jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 1 | /* |
| 2 | ****************************************************************************** |
| 3 | * |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 4 | * Copyright (C) 1999-2014, International Business Machines |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 5 | * Corporation and others. All Rights Reserved. |
| 6 | * |
| 7 | ****************************************************************************** |
| 8 | * file name: udata.cpp |
| 9 | * encoding: US-ASCII |
| 10 | * tab size: 8 (not used) |
| 11 | * indentation:4 |
| 12 | * |
| 13 | * created on: 1999oct25 |
| 14 | * created by: Markus W. Scherer |
| 15 | */ |
| 16 | |
| 17 | #include "unicode/utypes.h" /* U_PLATFORM etc. */ |
| 18 | |
| 19 | #ifdef __GNUC__ |
| 20 | /* if gcc |
| 21 | #define ATTRIBUTE_WEAK __attribute__ ((weak)) |
| 22 | might have to #include some other header |
| 23 | */ |
| 24 | #endif |
| 25 | |
| 26 | #include "unicode/putil.h" |
| 27 | #include "unicode/udata.h" |
| 28 | #include "unicode/uversion.h" |
| 29 | #include "charstr.h" |
| 30 | #include "cmemory.h" |
| 31 | #include "cstring.h" |
| 32 | #include "putilimp.h" |
| 33 | #include "uassert.h" |
| 34 | #include "ucln_cmn.h" |
| 35 | #include "ucmndata.h" |
| 36 | #include "udatamem.h" |
| 37 | #include "uhash.h" |
| 38 | #include "umapfile.h" |
| 39 | #include "umutex.h" |
| 40 | |
| 41 | /*********************************************************************** |
| 42 | * |
| 43 | * Notes on the organization of the ICU data implementation |
| 44 | * |
| 45 | * All of the public API is defined in udata.h |
| 46 | * |
| 47 | * The implementation is split into several files... |
| 48 | * |
| 49 | * - udata.c (this file) contains higher level code that knows about |
| 50 | * the search paths for locating data, caching opened data, etc. |
| 51 | * |
| 52 | * - umapfile.c contains the low level platform-specific code for actually loading |
| 53 | * (memory mapping, file reading, whatever) data into memory. |
| 54 | * |
| 55 | * - ucmndata.c deals with the tables of contents of ICU data items within |
| 56 | * an ICU common format data file. The implementation includes |
| 57 | * an abstract interface and support for multiple TOC formats. |
| 58 | * All knowledge of any specific TOC format is encapsulated here. |
| 59 | * |
| 60 | * - udatamem.c has code for managing UDataMemory structs. These are little |
| 61 | * descriptor objects for blocks of memory holding ICU data of |
| 62 | * various types. |
| 63 | */ |
| 64 | |
| 65 | /* configuration ---------------------------------------------------------- */ |
| 66 | |
| 67 | /* If you are excruciatingly bored turn this on .. */ |
| 68 | /* #define UDATA_DEBUG 1 */ |
| 69 | |
| 70 | #if defined(UDATA_DEBUG) |
| 71 | # include <stdio.h> |
| 72 | #endif |
| 73 | |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 74 | U_NAMESPACE_USE |
| 75 | |
| 76 | /* |
| 77 | * Forward declarations |
| 78 | */ |
| 79 | static UDataMemory *udata_findCachedData(const char *path); |
| 80 | |
| 81 | /*********************************************************************** |
| 82 | * |
| 83 | * static (Global) data |
| 84 | * |
| 85 | ************************************************************************/ |
| 86 | |
| 87 | /* |
| 88 | * Pointers to the common ICU data. |
| 89 | * |
| 90 | * We store multiple pointers to ICU data packages and iterate through them |
| 91 | * when looking for a data item. |
| 92 | * |
| 93 | * It is possible to combine this with dependency inversion: |
| 94 | * One or more data package libraries may export |
| 95 | * functions that each return a pointer to their piece of the ICU data, |
| 96 | * and this file would import them as weak functions, without a |
| 97 | * strong linker dependency from the common library on the data library. |
| 98 | * |
| 99 | * Then we can have applications depend on only that part of ICU's data |
| 100 | * that they really need, reducing the size of binaries that take advantage |
| 101 | * of this. |
| 102 | */ |
| 103 | static UDataMemory *gCommonICUDataArray[10] = { NULL }; |
| 104 | |
| 105 | static UBool gHaveTriedToLoadCommonData = FALSE; /* See extendICUData(). */ |
| 106 | |
| 107 | static UHashtable *gCommonDataCache = NULL; /* Global hash table of opened ICU data files. */ |
| 108 | static icu::UInitOnce gCommonDataCacheInitOnce = U_INITONCE_INITIALIZER; |
| 109 | |
| 110 | static UDataFileAccess gDataFileAccess = UDATA_DEFAULT_ACCESS; |
| 111 | |
| 112 | static UBool U_CALLCONV |
| 113 | udata_cleanup(void) |
| 114 | { |
| 115 | int32_t i; |
| 116 | |
| 117 | if (gCommonDataCache) { /* Delete the cache of user data mappings. */ |
| 118 | uhash_close(gCommonDataCache); /* Table owns the contents, and will delete them. */ |
| 119 | gCommonDataCache = NULL; /* Cleanup is not thread safe. */ |
| 120 | } |
| 121 | gCommonDataCacheInitOnce.reset(); |
| 122 | |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 123 | for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray) && gCommonICUDataArray[i] != NULL; ++i) { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 124 | udata_close(gCommonICUDataArray[i]); |
| 125 | gCommonICUDataArray[i] = NULL; |
| 126 | } |
| 127 | gHaveTriedToLoadCommonData = FALSE; |
| 128 | |
| 129 | return TRUE; /* Everything was cleaned up */ |
| 130 | } |
| 131 | |
| 132 | static UBool U_CALLCONV |
| 133 | findCommonICUDataByName(const char *inBasename) |
| 134 | { |
| 135 | UBool found = FALSE; |
| 136 | int32_t i; |
| 137 | |
| 138 | UDataMemory *pData = udata_findCachedData(inBasename); |
| 139 | if (pData == NULL) |
| 140 | return FALSE; |
| 141 | |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 142 | for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 143 | if ((gCommonICUDataArray[i] != NULL) && (gCommonICUDataArray[i]->pHeader == pData->pHeader)) { |
| 144 | /* The data pointer is already in the array. */ |
| 145 | found = TRUE; |
| 146 | break; |
| 147 | } |
| 148 | } |
| 149 | |
| 150 | return found; |
| 151 | } |
| 152 | |
| 153 | |
| 154 | /* |
| 155 | * setCommonICUData. Set a UDataMemory to be the global ICU Data |
| 156 | */ |
| 157 | static UBool |
| 158 | setCommonICUData(UDataMemory *pData, /* The new common data. Belongs to caller, we copy it. */ |
| 159 | UBool warn, /* If true, set USING_DEFAULT warning if ICUData was */ |
| 160 | /* changed by another thread before we got to it. */ |
| 161 | UErrorCode *pErr) |
| 162 | { |
| 163 | UDataMemory *newCommonData = UDataMemory_createNewInstance(pErr); |
| 164 | int32_t i; |
| 165 | UBool didUpdate = FALSE; |
| 166 | if (U_FAILURE(*pErr)) { |
| 167 | return FALSE; |
| 168 | } |
| 169 | |
| 170 | /* For the assignment, other threads must cleanly see either the old */ |
| 171 | /* or the new, not some partially initialized new. The old can not be */ |
