|
1 /* |
|
2 ****************************************************************************** |
|
3 * |
|
4 * Copyright (C) 1999-2013, International Business Machines |
|
5 * Corporation and others. All Rights Reserved. |
|
6 * |
|
7 ****************************************************************************** |
|
8 * file name: udata.cpp |
|
9 * encoding: US-ASCII |
|
10 * tab size: 8 (not used) |
|
11 * indentation:4 |
|
12 * |
|
13 * created on: 1999oct25 |
|
14 * created by: Markus W. Scherer |
|
15 */ |
|
16 |
|
17 #include "unicode/utypes.h" /* U_PLATFORM etc. */ |
|
18 |
|
19 #ifdef __GNUC__ |
|
20 /* if gcc |
|
21 #define ATTRIBUTE_WEAK __attribute__ ((weak)) |
|
22 might have to #include some other header |
|
23 */ |
|
24 #endif |
|
25 |
|
26 #include "unicode/putil.h" |
|
27 #include "unicode/udata.h" |
|
28 #include "unicode/uversion.h" |
|
29 #include "charstr.h" |
|
30 #include "cmemory.h" |
|
31 #include "cstring.h" |
|
32 #include "putilimp.h" |
|
33 #include "uassert.h" |
|
34 #include "ucln_cmn.h" |
|
35 #include "ucmndata.h" |
|
36 #include "udatamem.h" |
|
37 #include "uhash.h" |
|
38 #include "umapfile.h" |
|
39 #include "umutex.h" |
|
40 |
|
41 /*********************************************************************** |
|
42 * |
|
43 * Notes on the organization of the ICU data implementation |
|
44 * |
|
45 * All of the public API is defined in udata.h |
|
46 * |
|
47 * The implementation is split into several files... |
|
48 * |
|
49 * - udata.c (this file) contains higher level code that knows about |
|
50 * the search paths for locating data, caching opened data, etc. |
|
51 * |
|
52 * - umapfile.c contains the low level platform-specific code for actually loading |
|
53 * (memory mapping, file reading, whatever) data into memory. |
|
54 * |
|
55 * - ucmndata.c deals with the tables of contents of ICU data items within |
|
56 * an ICU common format data file. The implementation includes |
|
57 * an abstract interface and support for multiple TOC formats. |
|
58 * All knowledge of any specific TOC format is encapsulated here. |
|
59 * |
|
60 * - udatamem.c has code for managing UDataMemory structs. These are little |
|
61 * descriptor objects for blocks of memory holding ICU data of |
|
62 * various types. |
|
63 */ |
|
64 |
|
65 /* configuration ---------------------------------------------------------- */ |
|
66 |
|
67 /* If you are excruciatingly bored turn this on .. */ |
|
68 /* #define UDATA_DEBUG 1 */ |
|
69 |
|
70 #if defined(UDATA_DEBUG) |
|
71 # include <stdio.h> |
|
72 #endif |
|
73 |
|
74 #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0])) |
|
75 |
|
76 U_NAMESPACE_USE |
|
77 |
|
78 /* |
|
79 * Forward declarations |
|
80 */ |
|
81 static UDataMemory *udata_findCachedData(const char *path); |
|
82 |
|
83 /*********************************************************************** |
|
84 * |
|
85 * static (Global) data |
|
86 * |
|
87 ************************************************************************/ |
|
88 |
|
89 /* |
|
90 * Pointers to the common ICU data. |
|
91 * |
|
92 * We store multiple pointers to ICU data packages and iterate through them |
|
93 * when looking for a data item. |
|
94 * |
|
95 * It is possible to combine this with dependency inversion: |
|
96 * One or more data package libraries may export |
|
97 * functions that each return a pointer to their piece of the ICU data, |
|
98 * and this file would import them as weak functions, without a |
|
99 * strong linker dependency from the common library on the data library. |
|
100 * |
|
101 * Then we can have applications depend on only that part of ICU's data |
|
102 * that they really need, reducing the size of binaries that take advantage |
|
103 * of this. |
|
104 */ |
|
105 static UDataMemory *gCommonICUDataArray[10] = { NULL }; |
|
106 |
|
107 static UBool gHaveTriedToLoadCommonData = FALSE; /* See extendICUData(). */ |
|
108 |
|
109 static UHashtable *gCommonDataCache = NULL; /* Global hash table of opened ICU data files. */ |
|
110 static icu::UInitOnce gCommonDataCacheInitOnce = U_INITONCE_INITIALIZER; |
|
111 |
|
112 static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; |
|
113 |
|
114 static UBool U_CALLCONV |
|
115 udata_cleanup(void) |
|
116 { |
|
117 int32_t i; |
|
118 |
|
119 if (gCommonDataCache) { /* Delete the cache of user data mappings. */ |
|
120 uhash_close(gCommonDataCache); /* Table owns the contents, and will delete them. */ |
|
121 gCommonDataCache = NULL; /* Cleanup is not thread safe. */ |
|
122 } |
|
123 gCommonDataCacheInitOnce.reset(); |
|
124 |
|
125 for (i = 0; i < LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != NULL; ++i) { |
|
126 udata_close(gCommonICUDataArray[i]); |
|
127 gCommonICUDataArray[i] = NULL; |
|
128 } |
|
129 gHaveTriedToLoadCommonData = FALSE; |
|
130 |
|
131 return TRUE; /* Everything was cleaned up */ |
|
132 } |
|
133 |
|
134 static UBool U_CALLCONV |
|
135 findCommonICUDataByName(const char *inBasename) |
|
136 { |
|
137 UBool found = FALSE; |
|
138 int32_t i; |
|
139 |
|
140 UDataMemory *pData = udata_findCachedData(inBasename); |
|
141 if (pData == NULL) |
|
142 return FALSE; |
|
143 |
|
144 for (i = 0; i < LENGTHOF(gCommonICUDataArray); ++i) { |
|
145 if ((gCommonICUDataArray[i] != NULL) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) { |
|
146 /* The data pointer is already in the array. */ |
|
147 found = TRUE; |
|
148 break; |
|
149 } |
|
150 } |
|
151 |
|
152 return found; |
|
153 } |
|
154 |
|
155 |
|
156 /* |
|
157 * setCommonICUData. Set a UDataMemory to be the global ICU Data |
|
158 */ |
|
159 static UBool |
|
160 setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to caller, we copy it. */ |
|
161 UBool warn, /* If true, set USING_DEFAULT warning if ICUData was */ |
|
162 /* changed by another thread before we got to it. */ |
|
163 UErrorCode *pErr) |
|
164 { |
|
165 UDataMemory *newCommonData = UDataMemory_createNewInstance(pErr); |
|
166 int32_t i; |
|
167 UBool didUpdate = FALSE; |
|
168 if (U_FAILURE(*pErr)) { |
|
169 return FALSE; |
|
170 } |
|
171 |
|
172 /* For the assignment, other threads must cleanly see either the old */ |
|
173 /* or the new, not some partially initialized new. The old can not be */ |
|
174 /* deleted - someone may still have a pointer to it lying around in */ |
|
175 /* their locals. */ |
