1 | /*------------------------------------------------------------------------- |
2 | * |
3 | * ts_parse.c |
4 | * main parse functions for tsearch |
5 | * |
6 | * Portions Copyright (c) 1996-2019, PostgreSQL Global Development Group |
7 | * |
8 | * |
9 | * IDENTIFICATION |
10 | * src/backend/tsearch/ts_parse.c |
11 | * |
12 | *------------------------------------------------------------------------- |
13 | */ |
14 | |
15 | #include "postgres.h" |
16 | |
17 | #include "tsearch/ts_cache.h" |
18 | #include "tsearch/ts_utils.h" |
19 | |
20 | #define IGNORE_LONGLEXEME 1 |
21 | |
22 | /* |
23 | * Lexize subsystem |
24 | */ |
25 | |
26 | typedef struct ParsedLex |
27 | { |
28 | int type; |
29 | char *lemm; |
30 | int lenlemm; |
31 | struct ParsedLex *next; |
32 | } ParsedLex; |
33 | |
34 | typedef struct ListParsedLex |
35 | { |
36 | ParsedLex *head; |
37 | ParsedLex *tail; |
38 | } ListParsedLex; |
39 | |
40 | typedef struct |
41 | { |
42 | TSConfigCacheEntry *cfg; |
43 | Oid curDictId; |
44 | int posDict; |
45 | DictSubState dictState; |
46 | ParsedLex *curSub; |
47 | ListParsedLex towork; /* current list to work */ |
48 | ListParsedLex waste; /* list of lexemes that already lexized */ |
49 | |
50 | /* |
51 | * fields to store last variant to lexize (basically, thesaurus or similar |
52 | * to, which wants several lexemes |
53 | */ |
54 | |
55 | ParsedLex *lastRes; |
56 | TSLexeme *tmpRes; |
57 | } LexizeData; |
58 | |
59 | static void |
60 | LexizeInit(LexizeData *ld, TSConfigCacheEntry *cfg) |
61 | { |
62 | ld->cfg = cfg; |
63 | ld->curDictId = InvalidOid; |
64 | ld->posDict = 0; |
65 | ld->towork.head = ld->towork.tail = ld->curSub = NULL; |
66 | ld->waste.head = ld->waste.tail = NULL; |
67 | ld->lastRes = NULL; |
68 | ld->tmpRes = NULL; |
69 | } |
70 | |
71 | static void |
72 | LPLAddTail(ListParsedLex *list, ParsedLex *newpl) |
73 | { |
74 | if (list->tail) |
75 | { |
76 | list->tail->next = newpl; |
77 | list->tail = newpl; |
78 | } |
79 | else |
80 | list->head = list->tail = newpl; |
81 | newpl->next = NULL; |
82 | } |
83 | |
84 | static ParsedLex * |
85 | LPLRemoveHead(ListParsedLex *list) |
86 | { |
87 | ParsedLex *res = list->head; |
88 | |
89 | if (list->head) |
90 | list->head = list->head->next; |
91 | |
92 | if (list->head == NULL) |
93 | list->tail = NULL; |
94 | |
95 | return res; |
96 | } |
97 | |
98 | static void |
99 | LexizeAddLemm(LexizeData *ld, int type, char *lemm, int lenlemm) |
100 | { |
101 | ParsedLex *newpl = (ParsedLex *) palloc(sizeof(ParsedLex)); |
102 | |
103 | newpl->type = type; |
104 | newpl->lemm = lemm; |
105 | newpl->lenlemm = lenlemm; |
106 | LPLAddTail(&ld->towork, newpl); |
107 | ld->curSub = ld->towork.tail; |
108 | } |
109 | |
110 | static void |
111 | RemoveHead(LexizeData *ld) |
112 | { |
113 | LPLAddTail(&ld->waste, LPLRemoveHead(&ld->towork)); |
114 | |
115 | ld->posDict = 0; |
116 | } |
117 | |
118 | static void |
119 | setCorrLex(LexizeData *ld, ParsedLex **correspondLexem) |
120 | { |
121 | if (correspondLexem) |
