Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * dict_snowball.c
4 : * Snowball dictionary
5 : *
6 : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
7 : *
8 : * IDENTIFICATION
9 : * src/backend/snowball/dict_snowball.c
10 : *
11 : *-------------------------------------------------------------------------
12 : */
13 : #include "postgres.h"
14 :
15 : #include "catalog/pg_collation_d.h"
16 : #include "commands/defrem.h"
17 : #include "mb/pg_wchar.h"
18 : #include "tsearch/ts_public.h"
19 : #include "utils/formatting.h"
20 :
21 : /* Some platforms define MAXINT and/or MININT, causing conflicts */
22 : #ifdef MAXINT
23 : #undef MAXINT
24 : #endif
25 : #ifdef MININT
26 : #undef MININT
27 : #endif
28 :
29 : /* Now we can include the original Snowball snowball_runtime.h */
30 : #include "snowball/libstemmer/snowball_runtime.h"
31 : #include "snowball/libstemmer/stem_ISO_8859_1_basque.h"
32 : #include "snowball/libstemmer/stem_ISO_8859_1_catalan.h"
33 : #include "snowball/libstemmer/stem_ISO_8859_1_danish.h"
34 : #include "snowball/libstemmer/stem_ISO_8859_1_dutch.h"
35 : #include "snowball/libstemmer/stem_ISO_8859_1_dutch_porter.h"
36 : #include "snowball/libstemmer/stem_ISO_8859_1_english.h"
37 : #include "snowball/libstemmer/stem_ISO_8859_1_finnish.h"
38 : #include "snowball/libstemmer/stem_ISO_8859_1_french.h"
39 : #include "snowball/libstemmer/stem_ISO_8859_1_german.h"
40 : #include "snowball/libstemmer/stem_ISO_8859_1_indonesian.h"
41 : #include "snowball/libstemmer/stem_ISO_8859_1_irish.h"
42 : #include "snowball/libstemmer/stem_ISO_8859_1_italian.h"
43 : #include "snowball/libstemmer/stem_ISO_8859_1_norwegian.h"
44 : #include "snowball/libstemmer/stem_ISO_8859_1_porter.h"
45 : #include "snowball/libstemmer/stem_ISO_8859_1_portuguese.h"
46 : #include "snowball/libstemmer/stem_ISO_8859_1_spanish.h"
47 : #include "snowball/libstemmer/stem_ISO_8859_1_swedish.h"
48 : #include "snowball/libstemmer/stem_ISO_8859_2_hungarian.h"
49 : #include "snowball/libstemmer/stem_ISO_8859_2_polish.h"
50 : #include "snowball/libstemmer/stem_KOI8_R_russian.h"
51 : #include "snowball/libstemmer/stem_UTF_8_arabic.h"
52 : #include "snowball/libstemmer/stem_UTF_8_armenian.h"
53 : #include "snowball/libstemmer/stem_UTF_8_basque.h"
54 : #include "snowball/libstemmer/stem_UTF_8_catalan.h"
55 : #include "snowball/libstemmer/stem_UTF_8_danish.h"
56 : #include "snowball/libstemmer/stem_UTF_8_dutch.h"
57 : #include "snowball/libstemmer/stem_UTF_8_dutch_porter.h"
58 : #include "snowball/libstemmer/stem_UTF_8_english.h"
59 : #include "snowball/libstemmer/stem_UTF_8_esperanto.h"
60 : #include "snowball/libstemmer/stem_UTF_8_estonian.h"
61 : #include "snowball/libstemmer/stem_UTF_8_finnish.h"
62 : #include "snowball/libstemmer/stem_UTF_8_french.h"
63 : #include "snowball/libstemmer/stem_UTF_8_german.h"
64 : #include "snowball/libstemmer/stem_UTF_8_greek.h"
65 : #include "snowball/libstemmer/stem_UTF_8_hindi.h"
