blob: 80bea0631ef7dc21b69a4a3dc61fa6b5b5b5edcf [file] [log] [blame]
jshin@chromium.org6f31ac32014-03-26 22:15:14 +00001/*
2******************************************************************************
3*
Jungshik Shin508e9272015-07-16 15:36:55 -07004* Copyright (C) 1999-2015, International Business Machines
jshin@chromium.org6f31ac32014-03-26 22:15:14 +00005* 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))
22might 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"
Jungshik Shin508e9272015-07-16 15:36:55 -070032#include "mutex.h"
jshin@chromium.org6f31ac32014-03-26 22:15:14 +000033#include "putilimp.h"
34#include "uassert.h"
35#include "ucln_cmn.h"
36#include "ucmndata.h"
37#include "udatamem.h"
38#include "uhash.h"
39#include "umapfile.h"
40#include "umutex.h"
41
42/***********************************************************************
43*
44* Notes on the organization of the ICU data implementation
45*
46* All of the public API is defined in udata.h
47*
48* The implementation is split into several files...
49*
50* - udata.c (this file) contains higher level code that knows about
51* the search paths for locating data, caching opened data, etc.
52*
53* - umapfile.c contains the low level platform-specific code for actually loading
54* (memory mapping, file reading, whatever) data into memory.
55*
56* - ucmndata.c deals with the tables of contents of ICU data items within
57* an ICU common format data file. The implementation includes
58* an abstract interface and support for multiple TOC formats.
59* All knowledge of any specific TOC format is encapsulated here.
60*
61* - udatamem.c has code for managing UDataMemory structs. These are little
62* descriptor objects for blocks of memory holding ICU data of
63* various types.
64*/
65
66/* configuration ---------------------------------------------------------- */
67
68/* If you are excruciatingly bored turn this on .. */
69/* #define UDATA_DEBUG 1 */
70
71#if defined(UDATA_DEBUG)
72# include <stdio.h>
73#endif
74
jshin@chromium.org6f31ac32014-03-26 22:15:14 +000075U_NAMESPACE_USE
76
77/*
78 * Forward declarations
79 */
80static UDataMemory *udata_findCachedData(const char *path);
81
82/***********************************************************************
83*
84* static (Global) data
85*
86************************************************************************/
87
88/*
89 * Pointers to the common ICU data.
90 *
91 * We store multiple pointers to ICU data packages and iterate through them
92 * when looking for a data item.
93 *
94 * It is possible to combine this with dependency inversion:
95 * One or more data package libraries may export
96 * functions that each return a pointer to their piece of the ICU data,
97 * and this file would import them as weak functions, without a
98 * strong linker dependency from the common library on the data library.
99 *
100 * Then we can have applications depend on only that part of ICU's data
101 * that they really need, reducing the size of binaries that take advantage
102 * of this.
103 */
Jungshik Shin508e9272015-07-16 15:36:55 -0700104static UDataMemory *gCommonICUDataArray[10] = { NULL }; // Access protected by icu global mutex.
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000105
Jungshik Shin508e9272015-07-16 15:36:55 -0700106static u_atomic_int32_t gHaveTriedToLoadCommonData = ATOMIC_INT32_T_INITIALIZER(0); // See extendICUData().
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000107
108static UHashtable *gCommonDataCache = NULL; /* Global hash table of opened ICU data files. */
109static icu::UInitOnce gCommonDataCacheInitOnce = U_INITONCE_INITIALIZER;
110
Jungshik Shin508e9272015-07-16 15:36:55 -0700111static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; // Access not synchronized.
112 // Modifying is documented as thread-unsafe.
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000113
114static UBool U_CALLCONV
115udata_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
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -0800125 for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != NULL; ++i) {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000126 udata_close(gCommonICUDataArray[i]);
127 gCommonICUDataArray[i] = NULL;
128 }
Jungshik Shin508e9272015-07-16 15:36:55 -0700129 gHaveTriedToLoadCommonData = 0;
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000130
131 return TRUE; /* Everything was cleaned up */
132}
133
134static UBool U_CALLCONV
135findCommonICUDataByName(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
Jungshik Shin508e9272015-07-16 15:36:55 -0700144 {
145 Mutex lock;
146 for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) {
147 if ((gCommonICUDataArray[i] != NULL) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) {
148 /* The data pointer is already in the array. */
149 found = TRUE;
150 break;
151 }
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000152 }
153 }
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000154 return found;
155}
156
157
158/*
159 * setCommonICUData. Set a UDataMemory to be the global ICU Data
160 */
161static UBool
162setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to caller, we copy it. */
163 UBool warn, /* If true, set USING_DEFAULT warning if ICUData was */
164 /* changed by another thread before we got to it. */
165 UErrorCode *pErr)
166{
167 UDataMemory *newCommonData = UDataMemory_createNewInstance(pErr);
168 int32_t i;
169 UBool didUpdate = FALSE;
170 if (U_FAILURE(*pErr)) {
171 return FALSE;
172 }
173
174 /* For the assignment, other threads must cleanly see either the old */
175 /* or the new, not some partially initialized new. The old can not be */
176 /* deleted - someone may still have a pointer to it lying around in */
177 /* their locals. */
178 UDatamemory_assign(newCommonData, pData);
179 umtx_lock(NULL);
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -0800180 for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000181 if (gCommonICUDataArray[i] == NULL) {
182 gCommonICUDataArray[i] = newCommonData;
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000183 didUpdate = TRUE;
184 break;
185 } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) {
186 /* The same data pointer is already in the array. */
187 break;
188 }
189 }
190 umtx_unlock(NULL);
191
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -0800192 if (i == UPRV_LENGTHOF(gCommonICUDataArray) && warn) {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000193 *pErr = U_USING_DEFAULT_WARNING;
194 }
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -0800195 if (didUpdate) {
196 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup);
197 } else {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000198 uprv_free(newCommonData);
199 }
200 return didUpdate;
201}
202
203static UBool
204setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) {
205 UDataMemory tData;
206 UDataMemory_init(&tData);
207 UDataMemory_setData(&tData, pData);
208 udata_checkCommonData(&tData, pErrorCode);
209 return setCommonICUData(&tData, FALSE, pErrorCode);
210}
211
212static const char *
213findBasename(const char *path) {
214 const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR);
215 if(basename==NULL) {
216 return path;
217 } else {
218 return basename+1;
219 }
220}
221
222#ifdef UDATA_DEBUG
223static const char *
224packageNameFromPath(const char *path)
225{
226 if((path == NULL) || (*path == 0)) {
227 return U_ICUDATA_NAME;
228 }
229
230 path = findBasename(path);
231
232 if((path == NULL) || (*path == 0)) {
233 return U_ICUDATA_NAME;
234 }
235
236 return path;
237}
238#endif
239
240/*----------------------------------------------------------------------*
241 * *
242 * Cache for common data *
243 * Functions for looking up or adding entries to a cache of *
244 * data that has been previously opened. Avoids a potentially *
245 * expensive operation of re-opening the data for subsequent *
246 * uses. *
247 * *
248 * Data remains cached for the duration of the process. *
249 * *
250 *----------------------------------------------------------------------*/
251
252typedef struct DataCacheElement {
253 char *name;
254 UDataMemory *item;
255} DataCacheElement;
256
257
258
259/*
260 * Deleter function for DataCacheElements.
