OLD | NEW |
(Empty) | |
| 1 /* |
| 2 ****************************************************************************** |
| 3 * |
| 4 * Copyright (C) 1999-2010, 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_LINUX */ |
| 18 |
| 19 #ifdef U_LINUX |
| 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 "ucln_cmn.h" |
| 34 #include "ucmndata.h" |
| 35 #include "udatamem.h" |
| 36 #include "uhash.h" |
| 37 #include "umapfile.h" |
| 38 #include "umutex.h" |
| 39 |
| 40 /*********************************************************************** |
| 41 * |
| 42 * Notes on the organization of the ICU data implementation |
| 43 * |
| 44 * All of the public API is defined in udata.h |
| 45 * |
| 46 * The implementation is split into several files... |
| 47 * |
| 48 * - udata.c (this file) contains higher level code that knows about |
| 49 * the search paths for locating data, caching opened data, e
tc. |
| 50 * |
| 51 * - umapfile.c contains the low level platform-specific code for actual
ly loading |
| 52 * (memory mapping, file reading, whatever) data into memory. |
| 53 * |
| 54 * - ucmndata.c deals with the tables of contents of ICU data items with
in |
| 55 * an ICU common format data file. The implementation includ
es |
| 56 * an abstract interface and support for multiple TOC formats
. |
| 57 * All knowledge of any specific TOC format is encapsulated h
ere. |
| 58 * |
| 59 * - udatamem.c has code for managing UDataMemory structs. These are lit
tle |
| 60 * descriptor objects for blocks of memory holding ICU data o
f |
| 61 * various types. |
| 62 */ |
| 63 |
| 64 /* configuration ---------------------------------------------------------- */ |
| 65 |
| 66 /* If you are excruciatingly bored turn this on .. */ |
| 67 /* #define UDATA_DEBUG 1 */ |
| 68 |
| 69 #if defined(UDATA_DEBUG) |
| 70 # include <stdio.h> |
| 71 #endif |
| 72 |
| 73 #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0])) |
| 74 |
| 75 U_NAMESPACE_USE |
| 76 |
| 77 /*********************************************************************** |
| 78 * |
| 79 * static (Global) data |
| 80 * |
| 81 ************************************************************************/ |
| 82 |
| 83 /* |
| 84 * Pointers to the common ICU data. |
| 85 * |
| 86 * We store multiple pointers to ICU data packages and iterate through them |
| 87 * when looking for a data item. |
| 88 * |
| 89 * It is possible to combine this with dependency inversion: |
| 90 * One or more data package libraries may export |
| 91 * functions that each return a pointer to their piece of the ICU data, |
| 92 * and this file would import them as weak functions, without a |
| 93 * strong linker dependency from the common library on the data library. |
| 94 * |
| 95 * Then we can have applications depend on only that part of ICU's data |
| 96 * that they really need, reducing the size of binaries that take advantage |
| 97 * of this. |
| 98 */ |
| 99 static UDataMemory *gCommonICUDataArray[10] = { NULL }; |
| 100 |
| 101 static UBool gHaveTriedToLoadCommonData = FALSE; /* See extendICUData(). */ |
| 102 |
| 103 static UHashtable *gCommonDataCache = NULL; /* Global hash table of opened ICU
data files. */ |
| 104 |
| 105 static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; |
| 106 |
| 107 static UBool U_CALLCONV |
| 108 udata_cleanup(void) |
| 109 { |
| 110 int32_t i; |
| 111 |
| 112 if (gCommonDataCache) { /* Delete the cache of user data mapping
s. */ |
| 113 uhash_close(gCommonDataCache); /* Table owns the contents, and will d
elete them. */ |
| 114 gCommonDataCache = NULL; /* Cleanup is not thread safe.
*/ |
| 115 } |
| 116 |
| 117 for (i = 0; i < LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != N
ULL; ++i) { |
| 118 udata_close(gCommonICUDataArray[i]); |
| 119 gCommonICUDataArray[i] = NULL; |
| 120 } |
| 121 gHaveTriedToLoadCommonData = FALSE; |
| 122 |
| 123 return TRUE; /* Everything was cleaned up */ |
| 124 } |
| 125 |
| 126 |
| 127 |
| 128 |
| 129 /* |
| 130 * setCommonICUData. Set a UDataMemory to be the global ICU Data |
| 131 */ |
| 132 static UBool |
| 133 setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to ca
ller, we copy it. */ |
| 134 UBool warn, /* If true, set USING_DEFAULT warning
if ICUData was */ |
| 135 /* changed by another thread before
we got to it. */ |
| 136 UErrorCode *pErr) |
| 137 { |
| 138 UDataMemory *newCommonData = UDataMemory_createNewInstance(pErr); |
| 139 int32_t i; |
| 140 UBool didUpdate = FALSE; |
| 141 if (U_FAILURE(*pErr)) { |
| 142 return FALSE; |
| 143 } |
| 144 |
| 145 /* For the assignment, other threads must cleanly see either the old
*/ |
| 146 /* or the new, not some partially initialized new. The old can not be
*/ |
| 147 /* deleted - someone may still have a pointer to it lying around in
*/ |
| 148 /* their locals.
