granicus.if.org Git - postgresql/blob - src/port/chklocale.c

   1 /*-------------------------------------------------------------------------
   2  *
   3  * chklocale.c
   4  *              Functions for handling locale-related info
   5  *
   6  *
   7  * Copyright (c) 1996-2013, PostgreSQL Global Development Group
   8  *
   9  *
  10  * IDENTIFICATION
  11  *        src/port/chklocale.c
  12  *
  13  *-------------------------------------------------------------------------
  14  */
  15
  16 #ifndef FRONTEND
  17 #include "postgres.h"
  18 #else
  19 #include "postgres_fe.h"
  20 #endif
  21
  22 #include <locale.h>
  23 #ifdef HAVE_LANGINFO_H
  24 #include <langinfo.h>
  25 #endif
  26
  27 #include "mb/pg_wchar.h"
  28
  29
  30 /*
  31  * This table needs to recognize all the CODESET spellings for supported
  32  * backend encodings, as well as frontend-only encodings where possible
  33  * (the latter case is currently only needed for initdb to recognize
  34  * error situations).  On Windows, we rely on entries for codepage
  35  * numbers (CPnnn).
  36  *
  37  * Note that we search the table with pg_strcasecmp(), so variant
  38  * capitalizations don't need their own entries.
  39  */
  40 struct encoding_match
  41 {
  42         enum pg_enc pg_enc_code;
  43         const char *system_enc_name;
  44 };
  45
  46 static const struct encoding_match encoding_match_list[] = {
  47         {PG_EUC_JP, "EUC-JP"},
  48         {PG_EUC_JP, "eucJP"},
  49         {PG_EUC_JP, "IBM-eucJP"},
  50         {PG_EUC_JP, "sdeckanji"},
  51         {PG_EUC_JP, "CP20932"},
  52
  53         {PG_EUC_CN, "EUC-CN"},
  54         {PG_EUC_CN, "eucCN"},
  55         {PG_EUC_CN, "IBM-eucCN"},
  56         {PG_EUC_CN, "GB2312"},
  57         {PG_EUC_CN, "dechanzi"},
  58         {PG_EUC_CN, "CP20936"},
  59
  60         {PG_EUC_KR, "EUC-KR"},
  61         {PG_EUC_KR, "eucKR"},
  62         {PG_EUC_KR, "IBM-eucKR"},
  63         {PG_EUC_KR, "deckorean"},
  64         {PG_EUC_KR, "5601"},
  65         {PG_EUC_KR, "CP51949"},
  66
  67         {PG_EUC_TW, "EUC-TW"},
  68         {PG_EUC_TW, "eucTW"},
  69         {PG_EUC_TW, "IBM-eucTW"},
  70         {PG_EUC_TW, "cns11643"},
  71         /* No codepage for EUC-TW ? */
  72
  73         {PG_UTF8, "UTF-8"},
  74         {PG_UTF8, "utf8"},
  75         {PG_UTF8, "CP65001"},
  76
  77         {PG_LATIN1, "ISO-8859-1"},
  78         {PG_LATIN1, "ISO8859-1"},
  79         {PG_LATIN1, "iso88591"},
  80         {PG_LATIN1, "CP28591"},
  81
  82         {PG_LATIN2, "ISO-8859-2"},
  83         {PG_LATIN2, "ISO8859-2"},
  84         {PG_LATIN2, "iso88592"},
  85         {PG_LATIN2, "CP28592"},
  86
  87         {PG_LATIN3, "ISO-8859-3"},
  88         {PG_LATIN3, "ISO8859-3"},
  89         {PG_LATIN3, "iso88593"},
  90         {PG_LATIN3, "CP28593"},
  91
  92         {PG_LATIN4, "ISO-8859-4"},
  93         {PG_LATIN4, "ISO8859-4"},
  94         {PG_LATIN4, "iso88594"},
  95         {PG_LATIN4, "CP28594"},
  96
  97         {PG_LATIN5, "ISO-8859-9"},
  98         {PG_LATIN5, "ISO8859-9"},