|
176 UDatamemory_assign(newCommonData, pData); |
|
177 umtx_lock(NULL); |
|
178 for (i = 0; i < LENGTHOF(gCommonICUDataArray); ++i) { |
|
179 if (gCommonICUDataArray[i] == NULL) { |
|
180 gCommonICUDataArray[i] = newCommonData; |
|
181 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
|
182 didUpdate = TRUE; |
|
183 break; |
|
184 } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) { |
|
185 /* The same data pointer is already in the array. */ |
|
186 break; |
|
187 } |
|
188 } |
|
189 umtx_unlock(NULL); |
|
190 |
|
191 if (i == LENGTHOF(gCommonICUDataArray) && warn) { |
|
192 *pErr = U_USING_DEFAULT_WARNING; |
|
193 } |
|
194 if (!didUpdate) { |
|
195 uprv_free(newCommonData); |
|
196 } |
|
197 return didUpdate; |
|
198 } |
|
199 |
|
200 static UBool |
|
201 setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) { |
|
202 UDataMemory tData; |
|
203 UDataMemory_init(&tData); |
|
204 UDataMemory_setData(&tData, pData); |
|
205 udata_checkCommonData(&tData, pErrorCode); |
|
206 return setCommonICUData(&tData, FALSE, pErrorCode); |
|
207 } |
|
208 |
|
209 static const char * |
|
210 findBasename(const char *path) { |
|
211 const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR); |
|
212 if(basename==NULL) { |
|
213 return path; |
|
214 } else { |
|
215 return basename+1; |
|
216 } |
|
217 } |
|
218 |
|
219 #ifdef UDATA_DEBUG |
|
220 static const char * |
|
221 packageNameFromPath(const char *path) |
|
222 { |
|
223 if((path == NULL) || (*path == 0)) { |
|
224 return U_ICUDATA_NAME; |
|
225 } |
|
226 |
|
227 path = findBasename(path); |
|
228 |
|
229 if((path == NULL) || (*path == 0)) { |
|
230 return U_ICUDATA_NAME; |
|
231 } |
|
232 |
|
233 return path; |
|
234 } |
|
235 #endif |
|
236 |
|
237 /*----------------------------------------------------------------------* |
|
238 * * |
|
239 * Cache for common data * |
|
240 * Functions for looking up or adding entries to a cache of * |
|
241 * data that has been previously opened. Avoids a potentially * |
|
242 * expensive operation of re-opening the data for subsequent * |
|
243 * uses. * |
|
244 * * |
|
245 * Data remains cached for the duration of the process. * |
|
246 * * |
|
247 *----------------------------------------------------------------------*/ |
|
248 |
|
249 typedef struct DataCacheElement { |
|
250 char *name; |
|
251 UDataMemory *item; |
|
252 } DataCacheElement; |
|
253 |
|
254 |
|
255 |
|
256 /* |
|
257 * Deleter function for DataCacheElements. |
|
258 * udata cleanup function closes the hash table; hash table in turn calls back to |
|
259 * here for each entry. |
|
260 */ |
|
261 static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) { |
|
262 DataCacheElement *p = (DataCacheElement *)pDCEl; |
|
263 udata_close(p->item); /* unmaps storage */ |
|
264 uprv_free(p->name); /* delete the hash key string. */ |
|
265 uprv_free(pDCEl); /* delete 'this' */ |
|
266 } |
|
267 |
|
268 static void udata_initHashTable() { |
|
269 UErrorCode err = U_ZERO_ERROR; |
|
270 U_ASSERT(gCommonDataCache == NULL); |
|
271 gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err); |
|
272 if (U_FAILURE(err)) { |
|
273 // TODO: handle errors better. |
|
274 gCommonDataCache = NULL; |
|
275 } |
|
276 if (gCommonDataCache != NULL) { |
|
277 uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter); |
|
278 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
|
279 } |
|
280 } |
|
281 |
|
282 /* udata_getCacheHashTable() |
|
283 * Get the hash table used to store the data cache entries. |
|
284 * Lazy create it if it doesn't yet exist. |
|
285 */ |
|
286 static UHashtable *udata_getHashTable() { |
|
287 umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable); |
|
288 return gCommonDataCache; |
|
289 } |
|
290 |
|
291 |
|
292 |
|
293 static UDataMemory *udata_findCachedData(const char *path) |
|
294 { |
|
295 UHashtable *htable; |
|
296 UDataMemory *retVal = NULL; |
|
297 DataCacheElement *el; |
|
298 const char *baseName; |
|
299 |
|
300 baseName = findBasename(path); /* Cache remembers only the base name, not the full path. */ |
|
301 htable = udata_getHashTable(); |
|
302 umtx_lock(NULL); |
|
303 el = (DataCacheElement *)uhash_get(htable, baseName); |
|
304 umtx_unlock(NULL); |
|
305 if (el != NULL) { |
|
306 retVal = el->item; |
|
307 } |
|
308 #ifdef UDATA_DEBUG |
|
309 fprintf(stderr, "Cache: [%s] -> %p\n", baseName, retVal); |
|
310 #endif |
|
311 return retVal; |
|
312 } |
|
313 |
|
314 |
|
315 static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UErrorCode *pErr) { |
|
316 DataCacheElement *newElement; |
|
317 const char *baseName; |
|
318 int32_t nameLen; |
|
319 UHashtable *htable; |
|
320 DataCacheElement *oldValue = NULL; |
|
321 UErrorCode subErr = U_ZERO_ERROR; |
|
322 |
|
323 if (U_FAILURE(*pErr)) { |
|
324 return NULL; |
|
325 } |
|
326 |
|
327 /* Create a new DataCacheElement - the thingy we store in the hash table - |
|
328 * and copy the supplied path and UDataMemoryItems into it. |
|
329 */ |
|
330 newElement = (DataCacheElement *)uprv_malloc(sizeof(DataCacheElement)); |
|
331 if (newElement == NULL) { |
|
332 *pErr = U_MEMORY_ALLOCATION_ERROR; |
|
333 return NULL; |
|
334 } |
|
335 newElement->item = UDataMemory_createNewInstance(pErr); |
|
336 if (U_FAILURE(*pErr)) { |
|
337 uprv_free(newElement); |
|
338 return NULL; |
|
339 } |
|
340 UDatamemory_assign(newElement->item, item); |
|
341 |
|
342 baseName = findBasename(path); |
|
343 nameLen = (int32_t)uprv_strlen(baseName); |
|
344 newElement->name = (char *)uprv_malloc(nameLen+1); |
|
345 if (newElement->name == NULL) { |
|
346 *pErr = U_MEMORY_ALLOCATION_ERROR; |
|
347 uprv_free(newElement->item); |
|
348 uprv_free(newElement); |
|
349 return NULL; |
|
350 } |
|
351 uprv_strcpy(newElement->name, baseName); |
|
352 |
|
353 /* Stick the new DataCacheElement into the hash table. |
|
354 */ |
|
355 htable = udata_getHashTable(); |
|
356 umtx_lock(NULL); |
|
357 oldValue = (DataCacheElement *)uhash_get(htable, path); |
|
358 if (oldValue != NULL) { |
|
359 subErr = U_USING_DEFAULT_WARNING; |
|
360 } |
|
361 else { |
|
362 uhash_put( |
|
363 htable, |
|
364 newElement->name, /* Key */ |
|
365 newElement, /* Value */ |
|
366 &subErr); |
|
367 } |
|
368 umtx_unlock(NULL); |
|
369 |
|
370 #ifdef UDATA_DEBUG |
|
371 fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n", newElement->name, |
