
646 9125 : node = StopMiddle->node;
647 9125 : ptr++;
648 9125 : break;
649 : }
650 4745 : else if (StopMiddle->val < *ptr)
651 1610 : StopLow = StopMiddle + 1;
652 : else
653 3135 : StopHigh = StopMiddle;
654 : }
655 10180 : if (StopLow >= StopHigh)
656 1055 : break;
657 : }
658 1655 : return 0;
659 : }
660 :
661 : /*
662 : * Adds a new affix rule to the Affix field.
663 : *
664 : * Conf: current dictionary.
665 : * flag: affix flag ('\' in the below example).
666 : * flagflags: set of flags from the flagval field for this affix rule. This set
667 : * is listed after '/' character in the added string (repl).
668 : *
669 : * For example L flag in the hunspell_sample.affix:
670 : * SFX \ 0 Y/L [^Y]
671 : *
672 : * mask: condition for search ('[^Y]' in the above example).
673 : * find: stripping characters from beginning (at prefix) or end (at suffix)
674 : * of the word ('0' in the above example, 0 means that there is not
675 : * stripping character).
676 : * repl: adding string after stripping ('Y' in the above example).
677 : * type: FF_SUFFIX or FF_PREFIX.
678 : */
679 : static void
680 688 : NIAddAffix(IspellDict *Conf, const char *flag, char flagflags, const char *mask,
681 : const char *find, const char *repl, int type)
682 : {
683 : AFFIX *Affix;
684 :
685 688 : if (Conf->naffixes >= Conf->maffixes)
686 : {
687 83 : if (Conf->maffixes)
688 : {
689 0 : Conf->maffixes *= 2;
690 0 : Conf->Affix = (AFFIX *) repalloc(Conf->Affix, Conf->maffixes * sizeof(AFFIX));
691 : }
692 : else
693 : {
694 83 : Conf->maffixes = 16;
695 83 : Conf->Affix = palloc_array(AFFIX, Conf->maffixes);
696 : }
697 : }
698 :
699 688 : Affix = Conf->Affix + Conf->naffixes;
700 :
701 : /* This affix rule can be applied for words with any ending */
702 688 : if (strcmp(mask, ".") == 0 || *mask == '\0')
703 : {
704 166 : Affix->issimple = 1;
705 166 : Affix->isregis = 0;
706 : }
707 : /* This affix rule will use regis to search word ending */
708 522 : else if (RS_isRegis(mask))
709 : {
710 436 : Affix->issimple = 0;
711 436 : Affix->isregis = 1;
712 436 : RS_compile(&(Affix->reg.regis), (type == FF_SUFFIX),
713 436 : *mask ? mask : VoidString);
714 : }
715 : /* This affix rule will use regex_t to search word ending */
716 : else
717 : {
718 : int masklen;
719 : int wmasklen;
720 : int err;
721 : pg_wchar *wmask;
722 : char *tmask;
723 :
724 86 : Affix->issimple = 0;
725 86 : Affix->isregis = 0;
726 86 : tmask = (char *) tmpalloc(strlen(mask) + 3);
727 86 : if (type == FF_SUFFIX)
728 86 : sprintf(tmask, "%s$", mask);
729 : else
730 0 : sprintf(tmask, "^%s", mask);
731 :
732 86 : masklen = strlen(tmask);
733 86 : wmask = (pg_wchar *) tmpalloc((masklen + 1) * sizeof(pg_wchar));
734 86 : wmasklen = pg_mb2wchar_with_len(tmask, wmask, masklen);
735 :
736 : /*
737 : * The regex and all internal state created by pg_regcomp are
738 : * allocated in the dictionary's memory context, and will be freed
739 : * automatically when it is destroyed.
740 : */
741 86 : Affix->reg.pregex = palloc_object(regex_t);
742 86 : err = pg_regcomp(Affix->reg.pregex, wmask, wmasklen,
743 : REG_ADVANCED | REG_NOSUB,
744 : DEFAULT_COLLATION_OID);
745 86 : if (err)
746 : {
747 : char errstr[100];
748 :
749 0 : pg_regerror(err, Affix->reg.pregex, errstr, sizeof(errstr));
750 0 : ereport(ERROR,
751 : (errcode(ERRCODE_INVALID_REGULAR_EXPRESSION),
752 : errmsg("invalid regular expression: %s", errstr)));
753 : }
754 : }
755 :
756 688 : Affix->flagflags = flagflags;
757 688 : if ((Affix->flagflags & FF_COMPOUNDONLY) || (Affix->flagflags & FF_COMPOUNDPERMITFLAG))
758 : {
759 125 : if ((Affix->flagflags & FF_COMPOUNDFLAG) == 0)
760 125 : Affix->flagflags |= FF_COMPOUNDFLAG;
761 : }
762 688 : Affix->flag = cpstrdup(Conf, flag);
763 688 : Affix->type = type;
764 :
765 688 : Affix->find = (find && *find) ? cpstrdup(Conf, find) : VoidString;
766 688 : if ((Affix->replen = strlen(repl)) > 0)
767 666 : Affix->repl = cpstrdup(Conf, repl);
768 : else
769 22 : Affix->repl = VoidString;
770 688 : Conf->naffixes++;
771 688 : }
772 :
773 : /* Parsing states for parse_affentry() and friends */
774 : #define PAE_WAIT_MASK 0
775 : #define PAE_INMASK 1
776 : #define PAE_WAIT_FIND 2
777 : #define PAE_INFIND 3
778 : #define PAE_WAIT_REPL 4
779 : #define PAE_INREPL 5
780 : #define PAE_WAIT_TYPE 6
781 : #define PAE_WAIT_FLAG 7
782 :
783 : /*
784 : * Parse next space-separated field of an .affix file line.
785 : *
786 : * *str is the input pointer (will be advanced past field)
787 : * next is where to copy the field value to, with null termination
788 : *
789 : * The buffer at "next" must be of size BUFSIZ; we truncate the input to fit.
790 : *
791 : * Returns true if we found a field, false if not.
792 : */
793 : static bool
794 6463 : get_nextfield(char **str, char *next)
795 : {
796 6463 : int state = PAE_WAIT_MASK;
797 6463 : int avail = BUFSIZ;
798 :
799 27642 : while (**str)
800 : {
801 26883 : int clen = pg_mblen_cstr(*str);
802 :
803 26883 : if (state == PAE_WAIT_MASK)
804 : {
805 11921 : if (t_iseq(*str, '#'))
806 231 : return false;
807 11690 : else if (!isspace((unsigned char) **str))
808 : {
809 5473 : if (clen < avail)
810 : {
811 5473 : ts_copychar_with_len(next, *str, clen);
812 5473 : next += clen;
813 5473 : avail -= clen;
814 : }
815 5473 : state = PAE_INMASK;
816 : }
817 : }
818 : else /* state == PAE_INMASK */
819 : {
820 14962 : if (isspace((unsigned char) **str))
821 : {
822 5473 : *next = '\0';
823 5473 : return true;
824 : }
825 : else
826 : {
827 9489 : if (clen < avail)
828 : {
829 9489 : ts_copychar_with_len(next, *str, clen);
830 9489 : next += clen;
831 9489 : avail -= clen;
832 : }
833 : }
834 : }
835 21179 : *str += clen;
836 : }
837 :
838 759 : *next = '\0';
839 :
840 759 : return (state == PAE_INMASK); /* OK if we got a nonempty field */
841 : }
842 :
843 : /*
844 : * Parses entry of an .affix file of MySpell or Hunspell format.
845 : *
846 : * An .affix file entry has the following format:
847 : * - header
848 : * <type> <flag> <cross_flag> <flag_count>
849 : * - fields after header:
850 : * <type> <flag> <find> <replace> <mask>
851 : *
852 : * str is the input line
853 : * field values are returned to type etc, which must be buffers of size BUFSIZ.
854 : *
855 : * Returns number of fields found; any omitted fields are set to empty strings.
856 : */
857 : static int
858 1489 : parse_ooaffentry(char *str, char *type, char *flag, char *find,
859 : char *repl, char *mask)
860 : {
861 1489 : int state = PAE_WAIT_TYPE;
862 1489 : int fields_read = 0;
863 1489 : bool valid = false;
864 :
865 1489 : *type = *flag = *find = *repl = *mask = '\0';
866 :
867 6463 : while (*str)
868 : {
869 6463 : switch (state)
870 : {
871 1489 : case PAE_WAIT_TYPE:
872 1489 : valid = get_nextfield(&str, type);
873 1489 : state = PAE_WAIT_FLAG;
874 1489 : break;
875 1489 : case PAE_WAIT_FLAG:
876 1489 : valid = get_nextfield(&str, flag);
877 1489 : state = PAE_WAIT_FIND;
878 1489 : break;
879 1489 : case PAE_WAIT_FIND:
880 1489 : valid = get_nextfield(&str, find);
881 1489 : state = PAE_WAIT_REPL;
882 1489 : break;
883 998 : case PAE_WAIT_REPL:
884 998 : valid = get_nextfield(&str, repl);
885 998 : state = PAE_WAIT_MASK;
886 998 : break;
887 998 : case PAE_WAIT_MASK:
888 998 : valid = get_nextfield(&str, mask);
889 998 : state = -1; /* force loop exit */
890 998 : break;
891 0 : default:
892 0 : elog(ERROR, "unrecognized state in parse_ooaffentry: %d",
893 : state);
894 : break;
895 : }
896 6463 : if (valid)
897 5473 : fields_read++;
898 : else
899 990 : break; /* early EOL */
900 5473 : if (state < 0)
901 499 : break; /* got all fields */
902 : }
903 :
904 1489 : return fields_read;
905 : }
906 :
907 : /*
908 : * Parses entry of an .affix file of Ispell format
909 : *
910 : * An .affix file entry has the following format:
911 : * <mask> > [-<find>,]<replace>
912 : *
913 : * Output buffers mask, find, repl must be of length BUFSIZ;
914 : * we truncate the input to fit.
