1/*-------------------------------------------------------------------------
2 *
3 * chklocale.c
4 * Functions for handling locale-related info
5 *
6 *
7 * Copyright (c) 1996-2019, 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#ifdef HAVE_LANGINFO_H
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 */
39struct encoding_match
40{
41 enum pg_enc pg_enc_code;
42 const char *system_enc_name;
43};
44
45static 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 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 * Visual Studio 2015 should still be able to do the same, but the declaration
199 * of lc_codepage is missing in _locale_t, causing this code compilation to
200 * fail, hence this falls back instead on GetLocaleInfoEx. VS 2015 may be an
201 * exception and post-VS2015 versions should be able to handle properly the
202 * codepage number using _create_locale(). So, instead of the same logic as
203 * VS 2012 and VS 2013, this routine uses GetLocaleInfoEx to parse short
204 * locale names like "de-DE", "fr-FR", etc. If those cannot be parsed correctly
205 * process falls back to the pre-VS-2010 manual parsing done with
206 * using <Language>_<Country>.<CodePage> as a base.
207 *
208 * Returns a malloc()'d string for the caller to free.
209 */
210static char *
211win32_langinfo(const char *ctype)
212{
213 char *r = NULL;
214
215#if (_MSC_VER >= 1700) && (_MSC_VER < 1900)
216 _locale_t loct = NULL;
217
218 loct = _create_locale(LC_CTYPE, ctype);
219 if (loct != NULL)
220 {
221 r = malloc(16); /* excess */
222 if (r != NULL)
223 sprintf(r, "CP%u", loct->locinfo->lc_codepage);
224 _free_locale(loct);
225 }
226#else
227 char *codepage;
228
229#if (_MSC_VER >= 1900)
230 uint32 cp;
231 WCHAR wctype[LOCALE_NAME_MAX_LENGTH];
232
233 memset(wctype, 0, sizeof(wctype));
234 MultiByteToWideChar(CP_ACP, 0, ctype, -1, wctype, LOCALE_NAME_MAX_LENGTH);
235
236 if (GetLocaleInfoEx(wctype,
237 LOCALE_IDEFAULTANSICODEPAGE | LOCALE_RETURN_NUMBER,
238 (LPWSTR) &cp, sizeof(cp) / sizeof(WCHAR)) > 0)
239 {
240 r = malloc(16); /* excess */
241 if (r != NULL)
242 sprintf(r, "CP%u", cp);
243 }
244 else
245#endif
246 {
247 /*
248 * Locale format on Win32 is <Language>_<Country>.<CodePage> . For
249 * example, English_United States.1252.
250 */
251 codepage = strrchr(ctype, '.');
252 if (codepage != NULL)
253 {
254 int ln;
255
256 codepage++;
257 ln = strlen(codepage);
258 r = malloc(ln + 3);
259 if (r != NULL)
260 sprintf(r, "CP%s", codepage);
261 }
262
263 }
264#endif
265
266 return r;
267}
268
269#ifndef FRONTEND
270/*
271 * Given a Windows code page identifier, find the corresponding PostgreSQL
272 * encoding. Issue a warning and return -1 if none found.
273 */
274int
275pg_codepage_to_encoding(UINT cp)
276{
277 char sys[16];
278 int i;
279
280 sprintf(sys, "CP%u", cp);
281
282 /* Check the table */
283 for (i = 0; encoding_match_list[i].system_enc_name; i++)
284 if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
285 return encoding_match_list[i].pg_enc_code;
286
287 ereport(WARNING,
288 (errmsg("could not determine encoding for codeset \"%s\"", sys)));
289
290 return -1;
291}
292#endif
293#endif /* WIN32 */
294
295#if (defined(HAVE_LANGINFO_H) && defined(CODESET)) || defined(WIN32)
296
297/*
298 * Given a setting for LC_CTYPE, return the Postgres ID of the associated
299 * encoding, if we can determine it. Return -1 if we can't determine it.
300 *
301 * Pass in NULL to get the encoding for the current locale setting.
302 * Pass "" to get the encoding selected by the server's environment.