|
372 newElement->item, u_errorName(subErr), newElement->item->vFuncs); |
|
373 #endif |
|
374 |
|
375 if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) { |
|
376 *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. */ |
|
377 uprv_free(newElement->name); |
|
378 uprv_free(newElement->item); |
|
379 uprv_free(newElement); |
|
380 return oldValue ? oldValue->item : NULL; |
|
381 } |
|
382 |
|
383 return newElement->item; |
|
384 } |
|
385 |
|
386 /*----------------------------------------------------------------------*============== |
|
387 * * |
|
388 * Path management. Could be shared with other tools/etc if need be * |
|
389 * later on. * |
|
390 * * |
|
391 *----------------------------------------------------------------------*/ |
|
392 |
|
393 #define U_DATA_PATHITER_BUFSIZ 128 /* Size of local buffer for paths */ |
|
394 /* Overflow causes malloc of larger buf */ |
|
395 |
|
396 U_NAMESPACE_BEGIN |
|
397 |
|
398 class UDataPathIterator |
|
399 { |
|
400 public: |
|
401 UDataPathIterator(const char *path, const char *pkg, |
|
402 const char *item, const char *suffix, UBool doCheckLastFour, |
|
403 UErrorCode *pErrorCode); |
|
404 const char *next(UErrorCode *pErrorCode); |
|
405 |
|
406 private: |
|
407 const char *path; /* working path (u_icudata_Dir) */ |
|
408 const char *nextPath; /* path following this one */ |
|
409 const char *basename; /* item's basename (icudt22e_mt.res)*/ |
|
410 const char *suffix; /* item suffix (can be null) */ |
|
411 |
|
412 uint32_t basenameLen; /* length of basename */ |
|
413 |
|
414 CharString itemPath; /* path passed in with item name */ |
|
415 CharString pathBuffer; /* output path for this it'ion */ |
|
416 CharString packageStub; /* example: "/icudt28b". Will ignore that leaf in set paths. */ |
|
417 |
|
418 UBool checkLastFour; /* if TRUE then allow paths such as '/foo/myapp.dat' |
|
419 * to match, checks last 4 chars of suffix with |
|
420 * last 4 of path, then previous chars. */ |
|
421 }; |
|
422 |
|
423 /** |
|
424 * @param iter The iterator to be initialized. Its current state does not matter. |
|
425 * @param path The full pathname to be iterated over. If NULL, defaults to U_ICUDATA_NAME |
|
426 * @param pkg Package which is being searched for, ex "icudt28l". Will ignore leave directories such as /icudt28l |
|
427 * @param item Item to be searched for. Can include full path, such as /a/b/foo.dat |
|
428 * @param suffix Optional item suffix, if not-null (ex. ".dat") then 'path' can contain 'item' explicitly. |
|
429 * Ex: 'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar/baz' as item #2. |
|
430 * '/blarg/stuff.dat' would also be found. |
|
431 */ |
|
432 UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg, |
|
433 const char *item, const char *inSuffix, UBool doCheckLastFour, |
|
434 UErrorCode *pErrorCode) |
|
435 { |
|
436 #ifdef UDATA_DEBUG |
|
437 fprintf(stderr, "SUFFIX1=%s PATH=%s\n", inSuffix, inPath); |
|
438 #endif |
|
439 /** Path **/ |
|
440 if(inPath == NULL) { |
|
441 path = u_getDataDirectory(); |
|
442 } else { |
|
443 path = inPath; |
|
444 } |
|
445 |
|
446 /** Package **/ |
|
447 if(pkg != NULL) { |
|
448 packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode); |
|
449 #ifdef UDATA_DEBUG |
|
450 fprintf(stderr, "STUB=%s [%d]\n", packageStub.data(), packageStub.length()); |
|
451 #endif |
|
452 } |
|
453 |
|
454 /** Item **/ |
|
455 basename = findBasename(item); |
|
456 basenameLen = (int32_t)uprv_strlen(basename); |
|
457 |
|
458 /** Item path **/ |
|
459 if(basename == item) { |
|
460 nextPath = path; |
|
461 } else { |
|
462 itemPath.append(item, (int32_t)(basename-item), *pErrorCode); |
|
463 nextPath = itemPath.data(); |
|
464 } |
|
465 #ifdef UDATA_DEBUG |
|
466 fprintf(stderr, "SUFFIX=%s [%p]\n", inSuffix, inSuffix); |
|
467 #endif |
|
468 |
|
469 /** Suffix **/ |
|
470 if(inSuffix != NULL) { |
|
471 suffix = inSuffix; |
|
472 } else { |
|
473 suffix = ""; |
|
474 } |
|
475 |
|
476 checkLastFour = doCheckLastFour; |
|
477 |
|
478 /* pathBuffer will hold the output path strings returned by this iterator */ |
|
479 |
|
480 #ifdef UDATA_DEBUG |
|
481 fprintf(stderr, "%p: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%s], [nextpath=%s], [checklast4=%s]\n", |
|
482 iter, |
|
483 item, |
|
484 path, |
|
485 basename, |
|
486 suffix, |
|
487 itemPath.data(), |
|
488 nextPath, |
|
489 checkLastFour?"TRUE":"false"); |
|
490 #endif |
|
491 } |
|
492 |
|
493 /** |
|
494 * Get the next path on the list. |
|
495 * |
|
496 * @param iter The Iter to be used |
|
497 * @param len If set, pointer to the length of the returned path, for convenience. |
|
498 * @return Pointer to the next path segment, or NULL if there are no more. |
|
499 */ |
|
500 const char *UDataPathIterator::next(UErrorCode *pErrorCode) |
|
501 { |
|
502 if(U_FAILURE(*pErrorCode)) { |
|
503 return NULL; |
|
504 } |
|
505 |
|
506 const char *currentPath = NULL; |
|
507 int32_t pathLen = 0; |
|
508 const char *pathBasename; |
|
509 |
|
510 do |
|
511 { |
|
512 if( nextPath == NULL ) { |
|
513 break; |
|
514 } |
|
515 currentPath = nextPath; |
|
516 |
|
517 if(nextPath == itemPath.data()) { /* we were processing item's path. */ |
|
518 nextPath = path; /* start with regular path next tm. */ |
|
519 pathLen = (int32_t)uprv_strlen(currentPath); |
|
520 } else { |
|
521 /* fix up next for next time */ |
|
522 nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR); |
|
523 if(nextPath == NULL) { |
|
524 /* segment: entire path */ |
|
525 pathLen = (int32_t)uprv_strlen(currentPath); |
|
526 } else { |
|
527 /* segment: until next segment */ |
|
528 pathLen = (int32_t)(nextPath - currentPath); |
|
529 /* skip divider */ |
|
530 nextPath ++; |
|
531 } |
|
532 } |
|
533 |
|
534 if(pathLen == 0) { |
|
535 continue; |
|
536 } |
|
537 |
|
538 #ifdef UDATA_DEBUG |
|
539 fprintf(stderr, "rest of path (IDD) = %s\n", currentPath); |
|
540 fprintf(stderr, " "); |
|
541 { |
|
542 uint32_t qqq; |
|
543 for(qqq=0;qqq<pathLen;qqq++) |
|
544 { |
|
545 fprintf(stderr, " "); |
|
546 } |
|
547 |
|
548 fprintf(stderr, "^\n"); |
|
549 } |
|
550 #endif |
|
551 pathBuffer.clear().append(currentPath, pathLen, *pErrorCode); |
|
552 |
|
553 /* check for .dat files */ |
|
554 pathBasename = findBasename(pathBuffer.data()); |
|
555 |
|
556 if(checkLastFour == TRUE && |