66 : #include "snowball/libstemmer/stem_UTF_8_hungarian.h"
67 : #include "snowball/libstemmer/stem_UTF_8_indonesian.h"
68 : #include "snowball/libstemmer/stem_UTF_8_irish.h"
69 : #include "snowball/libstemmer/stem_UTF_8_italian.h"
70 : #include "snowball/libstemmer/stem_UTF_8_lithuanian.h"
71 : #include "snowball/libstemmer/stem_UTF_8_nepali.h"
72 : #include "snowball/libstemmer/stem_UTF_8_norwegian.h"
73 : #include "snowball/libstemmer/stem_UTF_8_polish.h"
74 : #include "snowball/libstemmer/stem_UTF_8_porter.h"
75 : #include "snowball/libstemmer/stem_UTF_8_portuguese.h"
76 : #include "snowball/libstemmer/stem_UTF_8_romanian.h"
77 : #include "snowball/libstemmer/stem_UTF_8_russian.h"
78 : #include "snowball/libstemmer/stem_UTF_8_serbian.h"
79 : #include "snowball/libstemmer/stem_UTF_8_spanish.h"
80 : #include "snowball/libstemmer/stem_UTF_8_swedish.h"
81 : #include "snowball/libstemmer/stem_UTF_8_tamil.h"
82 : #include "snowball/libstemmer/stem_UTF_8_turkish.h"
83 : #include "snowball/libstemmer/stem_UTF_8_yiddish.h"
84 :
85 70 : PG_MODULE_MAGIC_EXT(
86 : .name = "dict_snowball",
87 : .version = PG_VERSION
88 : );
89 :
90 68 : PG_FUNCTION_INFO_V1(dsnowball_init);
91 :
92 68 : PG_FUNCTION_INFO_V1(dsnowball_lexize);
93 :
94 : /* List of supported modules */
95 : typedef struct stemmer_module
96 : {
97 : const char *name;
98 : pg_enc enc;
99 : struct SN_env *(*create) (void);
100 : void (*close) (struct SN_env *);
101 : int (*stem) (struct SN_env *);
102 : } stemmer_module;
103 :
104 : /* Args: stemmer name, PG code for encoding, Snowball's name for encoding */
105 : #define STEMMER_MODULE(name,enc,senc) \
106 : {#name, enc, name##_##senc##_create_env, name##_##senc##_close_env, name##_##senc##_stem}
107 :
108 : static const stemmer_module stemmer_modules[] =
109 : {
110 : /*
111 : * Stemmers list from Snowball distribution
112 : */
113 : STEMMER_MODULE(basque, PG_LATIN1, ISO_8859_1),
114 : STEMMER_MODULE(catalan, PG_LATIN1, ISO_8859_1),
115 : STEMMER_MODULE(danish, PG_LATIN1, ISO_8859_1),
116 : STEMMER_MODULE(dutch, PG_LATIN1, ISO_8859_1),
117 : STEMMER_MODULE(dutch_porter, PG_LATIN1, ISO_8859_1),
118 : STEMMER_MODULE(english, PG_LATIN1, ISO_8859_1),
119 : STEMMER_MODULE(finnish, PG_LATIN1, ISO_8859_1),
120 : STEMMER_MODULE(french, PG_LATIN1, ISO_8859_1),
121 : STEMMER_MODULE(german, PG_LATIN1, ISO_8859_1),
122 : STEMMER_MODULE(indonesian, PG_LATIN1, ISO_8859_1),
123 : STEMMER_MODULE(irish, PG_LATIN1, ISO_8859_1),
124 : STEMMER_MODULE(italian, PG_LATIN1, ISO_8859_1),
125 : STEMMER_MODULE(norwegian, PG_LATIN1, ISO_8859_1),
126 : STEMMER_MODULE(porter, PG_LATIN1, ISO_8859_1),
127 : STEMMER_MODULE(portuguese, PG_LATIN1, ISO_8859_1),
128 : STEMMER_MODULE(spanish, PG_LATIN1, ISO_8859_1),
129 : STEMMER_MODULE(swedish, PG_LATIN1, ISO_8859_1),
130 : STEMMER_MODULE(hungarian, PG_LATIN2, ISO_8859_2),
131 : STEMMER_MODULE(polish, PG_LATIN2, ISO_8859_2),
132 : STEMMER_MODULE(russian, PG_KOI8R, KOI8_R),
133 : STEMMER_MODULE(arabic, PG_UTF8, UTF_8),