915 : */
916 : static bool
917 189 : parse_affentry(const char *str, char *mask, char *find, char *repl)
918 : {
919 189 : int state = PAE_WAIT_MASK;
920 189 : char *pmask = mask,
921 189 : *pfind = find,
922 189 : *prepl = repl;
923 189 : char *emask = mask + BUFSIZ;
924 189 : char *efind = find + BUFSIZ;
925 189 : char *erepl = repl + BUFSIZ;
926 :
927 189 : *mask = *find = *repl = '\0';
928 :
929 4968 : while (*str)
930 : {
931 4968 : int clen = pg_mblen_cstr(str);
932 :
933 4968 : if (state == PAE_WAIT_MASK)
934 : {
935 459 : if (t_iseq(str, '#'))
936 0 : return false;
937 459 : else if (!isspace((unsigned char) *str))
938 : {
939 189 : if (pmask < emask - clen)
940 189 : pmask += ts_copychar_with_len(pmask, str, clen);
941 189 : state = PAE_INMASK;
942 : }
943 : }
944 4509 : else if (state == PAE_INMASK)
945 : {
946 1836 : if (t_iseq(str, '>'))
947 : {
948 189 : *pmask = '\0';
949 189 : state = PAE_WAIT_FIND;
950 : }
951 1647 : else if (!isspace((unsigned char) *str))
952 : {
953 648 : if (pmask < emask - clen)
954 648 : pmask += ts_copychar_with_len(pmask, str, clen);
955 : }
956 : }
957 2673 : else if (state == PAE_WAIT_FIND)
958 : {
959 756 : if (t_iseq(str, '-'))
960 : {
961 27 : state = PAE_INFIND;
962 : }
963 729 : else if (t_isalpha_cstr(str) || t_iseq(str, '\'') /* english 's */ )
964 : {
965 162 : if (prepl < erepl - clen)
966 162 : prepl += ts_copychar_with_len(prepl, str, clen);
967 162 : state = PAE_INREPL;
968 : }
969 567 : else if (!isspace((unsigned char) *str))
970 0 : ereport(ERROR,
971 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
972 : errmsg("syntax error")));
973 : }
974 1917 : else if (state == PAE_INFIND)
975 : {
976 54 : if (t_iseq(str, ','))
977 : {
978 27 : *pfind = '\0';
979 27 : state = PAE_WAIT_REPL;
980 : }
981 27 : else if (t_isalpha_cstr(str))
982 : {
983 27 : if (pfind < efind - clen)
984 27 : pfind += ts_copychar_with_len(pfind, str, clen);
985 : }
986 0 : else if (!isspace((unsigned char) *str))
987 0 : ereport(ERROR,
988 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
989 : errmsg("syntax error")));
990 : }
991 1863 : else if (state == PAE_WAIT_REPL)
992 : {
993 27 : if (t_iseq(str, '-'))
994 : {
995 0 : break; /* void repl */
996 : }
997 27 : else if (t_isalpha_cstr(str))
998 : {
999 27 : if (prepl < erepl - clen)
1000 27 : prepl += ts_copychar_with_len(prepl, str, clen);
1001 27 : state = PAE_INREPL;
1002 : }
1003 0 : else if (!isspace((unsigned char) *str))
1004 0 : ereport(ERROR,
1005 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1006 : errmsg("syntax error")));
1007 : }
1008 1836 : else if (state == PAE_INREPL)
1009 : {
1010 1836 : if (t_iseq(str, '#'))
1011 : {
1012 189 : *prepl = '\0';
1013 189 : break;
1014 : }
1015 1647 : else if (t_isalpha_cstr(str))
1016 : {
1017 243 : if (prepl < erepl - clen)
1018 243 : prepl += ts_copychar_with_len(prepl, str, clen);
1019 : }
1020 1404 : else if (!isspace((unsigned char) *str))
1021 0 : ereport(ERROR,
1022 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1023 : errmsg("syntax error")));
1024 : }
1025 : else
1026 0 : elog(ERROR, "unrecognized state in parse_affentry: %d", state);
1027 :
1028 4779 : str += clen;
1029 : }
1030 :
1031 189 : *pmask = *pfind = *prepl = '\0';
1032 :
1033 189 : return (*mask && (*find || *repl));
1034 : }
1035 :
1036 : /*
1037 : * Sets a Hunspell options depending on flag type.
1038 : */
1039 : static void
1040 1839 : setCompoundAffixFlagValue(IspellDict *Conf, CompoundAffixFlag *entry,
1041 : char *s, uint32 val)
1042 : {
1043 1839 : if (Conf->flagMode == FM_NUM)
1044 : {
1045 : char *next;
1046 : int i;
1047 :
1048 399 : errno = 0;
1049 399 : i = strtol(s, &next, 10);
1050 399 : if (s == next || errno == ERANGE)
1051 0 : ereport(ERROR,
1052 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1053 : errmsg("invalid affix flag \"%s\"", s)));
1054 399 : if (i < 0 || i > FLAGNUM_MAXSIZE)
1055 0 : ereport(ERROR,
1056 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1057 : errmsg("affix flag \"%s\" is out of range", s)));
1058 :
1059 399 : entry->flag.i = i;
1060 : }
1061 : else
1062 1440 : entry->flag.s = cpstrdup(Conf, s);
1063 :
1064 1839 : entry->flagMode = Conf->flagMode;
1065 1839 : entry->value = val;
1066 1839 : }
1067 :
1068 : /*
1069 : * Sets up a correspondence for the affix parameter with the affix flag.
1070 : *
1071 : * Conf: current dictionary.
1072 : * s: affix flag in string.
1073 : * val: affix parameter.
1074 : */
1075 : static void
1076 223 : addCompoundAffixFlagValue(IspellDict *Conf, const char *s, uint32 val)
1077 : {
1078 : CompoundAffixFlag *newValue;
1079 : char sbuf[BUFSIZ];
1080 : char *sflag;
1081 :
1082 419 : while (*s && isspace((unsigned char) *s))
1083 196 : s += pg_mblen_cstr(s);
1084 :
1085 223 : if (!*s)
1086 0 : ereport(ERROR,
1087 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1088 : errmsg("syntax error")));
1089 :
1090 : /* Get flag without \n */
1091 223 : sflag = sbuf;
1092 660 : while (*s && !isspace((unsigned char) *s) && *s != '\n')
1093 : {
1094 437 : int clen = pg_mblen_cstr(s);
1095 :
1096 : /* Truncate the input to fit in BUFSIZ */
1097 437 : if (sflag < sbuf + BUFSIZ - clen)
1098 437 : sflag += ts_copychar_with_len(sflag, s, clen);
1099 437 : s += clen;
1100 : }
1101 223 : *sflag = '\0';
1102 :
1103 : /* Resize array or allocate memory for array CompoundAffixFlag */
1104 223 : if (Conf->nCompoundAffixFlag >= Conf->mCompoundAffixFlag)
1105 : {
1106 83 : if (Conf->mCompoundAffixFlag)
1107 : {
1108 0 : Conf->mCompoundAffixFlag *= 2;
1109 0 : Conf->CompoundAffixFlags = (CompoundAffixFlag *)
1110 0 : repalloc(Conf->CompoundAffixFlags,
1111 0 : Conf->mCompoundAffixFlag * sizeof(CompoundAffixFlag));
1112 : }
1113 : else
1114 : {
1115 83 : Conf->mCompoundAffixFlag = 10;
1116 83 : Conf->CompoundAffixFlags = (CompoundAffixFlag *)
1117 83 : tmpalloc(Conf->mCompoundAffixFlag * sizeof(CompoundAffixFlag));
1118 : }
1119 : }
1120 :
1121 223 : newValue = Conf->CompoundAffixFlags + Conf->nCompoundAffixFlag;
1122 :
1123 223 : setCompoundAffixFlagValue(Conf, newValue, sbuf, val);
1124 :
1125 223 : Conf->usecompound = true;
1126 223 : Conf->nCompoundAffixFlag++;
1127 223 : }
1128 :
1129 : /*
1130 : * Returns a set of affix parameters which correspondence to the set of affix
1131 : * flags s.
1132 : */
1133 : static int
1134 798 : getCompoundAffixFlagValue(IspellDict *Conf, const char *s)
1135 : {
1136 798 : uint32 flag = 0;
1137 : CompoundAffixFlag *found,
1138 : key;
1139 : char sflag[BUFSIZ];
1140 : const char *flagcur;
1141 :
1142 798 : if (Conf->nCompoundAffixFlag == 0)
1143 0 : return 0;
1144 :
1145 798 : flagcur = s;
1146 2414 : while (*flagcur)
1147 : {
1148 1620 : getNextFlagFromString(Conf, &flagcur, sflag);
1149 1616 : setCompoundAffixFlagValue(Conf, &key, sflag, 0);
1150 :
1151 : found = (CompoundAffixFlag *)
1152 1616 : bsearch(&key, Conf->CompoundAffixFlags,
1153 1616 : Conf->nCompoundAffixFlag, sizeof(CompoundAffixFlag),
1154 : cmpcmdflag);
1155 1616 : if (found != NULL)
1156 374 : flag |= found->value;
1157 : }
1158 :
1159 794 : return flag;
1160 : }
1161 :
1162 : /*
1163 : * Returns a flag set using the s parameter.
1164 : *
1165 : * If Conf->useFlagAliases is true then the s parameter is index of the
1166 : * Conf->AffixData array and function returns its entry.
1167 : * Else function returns the s parameter.
