TLA Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * dict_snowball.c
4 : * Snowball dictionary
5 : *
6 : * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
7 : *
8 : * IDENTIFICATION
9 : * src/backend/snowball/dict_snowball.c
10 : *
11 : *-------------------------------------------------------------------------
12 : */
13 : #include "postgres.h"
14 :
15 : #include "commands/defrem.h"
16 : #include "tsearch/ts_locale.h"
17 : #include "tsearch/ts_utils.h"
18 :
19 : /* Some platforms define MAXINT and/or MININT, causing conflicts */
20 : #ifdef MAXINT
21 : #undef MAXINT
22 : #endif
23 : #ifdef MININT
24 : #undef MININT
25 : #endif
26 :
27 : /* Now we can include the original Snowball header.h */
28 : #include "snowball/libstemmer/header.h"
29 : #include "snowball/libstemmer/stem_ISO_8859_1_basque.h"
30 : #include "snowball/libstemmer/stem_ISO_8859_1_catalan.h"
31 : #include "snowball/libstemmer/stem_ISO_8859_1_danish.h"
32 : #include "snowball/libstemmer/stem_ISO_8859_1_dutch.h"
33 : #include "snowball/libstemmer/stem_ISO_8859_1_english.h"
34 : #include "snowball/libstemmer/stem_ISO_8859_1_finnish.h"
35 : #include "snowball/libstemmer/stem_ISO_8859_1_french.h"
36 : #include "snowball/libstemmer/stem_ISO_8859_1_german.h"
37 : #include "snowball/libstemmer/stem_ISO_8859_1_indonesian.h"
38 : #include "snowball/libstemmer/stem_ISO_8859_1_irish.h"
39 : #include "snowball/libstemmer/stem_ISO_8859_1_italian.h"
40 : #include "snowball/libstemmer/stem_ISO_8859_1_norwegian.h"
41 : #include "snowball/libstemmer/stem_ISO_8859_1_porter.h"
42 : #include "snowball/libstemmer/stem_ISO_8859_1_portuguese.h"
43 : #include "snowball/libstemmer/stem_ISO_8859_1_spanish.h"
44 : #include "snowball/libstemmer/stem_ISO_8859_1_swedish.h"
45 : #include "snowball/libstemmer/stem_ISO_8859_2_hungarian.h"
46 : #include "snowball/libstemmer/stem_ISO_8859_2_romanian.h"
47 : #include "snowball/libstemmer/stem_KOI8_R_russian.h"
48 : #include "snowball/libstemmer/stem_UTF_8_arabic.h"
49 : #include "snowball/libstemmer/stem_UTF_8_armenian.h"
50 : #include "snowball/libstemmer/stem_UTF_8_basque.h"
51 : #include "snowball/libstemmer/stem_UTF_8_catalan.h"
52 : #include "snowball/libstemmer/stem_UTF_8_danish.h"
53 : #include "snowball/libstemmer/stem_UTF_8_dutch.h"
54 : #include "snowball/libstemmer/stem_UTF_8_english.h"
55 : #include "snowball/libstemmer/stem_UTF_8_finnish.h"
56 : #include "snowball/libstemmer/stem_UTF_8_french.h"
57 : #include "snowball/libstemmer/stem_UTF_8_german.h"
58 : #include "snowball/libstemmer/stem_UTF_8_greek.h"
59 : #include "snowball/libstemmer/stem_UTF_8_hindi.h"
60 : #include "snowball/libstemmer/stem_UTF_8_hungarian.h"
61 : #include "snowball/libstemmer/stem_UTF_8_indonesian.h"