| 172 | /* deleted - someone may still have a pointer to it lying around in */ |
| 173 | /* their locals. */ |
| 174 | UDatamemory_assign(newCommonData, pData); |
| 175 | umtx_lock(NULL); |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 176 | for (i = 0; i < UPRV_LENGTHOF(gCommonICUDataArray); ++i) { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 177 | if (gCommonICUDataArray[i] == NULL) { |
| 178 | gCommonICUDataArray[i] = newCommonData; |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 179 | didUpdate = TRUE; |
| 180 | break; |
| 181 | } else if (gCommonICUDataArray[i]->pHeader == pData->pHeader) { |
| 182 | /* The same data pointer is already in the array. */ |
| 183 | break; |
| 184 | } |
| 185 | } |
| 186 | umtx_unlock(NULL); |
| 187 | |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 188 | if (i == UPRV_LENGTHOF(gCommonICUDataArray) && warn) { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 189 | *pErr = U_USING_DEFAULT_WARNING; |
| 190 | } |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 191 | if (didUpdate) { |
| 192 | ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
| 193 | } else { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 194 | uprv_free(newCommonData); |
| 195 | } |
| 196 | return didUpdate; |
| 197 | } |
| 198 | |
| 199 | static UBool |
| 200 | setCommonICUDataPointer(const void *pData, UBool /*warn*/, UErrorCode *pErrorCode) { |
| 201 | UDataMemory tData; |
| 202 | UDataMemory_init(&tData); |
| 203 | UDataMemory_setData(&tData, pData); |
| 204 | udata_checkCommonData(&tData, pErrorCode); |
| 205 | return setCommonICUData(&tData, FALSE, pErrorCode); |
| 206 | } |
| 207 | |
| 208 | static const char * |
| 209 | findBasename(const char *path) { |
| 210 | const char *basename=uprv_strrchr(path, U_FILE_SEP_CHAR); |
| 211 | if(basename==NULL) { |
| 212 | return path; |
| 213 | } else { |
| 214 | return basename+1; |
| 215 | } |
| 216 | } |
| 217 | |
| 218 | #ifdef UDATA_DEBUG |
| 219 | static const char * |
| 220 | packageNameFromPath(const char *path) |
| 221 | { |
| 222 | if((path == NULL) || (*path == 0)) { |
| 223 | return U_ICUDATA_NAME; |
| 224 | } |
| 225 | |
| 226 | path = findBasename(path); |
| 227 | |
| 228 | if((path == NULL) || (*path == 0)) { |
| 229 | return U_ICUDATA_NAME; |
| 230 | } |
| 231 | |
| 232 | return path; |
| 233 | } |
| 234 | #endif |
| 235 | |
| 236 | /*----------------------------------------------------------------------* |
| 237 | * * |
| 238 | * Cache for common data * |
| 239 | * Functions for looking up or adding entries to a cache of * |
| 240 | * data that has been previously opened. Avoids a potentially * |
| 241 | * expensive operation of re-opening the data for subsequent * |
| 242 | * uses. * |
| 243 | * * |
| 244 | * Data remains cached for the duration of the process. * |
| 245 | * * |
| 246 | *----------------------------------------------------------------------*/ |
| 247 | |
| 248 | typedef struct DataCacheElement { |
| 249 | char *name; |
| 250 | UDataMemory *item; |
| 251 | } DataCacheElement; |
| 252 | |
| 253 | |
| 254 | |
| 255 | /* |
| 256 | * Deleter function for DataCacheElements. |
| 257 | * udata cleanup function closes the hash table; hash table in turn calls back to |
| 258 | * here for each entry. |
| 259 | */ |
| 260 | static void U_CALLCONV DataCacheElement_deleter(void *pDCEl) { |
| 261 | DataCacheElement *p = (DataCacheElement *)pDCEl; |
| 262 | udata_close(p->item); /* unmaps storage */ |
| 263 | uprv_free(p->name); /* delete the hash key string. */ |
| 264 | uprv_free(pDCEl); /* delete 'this' */ |
| 265 | } |
| 266 | |
| 267 | static void udata_initHashTable() { |
| 268 | UErrorCode err = U_ZERO_ERROR; |
| 269 | U_ASSERT(gCommonDataCache == NULL); |
| 270 | gCommonDataCache = uhash_open(uhash_hashChars, uhash_compareChars, NULL, &err); |
| 271 | if (U_FAILURE(err)) { |
| 272 | // TODO: handle errors better. |
| 273 | gCommonDataCache = NULL; |
| 274 | } |
| 275 | if (gCommonDataCache != NULL) { |
| 276 | uhash_setValueDeleter(gCommonDataCache, DataCacheElement_deleter); |
| 277 | ucln_common_registerCleanup(UCLN_COMMON_UDATA, udata_cleanup); |
| 278 | } |
| 279 | } |
| 280 | |
| 281 | /* udata_getCacheHashTable() |
| 282 | * Get the hash table used to store the data cache entries. |
| 283 | * Lazy create it if it doesn't yet exist. |
| 284 | */ |
| 285 | static UHashtable *udata_getHashTable() { |
| 286 | umtx_initOnce(gCommonDataCacheInitOnce, &udata_initHashTable); |
| 287 | return gCommonDataCache; |
| 288 | } |
| 289 | |
| 290 | |
| 291 | |
| 292 | static UDataMemory *udata_findCachedData(const char *path) |
| 293 | { |
| 294 | UHashtable *htable; |
| 295 | UDataMemory *retVal = NULL; |
| 296 | DataCacheElement *el; |
| 297 | const char *baseName; |
| 298 | |
| 299 | baseName = findBasename(path); /* Cache remembers only the base name, not the full path. */ |
| 300 | htable = udata_getHashTable(); |
| 301 | umtx_lock(NULL); |
| 302 | el = (DataCacheElement *)uhash_get(htable, baseName); |
| 303 | umtx_unlock(NULL); |
| 304 | if (el != NULL) { |
| 305 | retVal = el->item; |
| 306 | } |
| 307 | #ifdef UDATA_DEBUG |
| 308 | fprintf(stderr, "Cache: [%s] -> %p\n", baseName, retVal); |
| 309 | #endif |
| 310 | return retVal; |
| 311 | } |
| 312 | |
| 313 | |
| 314 | static UDataMemory *udata_cacheDataItem(const char *path, UDataMemory *item, UErrorCode *pErr) { |
| 315 | DataCacheElement *newElement; |
| 316 | const char *baseName; |
| 317 | int32_t nameLen; |
| 318 | UHashtable *htable; |
| 319 | DataCacheElement *oldValue = NULL; |
| 320 | UErrorCode subErr = U_ZERO_ERROR; |
| 321 | |
| 322 | if (U_FAILURE(*pErr)) { |
| 323 | return NULL; |
| 324 | } |
| 325 | |
| 326 | /* Create a new DataCacheElement - the thingy we store in the hash table - |
| 327 | * and copy the supplied path and UDataMemoryItems into it. |
| 328 | */ |
| 329 | newElement = (DataCacheElement *)uprv_malloc(sizeof(DataCacheElement)); |
| 330 | if (newElement == NULL) { |
| 331 | *pErr = U_MEMORY_ALLOCATION_ERROR; |
| 332 | return NULL; |
| 333 | } |
| 334 | newElement->item = UDataMemory_createNewInstance(pErr); |
| 335 | if (U_FAILURE(*pErr)) { |
| 336 | uprv_free(newElement); |
| 337 | return NULL; |
| 338 | } |
| 339 | UDatamemory_assign(newElement->item, item); |
| 340 | |
| 341 | baseName = findBasename(path); |
| 342 | nameLen = (int32_t)uprv_strlen(baseName); |
| 343 | newElement->name = (char *)uprv_malloc(nameLen+1); |
| 344 | if (newElement->name == NULL) { |
| 345 | *pErr = U_MEMORY_ALLOCATION_ERROR; |
| 346 | uprv_free(newElement->item); |
| 347 | uprv_free(newElement); |
| 348 | return NULL; |
| 349 | } |
| 350 | uprv_strcpy(newElement->name, baseName); |
| 351 | |
| 352 | /* Stick the new DataCacheElement into the hash table. |
| 353 | */ |
| 354 | htable = udata_getHashTable(); |