1168 : */
1169 : static const char *
1170 98 : getAffixFlagSet(IspellDict *Conf, char *s)
1171 : {
1172 98 : if (Conf->useFlagAliases && *s != '\0')
1173 : {
1174 : int curaffix;
1175 : char *end;
1176 :
1177 63 : errno = 0;
1178 63 : curaffix = strtol(s, &end, 10);
1179 63 : if (s == end || errno == ERANGE)
1180 0 : ereport(ERROR,
1181 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1182 : errmsg("invalid affix alias \"%s\"", s)));
1183 :
1184 63 : if (curaffix > 0 && curaffix < Conf->nAffixData)
1185 :
1186 : /*
1187 : * Do not subtract 1 from curaffix because empty string was added
1188 : * in NIImportOOAffixes
1189 : */
1190 63 : return Conf->AffixData[curaffix];
1191 0 : else if (curaffix > Conf->nAffixData)
1192 0 : ereport(ERROR,
1193 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1194 : errmsg("invalid affix alias \"%s\"", s)));
1195 0 : return VoidString;
1196 : }
1197 : else
1198 35 : return s;
1199 : }
1200 :
1201 : /*
1202 : * Import an affix file that follows MySpell or Hunspell format.
1203 : *
1204 : * Conf: current dictionary.
1205 : * filename: path to the .affix file.
1206 : */
1207 : static void
1208 56 : NIImportOOAffixes(IspellDict *Conf, const char *filename)
1209 : {
1210 : char type[BUFSIZ],
1211 56 : *ptype = NULL;
1212 : char sflag[BUFSIZ];
1213 : char mask[BUFSIZ],
1214 : *pmask;
1215 : char find[BUFSIZ],
1216 : *pfind;
1217 : char repl[BUFSIZ],
1218 : *prepl;
1219 56 : bool isSuffix = false;
1220 56 : int naffix = 0,
1221 56 : curaffix = 0;
1222 56 : int sflaglen = 0;
1223 56 : char flagflags = 0;
1224 : tsearch_readline_state trst;
1225 : char *recoded;
1226 :
1227 : /* read file to find any flag */
1228 56 : Conf->usecompound = false;
1229 56 : Conf->useFlagAliases = false;
1230 56 : Conf->flagMode = FM_CHAR;
1231 :
1232 56 : if (!tsearch_readline_begin(&trst, filename))
1233 0 : ereport(ERROR,
1234 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1235 : errmsg("could not open affix file \"%s\": %m",
1236 : filename)));
1237 :
1238 2194 : while ((recoded = tsearch_readline(&trst)) != NULL)
1239 : {
1240 2138 : if (*recoded == '\0' || isspace((unsigned char) *recoded) || t_iseq(recoded, '#'))
1241 : {
1242 649 : pfree(recoded);
1243 649 : continue;
1244 : }
1245 :
1246 1489 : if (STRNCMP(recoded, "COMPOUNDFLAG") == 0)
1247 56 : addCompoundAffixFlagValue(Conf, recoded + strlen("COMPOUNDFLAG"),
1248 : FF_COMPOUNDFLAG);
1249 1433 : else if (STRNCMP(recoded, "COMPOUNDBEGIN") == 0)
1250 21 : addCompoundAffixFlagValue(Conf, recoded + strlen("COMPOUNDBEGIN"),
1251 : FF_COMPOUNDBEGIN);
1252 1412 : else if (STRNCMP(recoded, "COMPOUNDLAST") == 0)
1253 0 : addCompoundAffixFlagValue(Conf, recoded + strlen("COMPOUNDLAST"),
1254 : FF_COMPOUNDLAST);
1255 : /* COMPOUNDLAST and COMPOUNDEND are synonyms */
1256 1412 : else if (STRNCMP(recoded, "COMPOUNDEND") == 0)
1257 21 : addCompoundAffixFlagValue(Conf, recoded + strlen("COMPOUNDEND"),
1258 : FF_COMPOUNDLAST);
1259 1391 : else if (STRNCMP(recoded, "COMPOUNDMIDDLE") == 0)
1260 21 : addCompoundAffixFlagValue(Conf, recoded + strlen("COMPOUNDMIDDLE"),
1261 : FF_COMPOUNDMIDDLE);
1262 1370 : else if (STRNCMP(recoded, "ONLYINCOMPOUND") == 0)
1263 56 : addCompoundAffixFlagValue(Conf, recoded + strlen("ONLYINCOMPOUND"),
1264 : FF_COMPOUNDONLY);
1265 1314 : else if (STRNCMP(recoded, "COMPOUNDPERMITFLAG") == 0)
1266 21 : addCompoundAffixFlagValue(Conf,
1267 21 : recoded + strlen("COMPOUNDPERMITFLAG"),
1268 : FF_COMPOUNDPERMITFLAG);
1269 1293 : else if (STRNCMP(recoded, "COMPOUNDFORBIDFLAG") == 0)
1270 0 : addCompoundAffixFlagValue(Conf,
1271 0 : recoded + strlen("COMPOUNDFORBIDFLAG"),
1272 : FF_COMPOUNDFORBIDFLAG);
1273 1293 : else if (STRNCMP(recoded, "FLAG") == 0)
1274 : {
1275 43 : char *s = recoded + strlen("FLAG");
1276 :
1277 86 : while (*s && isspace((unsigned char) *s))
1278 43 : s += pg_mblen_cstr(s);
1279 :
1280 43 : if (*s)
1281 : {
1282 43 : if (STRNCMP(s, "long") == 0)
1283 21 : Conf->flagMode = FM_LONG;
1284 22 : else if (STRNCMP(s, "num") == 0)
1285 22 : Conf->flagMode = FM_NUM;
1286 0 : else if (STRNCMP(s, "default") != 0)
1287 0 : ereport(ERROR,
1288 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1289 : errmsg("Ispell dictionary supports only "
1290 : "\"default\", \"long\", "
1291 : "and \"num\" flag values")));
1292 : }
1293 : }
1294 :
1295 1489 : pfree(recoded);
1296 : }
1297 56 : tsearch_readline_end(&trst);
1298 :
1299 56 : if (Conf->nCompoundAffixFlag > 1)
1300 56 : qsort(Conf->CompoundAffixFlags, Conf->nCompoundAffixFlag,
1301 : sizeof(CompoundAffixFlag), cmpcmdflag);
1302 :
1303 56 : if (!tsearch_readline_begin(&trst, filename))
1304 0 : ereport(ERROR,
1305 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1306 : errmsg("could not open affix file \"%s\": %m",
1307 : filename)));
1308 :
1309 2194 : while ((recoded = tsearch_readline(&trst)) != NULL)
1310 : {
1311 : int fields_read;
1312 :
1313 2138 : if (*recoded == '\0' || isspace((unsigned char) *recoded) || t_iseq(recoded, '#'))
1314 649 : goto nextline;
1315 :
1316 1489 : fields_read = parse_ooaffentry(recoded, type, sflag, find, repl, mask);
1317 :
1318 1489 : if (ptype)
1319 1433 : pfree(ptype);
1320 1489 : ptype = lowerstr_ctx(Conf, type);
1321 :
1322 : /* First try to parse AF parameter (alias compression) */
1323 1489 : if (STRNCMP(ptype, "af") == 0)
1324 : {
1325 : /* First line is the number of aliases */
1326 252 : if (!Conf->useFlagAliases)
1327 : {
1328 21 : Conf->useFlagAliases = true;
1329 21 : naffix = atoi(sflag);
1330 21 : if (naffix <= 0)
1331 0 : ereport(ERROR,
1332 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1333 : errmsg("invalid number of flag vector aliases")));
1334 :
1335 : /* Also reserve place for empty flag set */
1336 21 : naffix++;
1337 :
1338 21 : Conf->AffixData = palloc0_array(const char *, naffix);
1339 21 : Conf->lenAffixData = Conf->nAffixData = naffix;
1340 :
1341 : /* Add empty flag set into AffixData */
1342 21 : Conf->AffixData[curaffix] = VoidString;
1343 21 : curaffix++;
1344 : }
1345 : /* Other lines are aliases */
1346 : else
1347 : {
1348 231 : if (curaffix < naffix)
1349 : {
1350 231 : Conf->AffixData[curaffix] = cpstrdup(Conf, sflag);
1351 231 : curaffix++;
1352 : }
1353 : else
1354 0 : ereport(ERROR,
1355 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1356 : errmsg("number of aliases exceeds specified number %d",
1357 : naffix - 1)));
1358 : }
1359 252 : goto nextline;
1360 : }
1361 : /* Else try to parse prefixes and suffixes */
1362 1237 : if (fields_read < 4 ||
1363 998 : (STRNCMP(ptype, "sfx") != 0 && STRNCMP(ptype, "pfx") != 0))
1364 239 : goto nextline;
1365 :
1366 998 : sflaglen = strlen(sflag);
1367 998 : if (sflaglen == 0
1368 998 : || (sflaglen > 1 && Conf->flagMode == FM_CHAR)
1369 998 : || (sflaglen > 2 && Conf->flagMode == FM_LONG))
1370 0 : goto nextline;
1371 :
1372 : /*--------
1373 : * Affix header. For example:
1374 : * SFX \ N 1
1375 : *--------
1376 : */
1377 998 : if (fields_read == 4)
1378 : {
1379 499 : isSuffix = (STRNCMP(ptype, "sfx") == 0);
1380 499 : if (t_iseq(find, 'y') || t_iseq(find, 'Y'))
1381 345 : flagflags = FF_CROSSPRODUCT;
1382 : else
1383 154 : flagflags = 0;
1384 : }
1385 : /*--------
1386 : * Affix fields. For example:
1387 : * SFX \ 0 Y/L [^Y]
1388 : *--------
1389 : */
1390 : else
1391 : {
1392 : char *ptr;
1393 499 : int aflg = 0;
1394 :
1395 : /* Get flags after '/' (flags are case sensitive) */
1396 499 : if ((ptr = strchr(repl, '/')) != NULL)
1397 98 : aflg |= getCompoundAffixFlagValue(Conf,
1398 : getAffixFlagSet(Conf,
1399 : ptr + 1));
1400 : /* Get lowercased version of string before '/' */
1401 499 : prepl = lowerstr_ctx(Conf, repl);
1402 499 : if ((ptr = strchr(prepl, '/')) != NULL)
1403 98 : *ptr = '\0';
1404 499 : pfind = lowerstr_ctx(Conf, find);
1405 499 : pmask = lowerstr_ctx(Conf, mask);
1406 499 : if (t_iseq(find, '0'))
1407 421 : *pfind = '\0';
1408 499 : if (t_iseq(repl, '0'))
1409 22 : *prepl = '\0';
1410 :
1411 499 : NIAddAffix(Conf, sflag, flagflags | aflg, pmask, pfind, prepl,
1412 : isSuffix ? FF_SUFFIX : FF_PREFIX);
1413 499 : pfree(prepl);
1414 499 : pfree(pfind);
1415 499 : pfree(pmask);
1416 : }
1417 :
1418 2138 : nextline:
1419 2138 : pfree(recoded);
1420 : }
1421 :
1422 56 : tsearch_readline_end(&trst);
1423 56 : if (ptype)
1424 56 : pfree(ptype);
1425 56 : }
1426 :
1427 : /*
1428 : * import affixes
1429 : *
1430 : * Note caller must already have applied get_tsearch_config_filename
1431 : *
1432 : * This function is responsible for parsing ispell ("old format") affix files.
