icuSources/common/locutil.cpp

   1 // © 2016 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 /*
   4  *******************************************************************************
   5  * Copyright (C) 2002-2014, International Business Machines Corporation and
   6  * others. All Rights Reserved.
   7  *******************************************************************************
   8  */
   9 #include "unicode/utypes.h"
  10
  11 #if !UCONFIG_NO_SERVICE || !UCONFIG_NO_TRANSLITERATION
  12
  13 #include "unicode/resbund.h"
  14 #include "cmemory.h"
  15 #include "ustrfmt.h"
  16 #include "locutil.h"
  17 #include "charstr.h"
  18 #include "ucln_cmn.h"
  19 #include "uassert.h"
  20 #include "umutex.h"
  21
  22 // see LocaleUtility::getAvailableLocaleNames
  23 static icu::UInitOnce   LocaleUtilityInitOnce = U_INITONCE_INITIALIZER;
  24 static icu::Hashtable * LocaleUtility_cache = NULL;
  25
  26 #define UNDERSCORE_CHAR ((UChar)0x005f)
  27 #define AT_SIGN_CHAR    ((UChar)64)
  28 #define PERIOD_CHAR     ((UChar)46)
  29
  30 /*
  31  ******************************************************************
  32  */
  33
  34 /**
  35  * Release all static memory held by Locale Utility.
  36  */
  37 U_CDECL_BEGIN
  38 static UBool U_CALLCONV service_cleanup(void) {
  39     if (LocaleUtility_cache) {
  40         delete LocaleUtility_cache;
  41         LocaleUtility_cache = NULL;
  42     }
  43     return TRUE;
  44 }
  45
  46
  47 static void U_CALLCONV locale_utility_init(UErrorCode &status) {
  48     using namespace icu;
  49     U_ASSERT(LocaleUtility_cache == NULL);
  50     ucln_common_registerCleanup(UCLN_COMMON_SERVICE, service_cleanup);
  51     LocaleUtility_cache = new Hashtable(status);
  52     if (U_FAILURE(status)) {
  53         delete LocaleUtility_cache;
  54         LocaleUtility_cache = NULL;
  55         return;
  56     }
  57     if (LocaleUtility_cache == NULL) {
  58         status = U_MEMORY_ALLOCATION_ERROR;
  59         return;
  60     }
  61     LocaleUtility_cache->setValueDeleter(uhash_deleteHashtable);
  62 }
  63
  64 U_CDECL_END
  65
  66 U_NAMESPACE_BEGIN
  67
  68 UnicodeString&
  69 LocaleUtility::canonicalLocaleString(const UnicodeString* id, UnicodeString& result)
  70 {
  71   if (id == NULL) {
  72     result.setToBogus();
  73   } else {
  74     // Fix case only (no other changes) up to the first '@' or '.' or
  75     // end of string, whichever comes first.  In 3.0 I changed this to
  76     // stop at first '@' or '.'.  It used to run out to the end of
  77     // string.  My fix makes the tests pass but is probably
  78     // structurally incorrect.  See below.  [alan 3.0]
  79
  80     // TODO: Doug, you might want to revise this...
  81     result = *id;
  82     int32_t i = 0;
  83     int32_t end = result.indexOf(AT_SIGN_CHAR);
  84     int32_t n = result.indexOf(PERIOD_CHAR);
  85     if (n >= 0 && n < end) {
  86         end = n;
  87     }
  88     if (end < 0) {
  89         end = result.length();
  90     }
  91     n = result.indexOf(UNDERSCORE_CHAR);
  92     if (n < 0) {
  93       n = end;
  94     }
  95     for (; i < n; ++i) {
  96       UChar c = result.charAt(i);
  97       if (c >= 0x0041 && c <= 0x005a) {
  98         c += 0x20;
  99         result.setCharAt(i, c);
 100       }
 101     }
 102     for (n = end; i < n; ++i) {
 103       UChar c = result.charAt(i);
 104       if (c >= 0x0061 && c <= 0x007a) {
 105         c -= 0x20;
 106         result.setCharAt(i, c);
 107       }
 108     }
 109   }
 110   return result;
 111
 112 #if 0
 113     // This code does a proper full level 2 canonicalization of id.
 114     // It's nasty to go from UChar to char to char to UChar -- but
 115     // that's what you have to do to use the uloc_canonicalize
 116     // function on UnicodeStrings.
 117
 118     // I ended up doing the alternate fix (see above) not for
 119     // performance reasons, although performance will certainly be
 120     // better, but because doing a full level 2 canonicalization
 121     // causes some tests to fail.  [alan 3.0]
 122
 123     // TODO: Doug, you might want to revisit this...
 124     result.setToBogus();
 125     if (id != 0) {
 126         int32_t buflen = id->length() + 8; // space for NUL
 127         char* buf = (char*) uprv_malloc(buflen);
 128         char* canon = (buf == 0) ? 0 : (char*) uprv_malloc(buflen);
 129         if (buf != 0 && canon != 0) {
 130             U_ASSERT(id->extract(0, INT32_MAX, buf, buflen) < buflen);
 131             UErrorCode ec = U_ZERO_ERROR;
 132             uloc_canonicalize(buf, canon, buflen, &ec);
 133             if (U_SUCCESS(ec)) {
 134                 result = UnicodeString(canon);
 135             }
 136         }
 137         uprv_free(buf);
 138         uprv_free(canon);
 139     }
 140     return result;
 141 #endif
 142 }
 143
 144 Locale&
 145 LocaleUtility::initLocaleFromName(const UnicodeString& id, Locale& result)
 146 {
 147     enum { BUFLEN = 128 }; // larger than ever needed
 148
 149     if (id.isBogus() || id.length() >= BUFLEN) {
 150         result.setToBogus();
 151     } else {
 152         /*
 153          * We need to convert from a UnicodeString to char * in order to
 154          * create a Locale.