|
557 (pathLen>=4) && |
|
558 uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix, 4)==0 && /* suffix matches */ |
|
559 uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)==0 && /* base matches */ |
|
560 uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full len */ |
|
561 |
|
562 #ifdef UDATA_DEBUG |
|
563 fprintf(stderr, "Have %s file on the path: %s\n", suffix, pathBuffer.data()); |
|
564 #endif |
|
565 /* do nothing */ |
|
566 } |
|
567 else |
|
568 { /* regular dir path */ |
|
569 if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) { |
|
570 if((pathLen>=4) && |
|
571 uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat", 4) == 0) |
|
572 { |
|
573 #ifdef UDATA_DEBUG |
|
574 fprintf(stderr, "skipping non-directory .dat file %s\n", pathBuffer.data()); |
|
575 #endif |
|
576 continue; |
|
577 } |
|
578 |
|
579 /* Check if it is a directory with the same name as our package */ |
|
580 if(!packageStub.isEmpty() && |
|
581 (pathLen > packageStub.length()) && |
|
582 !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length(), packageStub.data())) { |
|
583 #ifdef UDATA_DEBUG |
|
584 fprintf(stderr, "Found stub %s (will add package %s of len %d)\n", packageStub.data(), basename, basenameLen); |
|
585 #endif |
|
586 pathBuffer.truncate(pathLen - packageStub.length()); |
|
587 } |
|
588 pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode); |
|
589 } |
|
590 |
|
591 /* + basename */ |
|
592 pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pErrorCode); |
|
593 |
|
594 if(*suffix) /* tack on suffix */ |
|
595 { |
|
596 pathBuffer.append(suffix, *pErrorCode); |
|
597 } |
|
598 } |
|
599 |
|
600 #ifdef UDATA_DEBUG |
|
601 fprintf(stderr, " --> %s\n", pathBuffer.data()); |
|
602 #endif |
|
603 |
|
604 return pathBuffer.data(); |
|
605 |
|
606 } while(path); |
|
607 |
|
608 /* fell way off the end */ |
|
609 return NULL; |
|
610 } |
|
611 |
|
612 U_NAMESPACE_END |
|
613 |
|
614 /* ==================================================================================*/ |
|
615 |
|
616 |
|
617 /*----------------------------------------------------------------------* |
|
618 * * |
|
619 * Add a static reference to the common data library * |
|
620 * Unless overridden by an explicit udata_setCommonData, this will be * |
|
621 * our common data. * |
|
622 * * |
|
623 *----------------------------------------------------------------------*/ |
|
624 extern "C" const DataHeader U_DATA_API U_ICUDATA_ENTRY_POINT; |
|
625 |
|
626 /* |
|
627 * This would be a good place for weak-linkage declarations of |
|
628 * partial-data-library access functions where each returns a pointer |
|
629 * to its data package, if it is linked in. |
|
630 */ |
|
631 /* |
|
632 extern const void *uprv_getICUData_collation(void) ATTRIBUTE_WEAK; |
|
633 extern const void *uprv_getICUData_conversion(void) ATTRIBUTE_WEAK; |
|
634 */ |
|
635 |
|
636 /*----------------------------------------------------------------------* |
|
637 * * |
|
638 * openCommonData Attempt to open a common format (.dat) file * |
|
639 * Map it into memory (if it's not there already) * |
|
640 * and return a UDataMemory object for it. * |
|
641 * * |
|
642 * If the requested data is already open and cached * |
|
643 * just return the cached UDataMem object. * |
|
644 * * |
|
645 *----------------------------------------------------------------------*/ |
|
646 static UDataMemory * |
|
647 openCommonData(const char *path, /* Path from OpenChoice? */ |
|
648 int32_t commonDataIndex, /* ICU Data (index >= 0) if path == NULL */ |
|
649 UErrorCode *pErrorCode) |
|
650 { |
|
651 UDataMemory tData; |
|
652 const char *pathBuffer; |
|
653 const char *inBasename; |
|
654 |
|
655 if (U_FAILURE(*pErrorCode)) { |
|
656 return NULL; |
|
657 } |
|
658 |
|
659 UDataMemory_init(&tData); |
|
660 |
|
661 /* ??????? TODO revisit this */ |
|
662 if (commonDataIndex >= 0) { |
|
663 /* "mini-cache" for common ICU data */ |
|
664 if(commonDataIndex >= LENGTHOF(gCommonICUDataArray)) { |
|
665 return NULL; |
|
666 } |
|
667 if(gCommonICUDataArray[commonDataIndex] == NULL) { |
|
668 int32_t i; |
|
669 for(i = 0; i < commonDataIndex; ++i) { |
|
670 if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT) { |
|
671 /* The linked-in data is already in the list. */ |
|
672 return NULL; |
|
673 } |
|
674 } |
|
675 |
|
676 /* Add the linked-in data to the list. */ |
|
677 /* |
|
678 * This is where we would check and call weakly linked partial-data-library |
|
679 * access functions. |
|
680 */ |
|
681 /* |
|
682 if (uprv_getICUData_collation) { |
|
683 setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErrorCode); |
|
684 } |
|
685 if (uprv_getICUData_conversion) { |
|
686 setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pErrorCode); |
|
687 } |
|
688 */ |
|
689 setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, FALSE, pErrorCode); |
|
690 } |
|
691 return gCommonICUDataArray[commonDataIndex]; |
|
692 } |
|
693 |
|
694 |
|
695 /* request is NOT for ICU Data. */ |
|
696 |
|
697 /* Find the base name portion of the supplied path. */ |
|
698 /* inBasename will be left pointing somewhere within the original path string. */ |
|
699 inBasename = findBasename(path); |
|
700 #ifdef UDATA_DEBUG |
|
701 fprintf(stderr, "inBasename = %s\n", inBasename); |
|
702 #endif |
|
703 |
|
704 if(*inBasename==0) { |
|
705 /* no basename. This will happen if the original path was a directory name, */ |
|
706 /* like "a/b/c/". (Fallback to separate files will still work.) */ |
|
707 #ifdef UDATA_DEBUG |
|
708 fprintf(stderr, "ocd: no basename in %s, bailing.\n", path); |
|
709 #endif |
|
710 *pErrorCode=U_FILE_ACCESS_ERROR; |
|
711 return NULL; |
|
712 } |
|
713 |
|
714 /* Is the requested common data file already open and cached? */ |
|
715 /* Note that the cache is keyed by the base name only. The rest of the path, */ |
|
716 /* if any, is not considered. */ |
|
717 { |
|
718 UDataMemory *dataToReturn = udata_findCachedData(inBasename); |
|
719 if (dataToReturn != NULL) { |
|
720 return dataToReturn; |
|
721 } |
|
722 } |
|
723 |
|
724 /* Requested item is not in the cache. |
|
725 * Hunt it down, trying all the path locations |
|
726 */ |
|
727 |
|
728 UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat", TRUE, pErrorCode); |
|
729 |
|