62 : #include "snowball/libstemmer/stem_UTF_8_irish.h"
63 : #include "snowball/libstemmer/stem_UTF_8_italian.h"
64 : #include "snowball/libstemmer/stem_UTF_8_lithuanian.h"
65 : #include "snowball/libstemmer/stem_UTF_8_nepali.h"
66 : #include "snowball/libstemmer/stem_UTF_8_norwegian.h"
67 : #include "snowball/libstemmer/stem_UTF_8_porter.h"
68 : #include "snowball/libstemmer/stem_UTF_8_portuguese.h"
69 : #include "snowball/libstemmer/stem_UTF_8_romanian.h"
70 : #include "snowball/libstemmer/stem_UTF_8_russian.h"
71 : #include "snowball/libstemmer/stem_UTF_8_serbian.h"
72 : #include "snowball/libstemmer/stem_UTF_8_spanish.h"
73 : #include "snowball/libstemmer/stem_UTF_8_swedish.h"
74 : #include "snowball/libstemmer/stem_UTF_8_tamil.h"
75 : #include "snowball/libstemmer/stem_UTF_8_turkish.h"
76 : #include "snowball/libstemmer/stem_UTF_8_yiddish.h"
77 :
78 CBC 319 : PG_MODULE_MAGIC;
79 :
80 319 : PG_FUNCTION_INFO_V1(dsnowball_init);
81 :
82 319 : PG_FUNCTION_INFO_V1(dsnowball_lexize);
83 :
84 : /* List of supported modules */
85 : typedef struct stemmer_module
86 : {
87 : const char *name;
88 : pg_enc enc;
89 : struct SN_env *(*create) (void);
90 : void (*close) (struct SN_env *);
91 : int (*stem) (struct SN_env *);
92 : } stemmer_module;
93 :
94 : /* Args: stemmer name, PG code for encoding, Snowball's name for encoding */
95 : #define STEMMER_MODULE(name,enc,senc) \
96 : {#name, enc, name##_##senc##_create_env, name##_##senc##_close_env, name##_##senc##_stem}
97 :
98 : static const stemmer_module stemmer_modules[] =
99 : {
100 : /*
101 : * Stemmers list from Snowball distribution
102 : */
103 : STEMMER_MODULE(basque, PG_LATIN1, ISO_8859_1),
104 : STEMMER_MODULE(catalan, PG_LATIN1, ISO_8859_1),
105 : STEMMER_MODULE(danish, PG_LATIN1, ISO_8859_1),
106 : STEMMER_MODULE(dutch, PG_LATIN1, ISO_8859_1),
107 : STEMMER_MODULE(english, PG_LATIN1, ISO_8859_1),
108 : STEMMER_MODULE(finnish, PG_LATIN1, ISO_8859_1),
109 : STEMMER_MODULE(french, PG_LATIN1, ISO_8859_1),
110 : STEMMER_MODULE(german, PG_LATIN1, ISO_8859_1),
111 : STEMMER_MODULE(indonesian, PG_LATIN1, ISO_8859_1),
112 : STEMMER_MODULE(irish, PG_LATIN1, ISO_8859_1),
113 : STEMMER_MODULE(italian, PG_LATIN1, ISO_8859_1),
114 : STEMMER_MODULE(norwegian, PG_LATIN1, ISO_8859_1),
115 : STEMMER_MODULE(porter, PG_LATIN1, ISO_8859_1),
116 : STEMMER_MODULE(portuguese, PG_LATIN1, ISO_8859_1),
117 : STEMMER_MODULE(spanish, PG_LATIN1, ISO_8859_1),
118 : STEMMER_MODULE(swedish, PG_LATIN1, ISO_8859_1),
119 : STEMMER_MODULE(hungarian, PG_LATIN2, ISO_8859_2),
120 : STEMMER_MODULE(romanian, PG_LATIN2, ISO_8859_2),
121 : STEMMER_MODULE(russian, PG_KOI8R, KOI8_R),
122 : STEMMER_MODULE(arabic, PG_UTF8, UTF_8),
123 : STEMMER_MODULE(armenian, PG_UTF8, UTF_8),
