granicus.if.org Git - postgresql/blob - src/backend/snowball/dict_snowball.c

   1 /*-------------------------------------------------------------------------
   2  *
   3  * dict_snowball.c
   4  *              Snowball dictionary
   5  *
   6  * Portions Copyright (c) 1996-2014, 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_danish.h"
  30 #include "snowball/libstemmer/stem_ISO_8859_1_dutch.h"
  31 #include "snowball/libstemmer/stem_ISO_8859_1_english.h"
  32 #include "snowball/libstemmer/stem_ISO_8859_1_finnish.h"
  33 #include "snowball/libstemmer/stem_ISO_8859_1_french.h"
  34 #include "snowball/libstemmer/stem_ISO_8859_1_german.h"
  35 #include "snowball/libstemmer/stem_ISO_8859_1_hungarian.h"
  36 #include "snowball/libstemmer/stem_ISO_8859_1_italian.h"
  37 #include "snowball/libstemmer/stem_ISO_8859_1_norwegian.h"
  38 #include "snowball/libstemmer/stem_ISO_8859_1_porter.h"
  39 #include "snowball/libstemmer/stem_ISO_8859_1_portuguese.h"
  40 #include "snowball/libstemmer/stem_ISO_8859_1_spanish.h"
  41 #include "snowball/libstemmer/stem_ISO_8859_1_swedish.h"
  42 #include "snowball/libstemmer/stem_ISO_8859_2_romanian.h"
  43 #include "snowball/libstemmer/stem_KOI8_R_russian.h"
  44 #include "snowball/libstemmer/stem_UTF_8_danish.h"
  45 #include "snowball/libstemmer/stem_UTF_8_dutch.h"
  46 #include "snowball/libstemmer/stem_UTF_8_english.h"
  47 #include "snowball/libstemmer/stem_UTF_8_finnish.h"
  48 #include "snowball/libstemmer/stem_UTF_8_french.h"
  49 #include "snowball/libstemmer/stem_UTF_8_german.h"
  50 #include "snowball/libstemmer/stem_UTF_8_hungarian.h"
  51 #include "snowball/libstemmer/stem_UTF_8_italian.h"
  52 #include "snowball/libstemmer/stem_UTF_8_norwegian.h"
  53 #include "snowball/libstemmer/stem_UTF_8_porter.h"
  54 #include "snowball/libstemmer/stem_UTF_8_portuguese.h"
  55 #include "snowball/libstemmer/stem_UTF_8_romanian.h"
  56 #include "snowball/libstemmer/stem_UTF_8_russian.h"
  57 #include "snowball/libstemmer/stem_UTF_8_spanish.h"
  58 #include "snowball/libstemmer/stem_UTF_8_swedish.h"
  59 #include "snowball/libstemmer/stem_UTF_8_turkish.h"
  60
  61 PG_MODULE_MAGIC;
  62
  63 PG_FUNCTION_INFO_V1(dsnowball_init);
  64 Datum           dsnowball_init(PG_FUNCTION_ARGS);
  65
  66 PG_FUNCTION_INFO_V1(dsnowball_lexize);
  67 Datum           dsnowball_lexize(PG_FUNCTION_ARGS);
  68
  69 /* List of supported modules */
  70 typedef struct stemmer_module
  71 {
  72         const char *name;
  73         pg_enc          enc;
  74         struct SN_env *(*create) (void);
  75         void            (*close) (struct SN_env *);
  76         int                     (*stem) (struct SN_env *);
  77 } stemmer_module;
  78
  79 static const stemmer_module stemmer_modules[] =
  80 {
  81         /*
  82          * Stemmers list from Snowball distribution
  83          */
  84         {"danish", PG_LATIN1, danish_ISO_8859_1_create_env, danish_ISO_8859_1_close_env, danish_ISO_8859_1_stem},
  85         {"dutch", PG_LATIN1, dutch_ISO_8859_1_create_env, dutch_ISO_8859_1_close_env, dutch_ISO_8859_1_stem},
  86         {"english", PG_LATIN1, english_ISO_8859_1_create_env, english_ISO_8859_1_close_env, english_ISO_8859_1_stem},
  87         {"finnish", PG_LATIN1, finnish_ISO_8859_1_create_env, finnish_ISO_8859_1_close_env, finnish_ISO_8859_1_stem},
  88         {"french", PG_LATIN1, french_ISO_8859_1_create_env, french_ISO_8859_1_close_env, french_ISO_8859_1_stem},
  89         {"german", PG_LATIN1, german_ISO_8859_1_create_env, german_ISO_8859_1_close_env, german_ISO_8859_1_stem},
  90         {"hungarian", PG_LATIN1, hungarian_ISO_8859_1_create_env, hungarian_ISO_8859_1_close_env, hungarian_ISO_8859_1_stem},
  91         {"italian", PG_LATIN1, italian_ISO_8859_1_create_env, italian_ISO_8859_1_close_env, italian_ISO_8859_1_stem},
  92         {"norwegian", PG_LATIN1, norwegian_ISO_8859_1_create_env, norwegian_ISO_8859_1_close_env, norwegian_ISO_8859_1_stem},
  93         {"porter", PG_LATIN1, porter_ISO_8859_1_create_env, porter_ISO_8859_1_close_env, porter_ISO_8859_1_stem},