*/ |
| 149 UDatamemory_assign(newCommonData, pData); |
| 150 umtx_lock(NULL); |
| 151 for (i = 0; i < LENGTHOF(gCommonICUDataArray); ++i) { |
| 152 if (gCommonICUDataArray[i] == NULL) { |
| 153 gCommonICUDataArray[i] = newCommonData; |
| 154 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
| 155 didUpdate = TRUE; |
| 156 break; |
| 157 } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) { |
| 158 /* The same data pointer is already in the array. */ |
| 159 break; |
| 160 } |
| 161 } |
| 162 umtx_unlock(NULL); |
| 163 |
| 164 if (i == LENGTHOF(gCommonICUDataArray) && warn) { |
| 165 *pErr = U_USING_DEFAULT_WARNING; |
| 166 } |
| 167 if (!didUpdate) { |
| 168 uprv_free(newCommonData); |
| 169 } |
| 170 return didUpdate; |
| 171 } |
| 172 |
| 173 static UBool |
| 174 setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCod
e) { |
| 175 UDataMemory tData; |
| 176 UDataMemory_init(&tData); |
| 177 tData.pHeader = (const DataHeader *)pData; |
| 178 udata_checkCommonData(&tData, pErrorCode); |
| 179 return setCommonICUData(&tData, FALSE, pErrorCode); |
| 180 } |
| 181 |
| 182 static const char * |
| 183 findBasename(const char *path) { |
| 184 const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR); |
| 185 if(basename==NULL) { |
| 186 return path; |
| 187 } else { |
| 188 return basename+1; |
| 189 } |
| 190 } |
| 191 |
| 192 #ifdef UDATA_DEBUG |
| 193 static const char * |
| 194 packageNameFromPath(const char *path) |
| 195 { |
| 196 if((path == NULL) || (*path == 0)) { |
| 197 return U_ICUDATA_NAME; |
| 198 } |
| 199 |
| 200 path = findBasename(path); |
| 201 |
| 202 if((path == NULL) || (*path == 0)) { |
| 203 return U_ICUDATA_NAME; |
| 204 } |
| 205 |
| 206 return path; |
| 207 } |
| 208 #endif |
| 209 |
| 210 /*----------------------------------------------------------------------* |
| 211 * * |
| 212 * Cache for common data * |
| 213 * Functions for looking up or adding entries to a cache of * |
| 214 * data that has been previously opened. Avoids a potentially * |
| 215 * expensive operation of re-opening the data for subsequent * |
| 216 * uses. * |
| 217 * * |
| 218 * Data remains cached for the duration of the process. * |
| 219 * * |
| 220 *----------------------------------------------------------------------*/ |
| 221 |
| 222 typedef struct DataCacheElement { |
| 223 char *name; |
| 224 UDataMemory *item; |
| 225 } DataCacheElement; |
| 226 |
| 227 |
| 228 |
| 229 /* |
| 230 * Deleter function for DataCacheElements. |
| 231 * udata cleanup function closes the hash table; hash table in turn call
s back to |
| 232 * here for each entry. |
| 233 */ |
| 234 static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) { |
| 235 DataCacheElement *p = (DataCacheElement *)pDCEl; |
| 236 udata_close(p->item); /* unmaps storage */ |
| 237 uprv_free(p->name); /* delete the hash key string. */ |
| 238 uprv_free(pDCEl); /* delete 'this' */ |
| 239 } |
| 240 |
| 241 /* udata_getCacheHashTable() |
| 242 * Get the hash table used to store the data cache entries. |
| 243 * Lazy create it if it doesn't yet exist. |
| 244 */ |
| 245 static UHashtable *udata_getHashTable() { |
| 246 UErrorCode err = U_ZERO_ERROR; |
| 247 UBool cacheIsInitialized; |
| 248 UHashtable *tHT = NULL; |
| 249 |
| 250 UMTX_CHECK(NULL, (gCommonDataCache != NULL), cacheIsInitialized); |
| 251 |
| 252 if (cacheIsInitialized) { |
| 253 return gCommonDataCache; |
| 254 } |
| 255 |
| 256 tHT = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err); |
| 257 /* Check for null pointer. */ |
| 258 if (tHT == NULL) { |
| 259 return NULL; /* TODO: Handle this error better. */ |
| 260 } |
| 261 uhash_setValueDeleter(tHT, DataCacheElement_deleter); |
| 262 |
| 263 umtx_lock(NULL); |
| 264 if (gCommonDataCache == NULL) { |
| 265 gCommonDataCache = tHT; |
| 266 tHT = NULL; |
| 267 ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
| 268 } |
| 269 umtx_unlock(NULL); |
| 270 if (tHT != NULL) { |
| 271 uhash_close(tHT); |
| 272 } |
| 273 |
| 274 if (U_FAILURE(err)) { |
| 275 return NULL; /* TODO: handle this error better. */ |
| 276 } |
| 277 return gCommonDataCache; |
| 278 } |
| 279 |
| 280 |
| 281 |
| 282 static UDataMemory *udata_findCachedData(const char *path) |
| 283 { |
| 284 UHashtable *htable; |
| 285 UDataMemory *retVal = NULL; |
| 286 DataCacheElement *el; |
| 287 const char *baseName; |
| 288 |
| 289 baseName = findBasename(path); /* Cache remembers only the base name, not
the full path. */ |
| 290 htable = udata_getHashTable(); |
| 291 umtx_lock(NULL); |
| 292 el = (DataCacheElement *)uhash_get(htable, baseName); |
| 293 umtx_unlock(NULL); |
| 294 if (el != NULL) { |
| 295 retVal = el->item; |
| 296 } |
| 297 #ifdef UDATA_DEBUG |
| 298 fprintf(stderr, "Cache: [%s] -> %p\n", baseName, retVal); |
| 299 #endif |
| 300 return retVal; |
| 301 } |
| 302 |
| 303 |
| 304 static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UEr
rorCode *pErr) { |
| 305 DataCacheElement *newElement; |
| 306 const char *baseName; |
| 307 int32_t nameLen; |
| 308 UHashtable *htable; |
| 309 DataCacheElement *oldValue = NULL; |
| 310 UErrorCode subErr = U_ZERO_ERROR; |
| 311 |
| 312 if (U_FAILURE(*pErr)) { |
| 313 return NULL; |
| 314 } |
| 315 |
| 316 /* Create a new DataCacheElement - the thingy we store in the hash table - |
| 317 * and copy the supplied path and UDataMemoryItems into it. |
| 318 */ |
| 319 newElement = (DataCacheElement *)uprv_malloc(sizeof(DataCacheElement)); |
| 320 if (newElement == NULL) { |
| 321 *pErr = U_MEMORY_ALLOCATION_ERROR; |
| 322 return NULL; |
| 323 } |
| 324 newElement->item = UDataMemory_createNewInstance(pErr); |
| 325 if (U_FAILURE(*pErr)) { |
| 326 uprv_free(newElement); |
| 327 return NULL; |
| 328 } |
| 329 UDatamemory_assign(newElement->item, item); |
| 330 |
| 331 baseName = findBasename(path); |
| 332 nameLen = (int32_t)uprv_strlen(baseName); |
| 333 newElement->name = (char *)uprv_malloc(nameLen+1); |
| 334 if (newElement->name == NULL) { |
| 335 *pErr = U_MEMORY_ALLOCATION_ERROR; |
| 336 uprv_free(newElement->item); |
| 337 uprv_free(newElement); |
| 338 return NULL; |
| 339 } |
| 340 uprv_strcpy(newElement->name, baseName); |
| 341 |
| 342 /* Stick the new DataCacheElement into the hash table. |
| 343 */ |
| 344 htable = udata_getHashTable(); |
| 345 umtx_lock(NULL); |
| 346 oldValue = (DataCacheElement *)uhash_get(htable, path); |