122 | { |
123 | *correspondLexem = ld->waste.head; |
124 | } |
125 | else |
126 | { |
127 | ParsedLex *tmp, |
128 | *ptr = ld->waste.head; |
129 | |
130 | while (ptr) |
131 | { |
132 | tmp = ptr->next; |
133 | pfree(ptr); |
134 | ptr = tmp; |
135 | } |
136 | } |
137 | ld->waste.head = ld->waste.tail = NULL; |
138 | } |
139 | |
140 | static void |
141 | moveToWaste(LexizeData *ld, ParsedLex *stop) |
142 | { |
143 | bool go = true; |
144 | |
145 | while (ld->towork.head && go) |
146 | { |
147 | if (ld->towork.head == stop) |
148 | { |
149 | ld->curSub = stop->next; |
150 | go = false; |
151 | } |
152 | RemoveHead(ld); |
153 | } |
154 | } |
155 | |
156 | static void |
157 | setNewTmpRes(LexizeData *ld, ParsedLex *lex, TSLexeme *res) |
158 | { |
159 | if (ld->tmpRes) |
160 | { |
161 | TSLexeme *ptr; |
162 | |
163 | for (ptr = ld->tmpRes; ptr->lexeme; ptr++) |
164 | pfree(ptr->lexeme); |
165 | pfree(ld->tmpRes); |
166 | } |
167 | ld->tmpRes = res; |
168 | ld->lastRes = lex; |
169 | } |
170 | |
171 | static TSLexeme * |
172 | LexizeExec(LexizeData *ld, ParsedLex **correspondLexem) |
173 | { |
174 | int i; |
175 | ListDictionary *map; |
176 | TSDictionaryCacheEntry *dict; |
177 | TSLexeme *res; |
178 | |
179 | if (ld->curDictId == InvalidOid) |
180 | { |
181 | /* |
182 | * usual mode: dictionary wants only one word, but we should keep in |
183 | * mind that we should go through all stack |
184 | */ |
185 | |
186 | while (ld->towork.head) |
187 | { |
188 | ParsedLex *curVal = ld->towork.head; |
189 | char *curValLemm = curVal->lemm; |
190 | int curValLenLemm = curVal->lenlemm; |
191 | |
192 | map = ld->cfg->map + curVal->type; |
193 | |
194 | if (curVal->type == 0 || curVal->type >= ld->cfg->lenmap || map->len == 0) |
195 | { |
196 | /* skip this type of lexeme */ |
197 | RemoveHead(ld); |
198 | continue; |
199 | } |
200 | |
201 | for (i = ld->posDict; i < map->len; i++) |
202 | { |
203 | dict = lookup_ts_dictionary_cache(map->dictIds[i]); |
204 | |
205 | ld->dictState.isend = ld->dictState.getnext = false; |
206 | ld->dictState.private_state = NULL; |
207 | res = (TSLexeme *) DatumGetPointer(FunctionCall4( |
208 | &(dict->lexize), |
209 | PointerGetDatum(dict->dictData), |
210 | PointerGetDatum(curValLemm), |
211 | Int32GetDatum(curValLenLemm), |
212 | PointerGetDatum(&ld->dictState) |
213 | )); |
214 | |
215 | if (ld->dictState.getnext) |
216 | { |
217 | /* |
218 | * dictionary wants next word, so setup and store current |
219 | * position and go to multiword mode |
220 | */ |
221 | |
222 | ld->curDictId = DatumGetObjectId(map->dictIds[i]); |
223 | ld->posDict = i + 1; |
224 | ld->curSub = curVal->next; |
225 | if (res) |
226 | setNewTmpRes(ld, curVal, res); |
227 | return LexizeExec(ld, correspondLexem); |
228 | } |
229 | |
230 | if (!res) /* dictionary doesn't know this lexeme */ |
231 | continue; |
232 | |
233 | if (res->flags & TSL_FILTER) |
234 | { |
235 | curValLemm = res->lexeme; |
236 | curValLenLemm = strlen(res->lexeme); |
237 | continue; |