124 : STEMMER_MODULE(basque, PG_UTF8, UTF_8),
125 : STEMMER_MODULE(catalan, PG_UTF8, UTF_8),
126 : STEMMER_MODULE(danish, PG_UTF8, UTF_8),
127 : STEMMER_MODULE(dutch, PG_UTF8, UTF_8),
128 : STEMMER_MODULE(english, PG_UTF8, UTF_8),
129 : STEMMER_MODULE(finnish, PG_UTF8, UTF_8),
130 : STEMMER_MODULE(french, PG_UTF8, UTF_8),
131 : STEMMER_MODULE(german, PG_UTF8, UTF_8),
132 : STEMMER_MODULE(greek, PG_UTF8, UTF_8),
133 : STEMMER_MODULE(hindi, PG_UTF8, UTF_8),
134 : STEMMER_MODULE(hungarian, PG_UTF8, UTF_8),
135 : STEMMER_MODULE(indonesian, PG_UTF8, UTF_8),
136 : STEMMER_MODULE(irish, PG_UTF8, UTF_8),
137 : STEMMER_MODULE(italian, PG_UTF8, UTF_8),
138 : STEMMER_MODULE(lithuanian, PG_UTF8, UTF_8),
139 : STEMMER_MODULE(nepali, PG_UTF8, UTF_8),
140 : STEMMER_MODULE(norwegian, PG_UTF8, UTF_8),
141 : STEMMER_MODULE(porter, PG_UTF8, UTF_8),
142 : STEMMER_MODULE(portuguese, PG_UTF8, UTF_8),
143 : STEMMER_MODULE(romanian, PG_UTF8, UTF_8),
144 : STEMMER_MODULE(russian, PG_UTF8, UTF_8),
145 : STEMMER_MODULE(serbian, PG_UTF8, UTF_8),
146 : STEMMER_MODULE(spanish, PG_UTF8, UTF_8),
147 : STEMMER_MODULE(swedish, PG_UTF8, UTF_8),
148 : STEMMER_MODULE(tamil, PG_UTF8, UTF_8),
149 : STEMMER_MODULE(turkish, PG_UTF8, UTF_8),
150 : STEMMER_MODULE(yiddish, PG_UTF8, UTF_8),
151 :
152 : /*
153 : * Stemmer with PG_SQL_ASCII encoding should be valid for any server
154 : * encoding
155 : */
156 : STEMMER_MODULE(english, PG_SQL_ASCII, ISO_8859_1),
157 :
158 : {NULL, 0, NULL, NULL, NULL} /* list end marker */
159 : };
160 :
161 :
162 : typedef struct DictSnowball
163 : {
164 : struct SN_env *z;
165 : StopList stoplist;
166 : bool needrecode; /* needs recoding before/after call stem */
167 : int (*stem) (struct SN_env *z);
168 :
169 : /*
170 : * snowball saves alloced memory between calls, so we should run it in our
171 : * private memory context. Note, init function is executed in long lived
172 : * context, so we just remember CurrentMemoryContext
173 : */
174 : MemoryContext dictCtx;
175 : } DictSnowball;
176 :
177 :
178 : static void
179 19 : locate_stem_module(DictSnowball *d, const char *lang)
180 : {
181 : const stemmer_module *m;
182 :
183 : /*
184 : * First, try to find exact match of stemmer module. Stemmer with
185 : * PG_SQL_ASCII encoding is treated as working with any server encoding
186 : */
187 494 : for (m = stemmer_modules; m->name; m++)
188 : {
189 627 : if ((m->enc == PG_SQL_ASCII || m->enc == GetDatabaseEncoding()) &&
190 133 : pg_strcasecmp(m->name, lang) == 0)
191 : {
192 19 : d->stem = m->stem;
193 19 : d->z = m->create();
194 19 : d->needrecode = false;
195 19 : return;
196 : }
197 : }
198 :
199 : /*
200 : * Second, try to find stemmer for needed language for UTF8 encoding.
