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 "unicode/uenum.h" |
15 | #include "cmemory.h" |
16 | #include "ustrfmt.h" |
17 | #include "locutil.h" |
18 | #include "charstr.h" |
19 | #include "ucln_cmn.h" |
20 | #include "uassert.h" |
21 | #include "umutex.h" |
22 | |
23 | // see LocaleUtility::getAvailableLocaleNames |
24 | static icu::UInitOnce LocaleUtilityInitOnce {}; |
25 | static icu::Hashtable * LocaleUtility_cache = nullptr; |
26 | |
27 | #define UNDERSCORE_CHAR ((char16_t)0x005f) |
28 | #define AT_SIGN_CHAR ((char16_t)64) |
29 | #define PERIOD_CHAR ((char16_t)46) |
30 | |
31 | /* |
32 | ****************************************************************** |
33 | */ |
34 | |
35 | /** |
36 | * Release all static memory held by Locale Utility. |
37 | */ |
38 | U_CDECL_BEGIN |
39 | static UBool U_CALLCONV service_cleanup() { |
40 | if (LocaleUtility_cache) { |
41 | delete LocaleUtility_cache; |
42 | LocaleUtility_cache = nullptr; |
43 | } |
44 | return true; |
45 | } |
46 | |
47 | |
48 | static void U_CALLCONV locale_utility_init(UErrorCode &status) { |
49 | using namespace icu; |
50 | U_ASSERT(LocaleUtility_cache == nullptr); |
51 | ucln_common_registerCleanup(UCLN_COMMON_SERVICE, service_cleanup); |
52 | LocaleUtility_cache = new Hashtable(status); |
53 | if (U_FAILURE(status)) { |
54 | delete LocaleUtility_cache; |
55 | LocaleUtility_cache = nullptr; |
56 | return; |
57 | } |
58 | if (LocaleUtility_cache == nullptr) { |
59 | status = U_MEMORY_ALLOCATION_ERROR; |
60 | return; |
61 | } |
62 | LocaleUtility_cache->setValueDeleter(uhash_deleteHashtable); |
63 | } |
64 | |
65 | U_CDECL_END |
66 | |
67 | U_NAMESPACE_BEGIN |
68 | |
69 | UnicodeString& |
70 | LocaleUtility::canonicalLocaleString(const UnicodeString* id, UnicodeString& result) |
71 | { |
72 | if (id == nullptr) { |
73 | result.setToBogus(); |
74 | } else { |
75 | // Fix case only (no other changes) up to the first '@' or '.' or |
76 | // end of string, whichever comes first. In 3.0 I changed this to |
77 | // stop at first '@' or '.'. It used to run out to the end of |
78 | // string. My fix makes the tests pass but is probably |
79 | // structurally incorrect. See below. [alan 3.0] |
80 | |
81 | // TODO: Doug, you might want to revise this... |
82 | result = *id; |
83 | int32_t i = 0; |
84 | int32_t end = result.indexOf(AT_SIGN_CHAR); |
85 | int32_t n = result.indexOf(PERIOD_CHAR); |
86 | if (n >= 0 && n < end) { |
87 | end = n; |
88 | } |
89 | if (end < 0) { |
90 | end = result.length(); |
91 | } |
92 | n = result.indexOf(UNDERSCORE_CHAR); |
93 | if (n < 0) { |
94 | n = end; |
95 | } |
96 | for (; i < n; ++i) { |
97 | char16_t c = result.charAt(i); |
98 | if (c >= 0x0041 && c <= 0x005a) { |
99 | c += 0x20; |
100 | result.setCharAt(i, c); |
101 | } |
102 | } |
103 | for (n = end; i < n; ++i) { |
104 | char16_t c = result.charAt(i); |
105 | if (c >= 0x0061 && c <= 0x007a) { |
106 | c -= 0x20; |
107 | result.setCharAt(i, c); |
108 | } |
109 | } |
110 | } |
111 | return result; |
112 | |
113 | #if 0 |
114 | // This code does a proper full level 2 canonicalization of id. |
115 | // It's nasty to go from char16_t to char to char to char16_t -- but |
116 | // that's what you have to do to use the uloc_canonicalize |
117 | // function on UnicodeStrings. |
118 | |
119 | // I ended up doing the alternate fix (see above) not for |
120 | // performance reasons, although performance will certainly be |
121 | // better, but because doing a full level 2 canonicalization |
122 | // causes some tests to fail. [alan 3.0] |
123 | |
124 | // TODO: Doug, you might want to revisit this... |
125 | result.setToBogus(); |
126 | if (id != 0) { |
127 | int32_t buflen = id->length() + 8; // space for NUL |
128 | char* buf = (char*) uprv_malloc(buflen); |
129 | char* canon = (buf == 0) ? 0 : (char*) uprv_malloc(buflen); |
130 | if (buf != 0 && canon != 0) { |
131 | U_ASSERT(id->extract(0, INT32_MAX, buf, buflen) < buflen); |
132 | UErrorCode ec = U_ZERO_ERROR; |
133 | uloc_canonicalize(buf, canon, buflen, &ec); |
134 | if (U_SUCCESS(ec)) { |
135 | result = UnicodeString(canon); |
136 | } |
137 | } |
138 | uprv_free(buf); |
139 | uprv_free(canon); |
140 | } |
141 | return result; |
142 | #endif |
143 | } |