| 347 if (oldValue != NULL) { |
| 348 subErr = U_USING_DEFAULT_WARNING; |
| 349 } |
| 350 else { |
| 351 uhash_put( |
| 352 htable, |
| 353 newElement->name, /* Key */ |
| 354 newElement, /* Value */ |
| 355 &subErr); |
| 356 } |
| 357 umtx_unlock(NULL); |
| 358 |
| 359 #ifdef UDATA_DEBUG |
| 360 fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n", newElement->name, |
| 361 newElement->item, u_errorName(subErr), newElement->item->vFuncs); |
| 362 #endif |
| 363 |
| 364 if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) { |
| 365 *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. *
/ |
| 366 uprv_free(newElement->name); |
| 367 uprv_free(newElement->item); |
| 368 uprv_free(newElement); |
| 369 return oldValue ? oldValue->item : NULL; |
| 370 } |
| 371 |
| 372 return newElement->item; |
| 373 } |
| 374 |
| 375 /*----------------------------------------------------------------------*=======
======= |
| 376 * * |
| 377 * Path management. Could be shared with other tools/etc if need be * |
| 378 * later on. * |
| 379 * * |
| 380 *----------------------------------------------------------------------*/ |
| 381 |
| 382 #define U_DATA_PATHITER_BUFSIZ 128 /* Size of local buffer for paths
*/ |
| 383 /* Overflow causes malloc of larger
buf */ |
| 384 |
| 385 U_NAMESPACE_BEGIN |
| 386 |
| 387 class UDataPathIterator |
| 388 { |
| 389 public: |
| 390 UDataPathIterator(const char *path, const char *pkg, |
| 391 const char *item, const char *suffix, UBool doCheckLastFou
r, |
| 392 UErrorCode *pErrorCode); |
| 393 const char *next(UErrorCode *pErrorCode); |
| 394 |
| 395 private: |
| 396 const char *path; /* working path (u_icudata_Di
r) */ |
| 397 const char *nextPath; /* path following this one */ |
| 398 const char *basename; /* item's basename (icudt22e_
mt.res)*/ |
| 399 const char *suffix; /* item suffix (can be null)
*/ |
| 400 |
| 401 uint32_t basenameLen; /* length of basename */ |
| 402 |
| 403 CharString itemPath; /* path passed in with item n
ame */ |
| 404 CharString pathBuffer; /* output path for this it'io
n */ |
| 405 CharString packageStub; /* example: "/icudt28b". Wil
l ignore that leaf in set paths. */ |
| 406 |
| 407 UBool checkLastFour; /* if TRUE then allow paths s
uch as '/foo/myapp.dat' |
| 408 * to match, checks last 4 ch
ars of suffix with |
| 409 * last 4 of path, then previ
ous chars. */ |
| 410 }; |
| 411 |
| 412 /** |
| 413 * @param iter The iterator to be initialized. Its current state does not matte
r. |
| 414 * @param path The full pathname to be iterated over. If NULL, defaults to U_I
CUDATA_NAME |
| 415 * @param pkg Package which is being searched for, ex "icudt28l". Will ignore
leave directories such as /icudt28l |
| 416 * @param item Item to be searched for. Can include full path, such as /a/b/fo
o.dat |
| 417 * @param suffix Optional item suffix, if not-null (ex. ".dat") then 'path' can
contain 'item' explicitly. |
| 418 * Ex: 'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar
/baz' as item #2. |
| 419 * '/blarg/stuff.dat' would also be found. |
| 420 */ |
| 421 UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg, |
| 422 const char *item, const char *inSuffix, UBo
ol doCheckLastFour, |
| 423 UErrorCode *pErrorCode) |
| 424 { |
| 425 #ifdef UDATA_DEBUG |
| 426 fprintf(stderr, "SUFFIX1=%s PATH=%s\n", inSuffix, inPath); |
| 427 #endif |
| 428 /** Path **/ |
| 429 if(inPath == NULL) { |
| 430 path = u_getDataDirectory(); |
| 431 } else { |
| 432 path = inPath; |
| 433 } |
| 434 |
| 435 /** Package **/ |
| 436 if(pkg != NULL) { |
| 437 packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode); |
| 438 #ifdef UDATA_DEBUG |
| 439 fprintf(stderr, "STUB=%s [%d]\n", packageStub.data(), packageStub.length()
); |
| 440 #endif |
| 441 } |
| 442 |
| 443 /** Item **/ |
| 444 basename = findBasename(item); |
| 445 basenameLen = (int32_t)uprv_strlen(basename); |
| 446 |
| 447 /** Item path **/ |
| 448 if(basename == item) { |
| 449 nextPath = path; |
| 450 } else { |
| 451 itemPath.append(item, (int32_t)(basename-item), *pErrorCode); |
| 452 nextPath = itemPath.data(); |
| 453 } |
| 454 #ifdef UDATA_DEBUG |
| 455 fprintf(stderr, "SUFFIX=%s [%p]\n", inSuffix, inSuffix); |
| 456 #endif |
| 457 |
| 458 /** Suffix **/ |
| 459 if(inSuffix != NULL) { |
| 460 suffix = inSuffix; |
| 461 } else { |
| 462 suffix = ""; |
| 463 } |
| 464 |
| 465 checkLastFour = doCheckLastFour; |
| 466 |
| 467 /* pathBuffer will hold the output path strings returned by this iterator */ |
| 468 |
| 469 #ifdef UDATA_DEBUG |
| 470 fprintf(stderr, "%p: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%
s], [nextpath=%s], [checklast4=%s]\n", |
| 471 iter, |
| 472 item, |
| 473 path, |
| 474 basename, |
| 475 suffix, |
| 476 itemPath.data(), |
| 477 nextPath, |
| 478 checkLastFour?"TRUE":"false"); |
| 479 #endif |
| 480 } |
| 481 |
| 482 /** |
| 483 * Get the next path on the list. |
| 484 * |
| 485 * @param iter The Iter to be used |
| 486 * @param len If set, pointer to the length of the returned path, for convenien
ce. |
| 487 * @return Pointer to the next path segment, or NULL if there are no more. |
| 488 */ |
| 489 const char *UDataPathIterator::next(UErrorCode *pErrorCode) |
| 490 { |
| 491 if(U_FAILURE(*pErrorCode)) { |
| 492 return NULL; |
| 493 } |
| 494 |
| 495 const char *currentPath = NULL; |
| 496 int32_t pathLen = 0; |
| 497 const char *pathBasename; |
| 498 |
| 499 do |
| 500 { |
| 501 if( nextPath == NULL ) { |
| 502 break; |
| 503 } |
| 504 currentPath = nextPath; |
| 505 |
| 506 if(nextPath == itemPath.data()) { /* we were processing item's path. */ |
| 507 nextPath = path; /* start with regular path next tm. */ |
| 508 pathLen = (int32_t)uprv_strlen(currentPath); |
| 509 } else { |
| 510 /* fix up next for next time */ |
| 511 nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR); |
| 512 if(nextPath == NULL) { |
| 513 /* segment: entire path */ |
| 514 pathLen = (int32_t)uprv_strlen(currentPath); |
| 515 } else { |
| 516 /* segment: until next segment */ |
| 517 pathLen = (int32_t)(nextPath - currentPath); |
| 518 /* skip divider */ |
| 519 nextPath ++; |
| 520 } |
| 521 } |
| 522 |
| 523 if(pathLen == 0) { |
| 524 continue; |
| 525 } |
| 526 |
| 527 #ifdef UDATA_DEBUG |
| 528 fprintf(stderr, "rest of path (IDD) = %s\n", currentPath); |