238 | } |
239 | |
240 | RemoveHead(ld); |
241 | setCorrLex(ld, correspondLexem); |
242 | return res; |
243 | } |
244 | |
245 | RemoveHead(ld); |
246 | } |
247 | } |
248 | else |
249 | { /* curDictId is valid */ |
250 | dict = lookup_ts_dictionary_cache(ld->curDictId); |
251 | |
252 | /* |
253 | * Dictionary ld->curDictId asks us about following words |
254 | */ |
255 | |
256 | while (ld->curSub) |
257 | { |
258 | ParsedLex *curVal = ld->curSub; |
259 | |
260 | map = ld->cfg->map + curVal->type; |
261 | |
262 | if (curVal->type != 0) |
263 | { |
264 | bool dictExists = false; |
265 | |
266 | if (curVal->type >= ld->cfg->lenmap || map->len == 0) |
267 | { |
268 | /* skip this type of lexeme */ |
269 | ld->curSub = curVal->next; |
270 | continue; |
271 | } |
272 | |
273 | /* |
274 | * We should be sure that current type of lexeme is recognized |
275 | * by our dictionary: we just check is it exist in list of |
276 | * dictionaries ? |
277 | */ |
278 | for (i = 0; i < map->len && !dictExists; i++) |
279 | if (ld->curDictId == DatumGetObjectId(map->dictIds[i])) |
280 | dictExists = true; |
281 | |
282 | if (!dictExists) |
283 | { |
284 | /* |
285 | * Dictionary can't work with current tpe of lexeme, |
286 | * return to basic mode and redo all stored lexemes |
287 | */ |
288 | ld->curDictId = InvalidOid; |
289 | return LexizeExec(ld, correspondLexem); |
290 | } |
291 | } |
292 | |
293 | ld->dictState.isend = (curVal->type == 0) ? true : false; |
294 | ld->dictState.getnext = false; |
295 | |
296 | res = (TSLexeme *) DatumGetPointer(FunctionCall4( |
297 | &(dict->lexize), |
298 | PointerGetDatum(dict->dictData), |
299 | PointerGetDatum(curVal->lemm), |
300 | Int32GetDatum(curVal->lenlemm), |
301 | PointerGetDatum(&ld->dictState) |
302 | )); |
303 | |
304 | if (ld->dictState.getnext) |
305 | { |
306 | /* Dictionary wants one more */ |
307 | ld->curSub = curVal->next; |
308 | if (res) |
309 | setNewTmpRes(ld, curVal, res); |
310 | continue; |
311 | } |
312 | |
313 | if (res || ld->tmpRes) |
314 | { |
315 | /* |
316 | * Dictionary normalizes lexemes, so we remove from stack all |
317 | * used lexemes, return to basic mode and redo end of stack |
318 | * (if it exists) |
319 | */ |
320 | if (res) |
321 | { |
322 | moveToWaste(ld, ld->curSub); |
323 | } |
324 | else |
325 | { |
326 | res = ld->tmpRes; |
327 | moveToWaste(ld, ld->lastRes); |
328 | } |
329 | |
330 | /* reset to initial state */ |
331 | ld->curDictId = InvalidOid; |
332 | ld->posDict = 0; |
333 | ld->lastRes = NULL; |
334 | ld->tmpRes = NULL; |
335 | setCorrLex(ld, correspondLexem); |
336 | return res; |
337 | } |
338 | |
339 | /* |
340 | * Dict don't want next lexem and didn't recognize anything, redo |
341 | * from ld->towork.head |
342 | */ |
343 | ld->curDictId = InvalidOid; |
344 | return LexizeExec(ld, correspondLexem); |
345 | } |
346 | } |
347 | |
348 | setCorrLex(ld, correspondLexem); |
349 | return NULL; |
350 | } |
351 | |
352 | /* |
353 | * Parse string and lexize words. |
354 | * |