| 355 | umtx_lock(NULL); |
| 356 | oldValue = (DataCacheElement *)uhash_get(htable, path); |
| 357 | if (oldValue != NULL) { |
| 358 | subErr = U_USING_DEFAULT_WARNING; |
| 359 | } |
| 360 | else { |
| 361 | uhash_put( |
| 362 | htable, |
| 363 | newElement->name, /* Key */ |
| 364 | newElement, /* Value */ |
| 365 | &subErr); |
| 366 | } |
| 367 | umtx_unlock(NULL); |
| 368 | |
| 369 | #ifdef UDATA_DEBUG |
| 370 | fprintf(stderr, "Cache: [%s] <<< %p : %s. vFunc=%p\n", newElement->name, |
| 371 | newElement->item, u_errorName(subErr), newElement->item->vFuncs); |
| 372 | #endif |
| 373 | |
| 374 | if (subErr == U_USING_DEFAULT_WARNING || U_FAILURE(subErr)) { |
| 375 | *pErr = subErr; /* copy sub err unto fillin ONLY if something happens. */ |
| 376 | uprv_free(newElement->name); |
| 377 | uprv_free(newElement->item); |
| 378 | uprv_free(newElement); |
| 379 | return oldValue ? oldValue->item : NULL; |
| 380 | } |
| 381 | |
| 382 | return newElement->item; |
| 383 | } |
| 384 | |
| 385 | /*----------------------------------------------------------------------*============== |
| 386 | * * |
| 387 | * Path management. Could be shared with other tools/etc if need be * |
| 388 | * later on. * |
| 389 | * * |
| 390 | *----------------------------------------------------------------------*/ |
| 391 | |
| 392 | #define U_DATA_PATHITER_BUFSIZ 128 /* Size of local buffer for paths */ |
| 393 | /* Overflow causes malloc of larger buf */ |
| 394 | |
| 395 | U_NAMESPACE_BEGIN |
| 396 | |
| 397 | class UDataPathIterator |
| 398 | { |
| 399 | public: |
| 400 | UDataPathIterator(const char *path, const char *pkg, |
| 401 | const char *item, const char *suffix, UBool doCheckLastFour, |
| 402 | UErrorCode *pErrorCode); |
| 403 | const char *next(UErrorCode *pErrorCode); |
| 404 | |
| 405 | private: |
| 406 | const char *path; /* working path (u_icudata_Dir) */ |
| 407 | const char *nextPath; /* path following this one */ |
| 408 | const char *basename; /* item's basename (icudt22e_mt.res)*/ |
| 409 | const char *suffix; /* item suffix (can be null) */ |
| 410 | |
| 411 | uint32_t basenameLen; /* length of basename */ |
| 412 | |
| 413 | CharString itemPath; /* path passed in with item name */ |
| 414 | CharString pathBuffer; /* output path for this it'ion */ |
| 415 | CharString packageStub; /* example: "/icudt28b". Will ignore that leaf in set paths. */ |
| 416 | |
| 417 | UBool checkLastFour; /* if TRUE then allow paths such as '/foo/myapp.dat' |
| 418 | * to match, checks last 4 chars of suffix with |
| 419 | * last 4 of path, then previous chars. */ |
| 420 | }; |
| 421 | |
| 422 | /** |
| 423 | * @param iter The iterator to be initialized. Its current state does not matter. |
| 424 | * @param path The full pathname to be iterated over. If NULL, defaults to U_ICUDATA_NAME |
| 425 | * @param pkg Package which is being searched for, ex "icudt28l". Will ignore leave directories such as /icudt28l |
| 426 | * @param item Item to be searched for. Can include full path, such as /a/b/foo.dat |
| 427 | * @param suffix Optional item suffix, if not-null (ex. ".dat") then 'path' can contain 'item' explicitly. |
| 428 | * Ex: 'stuff.dat' would be found in '/a/foo:/tmp/stuff.dat:/bar/baz' as item #2. |
| 429 | * '/blarg/stuff.dat' would also be found. |
| 430 | */ |
| 431 | UDataPathIterator::UDataPathIterator(const char *inPath, const char *pkg, |
| 432 | const char *item, const char *inSuffix, UBool doCheckLastFour, |
| 433 | UErrorCode *pErrorCode) |
| 434 | { |
| 435 | #ifdef UDATA_DEBUG |
| 436 | fprintf(stderr, "SUFFIX1=%s PATH=%s\n", inSuffix, inPath); |
| 437 | #endif |
| 438 | /** Path **/ |
| 439 | if(inPath == NULL) { |
| 440 | path = u_getDataDirectory(); |
| 441 | } else { |
| 442 | path = inPath; |
| 443 | } |
| 444 | |
| 445 | /** Package **/ |
| 446 | if(pkg != NULL) { |
| 447 | packageStub.append(U_FILE_SEP_CHAR, *pErrorCode).append(pkg, *pErrorCode); |
| 448 | #ifdef UDATA_DEBUG |
| 449 | fprintf(stderr, "STUB=%s [%d]\n", packageStub.data(), packageStub.length()); |
| 450 | #endif |
| 451 | } |
| 452 | |
| 453 | /** Item **/ |
| 454 | basename = findBasename(item); |
| 455 | basenameLen = (int32_t)uprv_strlen(basename); |
| 456 | |
| 457 | /** Item path **/ |
| 458 | if(basename == item) { |
| 459 | nextPath = path; |
| 460 | } else { |
| 461 | itemPath.append(item, (int32_t)(basename-item), *pErrorCode); |
| 462 | nextPath = itemPath.data(); |
| 463 | } |
| 464 | #ifdef UDATA_DEBUG |
| 465 | fprintf(stderr, "SUFFIX=%s [%p]\n", inSuffix, inSuffix); |
| 466 | #endif |
| 467 | |
| 468 | /** Suffix **/ |
| 469 | if(inSuffix != NULL) { |
| 470 | suffix = inSuffix; |
| 471 | } else { |
| 472 | suffix = ""; |
| 473 | } |
| 474 | |
| 475 | checkLastFour = doCheckLastFour; |
| 476 | |
| 477 | /* pathBuffer will hold the output path strings returned by this iterator */ |
| 478 | |
| 479 | #ifdef UDATA_DEBUG |
| 480 | fprintf(stderr, "%p: init %s -> [path=%s], [base=%s], [suff=%s], [itempath=%s], [nextpath=%s], [checklast4=%s]\n", |
| 481 | iter, |
| 482 | item, |
| 483 | path, |
| 484 | basename, |
| 485 | suffix, |
| 486 | itemPath.data(), |
| 487 | nextPath, |
| 488 | checkLastFour?"TRUE":"false"); |
| 489 | #endif |
| 490 | } |
| 491 | |
| 492 | /** |
| 493 | * Get the next path on the list. |
| 494 | * |
| 495 | * @param iter The Iter to be used |
| 496 | * @param len If set, pointer to the length of the returned path, for convenience. |
| 497 | * @return Pointer to the next path segment, or NULL if there are no more. |
| 498 | */ |
| 499 | const char *UDataPathIterator::next(UErrorCode *pErrorCode) |
| 500 | { |
| 501 | if(U_FAILURE(*pErrorCode)) { |
| 502 | return NULL; |
| 503 | } |
| 504 | |
| 505 | const char *currentPath = NULL; |
| 506 | int32_t pathLen = 0; |
| 507 | const char *pathBasename; |
| 508 | |
| 509 | do |
| 510 | { |
| 511 | if( nextPath == NULL ) { |
| 512 | break; |
| 513 | } |
| 514 | currentPath = nextPath; |
| 515 | |
| 516 | if(nextPath == itemPath.data()) { /* we were processing item's path. */ |
| 517 | nextPath = path; /* start with regular path next tm. */ |
| 518 | pathLen = (int32_t)uprv_strlen(currentPath); |
| 519 | } else { |
| 520 | /* fix up next for next time */ |
| 521 | nextPath = uprv_strchr(currentPath, U_PATH_SEP_CHAR); |
| 522 | if(nextPath == NULL) { |
| 523 | /* segment: entire path */ |
| 524 | pathLen = (int32_t)uprv_strlen(currentPath); |
| 525 | } else { |
| 526 | /* segment: until next segment */ |
| 527 | pathLen = (int32_t)(nextPath - currentPath); |
| 528 | /* skip divider */ |
| 529 | nextPath ++; |
| 530 | } |
| 531 | } |
| 532 | |
| 533 | if(pathLen == 0) { |
| 534 | continue; |
| 535 | } |
| 536 | |
| 537 | #ifdef UDATA_DEBUG |
| 538 | fprintf(stderr, "rest of path (IDD) = %s\n", currentPath); |
| 539 | fprintf(stderr, " "); |
| 540 | { |
| 541 | uint32_t qqq; |
