1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/icu/source/common/locutil.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,265 @@
1.4 +/*
1.5 + *******************************************************************************
1.6 + * Copyright (C) 2002-2011, International Business Machines Corporation and
1.7 + * others. All Rights Reserved.
1.8 + *******************************************************************************
1.9 + */
1.10 +#include "unicode/utypes.h"
1.11 +
1.12 +#if !UCONFIG_NO_SERVICE || !UCONFIG_NO_TRANSLITERATION
1.13 +
1.14 +#include "unicode/resbund.h"
1.15 +#include "cmemory.h"
1.16 +#include "ustrfmt.h"
1.17 +#include "locutil.h"
1.18 +#include "charstr.h"
1.19 +#include "ucln_cmn.h"
1.20 +#include "uassert.h"
1.21 +#include "umutex.h"
1.22 +
1.23 +// see LocaleUtility::getAvailableLocaleNames
1.24 +static icu::Hashtable * LocaleUtility_cache = NULL;
1.25 +
1.26 +#define UNDERSCORE_CHAR ((UChar)0x005f)
1.27 +#define AT_SIGN_CHAR ((UChar)64)
1.28 +#define PERIOD_CHAR ((UChar)46)
1.29 +
1.30 +/*
1.31 + ******************************************************************
1.32 + */
1.33 +
1.34 +/**
1.35 + * Release all static memory held by Locale Utility.
1.36 + */
1.37 +U_CDECL_BEGIN
1.38 +static UBool U_CALLCONV service_cleanup(void) {
1.39 + if (LocaleUtility_cache) {
1.40 + delete LocaleUtility_cache;
1.41 + LocaleUtility_cache = NULL;
1.42 + }
1.43 + return TRUE;
1.44 +}
1.45 +U_CDECL_END
1.46 +
1.47 +U_NAMESPACE_BEGIN
1.48 +
1.49 +UnicodeString&
1.50 +LocaleUtility::canonicalLocaleString(const UnicodeString* id, UnicodeString& result)
1.51 +{
1.52 + if (id == NULL) {
1.53 + result.setToBogus();
1.54 + } else {
1.55 + // Fix case only (no other changes) up to the first '@' or '.' or
1.56 + // end of string, whichever comes first. In 3.0 I changed this to
1.57 + // stop at first '@' or '.'. It used to run out to the end of
1.58 + // string. My fix makes the tests pass but is probably
1.59 + // structurally incorrect. See below. [alan 3.0]
1.60 +
1.61 + // TODO: Doug, you might want to revise this...
1.62 + result = *id;
1.63 + int32_t i = 0;
1.64 + int32_t end = result.indexOf(AT_SIGN_CHAR);
1.65 + int32_t n = result.indexOf(PERIOD_CHAR);
1.66 + if (n >= 0 && n < end) {
1.67 + end = n;
1.68 + }
1.69 + if (end < 0) {
1.70 + end = result.length();
1.71 + }
1.72 + n = result.indexOf(UNDERSCORE_CHAR);
1.73 + if (n < 0) {
1.74 + n = end;
1.75 + }
1.76 + for (; i < n; ++i) {
1.77 + UChar c = result.charAt(i);
1.78 + if (c >= 0x0041 && c <= 0x005a) {
1.79 + c += 0x20;
1.80 + result.setCharAt(i, c);
1.81 + }
1.82 + }
1.83 + for (n = end; i < n; ++i) {
1.84 + UChar c = result.charAt(i);
1.85 + if (c >= 0x0061 && c <= 0x007a) {
1.86 + c -= 0x20;
1.87 + result.setCharAt(i, c);
1.88 + }
1.89 + }
1.90 + }
1.91 + return result;
1.92 +
1.93 +#if 0
1.94 + // This code does a proper full level 2 canonicalization of id.
1.95 + // It's nasty to go from UChar to char to char to UChar -- but
1.96 + // that's what you have to do to use the uloc_canonicalize
1.97 + // function on UnicodeStrings.
1.98 +
1.99 + // I ended up doing the alternate fix (see above) not for
1.100 + // performance reasons, although performance will certainly be
1.101 + // better, but because doing a full level 2 canonicalization
1.102 + // causes some tests to fail. [alan 3.0]
1.103 +
1.104 + // TODO: Doug, you might want to revisit this...
1.105 + result.setToBogus();
1.106 + if (id != 0) {
1.107 + int32_t buflen = id->length() + 8; // space for NUL
1.108 + char* buf = (char*) uprv_malloc(buflen);
1.109 + char* canon = (buf == 0) ? 0 : (char*) uprv_malloc(buflen);
1.110 + if (buf != 0 && canon != 0) {
1.111 + U_ASSERT(id->extract(0, INT32_MAX, buf, buflen) < buflen);
1.112 + UErrorCode ec = U_ZERO_ERROR;
1.113 + uloc_canonicalize(buf, canon, buflen, &ec);
1.114 + if (U_SUCCESS(ec)) {
1.115 + result = UnicodeString(canon);
1.116 + }
1.117 + }
1.118 + uprv_free(buf);
1.119 + uprv_free(canon);
1.120 + }
1.121 + return result;
1.122 +#endif
1.123 +}
1.124 +
1.125 +Locale&
1.126 +LocaleUtility::initLocaleFromName(const UnicodeString& id, Locale& result)
1.127 +{
1.128 + enum { BUFLEN = 128 }; // larger than ever needed
1.129 +
1.130 + if (id.isBogus() || id.length() >= BUFLEN) {
1.131 + result.setToBogus();
1.132 + } else {
1.133 + /*
1.134 + * We need to convert from a UnicodeString to char * in order to
1.135 + * create a Locale.