261 * udata cleanup function closes the hash table; hash table in turn calls back to
262 * here for each entry.
263 */
264static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) {
265 DataCacheElement *p = (DataCacheElement *)pDCEl;
266 udata_close(p->item); /* unmaps storage */
267 uprv_free(p->name); /* delete the hash key string. */
268 uprv_free(pDCEl); /* delete 'this' */
269}
270
271static void udata_initHashTable() {
272 UErrorCode err = U_ZERO_ERROR;
273 U_ASSERT(gCommonDataCache == NULL);
274 gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err);
275 if (U_FAILURE(err)) {
276 // TODO: handle errors better.
277 gCommonDataCache = NULL;
278 }
279 if (gCommonDataCache != NULL) {
280 uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter);
281 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup);
282 }
283}
284
285 /* udata_getCacheHashTable()
286 * Get the hash table used to store the data cache entries.
287 * Lazy create it if it doesn't yet exist.
288 */
289static UHashtable *udata_getHashTable() {
290 umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable);
291 return gCommonDataCache;
292}
293
294
295
296static UDataMemory *udata_findCachedData(const char *path)
297{
298 UHashtable *htable;
299 UDataMemory *retVal = NULL;
300 DataCacheElement *el;
301 const char *baseName;
302
303 baseName = findBasename(path); /* Cache remembers only the base name, not the full path. */
304 htable = udata_getHashTable();
305 umtx_lock(NULL);
306 el = (DataCacheElement *)uhash_get(htable, baseName);
307 umtx_unlock(NULL);
308 if (el != NULL) {
309 retVal = el->item;
310 }
311#ifdef UDATA_DEBUG
312 fprintf(stderr, "Cache: [%s] -> %p\n", baseName, retVal);
313#endif
314 return retVal;
315}
316
317
318static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UErrorCode *pErr) {
319 DataCacheElement *newElement;
320 const char *baseName;
321 int32_t nameLen;
322 UHashtable *htable;
323 DataCacheElement *oldValue = NULL;
324 UErrorCode subErr = U_ZERO_ERROR;
325
326 if (U_FAILURE(*pErr)) {
327 return NULL;
328 }
329
330 /* Create a new DataCacheElement - the thingy we store in the hash table -
331 * and copy the supplied path and UDataMemoryItems into it.
332 */
333 newElement = (DataCacheElement *)uprv_malloc(sizeof(DataCacheElement));
334 if (newElement == NULL) {
335 *pErr = U_MEMORY_ALLOCATION_ERROR;
336 return NULL;
337 }
338 newElement->item = UDataMemory_createNewInstance(pErr);
339 if (U_FAILURE(*pErr)) {
340 uprv_free(newElement);
341 return NULL;
342 }
343 UDatamemory_assign(newElement->item, item);
344
345 baseName = findBasename(path);
346 nameLen = (int32_t)uprv_strlen(baseName);
347 newElement->name = (char *)uprv_malloc(nameLen+1);
348 if (newElement->name == NULL) {
349 *pErr = U_MEMORY_ALLOCATION_ERROR;
350 uprv_free(newElement->item);
351 uprv_free(newElement);
352 return NULL;
353 }
354 uprv_strcpy(newElement->name, baseName);
355
356 /* Stick the new DataCacheElement into the hash table.
357 */
358 htable = udata_getHashTable();
359 umtx_lock(NULL);
360 oldValue = (DataCacheElement *)uhash_get(htable, path);
361 if (oldValue != NULL) {
362 subErr = U_USING_DEFAULT_WARNING;
363 }
364 else {
365 uhash_put(
366 htable,
367 newElement->name, /* Key */
368 newElement, /* Value */
369 &subErr);
370 }
371 umtx_unlock(NULL);
372
373#ifdef UDATA_DEBUG
374 fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n", newElement->name,
375 newElement->item, u_errorName(subErr), newElement->item->vFuncs);
376#endif
377
378 if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) {
379 *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. */
380 uprv_free(newElement->name);
381 uprv_free(newElement->item);
382 uprv_free(newElement);
383 return oldValue ? oldValue->item : NULL;
384 }
385
386 return newElement->item;
387}
388
389/*----------------------------------------------------------------------*==============
390 * *
391 * Path management. Could be shared with other tools/etc if need be *
392 * later on. *
393 * *
394 *----------------------------------------------------------------------*/
395
396#define U_DATA_PATHITER_BUFSIZ 128 /* Size of local buffer for paths */
397 /* Overflow causes malloc of larger buf */
398
399U_NAMESPACE_BEGIN
400
401class UDataPathIterator
402{
403public:
404 UDataPathIterator(const char *path, const char *pkg,
405 const char *item, const char *suffix, UBool doCheckLastFour,
406 UErrorCode *pErrorCode);
407 const char *next(UErrorCode *pErrorCode);
408
409private:
410 const char *path; /* working path (u_icudata_Dir) */
411 const char *nextPath; /* path following this one */
412 const char *basename; /* item's basename (icudt22e_mt.res)*/
413 const char *suffix; /* item suffix (can be null) */
414
415 uint32_t basenameLen; /* length of basename */
416
417 CharString itemPath; /* path passed in with item name */
418 CharString pathBuffer; /* output path for this it'ion */
419 CharString packageStub; /* example: "/icudt28b". Will ignore that leaf in set paths. */
420
421 UBool checkLastFour; /* if TRUE then allow paths such as '/foo/myapp.dat'
422 * to match, checks last 4 chars of suffix with
423 * last 4 of path, then previous chars. */
424};
425
426/**
427 * @param iter The iterator to be initialized. Its current state does not matter.