| 542 | for(qqq=0;qqq<pathLen;qqq++) |
| 543 | { |
| 544 | fprintf(stderr, " "); |
| 545 | } |
| 546 | |
| 547 | fprintf(stderr, "^\n"); |
| 548 | } |
| 549 | #endif |
| 550 | pathBuffer.clear().append(currentPath, pathLen, *pErrorCode); |
| 551 | |
| 552 | /* check for .dat files */ |
| 553 | pathBasename = findBasename(pathBuffer.data()); |
| 554 | |
| 555 | if(checkLastFour == TRUE && |
| 556 | (pathLen>=4) && |
| 557 | uprv_strncmp(pathBuffer.data() +(pathLen-4), suffix, 4)==0 && /* suffix matches */ |
| 558 | uprv_strncmp(findBasename(pathBuffer.data()), basename, basenameLen)==0 && /* base matches */ |
| 559 | uprv_strlen(pathBasename)==(basenameLen+4)) { /* base+suffix = full len */ |
| 560 | |
| 561 | #ifdef UDATA_DEBUG |
| 562 | fprintf(stderr, "Have %s file on the path: %s\n", suffix, pathBuffer.data()); |
| 563 | #endif |
| 564 | /* do nothing */ |
| 565 | } |
| 566 | else |
| 567 | { /* regular dir path */ |
| 568 | if(pathBuffer[pathLen-1] != U_FILE_SEP_CHAR) { |
| 569 | if((pathLen>=4) && |
| 570 | uprv_strncmp(pathBuffer.data()+(pathLen-4), ".dat", 4) == 0) |
| 571 | { |
| 572 | #ifdef UDATA_DEBUG |
| 573 | fprintf(stderr, "skipping non-directory .dat file %s\n", pathBuffer.data()); |
| 574 | #endif |
| 575 | continue; |
| 576 | } |
| 577 | |
| 578 | /* Check if it is a directory with the same name as our package */ |
| 579 | if(!packageStub.isEmpty() && |
| 580 | (pathLen > packageStub.length()) && |
| 581 | !uprv_strcmp(pathBuffer.data() + pathLen - packageStub.length(), packageStub.data())) { |
| 582 | #ifdef UDATA_DEBUG |
| 583 | fprintf(stderr, "Found stub %s (will add package %s of len %d)\n", packageStub.data(), basename, basenameLen); |
| 584 | #endif |
| 585 | pathBuffer.truncate(pathLen - packageStub.length()); |
| 586 | } |
| 587 | pathBuffer.append(U_FILE_SEP_CHAR, *pErrorCode); |
| 588 | } |
| 589 | |
| 590 | /* + basename */ |
| 591 | pathBuffer.append(packageStub.data()+1, packageStub.length()-1, *pErrorCode); |
| 592 | |
| 593 | if(*suffix) /* tack on suffix */ |
| 594 | { |
| 595 | pathBuffer.append(suffix, *pErrorCode); |
| 596 | } |
| 597 | } |
| 598 | |
| 599 | #ifdef UDATA_DEBUG |
| 600 | fprintf(stderr, " --> %s\n", pathBuffer.data()); |
| 601 | #endif |
| 602 | |
| 603 | return pathBuffer.data(); |
| 604 | |
| 605 | } while(path); |
| 606 | |
| 607 | /* fell way off the end */ |
| 608 | return NULL; |
| 609 | } |
| 610 | |
| 611 | U_NAMESPACE_END |
| 612 | |
| 613 | /* ==================================================================================*/ |
| 614 | |
| 615 | |
| 616 | /*----------------------------------------------------------------------* |
| 617 | * * |
| 618 | * Add a static reference to the common data library * |
| 619 | * Unless overridden by an explicit udata_setCommonData, this will be * |
| 620 | * our common data. * |
| 621 | * * |
| 622 | *----------------------------------------------------------------------*/ |
Jungshik Shin (jungshik at google) | 5580ad6 | 2015-01-12 13:00:44 -0800 | [diff] [blame^] | 623 | extern "C" const ICU_Data_Header U_DATA_API U_ICUDATA_ENTRY_POINT; |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 624 | |
| 625 | /* |
| 626 | * This would be a good place for weak-linkage declarations of |
| 627 | * partial-data-library access functions where each returns a pointer |
| 628 | * to its data package, if it is linked in. |
| 629 | */ |
| 630 | /* |
| 631 | extern const void *uprv_getICUData_collation(void) ATTRIBUTE_WEAK; |
| 632 | extern const void *uprv_getICUData_conversion(void) ATTRIBUTE_WEAK; |
| 633 | */ |
| 634 | |
| 635 | /*----------------------------------------------------------------------* |
| 636 | * * |
| 637 | * openCommonData Attempt to open a common format (.dat) file * |
| 638 | * Map it into memory (if it's not there already) * |
| 639 | * and return a UDataMemory object for it. * |
| 640 | * * |
| 641 | * If the requested data is already open and cached * |
| 642 | * just return the cached UDataMem object. * |
| 643 | * * |
| 644 | *----------------------------------------------------------------------*/ |
| 645 | static UDataMemory * |
| 646 | openCommonData(const char *path, /* Path from OpenChoice? */ |
| 647 | int32_t commonDataIndex, /* ICU Data (index >= 0) if path == NULL */ |
| 648 | UErrorCode *pErrorCode) |
| 649 | { |
| 650 | UDataMemory tData; |
| 651 | const char *pathBuffer; |
| 652 | const char *inBasename; |
| 653 | |
| 654 | if (U_FAILURE(*pErrorCode)) { |
| 655 | return NULL; |
| 656 | } |
| 657 | |
| 658 | UDataMemory_init(&tData); |
| 659 | |
| 660 | /* ??????? TODO revisit this */ |
| 661 | if (commonDataIndex >= 0) { |
| 662 | /* "mini-cache" for common ICU data */ |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 663 | if(commonDataIndex >= UPRV_LENGTHOF(gCommonICUDataArray)) { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 664 | return NULL; |
| 665 | } |
| 666 | if(gCommonICUDataArray[commonDataIndex] == NULL) { |
| 667 | int32_t i; |
| 668 | for(i = 0; i < commonDataIndex; ++i) { |
Jungshik Shin (jungshik at google) | 5580ad6 | 2015-01-12 13:00:44 -0800 | [diff] [blame^] | 669 | if(gCommonICUDataArray[i]->pHeader == &U_ICUDATA_ENTRY_POINT.hdr) { |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 670 | /* The linked-in data is already in the list. */ |
| 671 | return NULL; |
| 672 | } |
| 673 | } |
| 674 | |
| 675 | /* Add the linked-in data to the list. */ |
| 676 | /* |
| 677 | * This is where we would check and call weakly linked partial-data-library |
| 678 | * access functions. |
| 679 | */ |
| 680 | /* |
| 681 | if (uprv_getICUData_collation) { |
| 682 | setCommonICUDataPointer(uprv_getICUData_collation(), FALSE, pErrorCode); |
| 683 | } |
| 684 | if (uprv_getICUData_conversion) { |
| 685 | setCommonICUDataPointer(uprv_getICUData_conversion(), FALSE, pErrorCode); |
| 686 | } |
| 687 | */ |
Jungshik Shin (jungshik at google) | 5580ad6 | 2015-01-12 13:00:44 -0800 | [diff] [blame^] | 688 | setCommonICUDataPointer(&U_ICUDATA_ENTRY_POINT.hdr, FALSE, pErrorCode); |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 689 | } |
| 690 | return gCommonICUDataArray[commonDataIndex]; |
| 691 | } |
| 692 | |
| 693 | |
| 694 | /* request is NOT for ICU Data. */ |
| 695 | |
| 696 | /* Find the base name portion of the supplied path. */ |
| 697 | /* inBasename will be left pointing somewhere within the original path string. */ |
| 698 | inBasename = findBasename(path); |
| 699 | #ifdef UDATA_DEBUG |
| 700 | fprintf(stderr, "inBasename = %s\n", inBasename); |
| 701 | #endif |
| 702 | |
| 703 | if(*inBasename==0) { |
| 704 | /* no basename. This will happen if the original path was a directory name, */ |
| 705 | /* like "a/b/c/". (Fallback to separate files will still work.) */ |
| 706 | #ifdef UDATA_DEBUG |
| 707 | fprintf(stderr, "ocd: no basename in %s, bailing.\n", path); |