  99         {PG_LATIN5, "iso88599"},
 100         {PG_LATIN5, "CP28599"},
 101
 102         {PG_LATIN6, "ISO-8859-10"},
 103         {PG_LATIN6, "ISO8859-10"},
 104         {PG_LATIN6, "iso885910"},
 105
 106         {PG_LATIN7, "ISO-8859-13"},
 107         {PG_LATIN7, "ISO8859-13"},
 108         {PG_LATIN7, "iso885913"},
 109
 110         {PG_LATIN8, "ISO-8859-14"},
 111         {PG_LATIN8, "ISO8859-14"},
 112         {PG_LATIN8, "iso885914"},
 113
 114         {PG_LATIN9, "ISO-8859-15"},
 115         {PG_LATIN9, "ISO8859-15"},
 116         {PG_LATIN9, "iso885915"},
 117         {PG_LATIN9, "CP28605"},
 118
 119         {PG_LATIN10, "ISO-8859-16"},
 120         {PG_LATIN10, "ISO8859-16"},
 121         {PG_LATIN10, "iso885916"},
 122
 123         {PG_KOI8R, "KOI8-R"},
 124         {PG_KOI8R, "CP20866"},
 125
 126         {PG_KOI8U, "KOI8-U"},
 127         {PG_KOI8U, "CP21866"},
 128
 129         {PG_WIN866, "CP866"},
 130         {PG_WIN874, "CP874"},
 131         {PG_WIN1250, "CP1250"},
 132         {PG_WIN1251, "CP1251"},
 133         {PG_WIN1251, "ansi-1251"},
 134         {PG_WIN1252, "CP1252"},
 135         {PG_WIN1253, "CP1253"},
 136         {PG_WIN1254, "CP1254"},
 137         {PG_WIN1255, "CP1255"},
 138         {PG_WIN1256, "CP1256"},
 139         {PG_WIN1257, "CP1257"},
 140         {PG_WIN1258, "CP1258"},
 141
 142         {PG_ISO_8859_5, "ISO-8859-5"},
 143         {PG_ISO_8859_5, "ISO8859-5"},
 144         {PG_ISO_8859_5, "iso88595"},
 145         {PG_ISO_8859_5, "CP28595"},
 146
 147         {PG_ISO_8859_6, "ISO-8859-6"},
 148         {PG_ISO_8859_6, "ISO8859-6"},
 149         {PG_ISO_8859_6, "iso88596"},
 150         {PG_ISO_8859_6, "CP28596"},
 151
 152         {PG_ISO_8859_7, "ISO-8859-7"},
 153         {PG_ISO_8859_7, "ISO8859-7"},
 154         {PG_ISO_8859_7, "iso88597"},
 155         {PG_ISO_8859_7, "CP28597"},
 156
 157         {PG_ISO_8859_8, "ISO-8859-8"},
 158         {PG_ISO_8859_8, "ISO8859-8"},
 159         {PG_ISO_8859_8, "iso88598"},
 160         {PG_ISO_8859_8, "CP28598"},
 161
 162         {PG_SJIS, "SJIS"},
 163         {PG_SJIS, "PCK"},
 164         {PG_SJIS, "CP932"},
 165
 166         {PG_BIG5, "BIG5"},
 167         {PG_BIG5, "BIG5HKSCS"},
 168         {PG_BIG5, "Big5-HKSCS"},
 169         {PG_BIG5, "CP950"},
 170
 171         {PG_GBK, "GBK"},
 172         {PG_GBK, "CP936"},
 173
 174         {PG_UHC, "UHC"},
 175         {PG_UHC, "CP949"},
 176
 177         {PG_JOHAB, "JOHAB"},
 178         {PG_JOHAB, "CP1361"},
 179
 180         {PG_GB18030, "GB18030"},
 181         {PG_GB18030, "CP54936"},
 182
 183         {PG_SHIFT_JIS_2004, "SJIS_2004"},
 184
 185         {PG_SQL_ASCII, "US-ASCII"},
 186
 187         {PG_SQL_ASCII, NULL}            /* end marker */
 188 };
 189
 190 #ifdef WIN32
 191 /*
 192  * On Windows, use CP<code page number> instead of the nl_langinfo() result
 193  *
 194  * Visual Studio 2012 expanded the set of valid LC_CTYPE values, so have its
 195  * locale machinery determine the code page.  See comments at IsoLocaleName().