| 529 fprintf(stderr, " "); |
| 530 { |
| 531 uint32_t qqq; |
| 532 for(qqq=0;qqq<pathLen;qqq++) |
| 533 { |
| 534 fprintf(stderr, " "); |
| 535 } |
| 536 |
| 537 fprintf(stderr, "^\n"); |
| 538 } |
| 539 #endif |
| 540 pathBuffer.clear().append(currentPath, pathLen, *pErrorCode); |
| 541 |
| 542 /* check for .dat files */ |
| 543 pathBasename = findBasename(pathBuffer.data()); |
| 544 |
| 545 if(checkLastFour == TRUE && |
| 546 (pathLen>=4) && |
| 547 uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix, 4)==0 && /* suff
ix matches */ |
| 548 uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)=
=0 && /* base matches */ |
| 549 uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full l
en */ |
| 550 |
| 551 #ifdef UDATA_DEBUG |
| 552 fprintf(stderr, "Have %s file on the path: %s\n", suffix, pathBuffer
.data()); |
| 553 #endif |
| 554 /* do nothing */ |
| 555 } |
| 556 else |
| 557 { /* regular dir path */ |
| 558 if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) { |
| 559 if((pathLen>=4) && |
| 560 uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat", 4) == 0) |
| 561 { |
| 562 #ifdef UDATA_DEBUG |
| 563 fprintf(stderr, "skipping non-directory .dat file %s\n", pat
hBuffer.data()); |
| 564 #endif |
| 565 continue; |
| 566 } |
| 567 |
| 568 /* Check if it is a directory with the same name as our package
*/ |
| 569 if(!packageStub.isEmpty() && |
| 570 (pathLen > packageStub.length()) && |
| 571 !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length
(), packageStub.data())) { |
| 572 #ifdef UDATA_DEBUG |
| 573 fprintf(stderr, "Found stub %s (will add package %s of len %d)
\n", packageStub.data(), basename, basenameLen); |
| 574 #endif |
| 575 pathBuffer.truncate(pathLen - packageStub.length()); |
| 576 } |
| 577 pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode); |
| 578 } |
| 579 |
| 580 /* + basename */ |
| 581 pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pEr
rorCode); |
| 582 |
| 583 if(*suffix) /* tack on suffix */ |
| 584 { |
| 585 pathBuffer.append(suffix, *pErrorCode); |
| 586 } |
| 587 } |
| 588 |
| 589 #ifdef UDATA_DEBUG |
| 590 fprintf(stderr, " --> %s\n", pathBuffer.data()); |
| 591 #endif |
| 592 |
| 593 return pathBuffer.data(); |
| 594 |
| 595 } while(path); |
| 596 |
| 597 /* fell way off the end */ |
| 598 return NULL; |
| 599 } |
| 600 |
| 601 U_NAMESPACE_END |
| 602 |
| 603 /* =============================================================================
=====*/ |
| 604 |
| 605 |
| 606 /*----------------------------------------------------------------------* |
| 607 * * |
| 608 * Add a static reference to the common data library * |
| 609 * Unless overridden by an explicit udata_setCommonData, this will be * |
| 610 * our common data. * |
| 611 * * |
| 612 *----------------------------------------------------------------------*/ |
| 613 extern "C" const DataHeader U_DATA_API U_ICUDATA_ENTRY_POINT; |
| 614 |
| 615 /* |
| 616 * This would be a good place for weak-linkage declarations of |
| 617 * partial-data-library access functions where each returns a pointer |
| 618 * to its data package, if it is linked in. |
| 619 */ |
| 620 /* |
| 621 extern const void *uprv_getICUData_collation(void) ATTRIBUTE_WEAK; |
| 622 extern const void *uprv_getICUData_conversion(void) ATTRIBUTE_WEAK; |
| 623 */ |
| 624 |
| 625 /*----------------------------------------------------------------------* |
| 626 * * |
| 627 * openCommonData Attempt to open a common format (.dat) file * |
| 628 * Map it into memory (if it's not there already) * |
| 629 * and return a UDataMemory object for it. * |
| 630 * * |
| 631 * If the requested data is already open and cached * |
| 632 * just return the cached UDataMem object. * |
| 633 * * |
| 634 *----------------------------------------------------------------------*/ |
| 635 static UDataMemory * |
| 636 openCommonData(const char *path, /* Path from OpenChoice? */ |
| 637 int32_t commonDataIndex, /* ICU Data (index >= 0) if path == N
ULL */ |
| 638 UErrorCode *pErrorCode) |
| 639 { |
| 640 UDataMemory tData; |
| 641 const char *pathBuffer; |
| 642 const char *inBasename; |
| 643 |
| 644 if (U_FAILURE(*pErrorCode)) { |
| 645 return NULL; |
| 646 } |
| 647 |
| 648 UDataMemory_init(&tData); |
| 649 |
| 650 /* ??????? TODO revisit this */ |
| 651 if (commonDataIndex >= 0) { |
| 652 /* "mini-cache" for common ICU data */ |
| 653 if(commonDataIndex >= LENGTHOF(gCommonICUDataArray)) { |
| 654 return NULL; |
| 655 } |
| 656 if(gCommonICUDataArray[commonDataIndex] == NULL) { |
| 657 int32_t i; |
| 658 for(i = 0; i < commonDataIndex; ++i) { |
| 659 if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT) { |
| 660 /* The linked-in data is already in the list. */ |
| 661 return NULL; |
| 662 } |
| 663 } |
| 664 |
| 665 /* Add the linked-in data to the list. */ |
| 666 /* |
| 667 * This is where we would check and call weakly linked partial-data-
library |
| 668 * access functions. |
| 669 */ |
| 670 /* |
| 671 if (uprv_getICUData_collation) { |
| 672 setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErr
orCode); |
| 673 } |
| 674 if (uprv_getICUData_conversion) { |
| 675 setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pEr
rorCode); |
| 676 } |
| 677 */ |
| 678 setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT, FALSE, pErrorCode); |
| 679 } |
| 680 return gCommonICUDataArray[commonDataIndex]; |
| 681 } |
| 682 |
| 683 |
| 684 /* request is NOT for ICU Data. */ |
| 685 |
| 686 /* Find the base name portion of the supplied path. */ |
| 687 /* inBasename will be left pointing somewhere within the original path str
ing. */ |
| 688 inBasename = findBasename(path); |
| 689 #ifdef UDATA_DEBUG |
| 690 fprintf(stderr, "inBasename = %s\n", inBasename); |
| 691 #endif |
| 692 |
| 693 if(*inBasename==0) { |
| 694 /* no basename. This will happen if the original path was a director
y name, */ |
| 695 /* like "a/b/c/". (Fallback to separate files will still work.)
*/ |
| 696 #ifdef UDATA_DEBUG |
| 697 fprintf(stderr, "ocd: no basename in %s, bailing.\n", path); |
| 698 #endif |
| 699 *pErrorCode=U_FILE_ACCESS_ERROR; |
| 700 return NULL; |
| 701 } |
| 702 |
| 703 /* Is the requested common data file already open and cached?
*/ |
| 704 /* Note that the cache is keyed by the base name only. The rest of the pat
h, */ |
| 705 /* if any, is not considered.