428 * @param path The full pathname to be iterated over. If NULL, defaults to U_ICUDATA_NAME
429 * @param pkg Package which is being searched for, ex "icudt28l". Will ignore leave directories such as /icudt28l
430 * @param item Item to be searched for. Can include full path, such as /a/b/foo.dat
431 * @param suffix Optional item suffix, if not-null (ex. ".dat") then 'path' can contain 'item' explicitly.
432 * Ex: 'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar/baz' as item #2.
433 * '/blarg/stuff.dat' would also be found.
434 */
435UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg,
436 const char *item, const char *inSuffix, UBool doCheckLastFour,
437 UErrorCode *pErrorCode)
438{
439#ifdef UDATA_DEBUG
440 fprintf(stderr, "SUFFIX1=%s PATH=%s\n", inSuffix, inPath);
441#endif
442 /** Path **/
443 if(inPath == NULL) {
444 path = u_getDataDirectory();
445 } else {
446 path = inPath;
447 }
448
449 /** Package **/
450 if(pkg != NULL) {
451 packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode);
452#ifdef UDATA_DEBUG
453 fprintf(stderr, "STUB=%s [%d]\n", packageStub.data(), packageStub.length());
454#endif
455 }
456
457 /** Item **/
458 basename = findBasename(item);
459 basenameLen = (int32_t)uprv_strlen(basename);
460
461 /** Item path **/
462 if(basename == item) {
463 nextPath = path;
464 } else {
465 itemPath.append(item, (int32_t)(basename-item), *pErrorCode);
466 nextPath = itemPath.data();
467 }
468#ifdef UDATA_DEBUG
469 fprintf(stderr, "SUFFIX=%s [%p]\n", inSuffix, inSuffix);
470#endif
471
472 /** Suffix **/
473 if(inSuffix != NULL) {
474 suffix = inSuffix;
475 } else {
476 suffix = "";
477 }
478
479 checkLastFour = doCheckLastFour;
480
481 /* pathBuffer will hold the output path strings returned by this iterator */
482
483#ifdef UDATA_DEBUG
484 fprintf(stderr, "%p: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%s], [nextpath=%s], [checklast4=%s]\n",
485 iter,
486 item,
487 path,
488 basename,
489 suffix,
490 itemPath.data(),
491 nextPath,
492 checkLastFour?"TRUE":"false");
493#endif
494}
495
496/**
497 * Get the next path on the list.
498 *
499 * @param iter The Iter to be used
500 * @param len If set, pointer to the length of the returned path, for convenience.
501 * @return Pointer to the next path segment, or NULL if there are no more.
502 */
503const char *UDataPathIterator::next(UErrorCode *pErrorCode)
504{
505 if(U_FAILURE(*pErrorCode)) {
506 return NULL;
507 }
508
509 const char *currentPath = NULL;
510 int32_t pathLen = 0;
511 const char *pathBasename;
512
513 do
514 {
515 if( nextPath == NULL ) {
516 break;
517 }
518 currentPath = nextPath;
519
520 if(nextPath == itemPath.data()) { /* we were processing item's path. */
521 nextPath = path; /* start with regular path next tm. */
522 pathLen = (int32_t)uprv_strlen(currentPath);
523 } else {
524 /* fix up next for next time */
525 nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR);
526 if(nextPath == NULL) {
527 /* segment: entire path */
528 pathLen = (int32_t)uprv_strlen(currentPath);
529 } else {
530 /* segment: until next segment */
531 pathLen = (int32_t)(nextPath - currentPath);
532 /* skip divider */
533 nextPath ++;
534 }
535 }
536
537 if(pathLen == 0) {
538 continue;
539 }
540
541#ifdef UDATA_DEBUG
542 fprintf(stderr, "rest of path (IDD) = %s\n", currentPath);
543 fprintf(stderr, " ");
544 {
545 uint32_t qqq;
546 for(qqq=0;qqq<pathLen;qqq++)
547 {
548 fprintf(stderr, " ");
549 }
550
551 fprintf(stderr, "^\n");
552 }
553#endif
554 pathBuffer.clear().append(currentPath, pathLen, *pErrorCode);
555
556 /* check for .dat files */
557 pathBasename = findBasename(pathBuffer.data());
558
559 if(checkLastFour == TRUE &&
560 (pathLen>=4) &&
561 uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix, 4)==0 && /* suffix matches */
562 uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)==0 && /* base matches */
563 uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full len */
564
565#ifdef UDATA_DEBUG
566 fprintf(stderr, "Have %s file on the path: %s\n", suffix, pathBuffer.data());
567#endif
568 /* do nothing */
569 }
570 else
571 { /* regular dir path */
572 if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) {
573 if((pathLen>=4) &&
574 uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat", 4) == 0)
575 {
576#ifdef UDATA_DEBUG
577 fprintf(stderr, "skipping non-directory .dat file %s\n", pathBuffer.data());
578#endif
579 continue;
580 }
581
582 /* Check if it is a directory with the same name as our package */
583 if(!packageStub.isEmpty() &&
584 (pathLen > packageStub.length()) &&
585 !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length(), packageStub.data())) {
586#ifdef UDATA_DEBUG
587 fprintf(stderr, "Found stub %s (will add package %s of len %d)\n", packageStub.data(), basename, basenameLen);
588#endif
589 pathBuffer.truncate(pathLen - packageStub.length());
590 }
591 pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode);
592 }
593
594 /* + basename */
595 pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pErrorCode);
596
597 if(*suffix) /* tack on suffix */
598 {
599 pathBuffer.append(suffix, *pErrorCode);
600 }
601 }
602
603#ifdef UDATA_DEBUG
604 fprintf(stderr, " --> %s\n", pathBuffer.data());
605#endif
606
607 return pathBuffer.data();
608
609 } while(path);
610
611 /* fell way off the end */
612 return NULL;
613}
614
615U_NAMESPACE_END
616
617/* ==================================================================================*/
618
619
620/*----------------------------------------------------------------------*
621 * *
622 * Add a static reference to the common data library *
623 * Unless overridden by an explicit udata_setCommonData, this will be *
624 * our common data. *
625 * *
626 *----------------------------------------------------------------------*/
Jungshik Shin (jungshik at google)5580ad62015-01-12 13:00:44 -0800627extern "C" const ICU_Data_Header U_DATA_API U_ICUDATA_ENTRY_POINT;
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000628
629/*
630 * This would be a good place for weak-linkage declarations of
631 * partial-data-library access functions where each returns a pointer
632 * to its data package, if it is linked in.