730 while((UDataMemory_isLoaded(&tData)==FALSE) && (pathBuffer = iter.next(pErrorCode)) != NULL) |
|
731 { |
|
732 #ifdef UDATA_DEBUG |
|
733 fprintf(stderr, "ocd: trying path %s - ", pathBuffer); |
|
734 #endif |
|
735 uprv_mapFile(&tData, pathBuffer); |
|
736 #ifdef UDATA_DEBUG |
|
737 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&tData)?"LOADED":"not loaded"); |
|
738 #endif |
|
739 } |
|
740 |
|
741 #if defined(OS390_STUBDATA) && defined(OS390BATCH) |
|
742 if (!UDataMemory_isLoaded(&tData)) { |
|
743 char ourPathBuffer[1024]; |
|
744 /* One more chance, for extendCommonData() */ |
|
745 uprv_strncpy(ourPathBuffer, path, 1019); |
|
746 ourPathBuffer[1019]=0; |
|
747 uprv_strcat(ourPathBuffer, ".dat"); |
|
748 uprv_mapFile(&tData, ourPathBuffer); |
|
749 } |
|
750 #endif |
|
751 |
|
752 if (!UDataMemory_isLoaded(&tData)) { |
|
753 /* no common data */ |
|
754 *pErrorCode=U_FILE_ACCESS_ERROR; |
|
755 return NULL; |
|
756 } |
|
757 |
|
758 /* we have mapped a file, check its header */ |
|
759 udata_checkCommonData(&tData, pErrorCode); |
|
760 |
|
761 |
|
762 /* Cache the UDataMemory struct for this .dat file, |
|
763 * so we won't need to hunt it down and map it again next time |
|
764 * something is needed from it. */ |
|
765 return udata_cacheDataItem(inBasename, &tData, pErrorCode); |
|
766 } |
|
767 |
|
768 |
|
769 /*----------------------------------------------------------------------* |
|
770 * * |
|
771 * extendICUData If the full set of ICU data was not loaded at * |
|
772 * program startup, load it now. This function will * |
|
773 * be called when the lookup of an ICU data item in * |
|
774 * the common ICU data fails. * |
|
775 * * |
|
776 * return true if new data is loaded, false otherwise.* |
|
777 * * |
|
778 *----------------------------------------------------------------------*/ |
|
779 static UBool extendICUData(UErrorCode *pErr) |
|
780 { |
|
781 UDataMemory *pData; |
|
782 UDataMemory copyPData; |
|
783 UBool didUpdate = FALSE; |
|
784 |
|
785 /* |
|
786 * There is a chance for a race condition here. |
|
787 * Normally, ICU data is loaded from a DLL or via mmap() and |
|
788 * setCommonICUData() will detect if the same address is set twice. |
|
789 * If ICU is built with data loading via fread() then the address will |
|
790 * be different each time the common data is loaded and we may add |
|
791 * multiple copies of the data. |
|
792 * In this case, use a mutex to prevent the race. |
|
793 * Use a specific mutex to avoid nested locks of the global mutex. |
|
794 */ |
|
795 #if MAP_IMPLEMENTATION==MAP_STDIO |
|
796 static UMutex extendICUDataMutex = U_MUTEX_INITIALIZER; |
|
797 umtx_lock(&extendICUDataMutex); |
|
798 #endif |
|
799 if(!gHaveTriedToLoadCommonData) { |
|
800 /* See if we can explicitly open a .dat file for the ICUData. */ |
|
801 pData = openCommonData( |
|
802 U_ICUDATA_NAME, /* "icudt20l" , for example. */ |
|
803 -1, /* Pretend we're not opening ICUData */ |
|
804 pErr); |
|
805 |
|
806 /* How about if there is no pData, eh... */ |
|
807 |
|
808 UDataMemory_init(©PData); |
|
809 if(pData != NULL) { |
|
810 UDatamemory_assign(©PData, pData); |
|
811 copyPData.map = 0; /* The mapping for this data is owned by the hash table */ |
|
812 copyPData.mapAddr = 0; /* which will unmap it when ICU is shut down. */ |
|
813 /* CommonICUData is also unmapped when ICU is shut down.*/ |
|
814 /* To avoid unmapping the data twice, zero out the map */ |
|
815 /* fields in the UDataMemory that we're assigning */ |
|
816 /* to CommonICUData. */ |
|
817 |
|
818 didUpdate = /* no longer using this result */ |
|
819 setCommonICUData(©PData,/* The new common data. */ |
|
820 FALSE, /* No warnings if write didn't happen */ |
|
821 pErr); /* setCommonICUData honors errors; NOP if error set */ |
|
822 } |
|
823 |
|
824 gHaveTriedToLoadCommonData = TRUE; |
|
825 } |
|
826 |
|
827 didUpdate = findCommonICUDataByName(U_ICUDATA_NAME); /* Return 'true' when a racing writes out the extended */ |
|
828 /* data after another thread has failed to see it (in openCommonData), so */ |
|
829 /* extended data can be examined. */ |
|
830 /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */ |
|
831 |
|
832 #if MAP_IMPLEMENTATION==MAP_STDIO |
|
833 umtx_unlock(&extendICUDataMutex); |
|
834 #endif |
|
835 return didUpdate; /* Return true if ICUData pointer was updated. */ |
|
836 /* (Could potentialy have been done by another thread racing */ |
|
837 /* us through here, but that's fine, we still return true */ |
|
838 /* so that current thread will also examine extended data. */ |
|
839 } |
|
840 |
|
841 /*----------------------------------------------------------------------* |
|
842 * * |
|
843 * udata_setCommonData * |
|
844 * * |
|
845 *----------------------------------------------------------------------*/ |
|
846 U_CAPI void U_EXPORT2 |
|
847 udata_setCommonData(const void *data, UErrorCode *pErrorCode) { |
|
848 UDataMemory dataMemory; |
|
849 |
|
850 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
|
851 return; |
|
852 } |
|
853 |
|
854 if(data==NULL) { |
|
855 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
|
856 return; |
|
857 } |
|
858 |
|
859 /* set the data pointer and test for validity */ |
|
860 UDataMemory_init(&dataMemory); |
|
861 UDataMemory_setData(&dataMemory, data); |
|
862 udata_checkCommonData(&dataMemory, pErrorCode); |
|
863 if (U_FAILURE(*pErrorCode)) {return;} |
|
864 |
|
865 /* we have good data */ |
|
866 /* Set it up as the ICU Common Data. */ |
|
867 setCommonICUData(&dataMemory, TRUE, pErrorCode); |
|
868 } |
|
869 |
|
870 /*--------------------------------------------------------------------------- |
|
871 * |
|
872 * udata_setAppData |
|
873 * |
|
874 *---------------------------------------------------------------------------- */ |
|
875 U_CAPI void U_EXPORT2 |
|
876 udata_setAppData(const char *path, const void *data, UErrorCode *err) |
|
877 { |
|
878 UDataMemory udm; |
|
879 |
|
880 if(err==NULL || U_FAILURE(*err)) { |
|
881 return; |
|
882 } |
|
883 if(data==NULL) { |
|
884 *err=U_ILLEGAL_ARGUMENT_ERROR; |
|
885 return; |
|
886 } |
|
887 |
|
888 UDataMemory_init(&udm); |
|
889 UDataMemory_setData(&udm, data); |
|
890 udata_checkCommonData(&udm, err); |
|
891 udata_cacheDataItem(path, &udm, err); |
|
892 } |
|
893 |
|
894 /*----------------------------------------------------------------------------* |