 196  * For other compilers, follow the locale's predictable format.
 197  *
 198  * Returns a malloc()'d string for the caller to free.
 199  */
 200 static char *
 201 win32_langinfo(const char *ctype)
 202 {
 203         char       *r = NULL;
 204
 205 #if (_MSC_VER >= 1700)
 206         _locale_t       loct = NULL;
 207
 208         loct = _create_locale(LC_CTYPE, ctype);
 209         if (loct != NULL)
 210         {
 211                 r = malloc(16);                 /* excess */
 212                 if (r != NULL)
 213                         sprintf(r, "CP%u", loct->locinfo->lc_codepage);
 214                 _free_locale(loct);
 215         }
 216 #else
 217         char       *codepage;
 218
 219         /*
 220          * Locale format on Win32 is <Language>_<Country>.<CodePage> . For
 221          * example, English_United States.1252.
 222          */
 223         codepage = strrchr(ctype, '.');
 224         if (codepage != NULL)
 225         {
 226                 int                     ln;
 227
 228                 codepage++;
 229                 ln = strlen(codepage);
 230                 r = malloc(ln + 3);
 231                 if (r != NULL)
 232                         sprintf(r, "CP%s", codepage);
 233         }
 234 #endif
 235
 236         return r;
 237 }
 238
 239 #ifndef FRONTEND
 240 /*
 241  * Given a Windows code page identifier, find the corresponding PostgreSQL
 242  * encoding.  Issue a warning and return -1 if none found.
 243  */
 244 int
 245 pg_codepage_to_encoding(UINT cp)
 246 {
 247         char            sys[16];
 248         int                     i;
 249
 250         sprintf(sys, "CP%u", cp);
 251
 252         /* Check the table */
 253         for (i = 0; encoding_match_list[i].system_enc_name; i++)
 254                 if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
 255                         return encoding_match_list[i].pg_enc_code;
 256
 257         ereport(WARNING,
 258                         (errmsg("could not determine encoding for codeset \"%s\"", sys),
 259                    errdetail("Please report this to <pgsql-bugs@postgresql.org>.")));
 260
 261         return -1;
 262 }
 263 #endif
 264 #endif   /* WIN32 */
 265
 266 #if (defined(HAVE_LANGINFO_H) && defined(CODESET)) || defined(WIN32)
 267
 268 /*
 269  * Given a setting for LC_CTYPE, return the Postgres ID of the associated
 270  * encoding, if we can determine it.  Return -1 if we can't determine it.
 271  *
 272  * Pass in NULL to get the encoding for the current locale setting.
 273  * Pass "" to get the encoding selected by the server's environment.
 274  *
 275  * If the result is PG_SQL_ASCII, callers should treat it as being compatible
 276  * with any desired encoding.
 277  *
 278  * If running in the backend and write_message is false, this function must
 279  * cope with the possibility that elog() and palloc() are not yet usable.