*/ |
| 706 { |
| 707 UDataMemory *dataToReturn = udata_findCachedData(inBasename); |
| 708 if (dataToReturn != NULL) { |
| 709 return dataToReturn; |
| 710 } |
| 711 } |
| 712 |
| 713 /* Requested item is not in the cache. |
| 714 * Hunt it down, trying all the path locations |
| 715 */ |
| 716 |
| 717 UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat", TRUE,
pErrorCode); |
| 718 |
| 719 while((UDataMemory_isLoaded(&tData)==FALSE) && (pathBuffer = iter.next(pErro
rCode)) != NULL) |
| 720 { |
| 721 #ifdef UDATA_DEBUG |
| 722 fprintf(stderr, "ocd: trying path %s - ", pathBuffer); |
| 723 #endif |
| 724 uprv_mapFile(&tData, pathBuffer); |
| 725 #ifdef UDATA_DEBUG |
| 726 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&tData)?"LOADED":"not loade
d"); |
| 727 #endif |
| 728 } |
| 729 |
| 730 #if defined(OS390_STUBDATA) && defined(OS390BATCH) |
| 731 if (!UDataMemory_isLoaded(&tData)) { |
| 732 char ourPathBuffer[1024]; |
| 733 /* One more chance, for extendCommonData() */ |
| 734 uprv_strncpy(ourPathBuffer, path, 1019); |
| 735 ourPathBuffer[1019]=0; |
| 736 uprv_strcat(ourPathBuffer, ".dat"); |
| 737 uprv_mapFile(&tData, ourPathBuffer); |
| 738 } |
| 739 #endif |
| 740 |
| 741 if (!UDataMemory_isLoaded(&tData)) { |
| 742 /* no common data */ |
| 743 *pErrorCode=U_FILE_ACCESS_ERROR; |
| 744 return NULL; |
| 745 } |
| 746 |
| 747 /* we have mapped a file, check its header */ |
| 748 udata_checkCommonData(&tData, pErrorCode); |
| 749 |
| 750 |
| 751 /* Cache the UDataMemory struct for this .dat file, |
| 752 * so we won't need to hunt it down and map it again next time |
| 753 * something is needed from it. */ |
| 754 return udata_cacheDataItem(inBasename, &tData, pErrorCode); |
| 755 } |
| 756 |
| 757 |
| 758 /*----------------------------------------------------------------------* |
| 759 * * |
| 760 * extendICUData If the full set of ICU data was not loaded at * |
| 761 * program startup, load it now. This function will * |
| 762 * be called when the lookup of an ICU data item in * |
| 763 * the common ICU data fails. * |
| 764 * * |
| 765 * return true if new data is loaded, false otherwise.* |
| 766 * * |
| 767 *----------------------------------------------------------------------*/ |
| 768 static UBool extendICUData(UErrorCode *pErr) |
| 769 { |
| 770 UDataMemory *pData; |
| 771 UDataMemory copyPData; |
| 772 UBool didUpdate = FALSE; |
| 773 |
| 774 /* |
| 775 * There is a chance for a race condition here. |
| 776 * Normally, ICU data is loaded from a DLL or via mmap() and |
| 777 * setCommonICUData() will detect if the same address is set twice. |
| 778 * If ICU is built with data loading via fread() then the address will |
| 779 * be different each time the common data is loaded and we may add |
| 780 * multiple copies of the data. |
| 781 * In this case, use a mutex to prevent the race. |
| 782 * Use a specific mutex to avoid nested locks of the global mutex. |
| 783 */ |
| 784 #if MAP_IMPLEMENTATION==MAP_STDIO |
| 785 static UMTX extendICUDataMutex = NULL; |
| 786 umtx_lock(&extendICUDataMutex); |
| 787 #endif |
| 788 if(!gHaveTriedToLoadCommonData) { |
| 789 gHaveTriedToLoadCommonData = TRUE; |
| 790 |
| 791 /* See if we can explicitly open a .dat file for the ICUData. */ |
| 792 pData = openCommonData( |
| 793 U_ICUDATA_NAME, /* "icudt20l" , for example.
*/ |
| 794 -1, /* Pretend we're not opening ICUD
ata */ |
| 795 pErr); |
| 796 |
| 797 /* How about if there is no pData, eh... */ |
| 798 |
| 799 UDataMemory_init(©PData); |
| 800 if(pData != NULL) { |
| 801 UDatamemory_assign(©PData, pData); |
| 802 copyPData.map = 0; /* The mapping for this data is owned
by the hash table */ |
| 803 copyPData.mapAddr = 0; /* which will unmap it when ICU is s
hut down. */ |
| 804 /* CommonICUData is also unmapped when
ICU is shut down.*/ |
| 805 /* To avoid unmapping the data twice,
zero out the map */ |
| 806 /* fields in the UDataMemory that we
're assigning */ |
| 807 /* to CommonICUData.
*/ |
| 808 |
| 809 didUpdate = |
| 810 setCommonICUData(©PData,/* The new common data.
*/ |
| 811 FALSE, /* No warnings if write didn't happen
*/ |
| 812 pErr); /* setCommonICUData honors errors; NO
P if error set */ |
| 813 } |
| 814 } |
| 815 #if MAP_IMPLEMENTATION==MAP_STDIO |
| 816 umtx_unlock(&extendICUDataMutex); |
| 817 #endif |
| 818 return didUpdate; /* Return true if ICUData pointer was update
d. */ |
| 819 /* (Could potentialy have been done by ano
ther thread racing */ |
| 820 /* us through here, but that's fine, we st
ill return true */ |
| 821 /* so that current thread will also examin
e extended data. */ |
| 822 } |
| 823 |
| 824 /*----------------------------------------------------------------------* |
| 825 * * |
| 826 * udata_setCommonData * |
| 827 * * |
| 828 *----------------------------------------------------------------------*/ |
| 829 U_CAPI void U_EXPORT2 |
| 830 udata_setCommonData(const void *data, UErrorCode *pErrorCode) { |
| 831 UDataMemory dataMemory; |
| 832 |
| 833 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 834 return; |
| 835 } |
| 836 |
| 837 if(data==NULL) { |
| 838 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
| 839 return; |
| 840 } |
| 841 |
| 842 /* set the data pointer and test for validity */ |
| 843 UDataMemory_init(&dataMemory); |
| 844 UDataMemory_setData(&dataMemory, data); |
| 845 udata_checkCommonData(&dataMemory, pErrorCode); |
| 846 if (U_FAILURE(*pErrorCode)) {return;} |
| 847 |
| 848 /* we have good data */ |
| 849 /* Set it up as the ICU Common Data. */ |
| 850 setCommonICUData(&dataMemory, TRUE, pErrorCode); |
| 851 } |
| 852 |
| 853 /*--------------------------------------------------------------------------- |
| 854 * |
| 855 * udata_setAppData |
| 856 * |
| 857 *---------------------------------------------------------------------------- *
/ |
| 858 U_CAPI void U_EXPORT2 |
| 859 udata_setAppData(const char *path, const void *data, UErrorCode *err) |
| 860 { |
| 861 UDataMemory udm; |
| 862 |
| 863 if(err==NULL || U_FAILURE(*err)) { |
| 864 return; |
| 865 } |
| 866 if(data==NULL) { |
| 867 *err=U_ILLEGAL_ARGUMENT_ERROR; |
| 868 return; |
| 869 } |
| 870 |
| 871 UDataMemory_init(&udm); |
| 872 UDataMemory_setData(&udm, data); |
| 873 udata_checkCommonData(&udm, err); |
| 874 udata_cacheDataItem(path, &udm, err); |
| 875 } |
| 876 |
| 877 /*----------------------------------------------------------------------------* |
| 878 * * |
| 879 * checkDataItem Given a freshly located/loaded data item, either * |
| 880 * an entry in a common file or a separately loaded file, * |
| 881 * sanity check its header, and see if the data is * |
| 882 * acceptable to the app. * |
| 883 * If the data is good, create and return a UDataMemory * |
| 884 * object that can be returned to the application. * |
| 885 * Return NULL on any sort of failure. * |
| 886 * * |
| 887 *----------------------------------------------------------------------------*/ |
| 888 static UDataMemory * |
| 889 checkDataItem |
| 890 ( |
| 891 const DataHeader *pHeader, /* The data item to be checked.