1433 : * If we realize that the file contains new-format commands, we pass off the
1434 : * work to NIImportOOAffixes(), which will re-read the whole file.
1435 : */
1436 : void
1437 83 : NIImportAffixes(IspellDict *Conf, const char *filename)
1438 : {
1439 83 : char *pstr = NULL;
1440 : char flag[BUFSIZ];
1441 : char mask[BUFSIZ];
1442 : char find[BUFSIZ];
1443 : char repl[BUFSIZ];
1444 : char *s;
1445 83 : bool suffixes = false;
1446 83 : bool prefixes = false;
1447 83 : char flagflags = 0;
1448 : tsearch_readline_state trst;
1449 83 : bool oldformat = false;
1450 83 : char *recoded = NULL;
1451 :
1452 83 : if (!tsearch_readline_begin(&trst, filename))
1453 0 : ereport(ERROR,
1454 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1455 : errmsg("could not open affix file \"%s\": %m",
1456 : filename)));
1457 :
1458 83 : Conf->usecompound = false;
1459 83 : Conf->useFlagAliases = false;
1460 83 : Conf->flagMode = FM_CHAR;
1461 :
1462 785 : while ((recoded = tsearch_readline(&trst)) != NULL)
1463 : {
1464 758 : pstr = str_tolower(recoded, strlen(recoded), DEFAULT_COLLATION_OID);
1465 :
1466 : /* Skip comments and empty lines */
1467 758 : if (*pstr == '#' || *pstr == '\n')
1468 243 : goto nextline;
1469 :
1470 515 : if (STRNCMP(pstr, "compoundwords") == 0)
1471 : {
1472 : /* Find case-insensitive L flag in non-lowercased string */
1473 27 : s = findchar2(recoded, 'l', 'L');
1474 27 : if (s)
1475 : {
1476 135 : while (*s && !isspace((unsigned char) *s))
1477 108 : s += pg_mblen_cstr(s);
1478 54 : while (*s && isspace((unsigned char) *s))
1479 27 : s += pg_mblen_cstr(s);
1480 :
1481 27 : if (*s && pg_mblen_cstr(s) == 1)
1482 : {
1483 27 : addCompoundAffixFlagValue(Conf, s, FF_COMPOUNDFLAG);
1484 27 : Conf->usecompound = true;
1485 : }
1486 27 : oldformat = true;
1487 27 : goto nextline;
1488 : }
1489 : }
1490 488 : if (STRNCMP(pstr, "suffixes") == 0)
1491 : {
1492 27 : suffixes = true;
1493 27 : prefixes = false;
1494 27 : oldformat = true;
1495 27 : goto nextline;
1496 : }
1497 461 : if (STRNCMP(pstr, "prefixes") == 0)
1498 : {
1499 27 : suffixes = false;
1500 27 : prefixes = true;
1501 27 : oldformat = true;
1502 27 : goto nextline;
1503 : }
1504 434 : if (STRNCMP(pstr, "flag") == 0)
1505 : {
1506 232 : s = recoded + 4; /* we need non-lowercased string */
1507 232 : flagflags = 0;
1508 :
1509 464 : while (*s && isspace((unsigned char) *s))
1510 232 : s += pg_mblen_cstr(s);
1511 :
1512 232 : if (*s == '*')
1513 : {
1514 135 : flagflags |= FF_CROSSPRODUCT;
1515 135 : s++;
1516 : }
1517 97 : else if (*s == '~')
1518 : {
1519 27 : flagflags |= FF_COMPOUNDONLY;
1520 27 : s++;
1521 : }
1522 :
1523 232 : if (*s == '\\')
1524 27 : s++;
1525 :
1526 : /*
1527 : * An old-format flag is a single ASCII character; we expect it to
1528 : * be followed by EOL, whitespace, or ':'. Otherwise this is a
1529 : * new-format flag command.
1530 : */
1531 232 : if (*s && pg_mblen_cstr(s) == 1)
1532 : {
1533 232 : flag[0] = *s++;
1534 232 : flag[1] = '\0';
1535 :
1536 232 : if (*s == '\0' || *s == '#' || *s == '\n' || *s == ':' ||
1537 43 : isspace((unsigned char) *s))
1538 : {
1539 189 : oldformat = true;
1540 189 : goto nextline;
1541 : }
1542 : }
1543 43 : goto isnewformat;
1544 : }
1545 202 : if (STRNCMP(recoded, "COMPOUNDFLAG") == 0 ||
1546 189 : STRNCMP(recoded, "COMPOUNDMIN") == 0 ||
1547 189 : STRNCMP(recoded, "PFX") == 0 ||
1548 189 : STRNCMP(recoded, "SFX") == 0)
1549 13 : goto isnewformat;
1550 :
1551 189 : if ((!suffixes) && (!prefixes))
1552 0 : goto nextline;
1553 :
1554 189 : if (!parse_affentry(pstr, mask, find, repl))
1555 0 : goto nextline;
1556 :
1557 189 : NIAddAffix(Conf, flag, flagflags, mask, find, repl, suffixes ? FF_SUFFIX : FF_PREFIX);
1558 :
1559 702 : nextline:
1560 702 : pfree(recoded);
1561 702 : pfree(pstr);
1562 : }
1563 27 : tsearch_readline_end(&trst);
1564 27 : return;
1565 :
1566 56 : isnewformat:
1567 56 : if (oldformat)
1568 0 : ereport(ERROR,
1569 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1570 : errmsg("affix file contains both old-style and new-style commands")));
1571 56 : tsearch_readline_end(&trst);
1572 :
1573 56 : NIImportOOAffixes(Conf, filename);
1574 : }
1575 :
1576 : /*
1577 : * Merges two affix flag sets and stores a new affix flag set into
1578 : * Conf->AffixData.
1579 : *
1580 : * Returns index of a new affix flag set.
1581 : */
1582 : static int
1583 41 : MergeAffix(IspellDict *Conf, int a1, int a2)
1584 : {
1585 : const char **ptr;
1586 :
1587 : Assert(a1 < Conf->nAffixData && a2 < Conf->nAffixData);
1588 :
1589 : /* Do not merge affix flags if one of affix flags is empty */
1590 41 : if (*Conf->AffixData[a1] == '\0')
1591 0 : return a2;
1592 41 : else if (*Conf->AffixData[a2] == '\0')
1593 0 : return a1;
1594 :
1595 : /* Double the size of AffixData if there's not enough space */
1596 41 : if (Conf->nAffixData + 1 >= Conf->lenAffixData)
1597 : {
1598 41 : Conf->lenAffixData *= 2;
1599 41 : Conf->AffixData = (const char **) repalloc(Conf->AffixData,
1600 41 : sizeof(char *) * Conf->lenAffixData);
1601 : }
1602 :
1603 41 : ptr = Conf->AffixData + Conf->nAffixData;
1604 41 : if (Conf->flagMode == FM_NUM)
1605 : {
1606 18 : char *p = cpalloc(strlen(Conf->AffixData[a1]) +
1607 : strlen(Conf->AffixData[a2]) +
1608 : 1 /* comma */ + 1 /* \0 */ );
1609 :
1610 18 : sprintf(p, "%s,%s", Conf->AffixData[a1], Conf->AffixData[a2]);
1611 18 : *ptr = p;
1612 : }
1613 : else
1614 : {
1615 23 : char *p = cpalloc(strlen(Conf->AffixData[a1]) +
1616 : strlen(Conf->AffixData[a2]) +
1617 : 1 /* \0 */ );
1618 :
1619 23 : sprintf(p, "%s%s", Conf->AffixData[a1], Conf->AffixData[a2]);
1620 23 : *ptr = p;
1621 : }
1622 41 : ptr++;
1623 41 : *ptr = NULL;
1624 41 : Conf->nAffixData++;
1625 :
1626 41 : return Conf->nAffixData - 1;
1627 : }
1628 :
1629 : /*
1630 : * Returns a set of affix parameters which correspondence to the set of affix
1631 : * flags with the given index.
1632 : */
1633 : static uint32
1634 700 : makeCompoundFlags(IspellDict *Conf, int affix)
1635 : {
1636 : Assert(affix < Conf->nAffixData);
1637 :
1638 700 : return (getCompoundAffixFlagValue(Conf, Conf->AffixData[affix]) &
1639 : FF_COMPOUNDFLAGMASK);
1640 : }
1641 :
1642 : /*
1643 : * Makes a prefix tree for the given level.
1644 : *
1645 : * Conf: current dictionary.
1646 : * low: lower index of the Conf->Spell array.
1647 : * high: upper index of the Conf->Spell array.
1648 : * level: current prefix tree level.