303 *
304 * If the result is PG_SQL_ASCII, callers should treat it as being compatible
305 * with any desired encoding.
306 *
307 * If running in the backend and write_message is false, this function must
308 * cope with the possibility that elog() and palloc() are not yet usable.
309 */
310int
311pg_get_encoding_from_locale(const char *ctype, bool write_message)
312{
313 char *sys;
314 int i;
315
316 /* Get the CODESET property, and also LC_CTYPE if not passed in */
317 if (ctype)
318 {
319 char *save;
320 char *name;
321
322 /* If locale is C or POSIX, we can allow all encodings */
323 if (pg_strcasecmp(ctype, "C") == 0 ||
324 pg_strcasecmp(ctype, "POSIX") == 0)
325 return PG_SQL_ASCII;
326
327 save = setlocale(LC_CTYPE, NULL);
328 if (!save)
329 return -1; /* setlocale() broken? */
330 /* must copy result, or it might change after setlocale */
331 save = strdup(save);
332 if (!save)
333 return -1; /* out of memory; unlikely */
334
335 name = setlocale(LC_CTYPE, ctype);
336 if (!name)
337 {
338 free(save);
339 return -1; /* bogus ctype passed in? */
340 }
341
342#ifndef WIN32
343 sys = nl_langinfo(CODESET);
344 if (sys)
345 sys = strdup(sys);
346#else
347 sys = win32_langinfo(name);
348#endif
349
350 setlocale(LC_CTYPE, save);
351 free(save);
352 }
353 else
354 {
355 /* much easier... */
356 ctype = setlocale(LC_CTYPE, NULL);
357 if (!ctype)
358 return -1; /* setlocale() broken? */
359
360 /* If locale is C or POSIX, we can allow all encodings */
361 if (pg_strcasecmp(ctype, "C") == 0 ||
362 pg_strcasecmp(ctype, "POSIX") == 0)
363 return PG_SQL_ASCII;
364
365#ifndef WIN32
366 sys = nl_langinfo(CODESET);
367 if (sys)
368 sys = strdup(sys);
369#else
370 sys = win32_langinfo(ctype);
371#endif
372 }
373
374 if (!sys)
375 return -1; /* out of memory; unlikely */
376
377 /* Check the table */
378 for (i = 0; encoding_match_list[i].system_enc_name; i++)
379 {
380 if (pg_strcasecmp(sys, encoding_match_list[i].system_enc_name) == 0)
381 {
382 free(sys);
383 return encoding_match_list[i].pg_enc_code;
384 }
385 }
386
387 /* Special-case kluges for particular platforms go here */
388
389#ifdef __darwin__
390
391 /*
392 * Current macOS has many locales that report an empty string for CODESET,
393 * but they all seem to actually use UTF-8.
394 */
395 if (strlen(sys) == 0)
396 {
397 free(sys);
398 return PG_UTF8;
399 }
400#endif
401
402 /*
403 * We print a warning if we got a CODESET string but couldn't recognize
404 * it. This means we need another entry in the table.
405 */
406 if (write_message)
407 {
408#ifdef FRONTEND
409 fprintf(stderr, _("could not determine encoding for locale \"%s\": codeset is \"%s\""),
410 ctype, sys);
411 /* keep newline separate so there's only one translatable string */
412 fputc('\n', stderr);
413#else
414 ereport(WARNING,
415 (errmsg("could not determine encoding for locale \"%s\": codeset is \"%s\"",
416 ctype, sys)));
417#endif
418 }
419
420 free(sys);
421 return -1;
422}
423#else /* (HAVE_LANGINFO_H && CODESET) || WIN32 */
424
425/*
426 * stub if no multi-language platform support
427 *
428 * Note: we could return -1 here, but that would have the effect of
429 * forcing users to specify an encoding to initdb on such platforms.
430 * It seems better to silently default to SQL_ASCII.
431 */
432int
433pg_get_encoding_from_locale(const char *ctype, bool write_message)
434{
435 return PG_SQL_ASCII;
436}
437
438#endif /* (HAVE_LANGINFO_H && CODESET) || WIN32 */
439