1.136 + *
1.137 + * Problem: Locale ID strings may contain '@' which is a variant
1.138 + * character and cannot be handled by invariant-character conversion.
1.139 + *
1.140 + * Hack: Since ICU code can handle locale IDs with multiple encodings
1.141 + * of '@' (at least for EBCDIC; it's not known to be a problem for
1.142 + * ASCII-based systems),
1.143 + * we use regular invariant-character conversion for everything else
1.144 + * and manually convert U+0040 into a compiler-char-constant '@'.
1.145 + * While this compilation-time constant may not match the runtime
1.146 + * encoding of '@', it should be one of the encodings which ICU
1.147 + * recognizes.
1.148 + *
1.149 + * There should be only at most one '@' in a locale ID.
1.150 + */
1.151 + char buffer[BUFLEN];
1.152 + int32_t prev, i;
1.153 + prev = 0;
1.154 + for(;;) {
1.155 + i = id.indexOf((UChar)0x40, prev);
1.156 + if(i < 0) {
1.157 + // no @ between prev and the rest of the string
1.158 + id.extract(prev, INT32_MAX, buffer + prev, BUFLEN - prev, US_INV);
1.159 + break; // done
1.160 + } else {
1.161 + // normal invariant-character conversion for text between @s
1.162 + id.extract(prev, i - prev, buffer + prev, BUFLEN - prev, US_INV);
1.163 + // manually "convert" U+0040 at id[i] into '@' at buffer[i]
1.164 + buffer[i] = '@';
1.165 + prev = i + 1;
1.166 + }
1.167 + }
1.168 + result = Locale::createFromName(buffer);
1.169 + }
1.170 + return result;
1.171 +}
1.172 +
1.173 +UnicodeString&
1.174 +LocaleUtility::initNameFromLocale(const Locale& locale, UnicodeString& result)
1.175 +{
1.176 + if (locale.isBogus()) {
1.177 + result.setToBogus();
1.178 + } else {
1.179 + result.append(UnicodeString(locale.getName(), -1, US_INV));
1.180 + }
1.181 + return result;
1.182 +}
1.183 +
1.184 +const Hashtable*
1.185 +LocaleUtility::getAvailableLocaleNames(const UnicodeString& bundleID)
1.186 +{
1.187 + // LocaleUtility_cache is a hash-of-hashes. The top-level keys
1.188 + // are path strings ('bundleID') passed to
1.189 + // ures_openAvailableLocales. The top-level values are
1.190 + // second-level hashes. The second-level keys are result strings
1.191 + // from ures_openAvailableLocales. The second-level values are
1.192 + // garbage ((void*)1 or other random pointer).
1.193 +
1.194 + UErrorCode status = U_ZERO_ERROR;
1.195 + Hashtable* cache;
1.196 + umtx_lock(NULL);
1.197 + cache = LocaleUtility_cache;
1.198 + umtx_unlock(NULL);
1.199 +
1.200 + if (cache == NULL) {
1.201 + cache = new Hashtable(status);
1.202 + if (cache == NULL || U_FAILURE(status)) {
1.203 + return NULL; // catastrophic failure; e.g. out of memory
1.204 + }
1.205 + cache->setValueDeleter(uhash_deleteHashtable);
1.206 + Hashtable* h; // set this to final LocaleUtility_cache value
1.207 + umtx_lock(NULL);
1.208 + h = LocaleUtility_cache;
1.209 + if (h == NULL) {
1.210 + LocaleUtility_cache = h = cache;
1.211 + cache = NULL;
1.212 + ucln_common_registerCleanup(UCLN_COMMON_SERVICE, service_cleanup);
1.213 + }
1.214 + umtx_unlock(NULL);
1.215 + if(cache != NULL) {
1.216 + delete cache;
1.217 + }
1.218 + cache = h;
1.219 + }
1.220 +
1.221 + U_ASSERT(cache != NULL);
1.222 +
1.223 + Hashtable* htp;
1.224 + umtx_lock(NULL);
1.225 + htp = (Hashtable*) cache->get(bundleID);
1.226 + umtx_unlock(NULL);
1.227 +
1.228 + if (htp == NULL) {
1.229 + htp = new Hashtable(status);
1.230 + if (htp && U_SUCCESS(status)) {
1.231 + CharString cbundleID;
1.232 + cbundleID.appendInvariantChars(bundleID, status);
1.233 + const char* path = cbundleID.isEmpty() ? NULL : cbundleID.data();
1.234 + UEnumeration *uenum = ures_openAvailableLocales(path, &status);
1.235 + for (;;) {
1.236 + const UChar* id = uenum_unext(uenum, NULL, &status);
1.237 + if (id == NULL) {
1.238 + break;
1.239 + }
1.240 + htp->put(UnicodeString(id), (void*)htp, status);
1.241 + }
1.242 + uenum_close(uenum);
1.243 + if (U_FAILURE(status)) {
1.244 + delete htp;
1.245 + return NULL;
1.246 + }
1.247 + umtx_lock(NULL);
1.248 + cache->put(bundleID, (void*)htp, status);
1.249 + umtx_unlock(NULL);
1.250 + }
1.251 + }
1.252 + return htp;
1.253 +}
1.254 +
1.255 +UBool
1.256 +LocaleUtility::isFallbackOf(const UnicodeString& root, const UnicodeString& child)
1.257 +{
1.258 + return child.indexOf(root) == 0 &&
1.259 + (child.length() == root.length() ||
1.260 + child.charAt(root.length()) == UNDERSCORE_CHAR);
1.261 +}
1.262 +
1.263 +U_NAMESPACE_END
1.264 +
1.265 +/* !UCONFIG_NO_SERVICE */
1.266 +#endif
1.267 +
1.268 +