 155          *
 156          * Problem: Locale ID strings may contain '@' which is a variant
 157          * character and cannot be handled by invariant-character conversion.
 158          *
 159          * Hack: Since ICU code can handle locale IDs with multiple encodings
 160          * of '@' (at least for EBCDIC; it's not known to be a problem for
 161          * ASCII-based systems),
 162          * we use regular invariant-character conversion for everything else
 163          * and manually convert U+0040 into a compiler-char-constant '@'.
 164          * While this compilation-time constant may not match the runtime
 165          * encoding of '@', it should be one of the encodings which ICU
 166          * recognizes.
 167          *
 168          * There should be only at most one '@' in a locale ID.
 169          */
 170         char buffer[BUFLEN];
 171         int32_t prev, i;
 172         prev = 0;
 173         for(;;) {
 174             i = id.indexOf((UChar)0x40, prev);
 175             if(i < 0) {
 176                 // no @ between prev and the rest of the string
 177                 id.extract(prev, INT32_MAX, buffer + prev, BUFLEN - prev, US_INV);
 178                 break; // done
 179             } else {
 180                 // normal invariant-character conversion for text between @s
 181                 id.extract(prev, i - prev, buffer + prev, BUFLEN - prev, US_INV);
 182                 // manually "convert" U+0040 at id[i] into '@' at buffer[i]
 183                 buffer[i] = '@';
 184                 prev = i + 1;
 185             }
 186         }
 187         result = Locale::createFromName(buffer);
 188     }
 189     return result;
 190 }
 191
 192 UnicodeString&
 193 LocaleUtility::initNameFromLocale(const Locale& locale, UnicodeString& result)
 194 {
 195     if (locale.isBogus()) {
 196         result.setToBogus();
 197     } else {
 198         result.append(UnicodeString(locale.getName(), -1, US_INV));
 199     }
 200     return result;
 201 }
 202
 203 const Hashtable*
 204 LocaleUtility::getAvailableLocaleNames(const UnicodeString& bundleID)
 205 {
 206     // LocaleUtility_cache is a hash-of-hashes.  The top-level keys
 207     // are path strings ('bundleID') passed to
 208     // ures_openAvailableLocales.  The top-level values are
 209     // second-level hashes.  The second-level keys are result strings
 210     // from ures_openAvailableLocales.  The second-level values are
 211     // garbage ((void*)1 or other random pointer).
 212
 213     UErrorCode status = U_ZERO_ERROR;
 214     umtx_initOnce(LocaleUtilityInitOnce, locale_utility_init, status);
 215     Hashtable *cache = LocaleUtility_cache;
 216     if (cache == NULL) {
 217         // Catastrophic failure.
 218         return NULL;
 219     }
 220
 221     Hashtable* htp;
 222     umtx_lock(NULL);
 223     htp = (Hashtable*) cache->get(bundleID);
 224     umtx_unlock(NULL);
 225
 226     if (htp == NULL) {
 227         htp = new Hashtable(status);
 228         if (htp && U_SUCCESS(status)) {
 229             CharString cbundleID;
 230             cbundleID.appendInvariantChars(bundleID, status);
 231             const char* path = cbundleID.isEmpty() ? NULL : cbundleID.data();
 232             UEnumeration *uenum = ures_openAvailableLocales(path, &status);
 233             for (;;) {
 234                 const UChar* id = uenum_unext(uenum, NULL, &status);
 235                 if (id == NULL) {
 236                     break;
 237                 }
 238                 htp->put(UnicodeString(id), (void*)htp, status);
 239             }
 240             uenum_close(uenum);
 241             if (U_FAILURE(status)) {
 242                 delete htp;
 243                 return NULL;
 244             }
 245             umtx_lock(NULL);
 246             Hashtable *t = static_cast<Hashtable *>(cache->get(bundleID));
 247             if (t != NULL) {
 248                 // Another thread raced through this code, creating the cache entry first.
 249                 // Discard ours and return theirs.
 250                 umtx_unlock(NULL);
 251                 delete htp;
 252                 htp = t;
 253             } else {
 254                 cache->put(bundleID, (void*)htp, status);
 255                 umtx_unlock(NULL);
 256             }
 257         }
 258     }
 259     return htp;
 260 }
 261
 262 UBool
 263 LocaleUtility::isFallbackOf(const UnicodeString& root, const UnicodeString& child)
 264 {
 265     return child.indexOf(root) == 0 &&
 266       (child.length() == root.length() ||
 267        child.charAt(root.length()) == UNDERSCORE_CHAR);
 268 }
 269
 270 U_NAMESPACE_END
 271
 272 /* !UCONFIG_NO_SERVICE */
 273 #endif
 274
 275