633 */
634/*
635extern const void *uprv_getICUData_collation(void) ATTRIBUTE_WEAK;
636extern const void *uprv_getICUData_conversion(void) ATTRIBUTE_WEAK;
637*/
638
639/*----------------------------------------------------------------------*
640 * *
641 * openCommonData Attempt to open a common format (.dat) file *
642 * Map it into memory (if it's not there already) *
643 * and return a UDataMemory object for it. *
644 * *
645 * If the requested data is already open and cached *
646 * just return the cached UDataMem object. *
647 * *
648 *----------------------------------------------------------------------*/
649static UDataMemory *
650openCommonData(const char *path, /* Path from OpenChoice? */
651 int32_t commonDataIndex, /* ICU Data (index >= 0) if path == NULL */
652 UErrorCode *pErrorCode)
653{
654 UDataMemory tData;
655 const char *pathBuffer;
656 const char *inBasename;
657
658 if (U_FAILURE(*pErrorCode)) {
659 return NULL;
660 }
661
662 UDataMemory_init(&tData);
663
664 /* ??????? TODO revisit this */
665 if (commonDataIndex >= 0) {
666 /* "mini-cache" for common ICU data */
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -0800667 if(commonDataIndex >= UPRV_LENGTHOF(gCommonICUDataArray)) {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000668 return NULL;
669 }
Jungshik Shin508e9272015-07-16 15:36:55 -0700670 {
671 Mutex lock;
672 if(gCommonICUDataArray[commonDataIndex] != NULL) {
673 return gCommonICUDataArray[commonDataIndex];
674 }
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000675 int32_t i;
676 for(i = 0; i < commonDataIndex; ++i) {
Jungshik Shin (jungshik at google)5580ad62015-01-12 13:00:44 -0800677 if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT.hdr) {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000678 /* The linked-in data is already in the list. */
679 return NULL;
680 }
681 }
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000682 }
Jungshik Shin508e9272015-07-16 15:36:55 -0700683
684 /* Add the linked-in data to the list. */
685 /*
686 * This is where we would check and call weakly linked partial-data-library
687 * access functions.
688 */
689 /*
690 if (uprv_getICUData_collation) {
691 setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErrorCode);
692 }
693 if (uprv_getICUData_conversion) {
694 setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pErrorCode);
695 }
696 */
697 setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT.hdr, FALSE, pErrorCode);
698 {
699 Mutex lock;
700 return gCommonICUDataArray[commonDataIndex];
701 }
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000702 }
703
704
705 /* request is NOT for ICU Data. */
706
707 /* Find the base name portion of the supplied path. */
708 /* inBasename will be left pointing somewhere within the original path string. */
709 inBasename = findBasename(path);
710#ifdef UDATA_DEBUG
711 fprintf(stderr, "inBasename = %s\n", inBasename);
712#endif
713
714 if(*inBasename==0) {
715 /* no basename. This will happen if the original path was a directory name, */
716 /* like "a/b/c/". (Fallback to separate files will still work.) */
717#ifdef UDATA_DEBUG
718 fprintf(stderr, "ocd: no basename in %s, bailing.\n", path);
719#endif
720 *pErrorCode=U_FILE_ACCESS_ERROR;
721 return NULL;
722 }
723
724 /* Is the requested common data file already open and cached? */
725 /* Note that the cache is keyed by the base name only. The rest of the path, */
726 /* if any, is not considered. */
727 {
728 UDataMemory *dataToReturn = udata_findCachedData(inBasename);
729 if (dataToReturn != NULL) {
730 return dataToReturn;
731 }
732 }
733
734 /* Requested item is not in the cache.
735 * Hunt it down, trying all the path locations
736 */
737
738 UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat", TRUE, pErrorCode);
739
740 while((UDataMemory_isLoaded(&tData)==FALSE) && (pathBuffer = iter.next(pErrorCode)) != NULL)
741 {
742#ifdef UDATA_DEBUG
743 fprintf(stderr, "ocd: trying path %s - ", pathBuffer);
744#endif
745 uprv_mapFile(&tData, pathBuffer);
746#ifdef UDATA_DEBUG
747 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&tData)?"LOADED":"not loaded");
748#endif
749 }
750
751#if defined(OS390_STUBDATA) && defined(OS390BATCH)
752 if (!UDataMemory_isLoaded(&tData)) {
753 char ourPathBuffer[1024];
754 /* One more chance, for extendCommonData() */
755 uprv_strncpy(ourPathBuffer, path, 1019);
756 ourPathBuffer[1019]=0;
757 uprv_strcat(ourPathBuffer, ".dat");
758 uprv_mapFile(&tData, ourPathBuffer);
759 }
760#endif
761
762 if (!UDataMemory_isLoaded(&tData)) {
763 /* no common data */
764 *pErrorCode=U_FILE_ACCESS_ERROR;
765 return NULL;
766 }
767
768 /* we have mapped a file, check its header */
769 udata_checkCommonData(&tData, pErrorCode);
770
771
772 /* Cache the UDataMemory struct for this .dat file,
773 * so we won't need to hunt it down and map it again next time
774 * something is needed from it. */
775 return udata_cacheDataItem(inBasename, &tData, pErrorCode);
776}
777
778
779/*----------------------------------------------------------------------*
780 * *
781 * extendICUData If the full set of ICU data was not loaded at *
782 * program startup, load it now. This function will *
783 * be called when the lookup of an ICU data item in *
784 * the common ICU data fails. *
785 * *
786 * return true if new data is loaded, false otherwise.*
787 * *
788 *----------------------------------------------------------------------*/
789static UBool extendICUData(UErrorCode *pErr)
790{
791 UDataMemory *pData;
792 UDataMemory copyPData;
793 UBool didUpdate = FALSE;
794
795 /*
796 * There is a chance for a race condition here.