1649 : */
1650 : static SPNode *
1651 2802 : mkSPNode(IspellDict *Conf, int low, int high, int level)
1652 : {
1653 : int i;
1654 2802 : int nchar = 0;
1655 2802 : char lastchar = '\0';
1656 : SPNode *rs;
1657 : SPNodeData *data;
1658 2802 : int lownew = low;
1659 :
1660 9209 : for (i = low; i < high; i++)
1661 6407 : if (Conf->Spell[i]->p.d.len > level && lastchar != Conf->Spell[i]->word[level])
1662 : {
1663 2747 : nchar++;
1664 2747 : lastchar = Conf->Spell[i]->word[level];
1665 : }
1666 :
1667 2802 : if (!nchar)
1668 401 : return NULL;
1669 :
1670 2401 : rs = (SPNode *) cpalloc0(SPNHDRSZ + nchar * sizeof(SPNodeData));
1671 2401 : rs->length = nchar;
1672 2401 : data = rs->data;
1673 :
1674 2401 : lastchar = '\0';
1675 8121 : for (i = low; i < high; i++)
1676 5732 : if (Conf->Spell[i]->p.d.len > level)
1677 : {
1678 4118 : if (lastchar != Conf->Spell[i]->word[level])
1679 : {
1680 2739 : if (lastchar)
1681 : {
1682 : /* Next level of the prefix tree */
1683 338 : data->node = mkSPNode(Conf, lownew, i, level + 1);
1684 330 : lownew = i;
1685 330 : data++;
1686 : }
1687 2731 : lastchar = Conf->Spell[i]->word[level];
1688 : }
1689 4110 : data->val = ((uint8 *) (Conf->Spell[i]->word))[level];
1690 4110 : if (Conf->Spell[i]->p.d.len == level + 1)
1691 : {
1692 659 : bool clearCompoundOnly = false;
1693 :
1694 659 : if (data->isword && data->affix != Conf->Spell[i]->p.d.affix)
1695 : {
1696 : /*
1697 : * MergeAffix called a few times. If one of word is
1698 : * allowed to be in compound word and another isn't, then
1699 : * clear FF_COMPOUNDONLY flag.
1700 : */
1701 :
1702 82 : clearCompoundOnly = (FF_COMPOUNDONLY & data->compoundflag
1703 41 : & makeCompoundFlags(Conf, Conf->Spell[i]->p.d.affix))
1704 : ? false : true;
1705 41 : data->affix = MergeAffix(Conf, data->affix, Conf->Spell[i]->p.d.affix);
1706 : }
1707 : else
1708 618 : data->affix = Conf->Spell[i]->p.d.affix;
1709 659 : data->isword = 1;
1710 :
1711 659 : data->compoundflag = makeCompoundFlags(Conf, data->affix);
1712 :
1713 655 : if ((data->compoundflag & FF_COMPOUNDONLY) &&
1714 0 : (data->compoundflag & FF_COMPOUNDFLAG) == 0)
1715 0 : data->compoundflag |= FF_COMPOUNDFLAG;
1716 :
1717 655 : if (clearCompoundOnly)
1718 41 : data->compoundflag &= ~FF_COMPOUNDONLY;
1719 : }
1720 : }
1721 :
1722 : /* Next level of the prefix tree */
1723 2389 : data->node = mkSPNode(Conf, lownew, high, level + 1);
1724 :
1725 2385 : return rs;
1726 : }
1727 :
1728 : /*
1729 : * Builds the Conf->Dictionary tree and AffixData from the imported dictionary
1730 : * and affixes.
1731 : */
1732 : void
1733 83 : NISortDictionary(IspellDict *Conf)
1734 : {
1735 : int i;
1736 : int naffix;
1737 : int curaffix;
1738 :
1739 : /* compress affixes */
1740 :
1741 : /*
1742 : * If we use flag aliases then we need to use Conf->AffixData filled in
1743 : * the NIImportOOAffixes().
1744 : */
1745 83 : if (Conf->useFlagAliases)
1746 : {
1747 164 : for (i = 0; i < Conf->nspell; i++)
1748 : {
1749 : char *end;
1750 :
1751 151 : if (*Conf->Spell[i]->p.flag != '\0')
1752 : {
1753 138 : errno = 0;
1754 138 : curaffix = strtol(Conf->Spell[i]->p.flag, &end, 10);
1755 138 : if (Conf->Spell[i]->p.flag == end || errno == ERANGE)
1756 4 : ereport(ERROR,
1757 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1758 : errmsg("invalid affix alias \"%s\"",
1759 : Conf->Spell[i]->p.flag)));
1760 134 : if (curaffix < 0 || curaffix >= Conf->nAffixData)
1761 4 : ereport(ERROR,
1762 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1763 : errmsg("invalid affix alias \"%s\"",
1764 : Conf->Spell[i]->p.flag)));
1765 130 : if (*end != '\0' && !isdigit((unsigned char) *end) && !isspace((unsigned char) *end))
1766 0 : ereport(ERROR,
1767 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
1768 : errmsg("invalid affix alias \"%s\"",
1769 : Conf->Spell[i]->p.flag)));
1770 : }
1771 : else
1772 : {
1773 : /*
1774 : * If Conf->Spell[i]->p.flag is empty, then get empty value of
1775 : * Conf->AffixData (0 index).
1776 : */
1777 13 : curaffix = 0;
1778 : }
1779 :
1780 143 : Conf->Spell[i]->p.d.affix = curaffix;
1781 143 : Conf->Spell[i]->p.d.len = strlen(Conf->Spell[i]->word);
1782 : }
1783 : }
1784 : /* Otherwise fill Conf->AffixData here */
1785 : else
1786 : {
1787 : /* Count the number of different flags used in the dictionary */
1788 62 : qsort(Conf->Spell, Conf->nspell, sizeof(SPELL *),
1789 : cmpspellaffix);
1790 :
1791 62 : naffix = 0;
1792 606 : for (i = 0; i < Conf->nspell; i++)
1793 : {
1794 544 : if (i == 0 ||
1795 482 : strcmp(Conf->Spell[i]->p.flag, Conf->Spell[i - 1]->p.flag) != 0)
1796 482 : naffix++;
1797 : }
1798 :
1799 : /*
1800 : * Fill in Conf->AffixData with the affixes that were used in the
1801 : * dictionary. Replace textual flag-field of Conf->Spell entries with
1802 : * indexes into Conf->AffixData array.
1803 : */
1804 62 : Conf->AffixData = palloc0_array(const char *, naffix);
1805 :
1806 62 : curaffix = -1;
1807 606 : for (i = 0; i < Conf->nspell; i++)
1808 : {
1809 544 : if (i == 0 ||
1810 482 : strcmp(Conf->Spell[i]->p.flag, Conf->AffixData[curaffix]) != 0)
1811 : {
1812 482 : curaffix++;
1813 : Assert(curaffix < naffix);
1814 482 : Conf->AffixData[curaffix] = cpstrdup(Conf,
1815 482 : Conf->Spell[i]->p.flag);
1816 : }
1817 :
1818 544 : Conf->Spell[i]->p.d.affix = curaffix;
1819 544 : Conf->Spell[i]->p.d.len = strlen(Conf->Spell[i]->word);
1820 : }
1821 :
1822 62 : Conf->lenAffixData = Conf->nAffixData = naffix;
1823 : }
1824 :
1825 : /* Start build a prefix tree */
1826 75 : qsort(Conf->Spell, Conf->nspell, sizeof(SPELL *), cmpspell);
1827 75 : Conf->Dictionary = mkSPNode(Conf, 0, Conf->nspell, 0);
1828 71 : }
1829 :
1830 : /*
1831 : * Makes a prefix tree for the given level using the repl string of an affix
1832 : * rule. Affixes with empty replace string do not include in the prefix tree.
1833 : * This affixes are included by mkVoidAffix().
1834 : *
1835 : * Conf: current dictionary.
1836 : * low: lower index of the Conf->Affix array.
1837 : * high: upper index of the Conf->Affix array.
1838 : * level: current prefix tree level.
1839 : * type: FF_SUFFIX or FF_PREFIX.
1840 : */
1841 : static AffixNode *
1842 1198 : mkANode(IspellDict *Conf, int low, int high, int level, int type)
1843 : {
1844 : int i;
1845 1198 : int nchar = 0;
1846 1198 : uint8 lastchar = '\0';
1847 : AffixNode *rs;
1848 : AffixNodeData *data;
1849 1198 : int lownew = low;
1850 : int naff;
1851 : AFFIX **aff;
1852 :
1853 3224 : for (i = low; i < high; i++)
1854 2026 : if (Conf->Affix[i].replen > level && lastchar != GETCHAR(Conf->Affix + i, level, type))
1855 : {
1856 1056 : nchar++;
1857 1056 : lastchar = GETCHAR(Conf->Affix + i, level, type);
1858 : }
1859 :
1860 1198 : if (!nchar)
1861 457 : return NULL;
1862 :
1863 741 : aff = (AFFIX **) tmpalloc(sizeof(AFFIX *) * (high - low + 1));
1864 741 : naff = 0;
1865 :
1866 741 : rs = (AffixNode *) cpalloc0(ANHRDSZ + nchar * sizeof(AffixNodeData));
1867 741 : rs->length = nchar;
1868 741 : data = rs->data;
1869 :
1870 741 : lastchar = '\0';
1871 2195 : for (i = low; i < high; i++)
1872 1454 : if (Conf->Affix[i].replen > level)
1873 : {
1874 1224 : if (lastchar != GETCHAR(Conf->Affix + i, level, type))
1875 : {
1876 1056 : if (lastchar)
1877 : {
1878 : /* Next level of the prefix tree */
1879 315 : data->node = mkANode(Conf, lownew, i, level + 1, type);
1880 315 : if (naff)
1881 : {
1882 71 : data->naff = naff;
1883 71 : data->aff = (AFFIX **) cpalloc(sizeof(AFFIX *) * naff);
1884 71 : memcpy(data->aff, aff, sizeof(AFFIX *) * naff);
1885 71 : naff = 0;
1886 : }
1887 315 : data++;
1888 315 : lownew = i;
1889 : }
1890 1056 : lastchar = GETCHAR(Conf->Affix + i, level, type);
1891 : }
1892 1224 : data->val = GETCHAR(Conf->Affix + i, level, type);
1893 1224 : if (Conf->Affix[i].replen == level + 1)
1894 : { /* affix stopped */
1895 554 : aff[naff++] = Conf->Affix + i;
1896 : }
1897 : }
1898 :
1899 : /* Next level of the prefix tree */
1900 741 : data->node = mkANode(Conf, lownew, high, level + 1, type);
1901 741 : if (naff)
1902 : {
1903 457 : data->naff = naff;
1904 457 : data->aff = (AFFIX **) cpalloc(sizeof(AFFIX *) * naff);
1905 457 : memcpy(data->aff, aff, sizeof(AFFIX *) * naff);
1906 457 : naff = 0;
1907 : }
1908 :
1909 741 : pfree(aff);
1910 :
1911 741 : return rs;
1912 : }
1913 :
1914 : /*
1915 : * Makes the root void node in the prefix tree. The root void node is created
1916 : * for affixes which have empty replace string ("repl" field).