*/ |
| 892 UDataMemoryIsAcceptable *isAcceptable, /* App's call-back function
*/ |
| 893 void *context, /* pass-thru param for above.
*/ |
| 894 const char *type, /* pass-thru param for above.
*/ |
| 895 const char *name, /* pass-thru param for above.
*/ |
| 896 UErrorCode *nonFatalErr, /* Error code if this data was not a
cceptable */ |
| 897 /* but openChoice should continue
with */ |
| 898 /* trying to get data from fallbac
k path. */ |
| 899 UErrorCode *fatalErr /* Bad error, caller should return i
mmediately */ |
| 900 ) |
| 901 { |
| 902 UDataMemory *rDataMem = NULL; /* the new UDataMemory, to be return
ed. */ |
| 903 |
| 904 if (U_FAILURE(*fatalErr)) { |
| 905 return NULL; |
| 906 } |
| 907 |
| 908 if(pHeader->dataHeader.magic1==0xda && |
| 909 pHeader->dataHeader.magic2==0x27 && |
| 910 (isAcceptable==NULL || isAcceptable(context, type, name, &pHeader->info)
) |
| 911 ) { |
| 912 rDataMem=UDataMemory_createNewInstance(fatalErr); |
| 913 if (U_FAILURE(*fatalErr)) { |
| 914 return NULL; |
| 915 } |
| 916 rDataMem->pHeader = pHeader; |
| 917 } else { |
| 918 /* the data is not acceptable, look further */ |
| 919 /* If we eventually find something good, this errorcode will be */ |
| 920 /* cleared out. */ |
| 921 *nonFatalErr=U_INVALID_FORMAT_ERROR; |
| 922 } |
| 923 return rDataMem; |
| 924 } |
| 925 |
| 926 /** |
| 927 * @return 0 if not loaded, 1 if loaded or err |
| 928 */ |
| 929 static UDataMemory *doLoadFromIndividualFiles(const char *pkgName, |
| 930 const char *dataPath, const char *tocEntryPathSuffix, |
| 931 /* following arguments are the same as doOpenChoice itself */ |
| 932 const char *path, const char *type, const char *name, |
| 933 UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 934 UErrorCode *subErrorCode, |
| 935 UErrorCode *pErrorCode) |
| 936 { |
| 937 const char *pathBuffer; |
| 938 UDataMemory dataMemory; |
| 939 UDataMemory *pEntryData; |
| 940 |
| 941 /* look in ind. files: package\nam.typ ========================= */ |
| 942 /* init path iterator for individual files */ |
| 943 UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, FALSE, p
ErrorCode); |
| 944 |
| 945 while((pathBuffer = iter.next(pErrorCode))) |
| 946 { |
| 947 #ifdef UDATA_DEBUG |
| 948 fprintf(stderr, "UDATA: trying individual file %s\n", pathBuffer); |
| 949 #endif |
| 950 if(uprv_mapFile(&dataMemory, pathBuffer)) |
| 951 { |
| 952 pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context
, type, name, subErrorCode, pErrorCode); |
| 953 if (pEntryData != NULL) { |
| 954 /* Data is good. |
| 955 * Hand off ownership of the backing memory to the user's UDataM
emory. |
| 956 * and return it. */ |
| 957 pEntryData->mapAddr = dataMemory.mapAddr; |
| 958 pEntryData->map = dataMemory.map; |
| 959 |
| 960 #ifdef UDATA_DEBUG |
| 961 fprintf(stderr, "** Mapped file: %s\n", pathBuffer); |
| 962 #endif |
| 963 return pEntryData; |
| 964 } |
| 965 |
| 966 /* the data is not acceptable, or some error occured. Either way, u
nmap the memory */ |
| 967 udata_close(&dataMemory); |
| 968 |
| 969 /* If we had a nasty error, bail out completely. */ |
| 970 if (U_FAILURE(*pErrorCode)) { |
| 971 return NULL; |
| 972 } |
| 973 |
| 974 /* Otherwise remember that we found data but didn't like it for some
reason */ |
| 975 *subErrorCode=U_INVALID_FORMAT_ERROR; |
| 976 } |
| 977 #ifdef UDATA_DEBUG |
| 978 fprintf(stderr, "%s\n", UDataMemory_isLoaded(&dataMemory)?"LOADED":"not
loaded"); |
| 979 #endif |
| 980 } |
| 981 return NULL; |
| 982 } |
| 983 |
| 984 /** |
| 985 * @return 0 if not loaded, 1 if loaded or err |
| 986 */ |
| 987 static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName
*/, |
| 988 const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const ch
ar *tocEntryName, |
| 989 /* following arguments are the same as doOpenChoice itself */ |
| 990 const char *path, const char *type, const char *name, |
| 991 UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 992 UErrorCode *subErrorCode, |
| 993 UErrorCode *pErrorCode) |
| 994 { |
| 995 UDataMemory *pEntryData; |
| 996 const DataHeader *pHeader; |
| 997 UDataMemory *pCommonData; |
| 998 int32_t commonDataIndex; |
| 999 /* try to get common data. The loop is for platforms such as the 390 that d
o |
| 1000 * not initially load the full set of ICU data. If the lookup of an ICU da
ta item |
| 1001 * fails, the full (but slower to load) set is loaded, the and the loop rep
eats, |
| 1002 * trying the lookup again. Once the full set of ICU data is loaded, the l
oop wont |
| 1003 * repeat because the full set will be checked the first time through. |
| 1004 * |
| 1005 * The loop also handles the fallback to a .dat file if the application lin
ked |
| 1006 * to the stub data library rather than a real library. |
| 1007 */ |
| 1008 for (commonDataIndex = isICUData ? 0 : -1;;) { |
| 1009 pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** sea
rch for pkg **/ |
| 1010 |
| 1011 if(U_SUCCESS(*subErrorCode) && pCommonData!=NULL) { |
| 1012 int32_t length; |
| 1013 |
| 1014 /* look up the data piece in the common data */ |
| 1015 pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &leng
th, subErrorCode); |
| 1016 #ifdef UDATA_DEBUG |
| 1017 fprintf(stderr, "%s: pHeader=%p - %s\n", tocEntryName, pHeader, u_er
rorName(*subErrorCode)); |
| 1018 #endif |
| 1019 |
| 1020 if(pHeader!=NULL) { |
| 1021 pEntryData = checkDataItem(pHeader, isAcceptable, context, type,
name, subErrorCode, pErrorCode); |
| 1022 #ifdef UDATA_DEBUG |
| 1023 fprintf(stderr, "pEntryData=%p\n", pEntryData); |
| 1024 #endif |
| 1025 if (U_FAILURE(*pErrorCode)) { |
| 1026 return NULL; |
| 1027 } |
| 1028 if (pEntryData != NULL) { |
| 1029 pEntryData->length = length; |
| 1030 return pEntryData; |
| 1031 } |
| 1032 } |
| 1033 } |
| 1034 /* Data wasn't found. If we were looking for an ICUData item and there
is |
| 1035 * more data available, load it and try again, |
| 1036 * otherwise break out of this loop. */ |
| 1037 if (!isICUData) { |
| 1038 return NULL; |
| 1039 } else if (pCommonData != NULL) { |
| 1040 ++commonDataIndex; /* try the next data package */ |
| 1041 } else if (extendICUData(subErrorCode)) { |