  94         {"portuguese", PG_LATIN1, portuguese_ISO_8859_1_create_env, portuguese_ISO_8859_1_close_env, portuguese_ISO_8859_1_stem},
  95         {"spanish", PG_LATIN1, spanish_ISO_8859_1_create_env, spanish_ISO_8859_1_close_env, spanish_ISO_8859_1_stem},
  96         {"swedish", PG_LATIN1, swedish_ISO_8859_1_create_env, swedish_ISO_8859_1_close_env, swedish_ISO_8859_1_stem},
  97         {"romanian", PG_LATIN2, romanian_ISO_8859_2_create_env, romanian_ISO_8859_2_close_env, romanian_ISO_8859_2_stem},
  98         {"russian", PG_KOI8R, russian_KOI8_R_create_env, russian_KOI8_R_close_env, russian_KOI8_R_stem},
  99         {"danish", PG_UTF8, danish_UTF_8_create_env, danish_UTF_8_close_env, danish_UTF_8_stem},
 100         {"dutch", PG_UTF8, dutch_UTF_8_create_env, dutch_UTF_8_close_env, dutch_UTF_8_stem},
 101         {"english", PG_UTF8, english_UTF_8_create_env, english_UTF_8_close_env, english_UTF_8_stem},
 102         {"finnish", PG_UTF8, finnish_UTF_8_create_env, finnish_UTF_8_close_env, finnish_UTF_8_stem},
 103         {"french", PG_UTF8, french_UTF_8_create_env, french_UTF_8_close_env, french_UTF_8_stem},
 104         {"german", PG_UTF8, german_UTF_8_create_env, german_UTF_8_close_env, german_UTF_8_stem},
 105         {"hungarian", PG_UTF8, hungarian_UTF_8_create_env, hungarian_UTF_8_close_env, hungarian_UTF_8_stem},
 106         {"italian", PG_UTF8, italian_UTF_8_create_env, italian_UTF_8_close_env, italian_UTF_8_stem},
 107         {"norwegian", PG_UTF8, norwegian_UTF_8_create_env, norwegian_UTF_8_close_env, norwegian_UTF_8_stem},
 108         {"porter", PG_UTF8, porter_UTF_8_create_env, porter_UTF_8_close_env, porter_UTF_8_stem},
 109         {"portuguese", PG_UTF8, portuguese_UTF_8_create_env, portuguese_UTF_8_close_env, portuguese_UTF_8_stem},
 110         {"romanian", PG_UTF8, romanian_UTF_8_create_env, romanian_UTF_8_close_env, romanian_UTF_8_stem},
 111         {"russian", PG_UTF8, russian_UTF_8_create_env, russian_UTF_8_close_env, russian_UTF_8_stem},
 112         {"spanish", PG_UTF8, spanish_UTF_8_create_env, spanish_UTF_8_close_env, spanish_UTF_8_stem},
 113         {"swedish", PG_UTF8, swedish_UTF_8_create_env, swedish_UTF_8_close_env, swedish_UTF_8_stem},
 114         {"turkish", PG_UTF8, turkish_UTF_8_create_env, turkish_UTF_8_close_env, turkish_UTF_8_stem},
 115
 116         /*
 117          * Stemmer with PG_SQL_ASCII encoding should be valid for any server
 118          * encoding
 119          */
 120         {"english", PG_SQL_ASCII, english_ISO_8859_1_create_env, english_ISO_8859_1_close_env, english_ISO_8859_1_stem},
 121
 122         {NULL, 0, NULL, NULL, NULL} /* list end marker */
 123 };
 124
 125
 126 typedef struct DictSnowball
 127 {
 128         struct SN_env *z;
 129         StopList        stoplist;
 130         bool            needrecode;             /* needs recoding before/after call stem */
 131         int                     (*stem) (struct SN_env * z);
 132
 133         /*
 134          * snowball saves alloced memory between calls, so we should run it in our
 135          * private memory context. Note, init function is executed in long lived
 136          * context, so we just remember CurrentMemoryContext
 137          */
 138         MemoryContext dictCtx;
 139 } DictSnowball;
 140
 141
 142 static void
 143 locate_stem_module(DictSnowball *d, char *lang)
 144 {
 145         const stemmer_module *m;
 146
 147         /*
 148          * First, try to find exact match of stemmer module. Stemmer with
 149          * PG_SQL_ASCII encoding is treated as working with any server encoding
 150          */
 151         for (m = stemmer_modules; m->name; m++)
 152         {
 153                 if ((m->enc == PG_SQL_ASCII || m->enc == GetDatabaseEncoding()) &&
 154                         pg_strcasecmp(m->name, lang) == 0)
 155                 {
 156                         d->stem = m->stem;
 157                         d->z = m->create();
 158                         d->needrecode = false;
 159                         return;
 160                 }
 161         }
 162
 163         /*
 164          * Second, try to find stemmer for needed language for UTF8 encoding.
