| 1 |
Add a variant of `locale_charset' that returns its result based solely on |
| 2 |
information from the environment. See |
| 3 |
http://lists.gnu.org/archive/html/guile-devel/2011-11/msg00040.html for the |
| 4 |
rationale. |
| 5 |
|
| 6 |
--- a/lib/localcharset.c |
| 7 |
+++ b/lib/localcharset.c |
| 8 |
@@ -544,3 +544,120 @@ locale_charset (void) |
| 9 |
|
| 10 |
return codeset; |
| 11 |
} |
| 12 |
+ |
| 13 |
+/* A variant of the above, without calls to `setlocale', `nl_langinfo', |
| 14 |
+ etc. */ |
| 15 |
+const char * |
| 16 |
+environ_locale_charset (void) |
| 17 |
+{ |
| 18 |
+ static char buf[2 + 10 + 1]; |
| 19 |
+ const char *codeset, *aliases; |
| 20 |
+ const char *locale = NULL; |
| 21 |
+ |
| 22 |
+ locale = getenv ("LC_ALL"); |
| 23 |
+ if (locale == NULL || locale[0] == '\0') |
| 24 |
+ { |
| 25 |
+ locale = getenv ("LC_CTYPE"); |
| 26 |
+ if (locale == NULL || locale[0] == '\0') |
| 27 |
+ locale = getenv ("LANG"); |
| 28 |
+ } |
| 29 |
+ |
| 30 |
+ if (locale != NULL && locale[0] != '\0') |
| 31 |
+ { |
| 32 |
+ /* If the locale name contains an encoding after the dot, return it. */ |
| 33 |
+ const char *dot = strchr (locale, '.'); |
| 34 |
+ |
| 35 |
+ if (dot != NULL) |
| 36 |
+ { |
| 37 |
+ const char *modifier; |
| 38 |
+ |
| 39 |
+ dot++; |
| 40 |
+ /* Look for the possible @... trailer and remove it, if any. */ |
| 41 |
+ modifier = strchr (dot, '@'); |
| 42 |
+ if (modifier == NULL) |
| 43 |
+ return dot; |
| 44 |
+ if (modifier - dot < sizeof (buf)) |
| 45 |
+ { |
| 46 |
+ memcpy (buf, dot, modifier - dot); |
| 47 |
+ buf [modifier - dot] = '\0'; |
| 48 |
+ return buf; |
| 49 |
+ } |
| 50 |
+ } |
| 51 |
+ else if (strcmp (locale, "C") == 0) |
| 52 |
+ { |
| 53 |
+ strcpy (buf, "ASCII"); |
| 54 |
+ return buf; |
| 55 |
+ } |
| 56 |
+ else |
| 57 |
+ codeset = ""; |
| 58 |
+ } |
| 59 |
+ else |
| 60 |
+ codeset = ""; |
| 61 |
+ |
| 62 |
+ /* Resolve alias. */ |
| 63 |
+ { |
| 64 |
+# ifdef alias_table_defined |
| 65 |
+ /* On some platforms, UTF-8 locales are the most frequently used ones. |
| 66 |
+ Speed up the common case and slow down the less common cases by |
| 67 |
+ testing for this case first. */ |
| 68 |
+# if defined __OpenBSD__ || (defined __APPLE__ && defined __MACH__) || defined __sun || defined __CYGWIN__ |
| 69 |
+ if (strcmp (codeset, "UTF-8") == 0) |
| 70 |
+ goto done_table_lookup; |
| 71 |
+ else |
| 72 |
+# endif |
| 73 |
+ { |
| 74 |
+ const struct table_entry * const table = alias_table; |
| 75 |
+ size_t const table_size = |
| 76 |
+ sizeof (alias_table) / sizeof (struct table_entry); |
| 77 |
+ /* The table is sorted. Perform a binary search. */ |
| 78 |
+ size_t hi = table_size; |
| 79 |
+ size_t lo = 0; |
| 80 |
+ while (lo < hi) |
| 81 |
+ { |
| 82 |
+ /* Invariant: |
| 83 |
+ for i < lo, strcmp (table[i].alias, codeset) < 0, |
| 84 |
+ for i >= hi, strcmp (table[i].alias, codeset) > 0. */ |
| 85 |
+ size_t mid = (hi + lo) >> 1; /* >= lo, < hi */ |
| 86 |
+ int cmp = strcmp (table[mid].alias, codeset); |
| 87 |
+ if (cmp < 0) |
| 88 |
+ lo = mid + 1; |
| 89 |
+ else if (cmp > 0) |
| 90 |
+ hi = mid; |
| 91 |
+ else |
| 92 |
+ { |
| 93 |
+ /* Found an i with |
| 94 |
+ strcmp (table[i].alias, codeset) == 0. */ |
| 95 |
+ codeset = table[mid].canonical; |
| 96 |
+ goto done_table_lookup; |
| 97 |
+ } |
| 98 |
+ } |
| 99 |
+ } |
| 100 |
+ if (0) |
| 101 |
+ done_table_lookup: ; |
| 102 |
+ else |
| 103 |
+# endif |
| 104 |
+ { |
| 105 |
+ /* Did not find it in the table. */ |
| 106 |
+ /* On Mac OS X, all modern locales use the UTF-8 encoding. |
| 107 |
+ BeOS and Haiku have a single locale, and it has UTF-8 encoding. */ |
| 108 |
+# if (defined __APPLE__ && defined __MACH__) || defined __BEOS__ || defined __HAIKU__ |
| 109 |
+ codeset = "UTF-8"; |
| 110 |
+# else |
| 111 |
+ /* Don't return an empty string. GNU libc and GNU libiconv interpret |
| 112 |
+ the empty string as denoting "the locale's character encoding", |
| 113 |
+ thus GNU libiconv would call this function a second time. */ |
| 114 |
+ if (codeset[0] == '\0') |
| 115 |
+ codeset = "ASCII"; |
| 116 |
+# endif |
| 117 |
+ } |
| 118 |
+ } |
| 119 |
+ |
| 120 |
+ /* Don't return an empty string. GNU libc and GNU libiconv interpret |
| 121 |
+ the empty string as denoting "the locale's character encoding", |
| 122 |
+ thus GNU libiconv would call this function a second time. */ |
| 123 |
+ if (codeset[0] == '\0') |
| 124 |
+ /* Default to Latin-1, for backward compatibility with Guile 1.8. */ |
| 125 |
+ codeset = "ISO-8859-1"; |
| 126 |
+ |
| 127 |
+ return codeset; |
| 128 |
+} |