|
895 * * |
|
896 * checkDataItem Given a freshly located/loaded data item, either * |
|
897 * an entry in a common file or a separately loaded file, * |
|
898 * sanity check its header, and see if the data is * |
|
899 * acceptable to the app. * |
|
900 * If the data is good, create and return a UDataMemory * |
|
901 * object that can be returned to the application. * |
|
902 * Return NULL on any sort of failure. * |
|
903 * * |
|
904 *----------------------------------------------------------------------------*/ |
|
905 static UDataMemory * |
|
906 checkDataItem |
|
907 ( |
|
908 const DataHeader *pHeader, /* The data item to be checked. */ |
|
909 UDataMemoryIsAcceptable *isAcceptable, /* App's call-back function */ |
|
910 void *context, /* pass-thru param for above. */ |
|
911 const char *type, /* pass-thru param for above. */ |
|
912 const char *name, /* pass-thru param for above. */ |
|
913 UErrorCode *nonFatalErr, /* Error code if this data was not acceptable */ |
|
914 /* but openChoice should continue with */ |
|
915 /* trying to get data from fallback path. */ |
|
916 UErrorCode *fatalErr /* Bad error, caller should return immediately */ |
|
917 ) |
|
918 { |
|
919 UDataMemory *rDataMem = NULL; /* the new UDataMemory, to be returned. */ |
|
920 |
|
921 if (U_FAILURE(*fatalErr)) { |
|
922 return NULL; |
|
923 } |
|
924 |
|
925 if(pHeader->dataHeader.magic1==0xda && |
|
926 pHeader->dataHeader.magic2==0x27 && |
|
927 (isAcceptable==NULL || isAcceptable(context, type, name, &pHeader->info)) |
|
928 ) { |
|
929 rDataMem=UDataMemory_createNewInstance(fatalErr); |
|
930 if (U_FAILURE(*fatalErr)) { |
|
931 return NULL; |
|
932 } |
|
933 rDataMem->pHeader = pHeader; |
|
934 } else { |
|
935 /* the data is not acceptable, look further */ |
|
936 /* If we eventually find something good, this errorcode will be */ |
|
937 /* cleared out. */ |
|
938 *nonFatalErr=U_INVALID_FORMAT_ERROR; |
|
939 } |
|
940 return rDataMem; |
|
941 } |
|
942 |
|
943 /** |
|
944 * @return 0 if not loaded, 1 if loaded or err |
|
945 */ |
|
946 static UDataMemory *doLoadFromIndividualFiles(const char *pkgName, |
|
947 const char *dataPath, const char *tocEntryPathSuffix, |
|
948 /* following arguments are the same as doOpenChoice itself */ |
|
949 const char *path, const char *type, const char *name, |
|
950 UDataMemoryIsAcceptable *isAcceptable, void *context, |
|
951 UErrorCode *subErrorCode, |
|
952 UErrorCode *pErrorCode) |
|
953 { |
|
954 const char *pathBuffer; |
|
955 UDataMemory dataMemory; |
|
956 UDataMemory *pEntryData; |
|
957 |
|
958 /* look in ind. files: package\nam.typ ========================= */ |
|
959 /* init path iterator for individual files */ |
|
960 UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, FALSE, pErrorCode); |
|
961 |
|
962 while((pathBuffer = iter.next(pErrorCode))) |
|
963 { |
|
964 #ifdef UDATA_DEBUG |
|
965 fprintf(stderr, "UDATA: trying individual file %s\n", pathBuffer); |
|
966 #endif |
|
967 if(uprv_mapFile(&dataMemory, pathBuffer)) |
|
968 { |
|
969 pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode); |
|
970 if (pEntryData != NULL) { |
|
971 /* Data is good. |
|
972 * Hand off ownership of the backing memory to the user's UDataMemory. |
|
973 * and return it. */ |
|
974 pEntryData->mapAddr = dataMemory.mapAddr; |
|
975 pEntryData->map = dataMemory.map; |
|
976 |
|
977 #ifdef UDATA_DEBUG |
|
978 fprintf(stderr, "** Mapped file: %s\n", pathBuffer); |
|
979 #endif |
|
980 return pEntryData; |
|
981 } |
|
982 |
|
983 /* the data is not acceptable, or some error occured. Either way, unmap the memory */ |
|
984 udata_close(&dataMemory); |
|
985 |
|
986 /* If we had a nasty error, bail out completely. */ |
|
987 if (U_FAILURE(*pErrorCode)) { |
|
988 return NULL; |
|
989 } |
|
990 |
|
991 /* Otherwise remember that we found data but didn't like it for some reason */ |
|
992 *subErrorCode=U_INVALID_FORMAT_ERROR; |
|
993 } |
|
994 #ifdef UDATA_DEBUG |
|
995 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&dataMemory)?"LOADED":"not loaded"); |
|
996 #endif |
|
997 } |
|
998 return NULL; |
|
999 } |
|
1000 |
|
1001 /** |
|
1002 * @return 0 if not loaded, 1 if loaded or err |
|
1003 */ |
|
1004 static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName*/, |
|
1005 const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const char *tocEntryName, |
|
1006 /* following arguments are the same as doOpenChoice itself */ |
|
1007 const char *path, const char *type, const char *name, |
|
1008 UDataMemoryIsAcceptable *isAcceptable, void *context, |
|
1009 UErrorCode *subErrorCode, |
|
1010 UErrorCode *pErrorCode) |
|
1011 { |
|
1012 UDataMemory *pEntryData; |
|
1013 const DataHeader *pHeader; |
|
1014 UDataMemory *pCommonData; |
|
1015 int32_t commonDataIndex; |
|
1016 UBool checkedExtendedICUData = FALSE; |
|
1017 /* try to get common data. The loop is for platforms such as the 390 that do |
|
1018 * not initially load the full set of ICU data. If the lookup of an ICU data item |
|
1019 * fails, the full (but slower to load) set is loaded, the and the loop repeats, |
|
1020 * trying the lookup again. Once the full set of ICU data is loaded, the loop wont |
|
1021 * repeat because the full set will be checked the first time through. |
|
1022 * |
|
1023 * The loop also handles the fallback to a .dat file if the application linked |
|
1024 * to the stub data library rather than a real library. |
|
1025 */ |
|
1026 for (commonDataIndex = isICUData ? 0 : -1;;) { |
|
1027 pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** search for pkg **/ |
|
1028 |
|
1029 if(U_SUCCESS(*subErrorCode) && pCommonData!=NULL) { |
|
1030 int32_t length; |
|
1031 |
|
1032 /* look up the data piece in the common data */ |
|
1033 pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &length, subErrorCode); |
|
1034 #ifdef UDATA_DEBUG |
|
1035 fprintf(stderr, "%s: pHeader=%p - %s\n", tocEntryName, pHeader, u_errorName(*subErrorCode)); |
|
1036 #endif |
|
1037 |
|
1038 if(pHeader!=NULL) { |
|
1039 pEntryData = checkDataItem(pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode); |
|
1040 #ifdef UDATA_DEBUG |
|
1041 fprintf(stderr, "pEntryData=%p\n", pEntryData); |
|
1042 #endif |
|
1043 if (U_FAILURE(*pErrorCode)) { |
|
1044 return NULL; |
|
1045 } |
|
1046 if (pEntryData != NULL) { |
|