1917 : */
1918 : static void
1919 142 : mkVoidAffix(IspellDict *Conf, bool issuffix, int startsuffix)
1920 : {
1921 : int i,
1922 142 : cnt = 0;
1923 142 : int start = (issuffix) ? startsuffix : 0;
1924 142 : int end = (issuffix) ? Conf->naffixes : startsuffix;
1925 142 : AffixNode *Affix = (AffixNode *) palloc0(ANHRDSZ + sizeof(AffixNodeData));
1926 :
1927 142 : Affix->length = 1;
1928 142 : Affix->isvoid = 1;
1929 :
1930 142 : if (issuffix)
1931 : {
1932 71 : Affix->data->node = Conf->Suffix;
1933 71 : Conf->Suffix = Affix;
1934 : }
1935 : else
1936 : {
1937 71 : Affix->data->node = Conf->Prefix;
1938 71 : Conf->Prefix = Affix;
1939 : }
1940 :
1941 : /* Count affixes with empty replace string */
1942 714 : for (i = start; i < end; i++)
1943 572 : if (Conf->Affix[i].replen == 0)
1944 18 : cnt++;
1945 :
1946 : /* There is not affixes with empty replace string */
1947 142 : if (cnt == 0)
1948 124 : return;
1949 :
1950 18 : Affix->data->aff = (AFFIX **) cpalloc(sizeof(AFFIX *) * cnt);
1951 18 : Affix->data->naff = (uint32) cnt;
1952 :
1953 18 : cnt = 0;
1954 144 : for (i = start; i < end; i++)
1955 126 : if (Conf->Affix[i].replen == 0)
1956 : {
1957 18 : Affix->data->aff[cnt] = Conf->Affix + i;
1958 18 : cnt++;
1959 : }
1960 : }
1961 :
1962 : /*
1963 : * Checks if the affixflag is used by dictionary. Conf->AffixData does not
1964 : * contain affixflag if this flag is not used actually by the .dict file.
1965 : *
1966 : * Conf: current dictionary.
1967 : * affixflag: affix flag.
1968 : *
1969 : * Returns true if the Conf->AffixData array contains affixflag, otherwise
1970 : * returns false.
1971 : */
1972 : static bool
1973 97 : isAffixInUse(IspellDict *Conf, const char *affixflag)
1974 : {
1975 : int i;
1976 :
1977 710 : for (i = 0; i < Conf->nAffixData; i++)
1978 695 : if (IsAffixFlagInUse(Conf, i, affixflag))
1979 82 : return true;
1980 :
1981 15 : return false;
1982 : }
1983 :
1984 : /*
1985 : * Builds Conf->Prefix and Conf->Suffix trees from the imported affixes.
1986 : */
1987 : void
1988 71 : NISortAffixes(IspellDict *Conf)
1989 : {
1990 : AFFIX *Affix;
1991 : size_t i;
1992 : CMPDAffix *ptr;
1993 71 : int firstsuffix = Conf->naffixes;
1994 :
1995 71 : if (Conf->naffixes == 0)
1996 0 : return;
1997 :
1998 : /* Store compound affixes in the Conf->CompoundAffix array */
1999 71 : if (Conf->naffixes > 1)
2000 71 : qsort(Conf->Affix, Conf->naffixes, sizeof(AFFIX), cmpaffix);
2001 71 : Conf->CompoundAffix = ptr = palloc_array(CMPDAffix, Conf->naffixes);
2002 71 : ptr->affix = NULL;
2003 :
2004 643 : for (i = 0; i < Conf->naffixes; i++)
2005 : {
2006 572 : Affix = &(((AFFIX *) Conf->Affix)[i]);
2007 572 : if (Affix->type == FF_SUFFIX && i < firstsuffix)
2008 71 : firstsuffix = i;
2009 :
2010 669 : if ((Affix->flagflags & FF_COMPOUNDFLAG) && Affix->replen > 0 &&
2011 97 : isAffixInUse(Conf, Affix->flag))
2012 : {
2013 82 : bool issuffix = (Affix->type == FF_SUFFIX);
2014 :
2015 82 : if (ptr == Conf->CompoundAffix ||
2016 52 : issuffix != (ptr - 1)->issuffix ||
2017 26 : strbncmp((const unsigned char *) (ptr - 1)->affix,
2018 26 : (const unsigned char *) Affix->repl,
2019 26 : (ptr - 1)->len))
2020 : {
2021 : /* leave only unique and minimal suffixes */
2022 69 : ptr->affix = Affix->repl;
2023 69 : ptr->len = Affix->replen;
2024 69 : ptr->issuffix = issuffix;
2025 69 : ptr++;
2026 : }
2027 : }
2028 : }
2029 71 : ptr->affix = NULL;
2030 71 : Conf->CompoundAffix = (CMPDAffix *) repalloc(Conf->CompoundAffix, sizeof(CMPDAffix) * (ptr - Conf->CompoundAffix + 1));
2031 :
2032 : /* Start build a prefix tree */
2033 71 : Conf->Prefix = mkANode(Conf, 0, firstsuffix, 0, FF_PREFIX);
2034 71 : Conf->Suffix = mkANode(Conf, firstsuffix, Conf->naffixes, 0, FF_SUFFIX);
2035 71 : mkVoidAffix(Conf, true, firstsuffix);
2036 71 : mkVoidAffix(Conf, false, firstsuffix);
2037 : }
2038 :
2039 : static AffixNodeData *
2040 3850 : FindAffixes(AffixNode *node, const char *word, int wrdlen, int *level, int type)
2041 : {
2042 : AffixNodeData *StopLow,
2043 : *StopHigh,
2044 : *StopMiddle;
2045 : uint8 symbol;
2046 :
2047 3850 : if (node->isvoid)
2048 : { /* search void affixes */
2049 3350 : if (node->data->naff)
2050 285 : return node->data;
2051 3065 : node = node->data->node;
2052 : }
2053 :
2054 4485 : while (node && *level < wrdlen)
2055 : {
2056 4465 : StopLow = node->data;
2057 4465 : StopHigh = node->data + node->length;
2058 9855 : while (StopLow < StopHigh)
2059 : {
2060 7395 : StopMiddle = StopLow + ((StopHigh - StopLow) >> 1);
2061 7395 : symbol = GETWCHAR(word, wrdlen, *level, type);
2062 :
2063 7395 : if (StopMiddle->val == symbol)
2064 : {
2065 2005 : (*level)++;
2066 2005 : if (StopMiddle->naff)
2067 1085 : return StopMiddle;
2068 920 : node = StopMiddle->node;
2069 920 : break;
2070 : }
2071 5390 : else if (StopMiddle->val < symbol)
2072 1340 : StopLow = StopMiddle + 1;
2073 : else
2074 4050 : StopHigh = StopMiddle;
2075 : }
2076 3380 : if (StopLow >= StopHigh)
2077 2460 : break;
2078 : }
2079 2480 : return NULL;
2080 : }
2081 :
2082 : /*
2083 : * Checks to see if affix applies to word, transforms word if so.
2084 : * The transformation consists of replacing Affix->replen leading or
2085 : * trailing bytes with the Affix->find string.
2086 : *
2087 : * word: input word
2088 : * len: length of input word
2089 : * Affix: affix to consider
2090 : * flagflags: context flags showing whether we are handling a compound word
2091 : * newword: output buffer (MUST be of length 2 * MAXNORMLEN)
2092 : * baselen: input/output argument
2093 : *
2094 : * If baselen isn't NULL, then *baselen is used to return the length of
2095 : * the non-changed part of the word when applying a suffix, and is used
2096 : * to detect whether the input contained only a prefix and suffix when
2097 : * later applying a prefix.
2098 : *
2099 : * Returns newword on success, or NULL if the affix can't be applied.
2100 : * On success, the modified word is stored into newword.
