src/port/chklocale.c

   1 /*-------------------------------------------------------------------------
   2  *
   3  * chklocale.c
   4  *              Functions for handling locale-related info
   5  *
   6  *
   7  * Copyright (c) 1996-2025, 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 #ifndef WIN32
  23 #include <langinfo.h>
  24 #endif
  25
  26 #include "mb/pg_wchar.h"
  27
  28
  29 /*
  30  * This table needs to recognize all the CODESET spellings for supported
  31  * backend encodings, as well as frontend-only encodings where possible
  32  * (the latter case is currently only needed for initdb to recognize
  33  * error situations).  On Windows, we rely on entries for codepage
  34  * numbers (CPnnn).
  35  *
  36  * Note that we search the table with pg_strcasecmp(), so variant
  37  * capitalizations don't need their own entries.
  38  */
  39 struct encoding_match
  40 {
  41         enum pg_enc pg_enc_code;
  42         const char *system_enc_name;
  43 };
  44
  45 static const struct encoding_match encoding_match_list[] = {
  46         {PG_EUC_JP, "EUC-JP"},
  47         {PG_EUC_JP, "eucJP"},
  48         {PG_EUC_JP, "IBM-eucJP"},
  49         {PG_EUC_JP, "sdeckanji"},
  50         {PG_EUC_JP, "CP20932"},
  51
  52         {PG_EUC_CN, "EUC-CN"},
  53         {PG_EUC_CN, "eucCN"},
  54         {PG_EUC_CN, "IBM-eucCN"},
  55         {PG_EUC_CN, "GB2312"},
  56         {PG_EUC_CN, "dechanzi"},
  57         {PG_EUC_CN, "CP20936"},
  58
  59         {PG_EUC_KR, "EUC-KR"},
  60         {PG_EUC_KR, "eucKR"},
  61         {PG_EUC_KR, "IBM-eucKR"},
  62         {PG_EUC_KR, "deckorean"},
  63         {PG_EUC_KR, "5601"},
  64         {PG_EUC_KR, "CP51949"},
  65
  66         {PG_EUC_TW, "EUC-TW"},
  67         {PG_EUC_TW, "eucTW"},
  68         {PG_EUC_TW, "IBM-eucTW"},
  69         {PG_EUC_TW, "cns11643"},
  70         /* No codepage for EUC-TW ? */
  71
  72         {PG_UTF8, "UTF-8"},
  73         {PG_UTF8, "utf8"},
  74         {PG_UTF8, "CP65001"},
  75
  76         {PG_LATIN1, "ISO-8859-1"},
  77         {PG_LATIN1, "ISO8859-1"},
  78         {PG_LATIN1, "iso88591"},
  79         {PG_LATIN1, "CP28591"},
  80
  81         {PG_LATIN2, "ISO-8859-2"},
  82         {PG_LATIN2, "ISO8859-2"},
  83         {PG_LATIN2, "iso88592"},
  84         {PG_LATIN2, "CP28592"},
  85
  86         {PG_LATIN3, "ISO-8859-3"},
  87         {PG_LATIN3, "ISO8859-3"},
  88         {PG_LATIN3, "iso88593"},
  89         {PG_LATIN3, "CP28593"},
  90
  91         {PG_LATIN4, "ISO-8859-4"},
  92         {PG_LATIN4, "ISO8859-4"},
  93         {PG_LATIN4, "iso88594"},
  94         {PG_LATIN4, "CP28594"},
  95
  96         {PG_LATIN5, "ISO-8859-9"},
  97         {PG_LATIN5, "ISO8859-9"},
  98         {PG_LATIN5, "iso88599"},
  99         {PG_LATIN5, "CP28599"},
 100
 101         {PG_LATIN6, "ISO-8859-10"},
 102         {PG_LATIN6, "ISO8859-10"},
 103         {PG_LATIN6, "iso885910"},
 104
 105         {PG_LATIN7, "ISO-8859-13"},
 106         {PG_LATIN7, "ISO8859-13"},
 107         {PG_LATIN7, "iso885913"},
 108
 109         {PG_LATIN8, "ISO-8859-14"},
 110         {PG_LATIN8, "ISO8859-14"},
 111         {PG_LATIN8, "iso885914"},
 112
 113         {PG_LATIN9, "ISO-8859-15"},
 114         {PG_LATIN9, "ISO8859-15"},
 115         {PG_LATIN9, "iso885915"},
 116         {PG_LATIN9, "CP28605"},
 117
 118         {PG_LATIN10, "ISO-8859-16"},
 119         {PG_LATIN10, "ISO8859-16"},
 120         {PG_LATIN10, "iso885916"},
 121
 122         {PG_KOI8R, "KOI8-R"},
 123         {PG_KOI8R, "CP20866"},
 124
 125         {PG_KOI8U, "KOI8-U"},
 126         {PG_KOI8U, "CP21866"},