797 * Normally, ICU data is loaded from a DLL or via mmap() and
798 * setCommonICUData() will detect if the same address is set twice.
799 * If ICU is built with data loading via fread() then the address will
800 * be different each time the common data is loaded and we may add
801 * multiple copies of the data.
802 * In this case, use a mutex to prevent the race.
803 * Use a specific mutex to avoid nested locks of the global mutex.
804 */
805#if MAP_IMPLEMENTATION==MAP_STDIO
806 static UMutex extendICUDataMutex = U_MUTEX_INITIALIZER;
807 umtx_lock(&extendICUDataMutex);
808#endif
Jungshik Shin508e9272015-07-16 15:36:55 -0700809 if(!umtx_loadAcquire(gHaveTriedToLoadCommonData)) {
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000810 /* See if we can explicitly open a .dat file for the ICUData. */
811 pData = openCommonData(
812 U_ICUDATA_NAME, /* "icudt20l" , for example. */
813 -1, /* Pretend we're not opening ICUData */
814 pErr);
815
816 /* How about if there is no pData, eh... */
817
818 UDataMemory_init(&copyPData);
819 if(pData != NULL) {
820 UDatamemory_assign(&copyPData, pData);
821 copyPData.map = 0; /* The mapping for this data is owned by the hash table */
822 copyPData.mapAddr = 0; /* which will unmap it when ICU is shut down. */
823 /* CommonICUData is also unmapped when ICU is shut down.*/
824 /* To avoid unmapping the data twice, zero out the map */
825 /* fields in the UDataMemory that we're assigning */
826 /* to CommonICUData. */
827
828 didUpdate = /* no longer using this result */
829 setCommonICUData(&copyPData,/* The new common data. */
830 FALSE, /* No warnings if write didn't happen */
831 pErr); /* setCommonICUData honors errors; NOP if error set */
832 }
833
Jungshik Shin508e9272015-07-16 15:36:55 -0700834 umtx_storeRelease(gHaveTriedToLoadCommonData, 1);
jshin@chromium.org6f31ac32014-03-26 22:15:14 +0000835 }
836
837 didUpdate = findCommonICUDataByName(U_ICUDATA_NAME); /* Return 'true' when a racing writes out the extended */
838 /* data after another thread has failed to see it (in openCommonData), so */
839 /* extended data can be examined. */
840 /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */
841
842#if MAP_IMPLEMENTATION==MAP_STDIO
843 umtx_unlock(&extendICUDataMutex);
844#endif
845 return didUpdate; /* Return true if ICUData pointer was updated. */
846 /* (Could potentialy have been done by another thread racing */
847 /* us through here, but that's fine, we still return true */
848 /* so that current thread will also examine extended data. */
849}
850
851/*----------------------------------------------------------------------*
852 * *
853 * udata_setCommonData *
854 * *
855 *----------------------------------------------------------------------*/
856U_CAPI void U_EXPORT2
857udata_setCommonData(const void *data, UErrorCode *pErrorCode) {
858 UDataMemory dataMemory;
859
860 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
861 return;
862 }
863
864 if(data==NULL) {
865 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
866 return;
867 }
868
869 /* set the data pointer and test for validity */
870 UDataMemory_init(&dataMemory);
871 UDataMemory_setData(&dataMemory, data);
872 udata_checkCommonData(&dataMemory, pErrorCode);
873 if (U_FAILURE(*pErrorCode)) {return;}
874
875 /* we have good data */
876 /* Set it up as the ICU Common Data. */
877 setCommonICUData(&dataMemory, TRUE, pErrorCode);
878}
879
880/*---------------------------------------------------------------------------
881 *
882 * udata_setAppData
883 *
884 *---------------------------------------------------------------------------- */
885U_CAPI void U_EXPORT2
886udata_setAppData(const char *path, const void *data, UErrorCode *err)
887{
888 UDataMemory udm;
889
890 if(err==NULL || U_FAILURE(*err)) {
891 return;
892 }
893 if(data==NULL) {
894 *err=U_ILLEGAL_ARGUMENT_ERROR;
895 return;
896 }
897
898 UDataMemory_init(&udm);
899 UDataMemory_setData(&udm, data);
900 udata_checkCommonData(&udm, err);
901 udata_cacheDataItem(path, &udm, err);
902}
903
904/*----------------------------------------------------------------------------*
905 * *
906 * checkDataItem Given a freshly located/loaded data item, either *
907 * an entry in a common file or a separately loaded file, *
908 * sanity check its header, and see if the data is *
909 * acceptable to the app. *
910 * If the data is good, create and return a UDataMemory *
911 * object that can be returned to the application. *
912 * Return NULL on any sort of failure. *
913 * *
914 *----------------------------------------------------------------------------*/
915static UDataMemory *
916checkDataItem
917(
918 const DataHeader *pHeader, /* The data item to be checked. */
919 UDataMemoryIsAcceptable *isAcceptable, /* App's call-back function */
920 void *context, /* pass-thru param for above. */
921 const char *type, /* pass-thru param for above. */
922 const char *name, /* pass-thru param for above. */
923 UErrorCode *nonFatalErr, /* Error code if this data was not acceptable */
924 /* but openChoice should continue with */
925 /* trying to get data from fallback path. */
926 UErrorCode *fatalErr /* Bad error, caller should return immediately */
927 )
928{
929 UDataMemory *rDataMem = NULL; /* the new UDataMemory, to be returned. */
930