| 708 | #endif |
| 709 | *pErrorCode=U_FILE_ACCESS_ERROR; |
| 710 | return NULL; |
| 711 | } |
| 712 | |
| 713 | /* Is the requested common data file already open and cached? */ |
| 714 | /* Note that the cache is keyed by the base name only. The rest of the path, */ |
| 715 | /* if any, is not considered. */ |
| 716 | { |
| 717 | UDataMemory *dataToReturn = udata_findCachedData(inBasename); |
| 718 | if (dataToReturn != NULL) { |
| 719 | return dataToReturn; |
| 720 | } |
| 721 | } |
| 722 | |
| 723 | /* Requested item is not in the cache. |
| 724 | * Hunt it down, trying all the path locations |
| 725 | */ |
| 726 | |
| 727 | UDataPathIterator iter(u_getDataDirectory(), inBasename, path, ".dat", TRUE, pErrorCode); |
| 728 | |
| 729 | while((UDataMemory_isLoaded(&tData)==FALSE) && (pathBuffer = iter.next(pErrorCode)) != NULL) |
| 730 | { |
| 731 | #ifdef UDATA_DEBUG |
| 732 | fprintf(stderr, "ocd: trying path %s - ", pathBuffer); |
| 733 | #endif |
| 734 | uprv_mapFile(&tData, pathBuffer); |
| 735 | #ifdef UDATA_DEBUG |
| 736 | fprintf(stderr, "%s\n", UDataMemory_isLoaded(&tData)?"LOADED":"not loaded"); |
| 737 | #endif |
| 738 | } |
| 739 | |
| 740 | #if defined(OS390_STUBDATA) && defined(OS390BATCH) |
| 741 | if (!UDataMemory_isLoaded(&tData)) { |
| 742 | char ourPathBuffer[1024]; |
| 743 | /* One more chance, for extendCommonData() */ |
| 744 | uprv_strncpy(ourPathBuffer, path, 1019); |
| 745 | ourPathBuffer[1019]=0; |
| 746 | uprv_strcat(ourPathBuffer, ".dat"); |
| 747 | uprv_mapFile(&tData, ourPathBuffer); |
| 748 | } |
| 749 | #endif |
| 750 | |
| 751 | if (!UDataMemory_isLoaded(&tData)) { |
| 752 | /* no common data */ |
| 753 | *pErrorCode=U_FILE_ACCESS_ERROR; |
| 754 | return NULL; |
| 755 | } |
| 756 | |
| 757 | /* we have mapped a file, check its header */ |
| 758 | udata_checkCommonData(&tData, pErrorCode); |
| 759 | |
| 760 | |
| 761 | /* Cache the UDataMemory struct for this .dat file, |
| 762 | * so we won't need to hunt it down and map it again next time |
| 763 | * something is needed from it. */ |
| 764 | return udata_cacheDataItem(inBasename, &tData, pErrorCode); |
| 765 | } |
| 766 | |
| 767 | |
| 768 | /*----------------------------------------------------------------------* |
| 769 | * * |
| 770 | * extendICUData If the full set of ICU data was not loaded at * |
| 771 | * program startup, load it now. This function will * |
| 772 | * be called when the lookup of an ICU data item in * |
| 773 | * the common ICU data fails. * |
| 774 | * * |
| 775 | * return true if new data is loaded, false otherwise.* |
| 776 | * * |
| 777 | *----------------------------------------------------------------------*/ |
| 778 | static UBool extendICUData(UErrorCode *pErr) |
| 779 | { |
| 780 | UDataMemory *pData; |
| 781 | UDataMemory copyPData; |
| 782 | UBool didUpdate = FALSE; |
| 783 | |
| 784 | /* |
| 785 | * There is a chance for a race condition here. |
| 786 | * Normally, ICU data is loaded from a DLL or via mmap() and |
| 787 | * setCommonICUData() will detect if the same address is set twice. |
| 788 | * If ICU is built with data loading via fread() then the address will |
| 789 | * be different each time the common data is loaded and we may add |
| 790 | * multiple copies of the data. |
| 791 | * In this case, use a mutex to prevent the race. |
| 792 | * Use a specific mutex to avoid nested locks of the global mutex. |
| 793 | */ |
| 794 | #if MAP_IMPLEMENTATION==MAP_STDIO |
| 795 | static UMutex extendICUDataMutex = U_MUTEX_INITIALIZER; |
| 796 | umtx_lock(&extendICUDataMutex); |
| 797 | #endif |
| 798 | if(!gHaveTriedToLoadCommonData) { |
| 799 | /* See if we can explicitly open a .dat file for the ICUData. */ |
| 800 | pData = openCommonData( |
| 801 | U_ICUDATA_NAME, /* "icudt20l" , for example. */ |
| 802 | -1, /* Pretend we're not opening ICUData */ |
| 803 | pErr); |
| 804 | |
| 805 | /* How about if there is no pData, eh... */ |
| 806 | |
| 807 | UDataMemory_init(©PData); |
| 808 | if(pData != NULL) { |
| 809 | UDatamemory_assign(©PData, pData); |
| 810 | copyPData.map = 0; /* The mapping for this data is owned by the hash table */ |
| 811 | copyPData.mapAddr = 0; /* which will unmap it when ICU is shut down. */ |
| 812 | /* CommonICUData is also unmapped when ICU is shut down.*/ |
| 813 | /* To avoid unmapping the data twice, zero out the map */ |
| 814 | /* fields in the UDataMemory that we're assigning */ |
| 815 | /* to CommonICUData. */ |
| 816 | |
| 817 | didUpdate = /* no longer using this result */ |
| 818 | setCommonICUData(©PData,/* The new common data. */ |
| 819 | FALSE, /* No warnings if write didn't happen */ |
| 820 | pErr); /* setCommonICUData honors errors; NOP if error set */ |
| 821 | } |
| 822 | |
| 823 | gHaveTriedToLoadCommonData = TRUE; |
| 824 | } |
| 825 | |
| 826 | didUpdate = findCommonICUDataByName(U_ICUDATA_NAME); /* Return 'true' when a racing writes out the extended */ |
| 827 | /* data after another thread has failed to see it (in openCommonData), so */ |
| 828 | /* extended data can be examined. */ |
| 829 | /* Also handles a race through here before gHaveTriedToLoadCommonData is set. */ |
| 830 | |
| 831 | #if MAP_IMPLEMENTATION==MAP_STDIO |
| 832 | umtx_unlock(&extendICUDataMutex); |
| 833 | #endif |
| 834 | return didUpdate; /* Return true if ICUData pointer was updated. */ |
| 835 | /* (Could potentialy have been done by another thread racing */ |
| 836 | /* us through here, but that's fine, we still return true */ |
| 837 | /* so that current thread will also examine extended data. */ |
| 838 | } |
| 839 | |
| 840 | /*----------------------------------------------------------------------* |
| 841 | * * |
| 842 | * udata_setCommonData * |
| 843 | * * |
| 844 | *----------------------------------------------------------------------*/ |
| 845 | U_CAPI void U_EXPORT2 |
| 846 | udata_setCommonData(const void *data, UErrorCode *pErrorCode) { |
| 847 | UDataMemory dataMemory; |
| 848 | |
| 849 | if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 850 | return; |
| 851 | } |
| 852 | |
| 853 | if(data==NULL) { |
| 854 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
| 855 | return; |
| 856 | } |
| 857 | |
| 858 | /* set the data pointer and test for validity */ |
| 859 | UDataMemory_init(&dataMemory); |
| 860 | UDataMemory_setData(&dataMemory, data); |
| 861 | udata_checkCommonData(&dataMemory, pErrorCode); |
| 862 | if (U_FAILURE(*pErrorCode)) {return;} |
| 863 | |
| 864 | /* we have good data */ |
| 865 | /* Set it up as the ICU Common Data. */ |
| 866 | setCommonICUData(&dataMemory, TRUE, pErrorCode); |
| 867 | } |
| 868 | |
| 869 | /*--------------------------------------------------------------------------- |
| 870 | * |
| 871 | * udata_setAppData |
| 872 | * |
| 873 | *---------------------------------------------------------------------------- */ |