 127
 128         {PG_WIN866, "CP866"},
 129         {PG_WIN874, "CP874"},
 130         {PG_WIN1250, "CP1250"},
 131         {PG_WIN1251, "CP1251"},
 132         {PG_WIN1251, "ansi-1251"},
 133         {PG_WIN1252, "CP1252"},
 134         {PG_WIN1253, "CP1253"},
 135         {PG_WIN1254, "CP1254"},
 136         {PG_WIN1255, "CP1255"},
 137         {PG_WIN1256, "CP1256"},
 138         {PG_WIN1257, "CP1257"},
 139         {PG_WIN1258, "CP1258"},
 140
 141         {PG_ISO_8859_5, "ISO-8859-5"},
 142         {PG_ISO_8859_5, "ISO8859-5"},
 143         {PG_ISO_8859_5, "iso88595"},
 144         {PG_ISO_8859_5, "CP28595"},
 145
 146         {PG_ISO_8859_6, "ISO-8859-6"},
 147         {PG_ISO_8859_6, "ISO8859-6"},
 148         {PG_ISO_8859_6, "iso88596"},
 149         {PG_ISO_8859_6, "CP28596"},
 150
 151         {PG_ISO_8859_7, "ISO-8859-7"},
 152         {PG_ISO_8859_7, "ISO8859-7"},
 153         {PG_ISO_8859_7, "iso88597"},
 154         {PG_ISO_8859_7, "CP28597"},
 155
 156         {PG_ISO_8859_8, "ISO-8859-8"},
 157         {PG_ISO_8859_8, "ISO8859-8"},
 158         {PG_ISO_8859_8, "iso88598"},
 159         {PG_ISO_8859_8, "CP28598"},
 160
 161         {PG_SJIS, "SJIS"},
 162         {PG_SJIS, "PCK"},
 163         {PG_SJIS, "CP932"},
 164         {PG_SJIS, "SHIFT_JIS"},
 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 CODESET.
 193  *
 194  * This routine uses GetLocaleInfoEx() to parse short locale names like
 195  * "de-DE", "fr-FR", etc.  If those cannot be parsed correctly process falls
 196  * back to the pre-VS-2010 manual parsing done with using
 197  * <Language>_<Country>.<CodePage> as a base.
 198  *
 199  * Returns a malloc()'d string for the caller to free.
 200  */
 201 static char *
 202 win32_get_codeset(const char *ctype)
 203 {
 204         char       *r = NULL;
 205         char       *codepage;
 206         uint32          cp;
 207         WCHAR           wctype[LOCALE_NAME_MAX_LENGTH];
 208
 209         memset(wctype, 0, sizeof(wctype));
 210         MultiByteToWideChar(CP_ACP, 0, ctype, -1, wctype, LOCALE_NAME_MAX_LENGTH);
 211
 212         if (GetLocaleInfoEx(wctype,
 213                                                 LOCALE_IDEFAULTANSICODEPAGE | LOCALE_RETURN_NUMBER,
 214                                                 (LPWSTR) &cp, sizeof(cp) / sizeof(WCHAR)) > 0)
 215         {
 216                 r = malloc(16);                 /* excess */
 217                 if (r != NULL)
 218                 {
 219                         /*
 220                          * If the return value is CP_ACP that means no ANSI code page is
 221                          * available, so only Unicode can be used for the locale.
 222                          */
 223                         if (cp == CP_ACP)
 224                                 strcpy(r, "utf8");
 225                         else
 226                                 sprintf(r, "CP%u", cp);
 227                 }
 228         }
 229         else
 230         {
 231                 /*
 232                  * Locale format on Win32 is <Language>_<Country>.<CodePage>.  For
 233                  * example, English_United States.1252.  If we see digits after the
 234                  * last dot, assume it's a codepage number.  Otherwise, we might be
 235                  * dealing with a Unix-style locale string; Windows' setlocale() will
 236                  * take those even though GetLocaleInfoEx() won't, so we end up here.
 237                  * In that case, just return what's after the last dot and hope we can
 238                  * find it in our table.