2101 : */
2102 : static char *
2103 1530 : CheckAffix(const char *word, size_t len, AFFIX *Affix, int flagflags, char *newword, int *baselen)
2104 : {
2105 : size_t keeplen,
2106 : findlen;
2107 :
2108 : /*
2109 : * Check compound allow flags
2110 : */
2111 :
2112 1530 : if (flagflags == 0)
2113 : {
2114 1055 : if (Affix->flagflags & FF_COMPOUNDONLY)
2115 110 : return NULL;
2116 : }
2117 475 : else if (flagflags & FF_COMPOUNDBEGIN)
2118 : {
2119 0 : if (Affix->flagflags & FF_COMPOUNDFORBIDFLAG)
2120 0 : return NULL;
2121 0 : if ((Affix->flagflags & FF_COMPOUNDBEGIN) == 0)
2122 0 : if (Affix->type == FF_SUFFIX)
2123 0 : return NULL;
2124 : }
2125 475 : else if (flagflags & FF_COMPOUNDMIDDLE)
2126 : {
2127 340 : if ((Affix->flagflags & FF_COMPOUNDMIDDLE) == 0 ||
2128 190 : (Affix->flagflags & FF_COMPOUNDFORBIDFLAG))
2129 150 : return NULL;
2130 : }
2131 135 : else if (flagflags & FF_COMPOUNDLAST)
2132 : {
2133 135 : if (Affix->flagflags & FF_COMPOUNDFORBIDFLAG)
2134 0 : return NULL;
2135 135 : if ((Affix->flagflags & FF_COMPOUNDLAST) == 0)
2136 125 : if (Affix->type == FF_PREFIX)
2137 0 : return NULL;
2138 : }
2139 :
2140 : /*
2141 : * Protect against output buffer overrun (len < Affix->replen would be
2142 : * caller error, but check anyway)
2143 : */
2144 : Assert(len == strlen(word));
2145 1270 : if (len < Affix->replen)
2146 0 : return NULL;
2147 1270 : keeplen = len - Affix->replen; /* how much of word we will keep */
2148 1270 : findlen = strlen(Affix->find);
2149 1270 : if (keeplen + findlen >= 2 * MAXNORMLEN)
2150 0 : return NULL;
2151 :
2152 : /*
2153 : * make replace pattern of affix
2154 : */
2155 1270 : if (Affix->type == FF_SUFFIX)
2156 : {
2157 870 : memcpy(newword, word, keeplen);
2158 870 : strcpy(newword + keeplen, Affix->find);
2159 870 : if (baselen) /* store length of non-changed part of word */
2160 870 : *baselen = keeplen;
2161 : }
2162 : else
2163 : {
2164 : /*
2165 : * if prefix is an all non-changed part's length then all word
2166 : * contains only prefix and suffix, so out
2167 : */
2168 400 : if (baselen && *baselen + findlen <= Affix->replen)
2169 0 : return NULL;
2170 400 : memcpy(newword, Affix->find, findlen);
2171 400 : strcpy(newword + findlen, word + Affix->replen);
2172 : }
2173 :
2174 : /*
2175 : * check resulting word
2176 : */
2177 1270 : if (Affix->issimple)
2178 400 : return newword;
2179 870 : else if (Affix->isregis)
2180 : {
2181 590 : if (RS_execute(&(Affix->reg.regis), newword))
2182 560 : return newword;
2183 : }
2184 : else
2185 : {
2186 : pg_wchar *data;
2187 : size_t data_len;
2188 : int newword_len;
2189 :
2190 : /* Convert data string to wide characters */
2191 280 : newword_len = strlen(newword);
2192 280 : data = palloc_array(pg_wchar, newword_len + 1);
2193 280 : data_len = pg_mb2wchar_with_len(newword, data, newword_len);
2194 :
2195 280 : if (pg_regexec(Affix->reg.pregex, data, data_len,
2196 : 0, NULL, 0, NULL, 0) == REG_OKAY)
2197 : {
2198 280 : pfree(data);
2199 280 : return newword;
2200 : }
2201 0 : pfree(data);
2202 : }
2203 :
2204 30 : return NULL;
2205 : }
2206 :
2207 : static int
2208 450 : addToResult(char **forms, char **cur, char *word)
2209 : {
2210 450 : if (cur - forms >= MAX_NORM - 1)
2211 0 : return 0;
2212 450 : if (forms == cur || strcmp(word, *(cur - 1)) != 0)
2213 : {
2214 450 : *cur = pstrdup(word);
2215 450 : *(cur + 1) = NULL;
2216 450 : return 1;
2217 : }
2218 :
2219 0 : return 0;
2220 : }
2221 :
2222 : static char **
2223 1255 : NormalizeSubWord(IspellDict *Conf, const char *word, int flag)
2224 : {
2225 1255 : AffixNodeData *suffix = NULL,
2226 1255 : *prefix = NULL;
2227 1255 : int slevel = 0,
2228 1255 : plevel = 0;
2229 1255 : int wrdlen = strlen(word),
2230 : swrdlen;
2231 : char **forms;
2232 : char **cur;
2233 1255 : char newword[2 * MAXNORMLEN] = "";
2234 1255 : char pnewword[2 * MAXNORMLEN] = "";
2235 1255 : AffixNode *snode = Conf->Suffix,
2236 : *pnode;
2237 : int i,
2238 : j;
2239 :
2240 1255 : if (wrdlen > MAXNORMLEN)
2241 0 : return NULL;
2242 1255 : cur = forms = palloc_array(char *, MAX_NORM);
2243 1255 : *cur = NULL;
2244 :
2245 :
2246 : /* Check that the word itself is normal form */
2247 1255 : if (FindWord(Conf, word, VoidString, flag))
2248 : {
2249 390 : *cur = pstrdup(word);
2250 390 : cur++;
2251 390 : *cur = NULL;
2252 : }
2253 :
2254 : /* Find all other NORMAL forms of the 'word' (check only prefix) */
2255 1255 : pnode = Conf->Prefix;
2256 1255 : plevel = 0;
2257 1435 : while (pnode)
2258 : {
2259 1255 : prefix = FindAffixes(pnode, word, wrdlen, &plevel, FF_PREFIX);
2260 1255 : if (!prefix)
2261 1075 : break;
2262 360 : for (j = 0; j < prefix->naff; j++)
2263 : {
2264 180 : if (CheckAffix(word, wrdlen, prefix->aff[j], flag, newword, NULL))
2265 : {
2266 : /* prefix success */
2267 160 : if (FindWord(Conf, newword, prefix->aff[j]->flag, flag))
2268 40 : cur += addToResult(forms, cur, newword);
2269 : }
2270 : }
2271 180 : pnode = prefix->node;
2272 : }
2273 :
2274 : /*
2275 : * Find all other NORMAL forms of the 'word' (check suffix and then
2276 : * prefix)
2277 : */
2278 2165 : while (snode)
2279 : {
2280 1755 : int baselen = 0;
2281 :
2282 : /* find possible suffix */
2283 1755 : suffix = FindAffixes(snode, word, wrdlen, &slevel, FF_SUFFIX);
2284 1755 : if (!suffix)
2285 845 : break;
2286 : /* foreach suffix check affix */
2287 1980 : for (i = 0; i < suffix->naff; i++)
2288 : {
2289 1070 : if (CheckAffix(word, wrdlen, suffix->aff[i], flag, newword, &baselen))
2290 : {
2291 : /* suffix success */
2292 840 : if (FindWord(Conf, newword, suffix->aff[i]->flag, flag))
2293 230 : cur += addToResult(forms, cur, newword);
2294 :
2295 : /* now we will look changed word with prefixes */
2296 840 : pnode = Conf->Prefix;
2297 840 : plevel = 0;
2298 840 : swrdlen = strlen(newword);
2299 1120 : while (pnode)
2300 : {
2301 840 : prefix = FindAffixes(pnode, newword, swrdlen, &plevel, FF_PREFIX);
2302 840 : if (!prefix)
2303 560 : break;
2304 560 : for (j = 0; j < prefix->naff; j++)
2305 : {
2306 280 : if (CheckAffix(newword, swrdlen, prefix->aff[j], flag, pnewword, &baselen))
2307 : {
2308 : /* prefix success */
2309 480 : const char *ff = (prefix->aff[j]->flagflags & suffix->aff[i]->flagflags & FF_CROSSPRODUCT) ?