| 874 | U_CAPI void U_EXPORT2 |
| 875 | udata_setAppData(const char *path, const void *data, UErrorCode *err) |
| 876 | { |
| 877 | UDataMemory udm; |
| 878 | |
| 879 | if(err==NULL || U_FAILURE(*err)) { |
| 880 | return; |
| 881 | } |
| 882 | if(data==NULL) { |
| 883 | *err=U_ILLEGAL_ARGUMENT_ERROR; |
| 884 | return; |
| 885 | } |
| 886 | |
| 887 | UDataMemory_init(&udm); |
| 888 | UDataMemory_setData(&udm, data); |
| 889 | udata_checkCommonData(&udm, err); |
| 890 | udata_cacheDataItem(path, &udm, err); |
| 891 | } |
| 892 | |
| 893 | /*----------------------------------------------------------------------------* |
| 894 | * * |
| 895 | * checkDataItem Given a freshly located/loaded data item, either * |
| 896 | * an entry in a common file or a separately loaded file, * |
| 897 | * sanity check its header, and see if the data is * |
| 898 | * acceptable to the app. * |
| 899 | * If the data is good, create and return a UDataMemory * |
| 900 | * object that can be returned to the application. * |
| 901 | * Return NULL on any sort of failure. * |
| 902 | * * |
| 903 | *----------------------------------------------------------------------------*/ |
| 904 | static UDataMemory * |
| 905 | checkDataItem |
| 906 | ( |
| 907 | const DataHeader *pHeader, /* The data item to be checked. */ |
| 908 | UDataMemoryIsAcceptable *isAcceptable, /* App's call-back function */ |
| 909 | void *context, /* pass-thru param for above. */ |
| 910 | const char *type, /* pass-thru param for above. */ |
| 911 | const char *name, /* pass-thru param for above. */ |
| 912 | UErrorCode *nonFatalErr, /* Error code if this data was not acceptable */ |
| 913 | /* but openChoice should continue with */ |
| 914 | /* trying to get data from fallback path. */ |
| 915 | UErrorCode *fatalErr /* Bad error, caller should return immediately */ |
| 916 | ) |
| 917 | { |
| 918 | UDataMemory *rDataMem = NULL; /* the new UDataMemory, to be returned. */ |
| 919 | |
| 920 | if (U_FAILURE(*fatalErr)) { |
| 921 | return NULL; |
| 922 | } |
| 923 | |
| 924 | if(pHeader->dataHeader.magic1==0xda && |
| 925 | pHeader->dataHeader.magic2==0x27 && |
| 926 | (isAcceptable==NULL || isAcceptable(context, type, name, &pHeader->info)) |
| 927 | ) { |
| 928 | rDataMem=UDataMemory_createNewInstance(fatalErr); |
| 929 | if (U_FAILURE(*fatalErr)) { |
| 930 | return NULL; |
| 931 | } |
| 932 | rDataMem->pHeader = pHeader; |
| 933 | } else { |
| 934 | /* the data is not acceptable, look further */ |
| 935 | /* If we eventually find something good, this errorcode will be */ |
| 936 | /* cleared out. */ |
| 937 | *nonFatalErr=U_INVALID_FORMAT_ERROR; |
| 938 | } |
| 939 | return rDataMem; |
| 940 | } |
| 941 | |
| 942 | /** |
| 943 | * @return 0 if not loaded, 1 if loaded or err |
| 944 | */ |
| 945 | static UDataMemory *doLoadFromIndividualFiles(const char *pkgName, |
| 946 | const char *dataPath, const char *tocEntryPathSuffix, |
| 947 | /* following arguments are the same as doOpenChoice itself */ |
| 948 | const char *path, const char *type, const char *name, |
| 949 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 950 | UErrorCode *subErrorCode, |
| 951 | UErrorCode *pErrorCode) |
| 952 | { |
| 953 | const char *pathBuffer; |
| 954 | UDataMemory dataMemory; |
| 955 | UDataMemory *pEntryData; |
| 956 | |
| 957 | /* look in ind. files: package\nam.typ ========================= */ |
| 958 | /* init path iterator for individual files */ |
| 959 | UDataPathIterator iter(dataPath, pkgName, path, tocEntryPathSuffix, FALSE, pErrorCode); |
| 960 | |
| 961 | while((pathBuffer = iter.next(pErrorCode))) |
| 962 | { |
| 963 | #ifdef UDATA_DEBUG |
| 964 | fprintf(stderr, "UDATA: trying individual file %s\n", pathBuffer); |
| 965 | #endif |
| 966 | if(uprv_mapFile(&dataMemory, pathBuffer)) |
| 967 | { |
| 968 | pEntryData = checkDataItem(dataMemory.pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode); |
| 969 | if (pEntryData != NULL) { |
| 970 | /* Data is good. |
| 971 | * Hand off ownership of the backing memory to the user's UDataMemory. |
| 972 | * and return it. */ |
| 973 | pEntryData->mapAddr = dataMemory.mapAddr; |
| 974 | pEntryData->map = dataMemory.map; |
| 975 | |
| 976 | #ifdef UDATA_DEBUG |
| 977 | fprintf(stderr, "** Mapped file: %s\n", pathBuffer); |
| 978 | #endif |
| 979 | return pEntryData; |
| 980 | } |
| 981 | |
| 982 | /* the data is not acceptable, or some error occured. Either way, unmap the memory */ |
| 983 | udata_close(&dataMemory); |
| 984 | |
| 985 | /* If we had a nasty error, bail out completely. */ |
| 986 | if (U_FAILURE(*pErrorCode)) { |
| 987 | return NULL; |
| 988 | } |
| 989 | |
| 990 | /* Otherwise remember that we found data but didn't like it for some reason */ |
| 991 | *subErrorCode=U_INVALID_FORMAT_ERROR; |
| 992 | } |
| 993 | #ifdef UDATA_DEBUG |
| 994 | fprintf(stderr, "%s\n", UDataMemory_isLoaded(&dataMemory)?"LOADED":"not loaded"); |
| 995 | #endif |
| 996 | } |
| 997 | return NULL; |
| 998 | } |
| 999 | |
| 1000 | /** |
| 1001 | * @return 0 if not loaded, 1 if loaded or err |
| 1002 | */ |
| 1003 | static UDataMemory *doLoadFromCommonData(UBool isICUData, const char * /*pkgName*/, |
| 1004 | const char * /*dataPath*/, const char * /*tocEntryPathSuffix*/, const char *tocEntryName, |
| 1005 | /* following arguments are the same as doOpenChoice itself */ |
| 1006 | const char *path, const char *type, const char *name, |
| 1007 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 1008 | UErrorCode *subErrorCode, |
| 1009 | UErrorCode *pErrorCode) |
| 1010 | { |
| 1011 | UDataMemory *pEntryData; |
| 1012 | const DataHeader *pHeader; |
| 1013 | UDataMemory *pCommonData; |
| 1014 | int32_t commonDataIndex; |
| 1015 | UBool checkedExtendedICUData = FALSE; |
| 1016 | /* try to get common data. The loop is for platforms such as the 390 that do |
| 1017 | * not initially load the full set of ICU data. If the lookup of an ICU data item |
| 1018 | * fails, the full (but slower to load) set is loaded, the and the loop repeats, |
| 1019 | * trying the lookup again. Once the full set of ICU data is loaded, the loop wont |
| 1020 | * repeat because the full set will be checked the first time through. |
| 1021 | * |
| 1022 | * The loop also handles the fallback to a .dat file if the application linked |
| 1023 | * to the stub data library rather than a real library. |
| 1024 | */ |
| 1025 | for (commonDataIndex = isICUData ? 0 : -1;;) { |
| 1026 | pCommonData=openCommonData(path, commonDataIndex, subErrorCode); /** search for pkg **/ |
| 1027 | |
| 1028 | if(U_SUCCESS(*subErrorCode) && pCommonData!=NULL) { |
| 1029 | int32_t length; |
| 1030 | |
| 1031 | /* look up the data piece in the common data */ |
| 1032 | pHeader=pCommonData->vFuncs->Lookup(pCommonData, tocEntryName, &length, subErrorCode); |
| 1033 | #ifdef UDATA_DEBUG |
