2 *******************************************************************************
3 * Copyright (C) 2002-2014, International Business Machines Corporation and
4 * others. All Rights Reserved.
5 *******************************************************************************
7 #include "unicode/utypes.h"
9 #if !UCONFIG_NO_SERVICE || !UCONFIG_NO_TRANSLITERATION
11 #include "unicode/resbund.h"
20 // see LocaleUtility::getAvailableLocaleNames
21 static icu::UInitOnce LocaleUtilityInitOnce
= U_INITONCE_INITIALIZER
;
22 static icu::Hashtable
* LocaleUtility_cache
= NULL
;
24 #define UNDERSCORE_CHAR ((UChar)0x005f)
25 #define AT_SIGN_CHAR ((UChar)64)
26 #define PERIOD_CHAR ((UChar)46)
29 ******************************************************************
33 * Release all static memory held by Locale Utility.
36 static UBool U_CALLCONV
service_cleanup(void) {
37 if (LocaleUtility_cache
) {
38 delete LocaleUtility_cache
;
39 LocaleUtility_cache
= NULL
;
45 static void U_CALLCONV
locale_utility_init(UErrorCode
&status
) {
47 U_ASSERT(LocaleUtility_cache
== NULL
);
48 ucln_common_registerCleanup(UCLN_COMMON_SERVICE
, service_cleanup
);
49 LocaleUtility_cache
= new Hashtable(status
);
50 if (U_FAILURE(status
)) {
51 delete LocaleUtility_cache
;
52 LocaleUtility_cache
= NULL
;
55 if (LocaleUtility_cache
== NULL
) {
56 status
= U_MEMORY_ALLOCATION_ERROR
;
59 LocaleUtility_cache
->setValueDeleter(uhash_deleteHashtable
);
67 LocaleUtility::canonicalLocaleString(const UnicodeString
* id
, UnicodeString
& result
)
72 // Fix case only (no other changes) up to the first '@' or '.' or
73 // end of string, whichever comes first. In 3.0 I changed this to
74 // stop at first '@' or '.'. It used to run out to the end of
75 // string. My fix makes the tests pass but is probably
76 // structurally incorrect. See below. [alan 3.0]
78 // TODO: Doug, you might want to revise this...
81 int32_t end
= result
.indexOf(AT_SIGN_CHAR
);
82 int32_t n
= result
.indexOf(PERIOD_CHAR
);
83 if (n
>= 0 && n
< end
) {
87 end
= result
.length();
89 n
= result
.indexOf(UNDERSCORE_CHAR
);
94 UChar c
= result
.charAt(i
);
95 if (c
>= 0x0041 && c
<= 0x005a) {
97 result
.setCharAt(i
, c
);
100 for (n
= end
; i
< n
; ++i
) {
101 UChar c
= result
.charAt(i
);
102 if (c
>= 0x0061 && c
<= 0x007a) {
104 result
.setCharAt(i
, c
);
111 // This code does a proper full level 2 canonicalization of id.
112 // It's nasty to go from UChar to char to char to UChar -- but
113 // that's what you have to do to use the uloc_canonicalize
114 // function on UnicodeStrings.
116 // I ended up doing the alternate fix (see above) not for
117 // performance reasons, although performance will certainly be
118 // better, but because doing a full level 2 canonicalization
119 // causes some tests to fail. [alan 3.0]
121 // TODO: Doug, you might want to revisit this...
124 int32_t buflen
= id
->length() + 8; // space for NUL
125 char* buf
= (char*) uprv_malloc(buflen
);
126 char* canon
= (buf
== 0) ? 0 : (char*) uprv_malloc(buflen
);
127 if (buf
!= 0 && canon
!= 0) {
128 U_ASSERT(id
->extract(0, INT32_MAX
, buf
, buflen
) < buflen
);
129 UErrorCode ec
= U_ZERO_ERROR
;
130 uloc_canonicalize(buf
, canon
, buflen
, &ec
);
132 result
= UnicodeString(canon
);
143 LocaleUtility::initLocaleFromName(const UnicodeString
& id
, Locale
& result
)
145 enum { BUFLEN
= 128 }; // larger than ever needed
147 if (id
.isBogus() || id
.length() >= BUFLEN
) {
151 * We need to convert from a UnicodeString to char * in order to
154 * Problem: Locale ID strings may contain '@' which is a variant
155 * character and cannot be handled by invariant-character conversion.
157 * Hack: Since ICU code can handle locale IDs with multiple encodings
158 * of '@' (at least for EBCDIC; it's not known to be a problem for
159 * ASCII-based systems),
160 * we use regular invariant-character conversion for everything else
161 * and manually convert U+0040 into a compiler-char-constant '@'.
162 * While this compilation-time constant may not match the runtime
163 * encoding of '@', it should be one of the encodings which ICU
166 * There should be only at most one '@' in a locale ID.
172 i
= id
.indexOf((UChar
)0x40, prev
);
174 // no @ between prev and the rest of the string
175 id
.extract(prev
, INT32_MAX
, buffer
+ prev
, BUFLEN
- prev
, US_INV
);
178 // normal invariant-character conversion for text between @s
179 id
.extract(prev
, i
- prev
, buffer
+ prev
, BUFLEN
- prev
, US_INV
);
180 // manually "convert" U+0040 at id[i] into '@' at buffer[i]
185 result
= Locale::createFromName(buffer
);
191 LocaleUtility::initNameFromLocale(const Locale
& locale
, UnicodeString
& result
)
193 if (locale
.isBogus()) {
196 result
.append(UnicodeString(locale
.getName(), -1, US_INV
));
202 LocaleUtility::getAvailableLocaleNames(const UnicodeString
& bundleID
)
204 // LocaleUtility_cache is a hash-of-hashes. The top-level keys
205 // are path strings ('bundleID') passed to
206 // ures_openAvailableLocales. The top-level values are
207 // second-level hashes. The second-level keys are result strings
208 // from ures_openAvailableLocales. The second-level values are
209 // garbage ((void*)1 or other random pointer).
211 UErrorCode status
= U_ZERO_ERROR
;
212 umtx_initOnce(LocaleUtilityInitOnce
, locale_utility_init
, status
);
213 Hashtable
*cache
= LocaleUtility_cache
;
215 // Catastrophic failure.
221 htp
= (Hashtable
*) cache
->get(bundleID
);
225 htp
= new Hashtable(status
);
226 if (htp
&& U_SUCCESS(status
)) {
227 CharString cbundleID
;
228 cbundleID
.appendInvariantChars(bundleID
, status
);
229 const char* path
= cbundleID
.isEmpty() ? NULL
: cbundleID
.data();
230 UEnumeration
*uenum
= ures_openAvailableLocales(path
, &status
);
232 const UChar
* id
= uenum_unext(uenum
, NULL
, &status
);
236 htp
->put(UnicodeString(id
), (void*)htp
, status
);
239 if (U_FAILURE(status
)) {
244 Hashtable
*t
= static_cast<Hashtable
*>(cache
->get(bundleID
));
246 // Another thread raced through this code, creating the cache entry first.
247 // Discard ours and return theirs.
252 cache
->put(bundleID
, (void*)htp
, status
);
261 LocaleUtility::isFallbackOf(const UnicodeString
& root
, const UnicodeString
& child
)
263 return child
.indexOf(root
) == 0 &&
264 (child
.length() == root
.length() ||
265 child
.charAt(root
.length()) == UNDERSCORE_CHAR
);
270 /* !UCONFIG_NO_SERVICE */