355 | * prs will be filled in. |
356 | */ |
357 | void |
358 | parsetext(Oid cfgId, ParsedText *prs, char *buf, int buflen) |
359 | { |
360 | int type, |
361 | lenlemm; |
362 | char *lemm = NULL; |
363 | LexizeData ldata; |
364 | TSLexeme *norms; |
365 | TSConfigCacheEntry *cfg; |
366 | TSParserCacheEntry *prsobj; |
367 | void *prsdata; |
368 | |
369 | cfg = lookup_ts_config_cache(cfgId); |
370 | prsobj = lookup_ts_parser_cache(cfg->prsId); |
371 | |
372 | prsdata = (void *) DatumGetPointer(FunctionCall2(&prsobj->prsstart, |
373 | PointerGetDatum(buf), |
374 | Int32GetDatum(buflen))); |
375 | |
376 | LexizeInit(&ldata, cfg); |
377 | |
378 | do |
379 | { |
380 | type = DatumGetInt32(FunctionCall3(&(prsobj->prstoken), |
381 | PointerGetDatum(prsdata), |
382 | PointerGetDatum(&lemm), |
383 | PointerGetDatum(&lenlemm))); |
384 | |
385 | if (type > 0 && lenlemm >= MAXSTRLEN) |
386 | { |
387 | #ifdef IGNORE_LONGLEXEME |
388 | ereport(NOTICE, |
389 | (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED), |
390 | errmsg("word is too long to be indexed" ), |
391 | errdetail("Words longer than %d characters are ignored." , |
392 | MAXSTRLEN))); |
393 | continue; |
394 | #else |
395 | ereport(ERROR, |
396 | (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED), |
397 | errmsg("word is too long to be indexed" ), |
398 | errdetail("Words longer than %d characters are ignored." , |
399 | MAXSTRLEN))); |
400 | #endif |
401 | } |
402 | |
403 | LexizeAddLemm(&ldata, type, lemm, lenlemm); |
404 | |
405 | while ((norms = LexizeExec(&ldata, NULL)) != NULL) |
406 | { |
407 | TSLexeme *ptr = norms; |
408 | |
409 | prs->pos++; /* set pos */ |
410 | |
411 | while (ptr->lexeme) |
412 | { |
413 | if (prs->curwords == prs->lenwords) |
414 | { |
415 | prs->lenwords *= 2; |
416 | prs->words = (ParsedWord *) repalloc((void *) prs->words, prs->lenwords * sizeof(ParsedWord)); |
417 | } |
418 | |
419 | if (ptr->flags & TSL_ADDPOS) |
420 | prs->pos++; |
421 | prs->words[prs->curwords].len = strlen(ptr->lexeme); |
422 | prs->words[prs->curwords].word = ptr->lexeme; |
423 | prs->words[prs->curwords].nvariant = ptr->nvariant; |
424 | prs->words[prs->curwords].flags = ptr->flags & TSL_PREFIX; |
425 | prs->words[prs->curwords].alen = 0; |
426 | prs->words[prs->curwords].pos.pos = LIMITPOS(prs->pos); |
427 | ptr++; |
428 | prs->curwords++; |
429 | } |
430 | pfree(norms); |
431 | } |
432 | } while (type > 0); |
433 | |
434 | FunctionCall1(&(prsobj->prsend), PointerGetDatum(prsdata)); |
435 | } |
436 | |
437 | /* |
438 | * Headline framework |
439 | */ |
440 | static void |
441 | hladdword(HeadlineParsedText *prs, char *buf, int buflen, int type) |
442 | { |
443 | while (prs->curwords >= prs->lenwords) |
444 | { |
445 | prs->lenwords *= 2; |
446 | prs->words = (HeadlineWordEntry *) repalloc((void *) prs->words, prs->lenwords * sizeof(HeadlineWordEntry)); |
447 | } |
448 | memset(&(prs->words[prs->curwords]), 0, sizeof(HeadlineWordEntry)); |
449 | prs->words[prs->curwords].type = (uint8) type; |