 165          */
 166         for (m = stemmer_modules; m->name; m++)
 167         {
 168                 if (m->enc == PG_UTF8 && pg_strcasecmp(m->name, lang) == 0)
 169                 {
 170                         d->stem = m->stem;
 171                         d->z = m->create();
 172                         d->needrecode = true;
 173                         return;
 174                 }
 175         }
 176
 177         ereport(ERROR,
 178                         (errcode(ERRCODE_UNDEFINED_OBJECT),
 179                          errmsg("no Snowball stemmer available for language \"%s\" and encoding \"%s\"",
 180                                         lang, GetDatabaseEncodingName())));
 181 }
 182
 183 Datum
 184 dsnowball_init(PG_FUNCTION_ARGS)
 185 {
 186         List       *dictoptions = (List *) PG_GETARG_POINTER(0);
 187         DictSnowball *d;
 188         bool            stoploaded = false;
 189         ListCell   *l;
 190
 191         d = (DictSnowball *) palloc0(sizeof(DictSnowball));
 192
 193         foreach(l, dictoptions)
 194         {
 195                 DefElem    *defel = (DefElem *) lfirst(l);
 196
 197                 if (pg_strcasecmp("StopWords", defel->defname) == 0)
 198                 {
 199                         if (stoploaded)
 200                                 ereport(ERROR,
 201                                                 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
 202                                                  errmsg("multiple StopWords parameters")));
 203                         readstoplist(defGetString(defel), &d->stoplist, lowerstr);
 204                         stoploaded = true;
 205                 }
 206                 else if (pg_strcasecmp("Language", defel->defname) == 0)
 207                 {
 208                         if (d->stem)
 209                                 ereport(ERROR,
 210                                                 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
 211                                                  errmsg("multiple Language parameters")));
 212                         locate_stem_module(d, defGetString(defel));
 213                 }
 214                 else
 215                 {
 216                         ereport(ERROR,
 217                                         (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
 218                                          errmsg("unrecognized Snowball parameter: \"%s\"",
 219                                                         defel->defname)));
 220                 }
 221         }
 222
 223         if (!d->stem)
 224                 ereport(ERROR,
 225                                 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
 226                                  errmsg("missing Language parameter")));
 227
 228         d->dictCtx = CurrentMemoryContext;
 229
 230         PG_RETURN_POINTER(d);
 231 }
 232
 233 Datum
 234 dsnowball_lexize(PG_FUNCTION_ARGS)
 235 {
 236         DictSnowball *d = (DictSnowball *) PG_GETARG_POINTER(0);
 237         char       *in = (char *) PG_GETARG_POINTER(1);
 238         int32           len = PG_GETARG_INT32(2);
 239         char       *txt = lowerstr_with_len(in, len);
 240         TSLexeme   *res = palloc0(sizeof(TSLexeme) * 2);
 241
 242         if (*txt == '\0' || searchstoplist(&(d->stoplist), txt))
 243         {
 244                 pfree(txt);
 245         }
 246         else
 247         {
 248                 MemoryContext saveCtx;
 249
 250                 /*
 251                  * recode to utf8 if stemmer is utf8 and doesn't match server encoding
 252                  */
 253                 if (d->needrecode)
 254                 {
 255                         char       *recoded;
 256
 257                         recoded = pg_server_to_any(txt, strlen(txt), PG_UTF8);
 258                         if (recoded != txt)
 259                         {
 260                                 pfree(txt);
 261                                 txt = recoded;
 262                         }
 263                 }
 264
 265                 /* see comment about d->dictCtx */
 266                 saveCtx = MemoryContextSwitchTo(d->dictCtx);
 267                 SN_set_current(d->z, strlen(txt), (symbol *) txt);
 268                 d->stem(d->z);
 269                 MemoryContextSwitchTo(saveCtx);
 270
 271                 if (d->z->p && d->z->l)
 272                 {
 273                         txt = repalloc(txt, d->z->l + 1);
 274                         memcpy(txt, d->z->p, d->z->l);
 275                         txt[d->z->l] = '\0';
 276                 }
 277
 278                 /* back recode if needed */
 279                 if (d->needrecode)
 280                 {
 281                         char       *recoded;
 282
 283                         recoded = pg_any_to_server(txt, strlen(txt), PG_UTF8);
 284                         if (recoded != txt)
 285                         {
 286                                 pfree(txt);
 287                                 txt = recoded;
 288                         }
 289                 }
 290
 291                 res->lexeme = txt;
 292         }
 293
 294         PG_RETURN_POINTER(res);
 295 }