| 1042 /* try this data package slot again: it changed from NULL to non-NUL
L */ |
| 1043 } else { |
| 1044 return NULL; |
| 1045 } |
| 1046 } |
| 1047 } |
| 1048 |
| 1049 /* |
| 1050 * A note on the ownership of Mapped Memory |
| 1051 * |
| 1052 * For common format files, ownership resides with the UDataMemory object |
| 1053 * that lives in the cache of opened common data. These UDataMemorys are pri
vate |
| 1054 * to the udata implementation, and are never seen directly by users. |
| 1055 * |
| 1056 * The UDataMemory objects returned to users will have the address of some de
sired |
| 1057 * data within the mapped region, but they wont have the mapping info itself,
and thus |
| 1058 * won't cause anything to be removed from memory when they are closed. |
| 1059 * |
| 1060 * For individual data files, the UDataMemory returned to the user holds the |
| 1061 * information necessary to unmap the data on close. If the user independently |
| 1062 * opens the same data file twice, two completely independent mappings will be
made. |
| 1063 * (There is no cache of opened data items from individual files, only a cache
of |
| 1064 * opened Common Data files, that is, files containing a collection of data it
ems.) |
| 1065 * |
| 1066 * For common data passed in from the user via udata_setAppData() or |
| 1067 * udata_setCommonData(), ownership remains with the user. |
| 1068 * |
| 1069 * UDataMemory objects themselves, as opposed to the memory they describe, |
| 1070 * can be anywhere - heap, stack/local or global. |
| 1071 * They have a flag to indicate when they're heap allocated and thus |
| 1072 * must be deleted when closed. |
| 1073 */ |
| 1074 |
| 1075 |
| 1076 /*----------------------------------------------------------------------------* |
| 1077 * * |
| 1078 * main data loading functions * |
| 1079 * * |
| 1080 *----------------------------------------------------------------------------*/ |
| 1081 static UDataMemory * |
| 1082 doOpenChoice(const char *path, const char *type, const char *name, |
| 1083 UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 1084 UErrorCode *pErrorCode) |
| 1085 { |
| 1086 UDataMemory *retVal = NULL; |
| 1087 |
| 1088 const char *dataPath; |
| 1089 |
| 1090 int32_t tocEntrySuffixIndex; |
| 1091 const char *tocEntryPathSuffix; |
| 1092 UErrorCode subErrorCode=U_ZERO_ERROR; |
| 1093 const char *treeChar; |
| 1094 |
| 1095 UBool isICUData = FALSE; |
| 1096 |
| 1097 |
| 1098 /* Is this path ICU data? */ |
| 1099 if(path == NULL || |
| 1100 !strcmp(path, U_ICUDATA_ALIAS) || /* "ICUDATA" */ |
| 1101 !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-
" */ |
| 1102 uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) || |
| 1103 !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-
" */ |
| 1104 uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) { |
| 1105 isICUData = TRUE; |
| 1106 } |
| 1107 |
| 1108 #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR) /* Windows: try "foo\bar" and "fo
o/bar" */ |
| 1109 /* remap from alternate path char to the main one */ |
| 1110 CharString altSepPath; |
| 1111 if(path) { |
| 1112 if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != NULL) { |
| 1113 altSepPath.append(path, *pErrorCode); |
| 1114 char *p; |
| 1115 while((p=uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR))) { |
| 1116 *p = U_FILE_SEP_CHAR; |
| 1117 } |
| 1118 #if defined (UDATA_DEBUG) |
| 1119 fprintf(stderr, "Changed path from [%s] to [%s]\n", path, altSepPath
.s); |
| 1120 #endif |
| 1121 path = altSepPath.data(); |
| 1122 } |
| 1123 } |
| 1124 #endif |
| 1125 |
| 1126 CharString tocEntryName; /* entry name in tree format. ex: 'icudt28b/coll/a
r.res' */ |
| 1127 CharString tocEntryPath; /* entry name in path format. ex: 'icudt28b\\coll\
\ar.res' */ |
| 1128 |
| 1129 CharString pkgName; |
| 1130 CharString treeName; |
| 1131 |
| 1132 /* ======= Set up strings */ |
| 1133 if(path==NULL) { |
| 1134 pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
| 1135 } else { |
| 1136 const char *pkg; |
| 1137 const char *first; |
| 1138 pkg = uprv_strrchr(path, U_FILE_SEP_CHAR); |
| 1139 first = uprv_strchr(path, U_FILE_SEP_CHAR); |
| 1140 if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash
in the path- not a tree name */ |
| 1141 /* see if this is an /absolute/path/to/package path */ |
| 1142 if(pkg) { |
| 1143 pkgName.append(pkg+1, *pErrorCode); |
| 1144 } else { |
| 1145 pkgName.append(path, *pErrorCode); |
| 1146 } |
| 1147 } else { |
| 1148 treeChar = uprv_strchr(path, U_TREE_SEPARATOR); |
| 1149 if(treeChar) { |
| 1150 treeName.append(treeChar+1, *pErrorCode); /* following '-' */ |
| 1151 if(isICUData) { |
| 1152 pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
| 1153 } else { |
| 1154 pkgName.append(path, (int32_t)(treeChar-path), *pErrorCode); |
| 1155 if (first == NULL) { |
| 1156 /* |
| 1157 This user data has no path, but there is a tree name. |
| 1158 Look up the correct path from the data cache later. |
| 1159 */ |
| 1160 path = pkgName.data(); |
| 1161 } |
| 1162 } |
| 1163 } else { |
| 1164 if(isICUData) { |
| 1165 pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
| 1166 } else { |
| 1167 pkgName.append(path, *pErrorCode); |
| 1168 } |
| 1169 } |
| 1170 } |
| 1171 } |
| 1172 |
| 1173 #ifdef UDATA_DEBUG |
| 1174 fprintf(stderr, " P=%s T=%s\n", pkgName.data(), treeName.data()); |
| 1175 #endif |
| 1176 |
| 1177 /* setting up the entry name and file name |
| 1178 * Make up a full name by appending the type to the supplied |
| 1179 * name, assuming that a type was supplied. |
| 1180 */ |
| 1181 |
| 1182 /* prepend the package */ |
| 1183 tocEntryName.append(pkgName, *pErrorCode); |
| 1184 tocEntryPath.append(pkgName, *pErrorCode); |
| 1185 tocEntrySuffixIndex = tocEntryName.length(); |
| 1186 |
| 1187 if(!treeName.isEmpty()) { |
| 1188 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName,
*pErrorCode); |
| 1189 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErr
orCode); |
| 1190 } |
| 1191 |
| 1192 tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pError
Code); |
| 1193 tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
| 1194 if(type!=NULL && *type!=0) { |