| 1034 | fprintf(stderr, "%s: pHeader=%p - %s\n", tocEntryName, pHeader, u_errorName(*subErrorCode)); |
| 1035 | #endif |
| 1036 | |
| 1037 | if(pHeader!=NULL) { |
| 1038 | pEntryData = checkDataItem(pHeader, isAcceptable, context, type, name, subErrorCode, pErrorCode); |
| 1039 | #ifdef UDATA_DEBUG |
| 1040 | fprintf(stderr, "pEntryData=%p\n", pEntryData); |
| 1041 | #endif |
| 1042 | if (U_FAILURE(*pErrorCode)) { |
| 1043 | return NULL; |
| 1044 | } |
| 1045 | if (pEntryData != NULL) { |
| 1046 | pEntryData->length = length; |
| 1047 | return pEntryData; |
| 1048 | } |
| 1049 | } |
| 1050 | } |
| 1051 | /* Data wasn't found. If we were looking for an ICUData item and there is |
| 1052 | * more data available, load it and try again, |
| 1053 | * otherwise break out of this loop. */ |
| 1054 | if (!isICUData) { |
| 1055 | return NULL; |
| 1056 | } else if (pCommonData != NULL) { |
| 1057 | ++commonDataIndex; /* try the next data package */ |
| 1058 | } else if ((!checkedExtendedICUData) && extendICUData(subErrorCode)) { |
| 1059 | checkedExtendedICUData = TRUE; |
| 1060 | /* try this data package slot again: it changed from NULL to non-NULL */ |
| 1061 | } else { |
| 1062 | return NULL; |
| 1063 | } |
| 1064 | } |
| 1065 | } |
| 1066 | |
| 1067 | /* |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 1068 | * Identify the Time Zone resources that are subject to special override data loading. |
| 1069 | */ |
| 1070 | static UBool isTimeZoneFile(const char *name, const char *type) { |
| 1071 | return ((uprv_strcmp(type, "res") == 0) && |
| 1072 | (uprv_strcmp(name, "zoneinfo64") == 0 || |
| 1073 | uprv_strcmp(name, "timezoneTypes") == 0 || |
| 1074 | uprv_strcmp(name, "windowsZones") == 0 || |
| 1075 | uprv_strcmp(name, "metaZones") == 0)); |
| 1076 | } |
| 1077 | |
| 1078 | /* |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 1079 | * A note on the ownership of Mapped Memory |
| 1080 | * |
| 1081 | * For common format files, ownership resides with the UDataMemory object |
| 1082 | * that lives in the cache of opened common data. These UDataMemorys are private |
| 1083 | * to the udata implementation, and are never seen directly by users. |
| 1084 | * |
| 1085 | * The UDataMemory objects returned to users will have the address of some desired |
| 1086 | * data within the mapped region, but they wont have the mapping info itself, and thus |
| 1087 | * won't cause anything to be removed from memory when they are closed. |
| 1088 | * |
| 1089 | * For individual data files, the UDataMemory returned to the user holds the |
| 1090 | * information necessary to unmap the data on close. If the user independently |
| 1091 | * opens the same data file twice, two completely independent mappings will be made. |
| 1092 | * (There is no cache of opened data items from individual files, only a cache of |
| 1093 | * opened Common Data files, that is, files containing a collection of data items.) |
| 1094 | * |
| 1095 | * For common data passed in from the user via udata_setAppData() or |
| 1096 | * udata_setCommonData(), ownership remains with the user. |
| 1097 | * |
| 1098 | * UDataMemory objects themselves, as opposed to the memory they describe, |
| 1099 | * can be anywhere - heap, stack/local or global. |
| 1100 | * They have a flag to indicate when they're heap allocated and thus |
| 1101 | * must be deleted when closed. |
| 1102 | */ |
| 1103 | |
| 1104 | |
| 1105 | /*----------------------------------------------------------------------------* |
| 1106 | * * |
| 1107 | * main data loading functions * |
| 1108 | * * |
| 1109 | *----------------------------------------------------------------------------*/ |
| 1110 | static UDataMemory * |
| 1111 | doOpenChoice(const char *path, const char *type, const char *name, |
| 1112 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 1113 | UErrorCode *pErrorCode) |
| 1114 | { |
| 1115 | UDataMemory *retVal = NULL; |
| 1116 | |
| 1117 | const char *dataPath; |
| 1118 | |
| 1119 | int32_t tocEntrySuffixIndex; |
| 1120 | const char *tocEntryPathSuffix; |
| 1121 | UErrorCode subErrorCode=U_ZERO_ERROR; |
| 1122 | const char *treeChar; |
| 1123 | |
| 1124 | UBool isICUData = FALSE; |
| 1125 | |
| 1126 | |
| 1127 | /* Is this path ICU data? */ |
| 1128 | if(path == NULL || |
| 1129 | !strcmp(path, U_ICUDATA_ALIAS) || /* "ICUDATA" */ |
| 1130 | !uprv_strncmp(path, U_ICUDATA_NAME U_TREE_SEPARATOR_STRING, /* "icudt26e-" */ |
| 1131 | uprv_strlen(U_ICUDATA_NAME U_TREE_SEPARATOR_STRING)) || |
| 1132 | !uprv_strncmp(path, U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING, /* "ICUDATA-" */ |
| 1133 | uprv_strlen(U_ICUDATA_ALIAS U_TREE_SEPARATOR_STRING))) { |
| 1134 | isICUData = TRUE; |
| 1135 | } |
| 1136 | |
| 1137 | #if (U_FILE_SEP_CHAR != U_FILE_ALT_SEP_CHAR) /* Windows: try "foo\bar" and "foo/bar" */ |
| 1138 | /* remap from alternate path char to the main one */ |
| 1139 | CharString altSepPath; |
| 1140 | if(path) { |
| 1141 | if(uprv_strchr(path,U_FILE_ALT_SEP_CHAR) != NULL) { |
| 1142 | altSepPath.append(path, *pErrorCode); |
| 1143 | char *p; |
| 1144 | while((p=uprv_strchr(altSepPath.data(), U_FILE_ALT_SEP_CHAR))) { |
| 1145 | *p = U_FILE_SEP_CHAR; |
| 1146 | } |
| 1147 | #if defined (UDATA_DEBUG) |
| 1148 | fprintf(stderr, "Changed path from [%s] to [%s]\n", path, altSepPath.s); |
| 1149 | #endif |
| 1150 | path = altSepPath.data(); |
| 1151 | } |
| 1152 | } |
| 1153 | #endif |
| 1154 | |
| 1155 | CharString tocEntryName; /* entry name in tree format. ex: 'icudt28b/coll/ar.res' */ |
| 1156 | CharString tocEntryPath; /* entry name in path format. ex: 'icudt28b\\coll\\ar.res' */ |
| 1157 | |
| 1158 | CharString pkgName; |
| 1159 | CharString treeName; |
| 1160 | |
| 1161 | /* ======= Set up strings */ |
| 1162 | if(path==NULL) { |
| 1163 | pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
| 1164 | } else { |
| 1165 | const char *pkg; |
| 1166 | const char *first; |
| 1167 | pkg = uprv_strrchr(path, U_FILE_SEP_CHAR); |
| 1168 | first = uprv_strchr(path, U_FILE_SEP_CHAR); |
| 1169 | if(uprv_pathIsAbsolute(path) || (pkg != first)) { /* more than one slash in the path- not a tree name */ |
| 1170 | /* see if this is an /absolute/path/to/package path */ |
| 1171 | if(pkg) { |
| 1172 | pkgName.append(pkg+1, *pErrorCode); |
| 1173 | } else { |
| 1174 | pkgName.append(path, *pErrorCode); |
| 1175 | } |
| 1176 | } else { |
| 1177 | treeChar = uprv_strchr(path, U_TREE_SEPARATOR); |
| 1178 | if(treeChar) { |
| 1179 | treeName.append(treeChar+1, *pErrorCode); /* following '-' */ |
| 1180 | if(isICUData) { |
| 1181 | pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
| 1182 | } else { |
| 1183 | pkgName.append(path, (int32_t)(treeChar-path), *pErrorCode); |
| 1184 | if (first == NULL) { |
| 1185 | /* |
| 1186 | This user data has no path, but there is a tree name. |
| 1187 | Look up the correct path from the data cache later. |