 239                  */
 240                 codepage = strrchr(ctype, '.');
 241                 if (codepage != NULL)
 242                 {
 243                         size_t          ln;
 244
 245                         codepage++;
 246                         ln = strlen(codepage);
 247                         r = malloc(ln + 3);
 248                         if (r != NULL)
 249                         {
 250                                 if (strspn(codepage, "0123456789") == ln)
 251                                         sprintf(r, "CP%s", codepage);
 252                                 else
 253                                         strcpy(r, codepage);
 254                         }
 255                 }
 256         }
 257
 258         return r;
 259 }
 260
 261 #ifndef FRONTEND
 262 /*
 263  * Given a Windows code page identifier, find the corresponding PostgreSQL
 264  * encoding.  Issue a warning and return -1 if none found.
 265  */
 266 int
 267 pg_codepage_to_encoding(UINT cp)
 268 {
 269         char            sys[16];
 270         int                     i;
 271
 272         sprintf(sys, "CP%u", cp);
 273
 274         /* Check the table */
 275         for (i = 0; encoding_match_list[i].system_enc_name; i++)
 276                 if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
 277                         return encoding_match_list[i].pg_enc_code;
 278
 279         ereport(WARNING,
 280                         (errmsg("could not determine encoding for codeset \"%s\"", sys)));
 281
 282         return -1;
 283 }
 284 #endif
 285 #endif                                                  /* WIN32 */
 286
 287 /*
 288  * Given a setting for LC_CTYPE, return the Postgres ID of the associated
 289  * encoding, if we can determine it.  Return -1 if we can't determine it.
 290  *
 291  * Pass in NULL to get the encoding for the current locale setting.
 292  * Pass "" to get the encoding selected by the server's environment.
 293  *
 294  * If the result is PG_SQL_ASCII, callers should treat it as being compatible
 295  * with any desired encoding.
 296  *
 297  * If running in the backend and write_message is false, this function must
 298  * cope with the possibility that elog() and palloc() are not yet usable.
 299  */
 300 int
 301 pg_get_encoding_from_locale(const char *ctype, bool write_message)
 302 {
 303         char       *sys;
 304         int                     i;
 305
 306 #ifndef WIN32
 307         locale_t        loc;
 308 #endif
 309
 310         /* Get the CODESET property, and also LC_CTYPE if not passed in */
 311         if (!ctype)
 312                 ctype = setlocale(LC_CTYPE, NULL);
 313
 314
 315         /* If locale is C or POSIX, we can allow all encodings */
 316         if (pg_strcasecmp(ctype, "C") == 0 ||
 317                 pg_strcasecmp(ctype, "POSIX") == 0)
 318                 return PG_SQL_ASCII;
 319
 320
 321 #ifndef WIN32
 322         loc = newlocale(LC_CTYPE_MASK, ctype, (locale_t) 0);
 323         if (loc == (locale_t) 0)
 324                 return -1;                              /* bogus ctype passed in? */
 325
 326         sys = nl_langinfo_l(CODESET, loc);
 327         if (sys)
 328                 sys = strdup(sys);
 329
 330         freelocale(loc);
 331 #else
 332         sys = win32_get_codeset(ctype);
 333 #endif
 334
 335         if (!sys)
 336                 return -1;                              /* out of memory; unlikely */
 337
 338         /* Check the table */
 339         for (i = 0; encoding_match_list[i].system_enc_name; i++)
 340         {
 341                 if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
 342                 {
 343                         free(sys);
 344                         return encoding_match_list[i].pg_enc_code;
 345                 }
 346         }
 347
 348         /* Special-case kluges for particular platforms go here */
 349
 350 #ifdef __darwin__
 351
 352         /*
 353          * Current macOS has many locales that report an empty string for CODESET,
 354          * but they all seem to actually use UTF-8.
 355          */
 356         if (strlen(sys) == 0)
 357         {
 358                 free(sys);
 359                 return PG_UTF8;
 360         }
 361 #endif
 362
 363         /*
 364          * We print a warning if we got a CODESET string but couldn't recognize
 365          * it.  This means we need another entry in the table.
 366          */
 367         if (write_message)
 368         {
 369 #ifdef FRONTEND
 370                 fprintf(stderr, _("could not determine encoding for locale \"%s\": codeset is \"%s\""),
 371                                 ctype, sys);
 372                 /* keep newline separate so there's only one translatable string */
 373                 fputc('\n', stderr);
 374 #else
 375                 ereport(WARNING,
 376                                 (errmsg("could not determine encoding for locale \"%s\": codeset is \"%s\"",
 377                                                 ctype, sys)));
 378 #endif
 379         }
 380
 381         free(sys);
 382         return -1;
 383 }