450 | prs->words[prs->curwords].len = buflen; |
451 | prs->words[prs->curwords].word = palloc(buflen); |
452 | memcpy(prs->words[prs->curwords].word, buf, buflen); |
453 | prs->curwords++; |
454 | } |
455 | |
456 | static void |
457 | hlfinditem(HeadlineParsedText *prs, TSQuery query, int32 pos, char *buf, int buflen) |
458 | { |
459 | int i; |
460 | QueryItem *item = GETQUERY(query); |
461 | HeadlineWordEntry *word; |
462 | |
463 | while (prs->curwords + query->size >= prs->lenwords) |
464 | { |
465 | prs->lenwords *= 2; |
466 | prs->words = (HeadlineWordEntry *) repalloc((void *) prs->words, prs->lenwords * sizeof(HeadlineWordEntry)); |
467 | } |
468 | |
469 | word = &(prs->words[prs->curwords - 1]); |
470 | word->pos = LIMITPOS(pos); |
471 | for (i = 0; i < query->size; i++) |
472 | { |
473 | if (item->type == QI_VAL && |
474 | tsCompareString(GETOPERAND(query) + item->qoperand.distance, item->qoperand.length, |
475 | buf, buflen, item->qoperand.prefix) == 0) |
476 | { |
477 | if (word->item) |
478 | { |
479 | memcpy(&(prs->words[prs->curwords]), word, sizeof(HeadlineWordEntry)); |
480 | prs->words[prs->curwords].item = &item->qoperand; |
481 | prs->words[prs->curwords].repeated = 1; |
482 | prs->curwords++; |
483 | } |
484 | else |
485 | word->item = &item->qoperand; |
486 | } |
487 | item++; |
488 | } |
489 | } |
490 | |
491 | static void |
492 | addHLParsedLex(HeadlineParsedText *prs, TSQuery query, ParsedLex *lexs, TSLexeme *norms) |
493 | { |
494 | ParsedLex *tmplexs; |
495 | TSLexeme *ptr; |
496 | int32 savedpos; |
497 | |
498 | while (lexs) |
499 | { |
500 | if (lexs->type > 0) |
501 | hladdword(prs, lexs->lemm, lexs->lenlemm, lexs->type); |
502 | |
503 | ptr = norms; |
504 | savedpos = prs->vectorpos; |
505 | while (ptr && ptr->lexeme) |
506 | { |
507 | if (ptr->flags & TSL_ADDPOS) |
508 | savedpos++; |
509 | hlfinditem(prs, query, savedpos, ptr->lexeme, strlen(ptr->lexeme)); |
510 | ptr++; |
511 | } |
512 | |
513 | tmplexs = lexs->next; |
514 | pfree(lexs); |
515 | lexs = tmplexs; |
516 | } |
517 | |
518 | if (norms) |
519 | { |
520 | ptr = norms; |
521 | while (ptr->lexeme) |
522 | { |
523 | if (ptr->flags & TSL_ADDPOS) |
524 | prs->vectorpos++; |
525 | pfree(ptr->lexeme); |
526 | ptr++; |
527 | } |
528 | pfree(norms); |
529 | } |
530 | } |
531 | |
532 | void |
533 | hlparsetext(Oid cfgId, HeadlineParsedText *prs, TSQuery query, char *buf, int buflen) |
534 | { |
535 | int type, |
536 | lenlemm; |
537 | char *lemm = NULL; |
538 | LexizeData ldata; |
539 | TSLexeme *norms; |
540 | ParsedLex *lexs; |
541 | TSConfigCacheEntry *cfg; |
542 | TSParserCacheEntry *prsobj; |
543 | void *prsdata; |
544 | |
545 | cfg = lookup_ts_config_cache(cfgId); |
546 | prsobj = lookup_ts_parser_cache(cfg->prsId); |
547 | |
548 | prsdata = (void *) DatumGetPointer(FunctionCall2(&(prsobj->prsstart), |
549 | PointerGetDatum(buf), |
550 | Int32GetDatum(buflen))); |
551 | |
552 | LexizeInit(&ldata, cfg); |
553 | |
554 | do |
555 | { |
556 | type = DatumGetInt32(FunctionCall3(&(prsobj->prstoken), |