931 if (U_FAILURE(*fatalErr)) {
932 return NULL;
933 }
934
935 if(pHeader->dataHeader.magic1==0xda &&
936 pHeader->dataHeader.magic2==0x27 &&
937 (isAcceptable==NULL || isAcceptable(context, type, name, &pHeader->info))
938 ) {
939 rDataMem=UDataMemory_createNewInstance(fatalErr);
940 if (U_FAILURE(*fatalErr)) {
941 return NULL;
942 }
943 rDataMem->pHeader = pHeader;
944 } else {
945 /* the data is not acceptable, look further */
946 /* If we eventually find something good, this errorcode will be */
947 /* cleared out. */
948 *nonFatalErr=U_INVALID_FORMAT_ERROR;
949 }
950 return rDataMem;
951}
952
953/**
954 * @return 0 if not loaded, 1 if loaded or err
955 */
956static UDataMemory *doLoadFromIndividualFiles(const char *pkgName,
957 const char *dataPath, const char *tocEntryPathSuffix,
958 /* following arguments are the same as doOpenChoice itself */
959 const char *path, const char *type, const char *name,
960 UDataMemoryIsAcceptable *isAcceptable, void *context,
961 UErrorCode *subErrorCode,
962 UErrorCode *pErrorCode)
963{
964 const char *pathBuffer;
965 UDataMemory dataMemory;
966 UDataMemory *pEntryData;
967
968 /* look in ind. files: package\nam.typ ========================= */
969 /* init path iterator for individual files */
970 UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, FALSE, pErrorCode);
971
972 while((pathBuffer = iter.next(pErrorCode)))
973 {
974#ifdef UDATA_DEBUG
975 fprintf(stderr, "UDATA: trying individual file %s\n", pathBuffer);
976#endif
977 if(uprv_mapFile(&dataMemory, pathBuffer))
978 {
979 pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode);
980 if (pEntryData != NULL) {
981 /* Data is good.
982 * Hand off ownership of the backing memory to the user's UDataMemory.
983 * and return it. */
984 pEntryData->mapAddr = dataMemory.mapAddr;
985 pEntryData->map = dataMemory.map;
986
987#ifdef UDATA_DEBUG
988 fprintf(stderr, "** Mapped file: %s\n", pathBuffer);
989#endif
990 return pEntryData;
991 }
992
993 /* the data is not acceptable, or some error occured. Either way, unmap the memory */
994 udata_close(&dataMemory);
995
996 /* If we had a nasty error, bail out completely. */
997 if (U_FAILURE(*pErrorCode)) {
998 return NULL;
999 }
1000
1001 /* Otherwise remember that we found data but didn't like it for some reason */
1002 *subErrorCode=U_INVALID_FORMAT_ERROR;
1003 }
1004#ifdef UDATA_DEBUG
1005 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&dataMemory)?"LOADED":"not loaded");
1006#endif
1007 }
1008 return NULL;
1009}
1010
1011/**
1012 * @return 0 if not loaded, 1 if loaded or err
1013 */
1014static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName*/,
1015 const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const char *tocEntryName,
1016 /* following arguments are the same as doOpenChoice itself */
1017 const char *path, const char *type, const char *name,
1018 UDataMemoryIsAcceptable *isAcceptable, void *context,
1019 UErrorCode *subErrorCode,
1020 UErrorCode *pErrorCode)
1021{
1022 UDataMemory *pEntryData;
1023 const DataHeader *pHeader;
1024 UDataMemory *pCommonData;
1025 int32_t commonDataIndex;
1026 UBool checkedExtendedICUData = FALSE;
1027 /* try to get common data. The loop is for platforms such as the 390 that do
1028 * not initially load the full set of ICU data. If the lookup of an ICU data item
1029 * fails, the full (but slower to load) set is loaded, the and the loop repeats,
1030 * trying the lookup again. Once the full set of ICU data is loaded, the loop wont
1031 * repeat because the full set will be checked the first time through.
1032 *
1033 * The loop also handles the fallback to a .dat file if the application linked
1034 * to the stub data library rather than a real library.
1035 */
1036 for (commonDataIndex = isICUData ? 0 : -1;;) {
1037 pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** search for pkg **/
1038
1039 if(U_SUCCESS(*subErrorCode) && pCommonData!=NULL) {
1040 int32_t length;
1041
1042 /* look up the data piece in the common data */
1043 pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &length, subErrorCode);
1044#ifdef UDATA_DEBUG
1045 fprintf(stderr, "%s: pHeader=%p - %s\n", tocEntryName, pHeader, u_errorName(*subErrorCode));
1046#endif
1047
1048 if(pHeader!=NULL) {
1049 pEntryData = checkDataItem(pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode);
1050#ifdef UDATA_DEBUG
1051 fprintf(stderr, "pEntryData=%p\n", pEntryData);
1052#endif
1053 if (U_FAILURE(*pErrorCode)) {
1054 return NULL;
1055 }
1056 if (pEntryData != NULL) {
1057 pEntryData->length = length;
1058 return pEntryData;
1059 }
1060 }
1061 }
1062 /* Data wasn't found. If we were looking for an ICUData item and there is
1063 * more data available, load it and try again,
1064 * otherwise break out of this loop. */
1065 if (!isICUData) {
1066 return NULL;
1067 } else if (pCommonData != NULL) {
1068 ++commonDataIndex; /* try the next data package */
1069 } else if ((!checkedExtendedICUData) && extendICUData(subErrorCode)) {
1070 checkedExtendedICUData = TRUE;
1071 /* try this data package slot again: it changed from NULL to non-NULL */
1072 } else {
1073 return NULL;
1074 }
1075 }
1076}
1077
1078/*
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -08001079 * Identify the Time Zone resources that are subject to special override data loading.