1047 pEntryData->length = length; |
|
1048 return pEntryData; |
|
1049 } |
|
1050 } |
|
1051 } |
|
1052 /* Data wasn't found. If we were looking for an ICUData item and there is |
|
1053 * more data available, load it and try again, |
|
1054 * otherwise break out of this loop. */ |
|
1055 if (!isICUData) { |
|
1056 return NULL; |
|
1057 } else if (pCommonData != NULL) { |
|
1058 ++commonDataIndex; /* try the next data package */ |
|
1059 } else if ((!checkedExtendedICUData) && extendICUData(subErrorCode)) { |
|
1060 checkedExtendedICUData = TRUE; |
|
1061 /* try this data package slot again: it changed from NULL to non-NULL */ |
|
1062 } else { |
|
1063 return NULL; |
|
1064 } |
|
1065 } |
|
1066 } |
|
1067 |
|
1068 /* |
|
1069 * A note on the ownership of Mapped Memory |
|
1070 * |
|
1071 * For common format files, ownership resides with the UDataMemory object |
|
1072 * that lives in the cache of opened common data. These UDataMemorys are private |
|
1073 * to the udata implementation, and are never seen directly by users. |
|
1074 * |
|
1075 * The UDataMemory objects returned to users will have the address of some desired |
|
1076 * data within the mapped region, but they wont have the mapping info itself, and thus |
|
1077 * won't cause anything to be removed from memory when they are closed. |
|
1078 * |
|
1079 * For individual data files, the UDataMemory returned to the user holds the |
|
1080 * information necessary to unmap the data on close. If the user independently |
|
1081 * opens the same data file twice, two completely independent mappings will be made. |
|
1082 * (There is no cache of opened data items from individual files, only a cache of |
|
1083 * opened Common Data files, that is, files containing a collection of data items.) |
|
1084 * |
|
1085 * For common data passed in from the user via udata_setAppData() or |
|
1086 * udata_setCommonData(), ownership remains with the user. |
|
1087 * |
|
1088 * UDataMemory objects themselves, as opposed to the memory they describe, |
|
1089 * can be anywhere - heap, stack/local or global. |
|
1090 * They have a flag to indicate when they're heap allocated and thus |
|
1091 * must be deleted when closed. |
|
1092 */ |
|
1093 |
|
1094 |
|
1095 /*----------------------------------------------------------------------------* |
|
1096 * * |
|
1097 * main data loading functions * |
|
1098 * * |
|
1099 *----------------------------------------------------------------------------*/ |
|
1100 static UDataMemory * |
|
1101 doOpenChoice(const char *path, const char *type, const char *name, |
|
1102 UDataMemoryIsAcceptable *isAcceptable, void *context, |
|
1103 UErrorCode *pErrorCode) |
|
1104 { |
|
1105 UDataMemory *retVal = NULL; |
|
1106 |
|
1107 const char *dataPath; |
|
1108 |
|
1109 int32_t tocEntrySuffixIndex; |
|
1110 const char *tocEntryPathSuffix; |
|
1111 UErrorCode subErrorCode=U_ZERO_ERROR; |
|
1112 const char *treeChar; |
|
1113 |
|
1114 UBool isICUData = FALSE; |
|
1115 |
|
1116 |
|
1117 /* Is this path ICU data? */ |
|
1118 if(path == NULL || |
|
1119 !strcmp(path, U_ICUDATA_ALIAS) || /* "ICUDATA" */ |
|
1120 !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-" */ |
|
1121 uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) || |
|
1122 !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-" */ |
|
1123 uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) { |
|
1124 isICUData = TRUE; |
|
1125 } |
|
1126 |
|
1127 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR) /* Windows: try "foo\bar" and "foo/bar" */ |
|
1128 /* remap from alternate path char to the main one */ |
|
1129 CharString altSepPath; |
|
1130 if(path) { |
|
1131 if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != NULL) { |
|
1132 altSepPath.append(path, *pErrorCode); |
|
1133 char *p; |
|
1134 while((p=uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR))) { |
|
1135 *p = U_FILE_SEP_CHAR; |
|
1136 } |
|
1137 #if defined (UDATA_DEBUG) |
|
1138 fprintf(stderr, "Changed path from [%s] to [%s]\n", path, altSepPath.s); |
|
1139 #endif |
|
1140 path = altSepPath.data(); |
|
1141 } |
|
1142 } |
|
1143 #endif |
|
1144 |
|
1145 CharString tocEntryName; /* entry name in tree format. ex: 'icudt28b/coll/ar.res' */ |
|
1146 CharString tocEntryPath; /* entry name in path format. ex: 'icudt28b\\coll\\ar.res' */ |
|
1147 |
|
1148 CharString pkgName; |
|
1149 CharString treeName; |
|
1150 |
|
1151 /* ======= Set up strings */ |
|
1152 if(path==NULL) { |
|
1153 pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
|
1154 } else { |
|
1155 const char *pkg; |
|
1156 const char *first; |
|
1157 pkg = uprv_strrchr(path, U_FILE_SEP_CHAR); |
|
1158 first = uprv_strchr(path, U_FILE_SEP_CHAR); |
|
1159 if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash in the path- not a tree name */ |
|
1160 /* see if this is an /absolute/path/to/package path */ |
|
1161 if(pkg) { |
|
1162 pkgName.append(pkg+1, *pErrorCode); |
|
1163 } else { |
|
1164 pkgName.append(path, *pErrorCode); |
|
1165 } |
|
1166 } else { |
|
1167 treeChar = uprv_strchr(path, U_TREE_SEPARATOR); |
|
1168 if(treeChar) { |
|
1169 treeName.append(treeChar+1, *pErrorCode); /* following '-' */ |
|
1170 if(isICUData) { |
|
1171 pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
|
1172 } else { |
|
1173 pkgName.append(path, (int32_t)(treeChar-path), *pErrorCode); |
|
1174 if (first == NULL) { |
|
1175 /* |
|
1176 This user data has no path, but there is a tree name. |
|
1177 Look up the correct path from the data cache later. |
|
1178 */ |
|
1179 path = pkgName.data(); |
|
1180 } |
|
1181 } |
|
1182 } else { |
|
1183 if(isICUData) { |
|
1184 pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
|
1185 } else { |
|
1186 pkgName.append(path, *pErrorCode); |
|
1187 } |
|
1188 } |
|
1189 } |
|
1190 } |
|
1191 |
|
1192 #ifdef UDATA_DEBUG |
|
1193 fprintf(stderr, " P=%s T=%s\n", pkgName.data(), treeName.data()); |
|
1194 #endif |
|
1195 |
|
1196 /* setting up the entry name and file name |
|
1197 * Make up a full name by appending the type to the supplied |
|
1198 * name, assuming that a type was supplied. |
|
1199 */ |
|
1200 |
|
1201 /* prepend the package */ |
|
1202 tocEntryName.append(pkgName, *pErrorCode); |
|
1203 tocEntryPath.append(pkgName, *pErrorCode); |
|
1204 tocEntrySuffixIndex = tocEntryName.length(); |
|
1205 |
|