 280  */
 281 int
 282 pg_get_encoding_from_locale(const char *ctype, bool write_message)
 283 {
 284         char       *sys;
 285         int                     i;
 286
 287         /* Get the CODESET property, and also LC_CTYPE if not passed in */
 288         if (ctype)
 289         {
 290                 char       *save;
 291                 char       *name;
 292
 293                 /* If locale is C or POSIX, we can allow all encodings */
 294                 if (pg_strcasecmp(ctype, "C") == 0 ||
 295                         pg_strcasecmp(ctype, "POSIX") == 0)
 296                         return PG_SQL_ASCII;
 297
 298                 save = setlocale(LC_CTYPE, NULL);
 299                 if (!save)
 300                         return -1;                      /* setlocale() broken? */
 301                 /* must copy result, or it might change after setlocale */
 302                 save = strdup(save);
 303                 if (!save)
 304                         return -1;                      /* out of memory; unlikely */
 305
 306                 name = setlocale(LC_CTYPE, ctype);
 307                 if (!name)
 308                 {
 309                         free(save);
 310                         return -1;                      /* bogus ctype passed in? */
 311                 }
 312
 313 #ifndef WIN32
 314                 sys = nl_langinfo(CODESET);
 315                 if (sys)
 316                         sys = strdup(sys);
 317 #else
 318                 sys = win32_langinfo(name);
 319 #endif
 320
 321                 setlocale(LC_CTYPE, save);
 322                 free(save);
 323         }
 324         else
 325         {
 326                 /* much easier... */
 327                 ctype = setlocale(LC_CTYPE, NULL);
 328                 if (!ctype)
 329                         return -1;                      /* setlocale() broken? */
 330
 331                 /* If locale is C or POSIX, we can allow all encodings */
 332                 if (pg_strcasecmp(ctype, "C") == 0 ||
 333                         pg_strcasecmp(ctype, "POSIX") == 0)
 334                         return PG_SQL_ASCII;
 335
 336 #ifndef WIN32
 337                 sys = nl_langinfo(CODESET);
 338                 if (sys)
 339                         sys = strdup(sys);
 340 #else
 341                 sys = win32_langinfo(ctype);
 342 #endif
 343         }
 344
 345         if (!sys)
 346                 return -1;                              /* out of memory; unlikely */
 347
 348         /* Check the table */
 349         for (i = 0; encoding_match_list[i].system_enc_name; i++)
 350         {
 351                 if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
 352                 {
 353                         free(sys);
 354                         return encoding_match_list[i].pg_enc_code;
 355                 }
 356         }
 357
 358         /* Special-case kluges for particular platforms go here */
 359
 360 #ifdef __darwin__
 361
 362         /*
 363          * Current OS X has many locales that report an empty string for CODESET,
 364          * but they all seem to actually use UTF-8.
 365          */
 366         if (strlen(sys) == 0)
 367         {
 368                 free(sys);
 369                 return PG_UTF8;
 370         }
 371 #endif
 372
 373         /*
 374          * We print a warning if we got a CODESET string but couldn't recognize
 375          * it.  This means we need another entry in the table.
 376          */
 377         if (write_message)
 378         {
 379 #ifdef FRONTEND
 380                 fprintf(stderr, _("could not determine encoding for locale \"%s\": codeset is \"%s\""),
 381                                 ctype, sys);
 382                 /* keep newline separate so there's only one translatable string */
 383                 fputc('\n', stderr);
 384 #else
 385                 ereport(WARNING,
 386                                 (errmsg("could not determine encoding for locale \"%s\": codeset is \"%s\"",
 387                                                 ctype, sys),
 388                    errdetail("Please report this to <pgsql-bugs@postgresql.org>.")));
 389 #endif
 390         }
 391
 392         free(sys);
 393         return -1;
 394 }
 395 #else                                                   /* (HAVE_LANGINFO_H && CODESET) || WIN32 */
 396
 397 /*
 398  * stub if no multi-language platform support
 399  *
 400  * Note: we could return -1 here, but that would have the effect of
 401  * forcing users to specify an encoding to initdb on such platforms.
 402  * It seems better to silently default to SQL_ASCII.
 403  */
 404 int
 405 pg_get_encoding_from_locale(const char *ctype, bool write_message)
 406 {
 407         return PG_SQL_ASCII;
 408 }
 409
 410 #endif   /* (HAVE_LANGINFO_H && CODESET) || WIN32 */