1080 */
1081static UBool isTimeZoneFile(const char *name, const char *type) {
1082 return ((uprv_strcmp(type, "res") == 0) &&
1083 (uprv_strcmp(name, "zoneinfo64") == 0 ||
1084 uprv_strcmp(name, "timezoneTypes") == 0 ||
1085 uprv_strcmp(name, "windowsZones") == 0 ||
1086 uprv_strcmp(name, "metaZones") == 0));
1087}
1088
1089/*
jshin@chromium.org6f31ac32014-03-26 22:15:14 +00001090 * A note on the ownership of Mapped Memory
1091 *
1092 * For common format files, ownership resides with the UDataMemory object
1093 * that lives in the cache of opened common data. These UDataMemorys are private
1094 * to the udata implementation, and are never seen directly by users.
1095 *
1096 * The UDataMemory objects returned to users will have the address of some desired
1097 * data within the mapped region, but they wont have the mapping info itself, and thus
1098 * won't cause anything to be removed from memory when they are closed.
1099 *
1100 * For individual data files, the UDataMemory returned to the user holds the
1101 * information necessary to unmap the data on close. If the user independently
1102 * opens the same data file twice, two completely independent mappings will be made.
1103 * (There is no cache of opened data items from individual files, only a cache of
1104 * opened Common Data files, that is, files containing a collection of data items.)
1105 *
1106 * For common data passed in from the user via udata_setAppData() or
1107 * udata_setCommonData(), ownership remains with the user.
1108 *
1109 * UDataMemory objects themselves, as opposed to the memory they describe,
1110 * can be anywhere - heap, stack/local or global.
1111 * They have a flag to indicate when they're heap allocated and thus
1112 * must be deleted when closed.
1113 */
1114
1115
1116/*----------------------------------------------------------------------------*
1117 * *
1118 * main data loading functions *
1119 * *
1120 *----------------------------------------------------------------------------*/
1121static UDataMemory *
1122doOpenChoice(const char *path, const char *type, const char *name,
1123 UDataMemoryIsAcceptable *isAcceptable, void *context,
1124 UErrorCode *pErrorCode)
1125{
1126 UDataMemory *retVal = NULL;
1127
1128 const char *dataPath;
1129
1130 int32_t tocEntrySuffixIndex;
1131 const char *tocEntryPathSuffix;
1132 UErrorCode subErrorCode=U_ZERO_ERROR;
1133 const char *treeChar;
1134
1135 UBool isICUData = FALSE;
1136
1137
1138 /* Is this path ICU data? */
1139 if(path == NULL ||
1140 !strcmp(path, U_ICUDATA_ALIAS) || /* "ICUDATA" */
1141 !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-" */
1142 uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) ||
1143 !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-" */
1144 uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) {
1145 isICUData = TRUE;
1146 }
1147
1148#if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR) /* Windows: try "foo\bar" and "foo/bar" */
1149 /* remap from alternate path char to the main one */
1150 CharString altSepPath;
1151 if(path) {
1152 if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != NULL) {
1153 altSepPath.append(path, *pErrorCode);
1154 char *p;
1155 while((p=uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR))) {
1156 *p = U_FILE_SEP_CHAR;
1157 }
1158#if defined (UDATA_DEBUG)
1159 fprintf(stderr, "Changed path from [%s] to [%s]\n", path, altSepPath.s);
1160#endif
1161 path = altSepPath.data();
1162 }
1163 }
1164#endif
1165
1166 CharString tocEntryName; /* entry name in tree format. ex: 'icudt28b/coll/ar.res' */
1167 CharString tocEntryPath; /* entry name in path format. ex: 'icudt28b\\coll\\ar.res' */
1168
1169 CharString pkgName;
1170 CharString treeName;
1171
1172 /* ======= Set up strings */
1173 if(path==NULL) {
1174 pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1175 } else {
1176 const char *pkg;
1177 const char *first;
1178 pkg = uprv_strrchr(path, U_FILE_SEP_CHAR);
1179 first = uprv_strchr(path, U_FILE_SEP_CHAR);
1180 if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash in the path- not a tree name */
1181 /* see if this is an /absolute/path/to/package path */
1182 if(pkg) {
1183 pkgName.append(pkg+1, *pErrorCode);
1184 } else {
1185 pkgName.append(path, *pErrorCode);
1186 }
1187 } else {
1188 treeChar = uprv_strchr(path, U_TREE_SEPARATOR);
1189 if(treeChar) {
1190 treeName.append(treeChar+1, *pErrorCode); /* following '-' */
1191 if(isICUData) {
1192 pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1193 } else {
1194 pkgName.append(path, (int32_t)(treeChar-path), *pErrorCode);
1195 if (first == NULL) {
1196 /*
1197 This user data has no path, but there is a tree name.
1198 Look up the correct path from the data cache later.
1199 */
1200 path = pkgName.data();
1201 }
1202 }
1203 } else {
1204 if(isICUData) {
1205 pkgName.append(U_ICUDATA_NAME, *pErrorCode);
1206 } else {
1207 pkgName.append(path, *pErrorCode);
1208 }
1209 }
1210 }
1211 }
1212
1213#ifdef UDATA_DEBUG
1214 fprintf(stderr, " P=%s T=%s\n", pkgName.data(), treeName.data());
1215#endif
1216
1217 /* setting up the entry name and file name
1218 * Make up a full name by appending the type to the supplied
1219 * name, assuming that a type was supplied.