557 | PointerGetDatum(prsdata), |
558 | PointerGetDatum(&lemm), |
559 | PointerGetDatum(&lenlemm))); |
560 | |
561 | if (type > 0 && lenlemm >= MAXSTRLEN) |
562 | { |
563 | #ifdef IGNORE_LONGLEXEME |
564 | ereport(NOTICE, |
565 | (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED), |
566 | errmsg("word is too long to be indexed" ), |
567 | errdetail("Words longer than %d characters are ignored." , |
568 | MAXSTRLEN))); |
569 | continue; |
570 | #else |
571 | ereport(ERROR, |
572 | (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED), |
573 | errmsg("word is too long to be indexed" ), |
574 | errdetail("Words longer than %d characters are ignored." , |
575 | MAXSTRLEN))); |
576 | #endif |
577 | } |
578 | |
579 | LexizeAddLemm(&ldata, type, lemm, lenlemm); |
580 | |
581 | do |
582 | { |
583 | if ((norms = LexizeExec(&ldata, &lexs)) != NULL) |
584 | { |
585 | prs->vectorpos++; |
586 | addHLParsedLex(prs, query, lexs, norms); |
587 | } |
588 | else |
589 | addHLParsedLex(prs, query, lexs, NULL); |
590 | } while (norms); |
591 | |
592 | } while (type > 0); |
593 | |
594 | FunctionCall1(&(prsobj->prsend), PointerGetDatum(prsdata)); |
595 | } |
596 | |
597 | text * |
598 | generateHeadline(HeadlineParsedText *prs) |
599 | { |
600 | text *out; |
601 | char *ptr; |
602 | int len = 128; |
603 | int numfragments = 0; |
604 | int16 infrag = 0; |
605 | |
606 | HeadlineWordEntry *wrd = prs->words; |
607 | |
608 | out = (text *) palloc(len); |
609 | ptr = ((char *) out) + VARHDRSZ; |
610 | |
611 | while (wrd - prs->words < prs->curwords) |
612 | { |
613 | while (wrd->len + prs->stopsellen + prs->startsellen + prs->fragdelimlen + (ptr - ((char *) out)) >= len) |
614 | { |
615 | int dist = ptr - ((char *) out); |
616 | |
617 | len *= 2; |
618 | out = (text *) repalloc(out, len); |
619 | ptr = ((char *) out) + dist; |
620 | } |
621 | |
622 | if (wrd->in && !wrd->repeated) |
623 | { |
624 | if (!infrag) |
625 | { |
626 | |
627 | /* start of a new fragment */ |
628 | infrag = 1; |
629 | numfragments++; |
630 | /* add a fragment delimiter if this is after the first one */ |
631 | if (numfragments > 1) |
632 | { |
633 | memcpy(ptr, prs->fragdelim, prs->fragdelimlen); |
634 | ptr += prs->fragdelimlen; |
635 | } |
636 | |
637 | } |
638 | if (wrd->replace) |
639 | { |
640 | *ptr = ' '; |
641 | ptr++; |
642 | } |
643 | else if (!wrd->skip) |
644 | { |
645 | if (wrd->selected) |
646 | { |
647 | memcpy(ptr, prs->startsel, prs->startsellen); |
648 | ptr += prs->startsellen; |
649 | } |
650 | memcpy(ptr, wrd->word, wrd->len); |
651 | ptr += wrd->len; |
652 | if (wrd->selected) |
653 | { |
654 | memcpy(ptr, prs->stopsel, prs->stopsellen); |
655 | ptr += prs->stopsellen; |
656 | } |
657 | } |
658 | } |
659 | else if (!wrd->repeated) |
660 | { |
661 | if (infrag) |
662 | infrag = 0; |
663 | pfree(wrd->word); |
664 | } |
665 | |
666 | wrd++; |
667 | } |
668 | |
669 | SET_VARSIZE(out, ptr - ((char *) out)); |
670 | return out; |
671 | } |
672 | |