134 : STEMMER_MODULE(armenian, PG_UTF8, UTF_8),
135 : STEMMER_MODULE(basque, PG_UTF8, UTF_8),
136 : STEMMER_MODULE(catalan, PG_UTF8, UTF_8),
137 : STEMMER_MODULE(danish, PG_UTF8, UTF_8),
138 : STEMMER_MODULE(dutch, PG_UTF8, UTF_8),
139 : STEMMER_MODULE(dutch_porter, PG_UTF8, UTF_8),
140 : STEMMER_MODULE(english, PG_UTF8, UTF_8),
141 : STEMMER_MODULE(esperanto, PG_UTF8, UTF_8),
142 : STEMMER_MODULE(estonian, PG_UTF8, UTF_8),
143 : STEMMER_MODULE(finnish, PG_UTF8, UTF_8),
144 : STEMMER_MODULE(french, PG_UTF8, UTF_8),
145 : STEMMER_MODULE(german, PG_UTF8, UTF_8),
146 : STEMMER_MODULE(greek, PG_UTF8, UTF_8),
147 : STEMMER_MODULE(hindi, PG_UTF8, UTF_8),
148 : STEMMER_MODULE(hungarian, PG_UTF8, UTF_8),
149 : STEMMER_MODULE(indonesian, PG_UTF8, UTF_8),
150 : STEMMER_MODULE(irish, PG_UTF8, UTF_8),
151 : STEMMER_MODULE(italian, PG_UTF8, UTF_8),
152 : STEMMER_MODULE(lithuanian, PG_UTF8, UTF_8),
153 : STEMMER_MODULE(nepali, PG_UTF8, UTF_8),
154 : STEMMER_MODULE(norwegian, PG_UTF8, UTF_8),
155 : STEMMER_MODULE(porter, PG_UTF8, UTF_8),
156 : STEMMER_MODULE(polish, PG_UTF8, UTF_8),
157 : STEMMER_MODULE(portuguese, PG_UTF8, UTF_8),
158 : STEMMER_MODULE(romanian, PG_UTF8, UTF_8),
159 : STEMMER_MODULE(russian, PG_UTF8, UTF_8),
160 : STEMMER_MODULE(serbian, PG_UTF8, UTF_8),
161 : STEMMER_MODULE(spanish, PG_UTF8, UTF_8),
162 : STEMMER_MODULE(swedish, PG_UTF8, UTF_8),
163 : STEMMER_MODULE(tamil, PG_UTF8, UTF_8),
164 : STEMMER_MODULE(turkish, PG_UTF8, UTF_8),
165 : STEMMER_MODULE(yiddish, PG_UTF8, UTF_8),
166 :
167 : /*
168 : * Stemmer with PG_SQL_ASCII encoding should be valid for any server
169 : * encoding
170 : */
171 : STEMMER_MODULE(english, PG_SQL_ASCII, ISO_8859_1),
172 :
173 : {NULL, 0, NULL, NULL, NULL} /* list end marker */
174 : };
175 :
176 :
177 : typedef struct DictSnowball
178 : {
179 : struct SN_env *z;
180 : StopList stoplist;
181 : bool needrecode; /* needs recoding before/after call stem */
182 : int (*stem) (struct SN_env *z);
183 :
184 : /*
185 : * snowball saves alloced memory between calls, so we should run it in our
186 : * private memory context. Note, init function is executed in long lived
187 : * context, so we just remember CurrentMemoryContext
188 : */
189 : MemoryContext dictCtx;
190 : } DictSnowball;
191 :
192 :
193 : static void
194 22 : locate_stem_module(DictSnowball *d, const char *lang)
195 : {
196 : const stemmer_module *m;
197 :
198 : /*
199 : * First, try to find exact match of stemmer module. Stemmer with
200 : * PG_SQL_ASCII encoding is treated as working with any server encoding
201 : */
202 616 : for (m = stemmer_modules; m->name; m++)
203 : {
204 792 : if ((m->enc == PG_SQL_ASCII || m->enc == GetDatabaseEncoding()) &&
205 176 : pg_strcasecmp(m->name, lang) == 0)
206 : {
207 22 : d->stem = m->stem;
208 22 : d->z = m->create();
209 22 : d->needrecode = false;
210 22 : return;
211 : }
212 : }
213 :
214 : /*
215 : * Second, try to find stemmer for needed language for UTF8 encoding.