1220 */
1221
1222 /* prepend the package */
1223 tocEntryName.append(pkgName, *pErrorCode);
1224 tocEntryPath.append(pkgName, *pErrorCode);
1225 tocEntrySuffixIndex = tocEntryName.length();
1226
1227 if(!treeName.isEmpty()) {
1228 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode);
1229 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode);
1230 }
1231
1232 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pErrorCode);
1233 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode);
1234 if(type!=NULL && *type!=0) {
1235 tocEntryName.append(".", *pErrorCode).append(type, *pErrorCode);
1236 tocEntryPath.append(".", *pErrorCode).append(type, *pErrorCode);
1237 }
1238 tocEntryPathSuffix = tocEntryPath.data()+tocEntrySuffixIndex; /* suffix starts here */
1239
1240#ifdef UDATA_DEBUG
1241 fprintf(stderr, " tocEntryName = %s\n", tocEntryName.data());
1242 fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data());
1243#endif
1244
1245 if(path == NULL) {
1246 path = COMMON_DATA_NAME; /* "icudt26e" */
1247 }
1248
1249 /************************ Begin loop looking for ind. files ***************/
1250#ifdef UDATA_DEBUG
1251 fprintf(stderr, "IND: inBasename = %s, pkg=%s\n", "(n/a)", packageNameFromPath(path));
1252#endif
1253
1254 /* End of dealing with a null basename */
1255 dataPath = u_getDataDirectory();
1256
Jungshik Shin (jungshik at google)0f8746a2015-01-08 15:46:45 -08001257 /**** Time zone individual files override */
1258 if (isTimeZoneFile(name, type) && isICUData) {
1259 const char *tzFilesDir = u_getTimeZoneFilesDirectory(pErrorCode);
1260 if (tzFilesDir[0] != 0) {
1261#ifdef UDATA_DEBUG
1262 fprintf(stderr, "Trying Time Zone Files directory = %s\n", tzFilesDir);
1263#endif
1264 retVal = doLoadFromIndividualFiles(/* pkgName.data() */ "", tzFilesDir, tocEntryPathSuffix,
1265 /* path */ "", type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1266 if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1267 return retVal;
1268 }
1269 }
1270 }
1271
jshin@chromium.org6f31ac32014-03-26 22:15:14 +00001272 /**** COMMON PACKAGE - only if packages are first. */
1273 if(gDataFileAccess == UDATA_PACKAGES_FIRST) {
1274#ifdef UDATA_DEBUG
1275 fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n");
1276#endif
1277 /* #2 */
1278 retVal = doLoadFromCommonData(isICUData,
1279 pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(),
1280 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1281 if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1282 return retVal;
1283 }
1284 }
1285
1286 /**** INDIVIDUAL FILES */
1287 if((gDataFileAccess==UDATA_PACKAGES_FIRST) ||
1288 (gDataFileAccess==UDATA_FILES_FIRST)) {
1289#ifdef UDATA_DEBUG
1290 fprintf(stderr, "Trying individual files\n");
1291#endif
1292 /* Check to make sure that there is a dataPath to iterate over */
1293 if ((dataPath && *dataPath) || !isICUData) {
1294 retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix,
1295 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1296 if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1297 return retVal;
1298 }
1299 }
1300 }
1301
1302 /**** COMMON PACKAGE */
1303 if((gDataFileAccess==UDATA_ONLY_PACKAGES) ||
1304 (gDataFileAccess==UDATA_FILES_FIRST)) {
1305#ifdef UDATA_DEBUG
1306 fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n");
1307#endif
1308 retVal = doLoadFromCommonData(isICUData,
1309 pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(),
1310 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1311 if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1312 return retVal;
1313 }
1314 }
1315
1316 /* Load from DLL. If we haven't attempted package load, we also haven't had any chance to
1317 try a DLL (static or setCommonData/etc) load.
1318 If we ever have a "UDATA_ONLY_FILES", add it to the or list here. */
1319 if(gDataFileAccess==UDATA_NO_FILES) {
1320#ifdef UDATA_DEBUG
1321 fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n");
1322#endif
1323 retVal = doLoadFromCommonData(isICUData,
1324 pkgName.data(), "", tocEntryPathSuffix, tocEntryName.data(),
1325 path, type, name, isAcceptable, context, &subErrorCode, pErrorCode);
1326 if((retVal != NULL) || U_FAILURE(*pErrorCode)) {
1327 return retVal;
1328 }
1329 }
1330
1331 /* data not found */
1332 if(U_SUCCESS(*pErrorCode)) {
1333 if(U_SUCCESS(subErrorCode)) {
1334 /* file not found */
1335 *pErrorCode=U_FILE_ACCESS_ERROR;
1336 } else {
1337 /* entry point not found or rejected */
1338 *pErrorCode=subErrorCode;
1339 }
1340 }
1341 return retVal;
1342}
1343
1344
1345
1346/* API ---------------------------------------------------------------------- */
1347
1348U_CAPI UDataMemory * U_EXPORT2
1349udata_open(const char *path, const char *type, const char *name,
1350 UErrorCode *pErrorCode) {
1351#ifdef UDATA_DEBUG
1352 fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type);
1353 fflush(stderr);
1354#endif
1355
1356 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
1357 return NULL;
1358 } else if(name==NULL || *name==0) {
1359 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1360 return NULL;
1361 } else {
1362 return doOpenChoice(path, type, name, NULL, NULL, pErrorCode);
1363 }
1364}
1365
1366
1367
1368U_CAPI UDataMemory * U_EXPORT2
1369udata_openChoice(const char *path, const char *type, const char *name,
1370 UDataMemoryIsAcceptable *isAcceptable, void *context,
1371 UErrorCode *pErrorCode) {
1372#ifdef UDATA_DEBUG
1373 fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type);
1374#endif
1375
1376 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
1377 return NULL;
1378 } else if(name==NULL || *name==0 || isAcceptable==NULL) {
1379 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1380 return NULL;
1381 } else {
1382 return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode);
1383 }
1384}
1385
1386
1387
1388U_CAPI void U_EXPORT2
1389udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) {
1390 if(pInfo!=NULL) {
1391 if(pData!=NULL && pData->pHeader!=NULL) {
1392 const UDataInfo *info=&pData->pHeader->info;
1393 uint16_t dataInfoSize=udata_getInfoSize(info);
1394 if(pInfo->size>dataInfoSize) {
1395 pInfo->size=dataInfoSize;
1396 }
1397 uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->size-2);
1398 if(info->isBigEndian!=U_IS_BIG_ENDIAN) {
1399 /* opposite endianness */
1400 uint16_t x=info->reservedWord;
1401 pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8));
1402 }
1403 } else {
1404 pInfo->size=0;
1405 }
1406 }
1407}
1408
1409
1410U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/)
1411{
Jungshik Shin508e9272015-07-16 15:36:55 -07001412 // Note: this function is documented as not thread safe.
jshin@chromium.org6f31ac32014-03-26 22:15:14 +00001413 gDataFileAccess = access;
1414}