2310 240 : VoidString : prefix->aff[j]->flag;
2311 :
2312 240 : if (FindWord(Conf, pnewword, ff, flag))
2313 180 : cur += addToResult(forms, cur, pnewword);
2314 : }
2315 : }
2316 280 : pnode = prefix->node;
2317 : }
2318 : }
2319 : }
2320 :
2321 910 : snode = suffix->node;
2322 : }
2323 :
2324 1255 : if (cur == forms)
2325 : {
2326 555 : pfree(forms);
2327 555 : return NULL;
2328 : }
2329 700 : return forms;
2330 : }
2331 :
2332 : typedef struct SplitVar
2333 : {
2334 : int nstem;
2335 : int lenstem;
2336 : char **stem;
2337 : struct SplitVar *next;
2338 : } SplitVar;
2339 :
2340 : static int
2341 5050 : CheckCompoundAffixes(CMPDAffix **ptr, const char *word, int len, bool CheckInPlace)
2342 : {
2343 : bool issuffix;
2344 :
2345 : /* in case CompoundAffix is null: */
2346 5050 : if (*ptr == NULL)
2347 0 : return -1;
2348 :
2349 5050 : if (CheckInPlace)
2350 : {
2351 9640 : while ((*ptr)->affix)
2352 : {
2353 5370 : if (len > (*ptr)->len && strncmp((*ptr)->affix, word, (*ptr)->len) == 0)
2354 : {
2355 50 : len = (*ptr)->len;
2356 50 : issuffix = (*ptr)->issuffix;
2357 50 : (*ptr)++;
2358 50 : return (issuffix) ? len : 0;
2359 : }
2360 5320 : (*ptr)++;
2361 : }
2362 : }
2363 : else
2364 : {
2365 : const char *affbegin;
2366 :
2367 1410 : while ((*ptr)->affix)
2368 : {
2369 785 : if (len > (*ptr)->len && (affbegin = strstr(word, (*ptr)->affix)) != NULL)
2370 : {
2371 105 : len = (*ptr)->len + (affbegin - word);
2372 105 : issuffix = (*ptr)->issuffix;
2373 105 : (*ptr)++;
2374 105 : return (issuffix) ? len : 0;
2375 : }
2376 680 : (*ptr)++;
2377 : }
2378 : }
2379 4895 : return -1;
2380 : }
2381 :
2382 : static SplitVar *
2383 1175 : CopyVar(SplitVar *s, int makedup)
2384 : {
2385 1175 : SplitVar *v = palloc_object(SplitVar);
2386 :
2387 1175 : v->next = NULL;
2388 1175 : if (s)
2389 : {
2390 : int i;
2391 :
2392 550 : v->lenstem = s->lenstem;
2393 550 : v->stem = palloc_array(char *, v->lenstem);
2394 550 : v->nstem = s->nstem;
2395 835 : for (i = 0; i < s->nstem; i++)
2396 285 : v->stem[i] = (makedup) ? pstrdup(s->stem[i]) : s->stem[i];
2397 : }
2398 : else
2399 : {
2400 625 : v->lenstem = 16;
2401 625 : v->stem = palloc_array(char *, v->lenstem);
2402 625 : v->nstem = 0;
2403 : }
2404 1175 : return v;
2405 : }
2406 :
2407 : static void
2408 1575 : AddStem(SplitVar *v, char *word)
2409 : {
2410 1575 : if (v->nstem >= v->lenstem)
2411 : {
2412 0 : v->lenstem *= 2;
2413 0 : v->stem = (char **) repalloc(v->stem, sizeof(char *) * v->lenstem);
2414 : }
2415 :
2416 1575 : v->stem[v->nstem] = word;
2417 1575 : v->nstem++;
2418 1575 : }
2419 :
2420 : static SplitVar *
2421 1100 : SplitToVariants(IspellDict *Conf, SPNode *snode, SplitVar *orig, const char *word, int wordlen, int startpos, int minpos)
2422 : {
2423 1100 : SplitVar *var = NULL;
2424 : SPNodeData *StopLow,
2425 : *StopHigh,
2426 1100 : *StopMiddle = NULL;
2427 1100 : SPNode *node = (snode) ? snode : Conf->Dictionary;
2428 1100 : int level = (snode) ? minpos : startpos; /* recursive
2429 : * minpos==level */
2430 : int lenaff;
2431 : CMPDAffix *caff;
2432 : char *notprobed;
2433 1100 : int compoundflag = 0;
2434 :
2435 : /* since this function recurses, it could be driven to stack overflow */
2436 1100 : check_stack_depth();
2437 :
2438 1100 : notprobed = (char *) palloc(wordlen);
2439 1100 : memset(notprobed, 1, wordlen);
2440 1100 : var = CopyVar(orig, 1);
2441 :
2442 6210 : while (level < wordlen)
2443 : {
2444 : /* find word with epenthetic or/and compound affix */
2445 5995 : caff = Conf->CompoundAffix;
2446 6150 : while (level > startpos && (lenaff = CheckCompoundAffixes(&caff, word + level, wordlen - level, (node) ? true : false)) >= 0)
2447 : {
2448 : /*
2449 : * there is one of compound affixes, so check word for existings
2450 : */
2451 : char buf[MAXNORMLEN];
2452 : char **subres;
2453 :
2454 155 : lenaff = level - startpos + lenaff;
2455 :
2456 155 : if (!notprobed[startpos + lenaff - 1])
2457 0 : continue;
2458 :
2459 155 : if (level + lenaff - 1 <= minpos)
2460 0 : continue;
2461 :
2462 155 : if (lenaff >= MAXNORMLEN)
2463 0 : continue; /* skip too big value */
2464 155 : if (lenaff > 0)
2465 155 : memcpy(buf, word + startpos, lenaff);
2466 155 : buf[lenaff] = '\0';
2467 :
2468 155 : if (level == 0)
2469 0 : compoundflag = FF_COMPOUNDBEGIN;
2470 155 : else if (level == wordlen - 1)
2471 0 : compoundflag = FF_COMPOUNDLAST;
2472 : else
2473 155 : compoundflag = FF_COMPOUNDMIDDLE;
2474 155 : subres = NormalizeSubWord(Conf, buf, compoundflag);
2475 155 : if (subres)
2476 : {
2477 : /* Yes, it was a word from dictionary */
2478 75 : SplitVar *new = CopyVar(var, 0);
2479 75 : SplitVar *ptr = var;
2480 75 : char **sptr = subres;
2481 :
2482 75 : notprobed[startpos + lenaff - 1] = 0;
2483 :
2484 150 : while (*sptr)
2485 : {
2486 75 : AddStem(new, *sptr);
2487 75 : sptr++;
2488 : }
2489 75 : pfree(subres);
2490 :
2491 75 : while (ptr->next)
2492 0 : ptr = ptr->next;
2493 75 : ptr->next = SplitToVariants(Conf, NULL, new, word, wordlen, startpos + lenaff, startpos + lenaff);
2494 :
2495 75 : pfree(new->stem);
2496 75 : pfree(new);
2497 : }
2498 : }
2499 :
2500 5995 : if (!node)
2501 625 : break;
2502 :
2503 5370 : StopLow = node->data;
2504 5370 : StopHigh = node->data + node->length;
2505 7245 : while (StopLow < StopHigh)
2506 : {
2507 6720 : StopMiddle = StopLow + ((StopHigh - StopLow) >> 1);
2508 6720 : if (StopMiddle->val == ((const uint8 *) (word))[level])
2509 4845 : break;
2510 1875 : else if (StopMiddle->val < ((const uint8 *) (word))[level])
2511 815 : StopLow = StopMiddle + 1;
2512 : else
2513 1060 : StopHigh = StopMiddle;
2514 : }
2515 :
2516 5370 : if (StopLow < StopHigh)
2517 : {
2518 4845 : if (startpos == 0)
2519 2725 : compoundflag = FF_COMPOUNDBEGIN;
2520 2120 : else if (level == wordlen - 1)
2521 240 : compoundflag = FF_COMPOUNDLAST;
2522 : else
2523 1880 : compoundflag = FF_COMPOUNDMIDDLE;
2524 :
2525 : /* find infinitive */
2526 4845 : if (StopMiddle->isword &&
2527 1280 : (StopMiddle->compoundflag & compoundflag) &&
2528 1060 : notprobed[level])
2529 : {
2530 : /* ok, we found full compoundallowed word */
2531 1060 : if (level > minpos)
2532 : {
2533 : /* and its length more than minimal */
2534 660 : if (wordlen == level + 1)
2535 : {
2536 : /* well, it was last word */
2537 260 : AddStem(var, pnstrdup(word + startpos, wordlen - startpos));
2538 260 : pfree(notprobed);
2539 260 : return var;
2540 : }
2541 : else
2542 400 : {
2543 : /* then we will search more big word at the same point */
2544 400 : SplitVar *ptr = var;
2545 :
2546 620 : while (ptr->next)
2547 220 : ptr = ptr->next;
2548 400 : ptr->next = SplitToVariants(Conf, node, var, word, wordlen, startpos, level);
2549 : /* we can find next word */
2550 400 : level++;
2551 400 : AddStem(var, pnstrdup(word + startpos, level - startpos));
2552 400 : node = Conf->Dictionary;
2553 400 : startpos = level;
2554 400 : continue;
2555 : }
2556 : }
2557 : }
2558 4185 : node = StopMiddle->node;
2559 : }
2560 : else
2561 525 : node = NULL;
2562 4710 : level++;
2563 : }
2564 :
2565 840 : AddStem(var, pnstrdup(word + startpos, wordlen - startpos));
2566 840 : pfree(notprobed);
2567 840 : return var;
2568 : }
2569 :
2570 : static void
2571 1095 : addNorm(TSLexeme **lres, TSLexeme **lcur, char *word, int flags, uint16 NVariant)
2572 : {
2573 1095 : if (*lres == NULL)
2574 505 : *lcur = *lres = palloc_array(TSLexeme, MAX_NORM);
2575 :
2576 1095 : if (*lcur - *lres < MAX_NORM - 1)
2577 : {
2578 1095 : (*lcur)->lexeme = word;
2579 1095 : (*lcur)->flags = flags;
2580 1095 : (*lcur)->nvariant = NVariant;
2581 1095 : (*lcur)++;
2582 1095 : (*lcur)->lexeme = NULL;
2583 : }
2584 1095 : }
2585 :
2586 : TSLexeme *
2587 625 : NINormalizeWord(IspellDict *Conf, const char *word)
2588 : {
2589 : char **res;
2590 625 : TSLexeme *lcur = NULL,
2591 625 : *lres = NULL;
2592 625 : uint16 NVariant = 1;
2593 :
2594 625 : res = NormalizeSubWord(Conf, word, 0);
2595 :
2596 625 : if (res)
2597 : {
2598 405 : char **ptr = res;
2599 :
2600 950 : while (*ptr && (lcur - lres) < MAX_NORM)
2601 : {
2602 545 : addNorm(&lres, &lcur, *ptr, 0, NVariant++);
2603 545 : ptr++;
2604 : }
2605 405 : pfree(res);
2606 : }
2607 :
2608 625 : if (Conf->usecompound)
2609 : {
2610 625 : int wordlen = strlen(word);
2611 : SplitVar *ptr,
2612 625 : *var = SplitToVariants(Conf, NULL, NULL, word, wordlen, 0, -1);
2613 : int i;
2614 :
2615 1725 : while (var)
2616 : {
2617 1100 : if (var->nstem > 1)
2618 : {
2619 475 : char **subres = NormalizeSubWord(Conf, var->stem[var->nstem - 1], FF_COMPOUNDLAST);
2620 :
2621 475 : if (subres)
2622 : {
2623 220 : char **subptr = subres;
2624 :
2625 440 : while (*subptr)
2626 : {
2627 550 : for (i = 0; i < var->nstem - 1; i++)
2628 : {
2629 330 : addNorm(&lres, &lcur, (subptr == subres) ? var->stem[i] : pstrdup(var->stem[i]), 0, NVariant);
2630 : }
2631 :
2632 220 : addNorm(&lres, &lcur, *subptr, 0, NVariant);
2633 220 : subptr++;
2634 220 : NVariant++;
2635 : }
2636 :
2637 220 : pfree(subres);
2638 220 : var->stem[0] = NULL;
2639 220 : pfree(var->stem[var->nstem - 1]);
2640 : }
2641 : }
2642 :
2643 2285 : for (i = 0; i < var->nstem && var->stem[i]; i++)
2644 1185 : pfree(var->stem[i]);
2645 1100 : ptr = var->next;
2646 1100 : pfree(var->stem);
2647 1100 : pfree(var);
2648 1100 : var = ptr;
2649 : }
2650 : }
2651 :
2652 625 : return lres;
2653 : }
|