144 | |
145 | Locale& |
146 | LocaleUtility::initLocaleFromName(const UnicodeString& id, Locale& result) |
147 | { |
148 | enum { BUFLEN = 128 }; // larger than ever needed |
149 | |
150 | if (id.isBogus() || id.length() >= BUFLEN) { |
151 | result.setToBogus(); |
152 | } else { |
153 | /* |
154 | * We need to convert from a UnicodeString to char * in order to |
155 | * create a Locale. |
156 | * |
157 | * Problem: Locale ID strings may contain '@' which is a variant |
158 | * character and cannot be handled by invariant-character conversion. |
159 | * |
160 | * Hack: Since ICU code can handle locale IDs with multiple encodings |
161 | * of '@' (at least for EBCDIC; it's not known to be a problem for |
162 | * ASCII-based systems), |
163 | * we use regular invariant-character conversion for everything else |
164 | * and manually convert U+0040 into a compiler-char-constant '@'. |
165 | * While this compilation-time constant may not match the runtime |
166 | * encoding of '@', it should be one of the encodings which ICU |
167 | * recognizes. |
168 | * |
169 | * There should be only at most one '@' in a locale ID. |
170 | */ |
171 | char buffer[BUFLEN]; |
172 | int32_t prev, i; |
173 | prev = 0; |
174 | for(;;) { |
175 | i = id.indexOf((char16_t)0x40, prev); |
176 | if(i < 0) { |
177 | // no @ between prev and the rest of the string |
178 | id.extract(prev, INT32_MAX, buffer + prev, BUFLEN - prev, US_INV); |
179 | break; // done |
180 | } else { |
181 | // normal invariant-character conversion for text between @s |
182 | id.extract(prev, i - prev, buffer + prev, BUFLEN - prev, US_INV); |
183 | // manually "convert" U+0040 at id[i] into '@' at buffer[i] |
184 | buffer[i] = '@'; |
185 | prev = i + 1; |
186 | } |
187 | } |
188 | result = Locale::createFromName(buffer); |
189 | } |
190 | return result; |
191 | } |
192 | |
193 | UnicodeString& |
194 | LocaleUtility::initNameFromLocale(const Locale& locale, UnicodeString& result) |
195 | { |
196 | if (locale.isBogus()) { |
197 | result.setToBogus(); |
198 | } else { |
199 | result.append(UnicodeString(locale.getName(), -1, US_INV)); |
200 | } |
201 | return result; |
202 | } |
203 | |
204 | const Hashtable* |
205 | LocaleUtility::getAvailableLocaleNames(const UnicodeString& bundleID) |
206 | { |
207 | // LocaleUtility_cache is a hash-of-hashes. The top-level keys |
208 | // are path strings ('bundleID') passed to |
209 | // ures_openAvailableLocales. The top-level values are |
210 | // second-level hashes. The second-level keys are result strings |
211 | // from ures_openAvailableLocales. The second-level values are |
212 | // garbage ((void*)1 or other random pointer). |
213 | |
214 | UErrorCode status = U_ZERO_ERROR; |
215 | umtx_initOnce(LocaleUtilityInitOnce, locale_utility_init, status); |
216 | Hashtable *cache = LocaleUtility_cache; |
217 | if (cache == nullptr) { |
218 | // Catastrophic failure. |
219 | return nullptr; |
220 | } |
221 | |
222 | Hashtable* htp; |
223 | umtx_lock(nullptr); |
224 | htp = (Hashtable*) cache->get(bundleID); |
225 | umtx_unlock(nullptr); |
226 | |
227 | if (htp == nullptr) { |
228 | htp = new Hashtable(status); |
229 | if (htp && U_SUCCESS(status)) { |
230 | CharString cbundleID; |
231 | cbundleID.appendInvariantChars(bundleID, status); |
232 | const char* path = cbundleID.isEmpty() ? nullptr : cbundleID.data(); |
233 | icu::LocalUEnumerationPointer uenum(ures_openAvailableLocales(path, &status)); |
234 | for (;;) { |
235 | const char16_t* id = uenum_unext(uenum.getAlias(), nullptr, &status); |
236 | if (id == nullptr) { |
237 | break; |
238 | } |
239 | htp->put(UnicodeString(id), (void*)htp, status); |
240 | } |
241 | if (U_FAILURE(status)) { |
242 | delete htp; |
243 | return nullptr; |
244 | } |
245 | umtx_lock(nullptr); |
246 | Hashtable *t = static_cast<Hashtable *>(cache->get(bundleID)); |
247 | if (t != nullptr) { |
248 | // Another thread raced through this code, creating the cache entry first. |
249 | // Discard ours and return theirs. |
250 | umtx_unlock(nullptr); |
251 | delete htp; |
252 | htp = t; |
253 | } else { |
254 | cache->put(bundleID, (void*)htp, status); |
255 | umtx_unlock(nullptr); |
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 | |
276 | |