1206 if(!treeName.isEmpty()) { |
|
1207 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode); |
|
1208 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode); |
|
1209 } |
|
1210 |
|
1211 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
|
1212 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
|
1213 if(type!=NULL && *type!=0) { |
|
1214 tocEntryName.append(".", *pErrorCode).append(type, *pErrorCode); |
|
1215 tocEntryPath.append(".", *pErrorCode).append(type, *pErrorCode); |
|
1216 } |
|
1217 tocEntryPathSuffix = tocEntryPath.data()+tocEntrySuffixIndex; /* suffix starts here */ |
|
1218 |
|
1219 #ifdef UDATA_DEBUG |
|
1220 fprintf(stderr, " tocEntryName = %s\n", tocEntryName.data()); |
|
1221 fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data()); |
|
1222 #endif |
|
1223 |
|
1224 if(path == NULL) { |
|
1225 path = COMMON_DATA_NAME; /* "icudt26e" */ |
|
1226 } |
|
1227 |
|
1228 /************************ Begin loop looking for ind. files ***************/ |
|
1229 #ifdef UDATA_DEBUG |
|
1230 fprintf(stderr, "IND: inBasename = %s, pkg=%s\n", "(n/a)", packageNameFromPath(path)); |
|
1231 #endif |
|
1232 |
|
1233 /* End of dealing with a null basename */ |
|
1234 dataPath = u_getDataDirectory(); |
|
1235 |
|
1236 /**** COMMON PACKAGE - only if packages are first. */ |
|
1237 if(gDataFileAccess == UDATA_PACKAGES_FIRST) { |
|
1238 #ifdef UDATA_DEBUG |
|
1239 fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n"); |
|
1240 #endif |
|
1241 /* #2 */ |
|
1242 retVal = doLoadFromCommonData(isICUData, |
|
1243 pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), |
|
1244 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
|
1245 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
|
1246 return retVal; |
|
1247 } |
|
1248 } |
|
1249 |
|
1250 /**** INDIVIDUAL FILES */ |
|
1251 if((gDataFileAccess==UDATA_PACKAGES_FIRST) || |
|
1252 (gDataFileAccess==UDATA_FILES_FIRST)) { |
|
1253 #ifdef UDATA_DEBUG |
|
1254 fprintf(stderr, "Trying individual files\n"); |
|
1255 #endif |
|
1256 /* Check to make sure that there is a dataPath to iterate over */ |
|
1257 if ((dataPath && *dataPath) || !isICUData) { |
|
1258 retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix, |
|
1259 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
|
1260 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
|
1261 return retVal; |
|
1262 } |
|
1263 } |
|
1264 } |
|
1265 |
|
1266 /**** COMMON PACKAGE */ |
|
1267 if((gDataFileAccess==UDATA_ONLY_PACKAGES) || |
|
1268 (gDataFileAccess==UDATA_FILES_FIRST)) { |
|
1269 #ifdef UDATA_DEBUG |
|
1270 fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n"); |
|
1271 #endif |
|
1272 retVal = doLoadFromCommonData(isICUData, |
|
1273 pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), |
|
1274 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
|
1275 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
|
1276 return retVal; |
|
1277 } |
|
1278 } |
|
1279 |
|
1280 /* Load from DLL. If we haven't attempted package load, we also haven't had any chance to |
|
1281 try a DLL (static or setCommonData/etc) load. |
|
1282 If we ever have a "UDATA_ONLY_FILES", add it to the or list here. */ |
|
1283 if(gDataFileAccess==UDATA_NO_FILES) { |
|
1284 #ifdef UDATA_DEBUG |
|
1285 fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n"); |
|
1286 #endif |
|
1287 retVal = doLoadFromCommonData(isICUData, |
|
1288 pkgName.data(), "", tocEntryPathSuffix, tocEntryName.data(), |
|
1289 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
|
1290 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
|
1291 return retVal; |
|
1292 } |
|
1293 } |
|
1294 |
|
1295 /* data not found */ |
|
1296 if(U_SUCCESS(*pErrorCode)) { |
|
1297 if(U_SUCCESS(subErrorCode)) { |
|
1298 /* file not found */ |
|
1299 *pErrorCode=U_FILE_ACCESS_ERROR; |
|
1300 } else { |
|
1301 /* entry point not found or rejected */ |
|
1302 *pErrorCode=subErrorCode; |
|
1303 } |
|
1304 } |
|
1305 return retVal; |
|
1306 } |
|
1307 |
|
1308 |
|
1309 |
|
1310 /* API ---------------------------------------------------------------------- */ |
|
1311 |
|
1312 U_CAPI UDataMemory * U_EXPORT2 |
|
1313 udata_open(const char *path, const char *type, const char *name, |
|
1314 UErrorCode *pErrorCode) { |
|
1315 #ifdef UDATA_DEBUG |
|
1316 fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type); |
|
1317 fflush(stderr); |
|
1318 #endif |
|
1319 |
|
1320 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
|
1321 return NULL; |
|
1322 } else if(name==NULL || *name==0) { |
|
1323 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
|
1324 return NULL; |
|
1325 } else { |
|
1326 return doOpenChoice(path, type, name, NULL, NULL, pErrorCode); |
|
1327 } |
|
1328 } |
|
1329 |
|
1330 |
|
1331 |
|
1332 U_CAPI UDataMemory * U_EXPORT2 |
|
1333 udata_openChoice(const char *path, const char *type, const char *name, |
|
1334 UDataMemoryIsAcceptable *isAcceptable, void *context, |
|
1335 UErrorCode *pErrorCode) { |
|
1336 #ifdef UDATA_DEBUG |
|
1337 fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type); |
|
1338 #endif |
|
1339 |
|
1340 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
|
1341 return NULL; |
|
1342 } else if(name==NULL || *name==0 || isAcceptable==NULL) { |
|
1343 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
|
1344 return NULL; |
|
1345 } else { |
|
1346 return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode); |
|
1347 } |
|
1348 } |
|
1349 |
|
1350 |
|
1351 |
|
1352 U_CAPI void U_EXPORT2 |
|
1353 udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) { |
|
1354 if(pInfo!=NULL) { |
|
1355 if(pData!=NULL && pData->pHeader!=NULL) { |
|
1356 const UDataInfo *info=&pData->pHeader->info; |
|
1357 uint16_t dataInfoSize=udata_getInfoSize(info); |
|
1358 if(pInfo->size>dataInfoSize) { |
|
1359 pInfo->size=dataInfoSize; |
|
1360 } |
|
1361 uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->size-2); |
|
1362 if(info->isBigEndian!=U_IS_BIG_ENDIAN) { |
|
1363 /* opposite endianness */ |
|
1364 uint16_t x=info->reservedWord; |
|
1365 pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8)); |
|
1366 } |
|
1367 } else { |
|
1368 pInfo->size=0; |
|
1369 } |
|
1370 } |
|
1371 } |
|
1372 |
|
1373 |
|
1374 U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/) |
|
1375 { |
|
1376 gDataFileAccess = access; |
|
1377 } |