| 1195 tocEntryName.append(".", *pErrorCode).append(type, *pErrorCode); |
| 1196 tocEntryPath.append(".", *pErrorCode).append(type, *pErrorCode); |
| 1197 } |
| 1198 tocEntryPathSuffix = tocEntryPath.data()+tocEntrySuffixIndex; /* suffix star
ts here */ |
| 1199 |
| 1200 #ifdef UDATA_DEBUG |
| 1201 fprintf(stderr, " tocEntryName = %s\n", tocEntryName.data()); |
| 1202 fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data()); |
| 1203 #endif |
| 1204 |
| 1205 if(path == NULL) { |
| 1206 path = COMMON_DATA_NAME; /* "icudt26e" */ |
| 1207 } |
| 1208 |
| 1209 /************************ Begin loop looking for ind. files ***************/ |
| 1210 #ifdef UDATA_DEBUG |
| 1211 fprintf(stderr, "IND: inBasename = %s, pkg=%s\n", "(n/a)", packageNameFromPa
th(path)); |
| 1212 #endif |
| 1213 |
| 1214 /* End of dealing with a null basename */ |
| 1215 dataPath = u_getDataDirectory(); |
| 1216 |
| 1217 /**** COMMON PACKAGE - only if packages are first. */ |
| 1218 if(gDataFileAccess == UDATA_PACKAGES_FIRST) { |
| 1219 #ifdef UDATA_DEBUG |
| 1220 fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n"); |
| 1221 #endif |
| 1222 /* #2 */ |
| 1223 retVal = doLoadFromCommonData(isICUData, |
| 1224 pkgName.data(), dataPath, tocEntryPathSuffix, tocEnt
ryName.data(), |
| 1225 path, type, name, isAcceptable, context, &subErrorCo
de, pErrorCode); |
| 1226 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1227 return retVal; |
| 1228 } |
| 1229 } |
| 1230 |
| 1231 /**** INDIVIDUAL FILES */ |
| 1232 if((gDataFileAccess==UDATA_PACKAGES_FIRST) || |
| 1233 (gDataFileAccess==UDATA_FILES_FIRST)) { |
| 1234 #ifdef UDATA_DEBUG |
| 1235 fprintf(stderr, "Trying individual files\n"); |
| 1236 #endif |
| 1237 /* Check to make sure that there is a dataPath to iterate over */ |
| 1238 if ((dataPath && *dataPath) || !isICUData) { |
| 1239 retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntr
yPathSuffix, |
| 1240 path, type, name, isAcceptable, context, &subErrorCo
de, pErrorCode); |
| 1241 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1242 return retVal; |
| 1243 } |
| 1244 } |
| 1245 } |
| 1246 |
| 1247 /**** COMMON PACKAGE */ |
| 1248 if((gDataFileAccess==UDATA_ONLY_PACKAGES) || |
| 1249 (gDataFileAccess==UDATA_FILES_FIRST)) { |
| 1250 #ifdef UDATA_DEBUG |
| 1251 fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIR
ST)\n"); |
| 1252 #endif |
| 1253 retVal = doLoadFromCommonData(isICUData, |
| 1254 pkgName.data(), dataPath, tocEntryPathSuffix, tocEnt
ryName.data(), |
| 1255 path, type, name, isAcceptable, context, &subErrorCo
de, pErrorCode); |
| 1256 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1257 return retVal; |
| 1258 } |
| 1259 } |
| 1260 |
| 1261 /* Load from DLL. If we haven't attempted package load, we also haven't had
any chance to |
| 1262 try a DLL (static or setCommonData/etc) load. |
| 1263 If we ever have a "UDATA_ONLY_FILES", add it to the or list here. */ |
| 1264 if(gDataFileAccess==UDATA_NO_FILES) { |
| 1265 #ifdef UDATA_DEBUG |
| 1266 fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n"); |
| 1267 #endif |
| 1268 retVal = doLoadFromCommonData(isICUData, |
| 1269 pkgName.data(), "", tocEntryPathSuffix, tocEntryName
.data(), |
| 1270 path, type, name, isAcceptable, context, &subErrorCo
de, pErrorCode); |
| 1271 if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1272 return retVal; |
| 1273 } |
| 1274 } |
| 1275 |
| 1276 /* data not found */ |
| 1277 if(U_SUCCESS(*pErrorCode)) { |
| 1278 if(U_SUCCESS(subErrorCode)) { |
| 1279 /* file not found */ |
| 1280 *pErrorCode=U_FILE_ACCESS_ERROR; |
| 1281 } else { |
| 1282 /* entry point not found or rejected */ |
| 1283 *pErrorCode=subErrorCode; |
| 1284 } |
| 1285 } |
| 1286 return retVal; |
| 1287 } |
| 1288 |
| 1289 |
| 1290 |
| 1291 /* API ---------------------------------------------------------------------- */ |
| 1292 |
| 1293 U_CAPI UDataMemory * U_EXPORT2 |
| 1294 udata_open(const char *path, const char *type, const char *name, |
| 1295 UErrorCode *pErrorCode) { |
| 1296 #ifdef UDATA_DEBUG |
| 1297 fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n", (path?path:"NULL"), n
ame, type); |
| 1298 fflush(stderr); |
| 1299 #endif |
| 1300 |
| 1301 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 1302 return NULL; |
| 1303 } else if(name==NULL || *name==0) { |
| 1304 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
| 1305 return NULL; |
| 1306 } else { |
| 1307 return doOpenChoice(path, type, name, NULL, NULL, pErrorCode); |
| 1308 } |
| 1309 } |
| 1310 |
| 1311 |
| 1312 |
| 1313 U_CAPI UDataMemory * U_EXPORT2 |
| 1314 udata_openChoice(const char *path, const char *type, const char *name, |
| 1315 UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 1316 UErrorCode *pErrorCode) { |
| 1317 #ifdef UDATA_DEBUG |
| 1318 fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n", (path?path:"NUL
L"), name, type); |
| 1319 #endif |
| 1320 |
| 1321 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 1322 return NULL; |
| 1323 } else if(name==NULL || *name==0 || isAcceptable==NULL) { |
| 1324 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
| 1325 return NULL; |
| 1326 } else { |
| 1327 return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode)
; |
| 1328 } |
| 1329 } |
| 1330 |
| 1331 |
| 1332 |
| 1333 U_CAPI void U_EXPORT2 |
| 1334 udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) { |
| 1335 if(pInfo!=NULL) { |
| 1336 if(pData!=NULL && pData->pHeader!=NULL) { |
| 1337 const UDataInfo *info=&pData->pHeader->info; |
| 1338 uint16_t dataInfoSize=udata_getInfoSize(info); |
| 1339 if(pInfo->size>dataInfoSize) { |
| 1340 pInfo->size=dataInfoSize; |
| 1341 } |
| 1342 uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->si
ze-2); |
| 1343 if(info->isBigEndian!=U_IS_BIG_ENDIAN) { |
| 1344 /* opposite endianness */ |
| 1345 uint16_t x=info->reservedWord; |
| 1346 pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8)); |
| 1347 } |
| 1348 } else { |
| 1349 pInfo->size=0; |
| 1350 } |
| 1351 } |
| 1352 } |
| 1353 |
| 1354 |
| 1355 U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /
*status*/) |
| 1356 { |
| 1357 gDataFileAccess = access; |
| 1358 } |
OLD | NEW |