201 : */
202 UBC 0 : for (m = stemmer_modules; m->name; m++)
203 : {
204 0 : if (m->enc == PG_UTF8 && pg_strcasecmp(m->name, lang) == 0)
205 : {
206 0 : d->stem = m->stem;
207 0 : d->z = m->create();
208 0 : d->needrecode = true;
209 0 : return;
210 : }
211 : }
212 :
213 0 : ereport(ERROR,
214 : (errcode(ERRCODE_UNDEFINED_OBJECT),
215 : errmsg("no Snowball stemmer available for language \"%s\" and encoding \"%s\"",
216 : lang, GetDatabaseEncodingName())));
217 : }
218 :
219 : Datum
220 CBC 19 : dsnowball_init(PG_FUNCTION_ARGS)
221 : {
222 19 : List *dictoptions = (List *) PG_GETARG_POINTER(0);
223 : DictSnowball *d;
224 19 : bool stoploaded = false;
225 : ListCell *l;
226 :
227 19 : d = (DictSnowball *) palloc0(sizeof(DictSnowball));
228 :
229 57 : foreach(l, dictoptions)
230 : {
231 38 : DefElem *defel = (DefElem *) lfirst(l);
232 :
233 38 : if (strcmp(defel->defname, "stopwords") == 0)
234 : {
235 19 : if (stoploaded)
236 UBC 0 : ereport(ERROR,
237 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
238 : errmsg("multiple StopWords parameters")));
239 CBC 19 : readstoplist(defGetString(defel), &d->stoplist, lowerstr);
240 19 : stoploaded = true;
241 : }
242 19 : else if (strcmp(defel->defname, "language") == 0)
243 : {
244 19 : if (d->stem)
245 UBC 0 : ereport(ERROR,
246 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
247 : errmsg("multiple Language parameters")));
248 CBC 19 : locate_stem_module(d, defGetString(defel));
249 : }
250 : else
251 : {
252 UBC 0 : ereport(ERROR,
253 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
254 : errmsg("unrecognized Snowball parameter: \"%s\"",
255 : defel->defname)));
256 : }
257 : }
258 :
259 CBC 19 : if (!d->stem)
260 UBC 0 : ereport(ERROR,
261 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
262 : errmsg("missing Language parameter")));
263 :
264 CBC 19 : d->dictCtx = CurrentMemoryContext;
265 :
266 19 : PG_RETURN_POINTER(d);
267 : }
268 :
269 : Datum
270 5135 : dsnowball_lexize(PG_FUNCTION_ARGS)
271 : {
272 5135 : DictSnowball *d = (DictSnowball *) PG_GETARG_POINTER(0);
273 5135 : char *in = (char *) PG_GETARG_POINTER(1);
274 5135 : int32 len = PG_GETARG_INT32(2);
275 5135 : char *txt = lowerstr_with_len(in, len);
276 5135 : TSLexeme *res = palloc0(sizeof(TSLexeme) * 2);
277 :
278 : /*
279 : * Do not pass strings exceeding 1000 bytes to the stemmer, as they're
280 : * surely not words in any human language. This restriction avoids
281 : * wasting cycles on stuff like base64-encoded data, and it protects us
282 : * against possible inefficiency or misbehavior in the stemmer. (For
283 : * example, the Turkish stemmer has an indefinite recursion, so it can
284 : * crash on long-enough strings.) However, Snowball dictionaries are
285 : * defined to recognize all strings, so we can't reject the string as an
286 : * unknown word.
287 : */
288 5135 : if (len > 1000)
289 : {
290 : /* return the lexeme lowercased, but otherwise unmodified */
291 UBC 0 : res->lexeme = txt;
292 : }
293 CBC 5135 : else if (*txt == '\0' || searchstoplist(&(d->stoplist), txt))
294 : {
295 : /* empty or stopword, so report as stopword */
296 1734 : pfree(txt);
297 : }
298 : else
299 : {
300 : MemoryContext saveCtx;
301 :
302 : /*
303 : * recode to utf8 if stemmer is utf8 and doesn't match server encoding
304 : */
305 3401 : if (d->needrecode)
306 : {
307 : char *recoded;
308 :
309 UBC 0 : recoded = pg_server_to_any(txt, strlen(txt), PG_UTF8);
310 0 : if (recoded != txt)
311 : {
312 0 : pfree(txt);
313 0 : txt = recoded;
314 : }
315 : }
316 :
317 : /* see comment about d->dictCtx */
318 CBC 3401 : saveCtx = MemoryContextSwitchTo(d->dictCtx);
319 3401 : SN_set_current(d->z, strlen(txt), (symbol *) txt);
320 3401 : d->stem(d->z);
321 3401 : MemoryContextSwitchTo(saveCtx);
322 :
323 3401 : if (d->z->p && d->z->l)
324 : {
325 3401 : txt = repalloc(txt, d->z->l + 1);
326 3401 : memcpy(txt, d->z->p, d->z->l);
327 3401 : txt[d->z->l] = '\0';
328 : }
329 :
330 : /* back recode if needed */
331 3401 : if (d->needrecode)
332 : {
333 : char *recoded;
334 :
335 UBC 0 : recoded = pg_any_to_server(txt, strlen(txt), PG_UTF8);
336 0 : if (recoded != txt)
337 : {
338 0 : pfree(txt);
339 0 : txt = recoded;
340 : }
341 : }
342 :
343 CBC 3401 : res->lexeme = txt;
344 : }
345 :
346 5135 : PG_RETURN_POINTER(res);
347 : }
|