216 : */
217 0 : for (m = stemmer_modules; m->name; m++)
218 : {
219 0 : if (m->enc == PG_UTF8 && pg_strcasecmp(m->name, lang) == 0)
220 : {
221 0 : d->stem = m->stem;
222 0 : d->z = m->create();
223 0 : d->needrecode = true;
224 0 : return;
225 : }
226 : }
227 :
228 0 : ereport(ERROR,
229 : (errcode(ERRCODE_UNDEFINED_OBJECT),
230 : errmsg("no Snowball stemmer available for language \"%s\" and encoding \"%s\"",
231 : lang, GetDatabaseEncodingName())));
232 : }
233 :
234 : Datum
235 22 : dsnowball_init(PG_FUNCTION_ARGS)
236 : {
237 22 : List *dictoptions = (List *) PG_GETARG_POINTER(0);
238 : DictSnowball *d;
239 22 : bool stoploaded = false;
240 : ListCell *l;
241 :
242 22 : d = palloc0_object(DictSnowball);
243 :
244 66 : foreach(l, dictoptions)
245 : {
246 44 : DefElem *defel = (DefElem *) lfirst(l);
247 :
248 44 : if (strcmp(defel->defname, "stopwords") == 0)
249 : {
250 22 : if (stoploaded)
251 0 : ereport(ERROR,
252 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
253 : errmsg("multiple StopWords parameters")));
254 22 : readstoplist(defGetString(defel), &d->stoplist, str_tolower);
255 22 : stoploaded = true;
256 : }
257 22 : else if (strcmp(defel->defname, "language") == 0)
258 : {
259 22 : if (d->stem)
260 0 : ereport(ERROR,
261 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
262 : errmsg("multiple Language parameters")));
263 22 : locate_stem_module(d, defGetString(defel));
264 : }
265 : else
266 : {
267 0 : ereport(ERROR,
268 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
269 : errmsg("unrecognized Snowball parameter: \"%s\"",
270 : defel->defname)));
271 : }
272 : }
273 :
274 22 : if (!d->stem)
275 0 : ereport(ERROR,
276 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
277 : errmsg("missing Language parameter")));
278 :
279 22 : d->dictCtx = CurrentMemoryContext;
280 :
281 22 : PG_RETURN_POINTER(d);
282 : }
283 :
284 : Datum
285 5159 : dsnowball_lexize(PG_FUNCTION_ARGS)
286 : {
287 5159 : DictSnowball *d = (DictSnowball *) PG_GETARG_POINTER(0);
288 5159 : char *in = (char *) PG_GETARG_POINTER(1);
289 5159 : int32 len = PG_GETARG_INT32(2);
290 5159 : char *txt = str_tolower(in, len, DEFAULT_COLLATION_OID);
291 5159 : TSLexeme *res = palloc0_array(TSLexeme, 2);
292 :
293 : /*
294 : * Do not pass strings exceeding 1000 bytes to the stemmer, as they're
295 : * surely not words in any human language. This restriction avoids
296 : * wasting cycles on stuff like base64-encoded data, and it protects us
297 : * against possible inefficiency or misbehavior in the stemmer. (For
298 : * example, the Turkish stemmer has an indefinite recursion, so it can
299 : * crash on long-enough strings.) However, Snowball dictionaries are
300 : * defined to recognize all strings, so we can't reject the string as an
301 : * unknown word.
302 : */
303 5159 : if (len > 1000)
304 : {
305 : /* return the lexeme lowercased, but otherwise unmodified */
306 0 : res->lexeme = txt;
307 : }
308 5159 : else if (*txt == '\0' || searchstoplist(&(d->stoplist), txt))
309 : {
310 : /* empty or stopword, so report as stopword */
311 1734 : pfree(txt);
312 : }
313 : else
314 : {
315 : MemoryContext saveCtx;
316 :
317 : /*
318 : * recode to utf8 if stemmer is utf8 and doesn't match server encoding
319 : */
320 3425 : if (d->needrecode)
321 : {
322 : char *recoded;
323 :
324 0 : recoded = pg_server_to_any(txt, strlen(txt), PG_UTF8);
325 0 : if (recoded != txt)
326 : {
327 0 : pfree(txt);
328 0 : txt = recoded;
329 : }
330 : }
331 :
332 : /* see comment about d->dictCtx */
333 3425 : saveCtx = MemoryContextSwitchTo(d->dictCtx);
334 3425 : SN_set_current(d->z, strlen(txt), (symbol *) txt);
335 3425 : d->stem(d->z);
336 3425 : MemoryContextSwitchTo(saveCtx);
337 :
338 3425 : if (d->z->p && d->z->l)
339 : {
340 3425 : txt = repalloc(txt, d->z->l + 1);
341 3425 : memcpy(txt, d->z->p, d->z->l);
342 3425 : txt[d->z->l] = '\0';
343 : }
344 :
345 : /* back recode if needed */
346 3425 : if (d->needrecode)
347 : {
348 : char *recoded;
349 :
350 0 : recoded = pg_any_to_server(txt, strlen(txt), PG_UTF8);
351 0 : if (recoded != txt)
352 : {
353 0 : pfree(txt);
354 0 : txt = recoded;
355 : }
356 : }
357 :
358 3425 : res->lexeme = txt;
359 : }
360 :
361 5159 : PG_RETURN_POINTER(res);
362 : }
|