| 1188 | */ |
| 1189 | path = pkgName.data(); |
| 1190 | } |
| 1191 | } |
| 1192 | } else { |
| 1193 | if(isICUData) { |
| 1194 | pkgName.append(U_ICUDATA_NAME, *pErrorCode); |
| 1195 | } else { |
| 1196 | pkgName.append(path, *pErrorCode); |
| 1197 | } |
| 1198 | } |
| 1199 | } |
| 1200 | } |
| 1201 | |
| 1202 | #ifdef UDATA_DEBUG |
| 1203 | fprintf(stderr, " P=%s T=%s\n", pkgName.data(), treeName.data()); |
| 1204 | #endif |
| 1205 | |
| 1206 | /* setting up the entry name and file name |
| 1207 | * Make up a full name by appending the type to the supplied |
| 1208 | * name, assuming that a type was supplied. |
| 1209 | */ |
| 1210 | |
| 1211 | /* prepend the package */ |
| 1212 | tocEntryName.append(pkgName, *pErrorCode); |
| 1213 | tocEntryPath.append(pkgName, *pErrorCode); |
| 1214 | tocEntrySuffixIndex = tocEntryName.length(); |
| 1215 | |
| 1216 | if(!treeName.isEmpty()) { |
| 1217 | tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode); |
| 1218 | tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(treeName, *pErrorCode); |
| 1219 | } |
| 1220 | |
| 1221 | tocEntryName.append(U_TREE_ENTRY_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
| 1222 | tocEntryPath.append(U_FILE_SEP_CHAR, *pErrorCode).append(name, *pErrorCode); |
| 1223 | if(type!=NULL && *type!=0) { |
| 1224 | tocEntryName.append(".", *pErrorCode).append(type, *pErrorCode); |
| 1225 | tocEntryPath.append(".", *pErrorCode).append(type, *pErrorCode); |
| 1226 | } |
| 1227 | tocEntryPathSuffix = tocEntryPath.data()+tocEntrySuffixIndex; /* suffix starts here */ |
| 1228 | |
| 1229 | #ifdef UDATA_DEBUG |
| 1230 | fprintf(stderr, " tocEntryName = %s\n", tocEntryName.data()); |
| 1231 | fprintf(stderr, " tocEntryPath = %s\n", tocEntryName.data()); |
| 1232 | #endif |
| 1233 | |
| 1234 | if(path == NULL) { |
| 1235 | path = COMMON_DATA_NAME; /* "icudt26e" */ |
| 1236 | } |
| 1237 | |
| 1238 | /************************ Begin loop looking for ind. files ***************/ |
| 1239 | #ifdef UDATA_DEBUG |
| 1240 | fprintf(stderr, "IND: inBasename = %s, pkg=%s\n", "(n/a)", packageNameFromPath(path)); |
| 1241 | #endif |
| 1242 | |
| 1243 | /* End of dealing with a null basename */ |
| 1244 | dataPath = u_getDataDirectory(); |
| 1245 | |
Jungshik Shin (jungshik at google) | 0f8746a | 2015-01-08 15:46:45 -0800 | [diff] [blame] | 1246 | /**** Time zone individual files override */ |
| 1247 | if (isTimeZoneFile(name, type) && isICUData) { |
| 1248 | const char *tzFilesDir = u_getTimeZoneFilesDirectory(pErrorCode); |
| 1249 | if (tzFilesDir[0] != 0) { |
| 1250 | #ifdef UDATA_DEBUG |
| 1251 | fprintf(stderr, "Trying Time Zone Files directory = %s\n", tzFilesDir); |
| 1252 | #endif |
| 1253 | retVal = doLoadFromIndividualFiles(/* pkgName.data() */ "", tzFilesDir, tocEntryPathSuffix, |
| 1254 | /* path */ "", type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
| 1255 | if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1256 | return retVal; |
| 1257 | } |
| 1258 | } |
| 1259 | } |
| 1260 | |
jshin@chromium.org | 6f31ac3 | 2014-03-26 22:15:14 +0000 | [diff] [blame] | 1261 | /**** COMMON PACKAGE - only if packages are first. */ |
| 1262 | if(gDataFileAccess == UDATA_PACKAGES_FIRST) { |
| 1263 | #ifdef UDATA_DEBUG |
| 1264 | fprintf(stderr, "Trying packages (UDATA_PACKAGES_FIRST)\n"); |
| 1265 | #endif |
| 1266 | /* #2 */ |
| 1267 | retVal = doLoadFromCommonData(isICUData, |
| 1268 | pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), |
| 1269 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
| 1270 | if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1271 | return retVal; |
| 1272 | } |
| 1273 | } |
| 1274 | |
| 1275 | /**** INDIVIDUAL FILES */ |
| 1276 | if((gDataFileAccess==UDATA_PACKAGES_FIRST) || |
| 1277 | (gDataFileAccess==UDATA_FILES_FIRST)) { |
| 1278 | #ifdef UDATA_DEBUG |
| 1279 | fprintf(stderr, "Trying individual files\n"); |
| 1280 | #endif |
| 1281 | /* Check to make sure that there is a dataPath to iterate over */ |
| 1282 | if ((dataPath && *dataPath) || !isICUData) { |
| 1283 | retVal = doLoadFromIndividualFiles(pkgName.data(), dataPath, tocEntryPathSuffix, |
| 1284 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
| 1285 | if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1286 | return retVal; |
| 1287 | } |
| 1288 | } |
| 1289 | } |
| 1290 | |
| 1291 | /**** COMMON PACKAGE */ |
| 1292 | if((gDataFileAccess==UDATA_ONLY_PACKAGES) || |
| 1293 | (gDataFileAccess==UDATA_FILES_FIRST)) { |
| 1294 | #ifdef UDATA_DEBUG |
| 1295 | fprintf(stderr, "Trying packages (UDATA_ONLY_PACKAGES || UDATA_FILES_FIRST)\n"); |
| 1296 | #endif |
| 1297 | retVal = doLoadFromCommonData(isICUData, |
| 1298 | pkgName.data(), dataPath, tocEntryPathSuffix, tocEntryName.data(), |
| 1299 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
| 1300 | if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1301 | return retVal; |
| 1302 | } |
| 1303 | } |
| 1304 | |
| 1305 | /* Load from DLL. If we haven't attempted package load, we also haven't had any chance to |
| 1306 | try a DLL (static or setCommonData/etc) load. |
| 1307 | If we ever have a "UDATA_ONLY_FILES", add it to the or list here. */ |
| 1308 | if(gDataFileAccess==UDATA_NO_FILES) { |
| 1309 | #ifdef UDATA_DEBUG |
| 1310 | fprintf(stderr, "Trying common data (UDATA_NO_FILES)\n"); |
| 1311 | #endif |
| 1312 | retVal = doLoadFromCommonData(isICUData, |
| 1313 | pkgName.data(), "", tocEntryPathSuffix, tocEntryName.data(), |
| 1314 | path, type, name, isAcceptable, context, &subErrorCode, pErrorCode); |
| 1315 | if((retVal != NULL) || U_FAILURE(*pErrorCode)) { |
| 1316 | return retVal; |
| 1317 | } |
| 1318 | } |
| 1319 | |
| 1320 | /* data not found */ |
| 1321 | if(U_SUCCESS(*pErrorCode)) { |
| 1322 | if(U_SUCCESS(subErrorCode)) { |
| 1323 | /* file not found */ |
| 1324 | *pErrorCode=U_FILE_ACCESS_ERROR; |
| 1325 | } else { |
| 1326 | /* entry point not found or rejected */ |
| 1327 | *pErrorCode=subErrorCode; |
| 1328 | } |
| 1329 | } |
| 1330 | return retVal; |
| 1331 | } |
| 1332 | |
| 1333 | |
| 1334 | |
| 1335 | /* API ---------------------------------------------------------------------- */ |
| 1336 | |
| 1337 | U_CAPI UDataMemory * U_EXPORT2 |
| 1338 | udata_open(const char *path, const char *type, const char *name, |
| 1339 | UErrorCode *pErrorCode) { |
| 1340 | #ifdef UDATA_DEBUG |
| 1341 | fprintf(stderr, "udata_open(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type); |
| 1342 | fflush(stderr); |
| 1343 | #endif |
| 1344 | |
| 1345 | if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 1346 | return NULL; |
| 1347 | } else if(name==NULL || *name==0) { |
| 1348 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
| 1349 | return NULL; |
| 1350 | } else { |
| 1351 | return doOpenChoice(path, type, name, NULL, NULL, pErrorCode); |
| 1352 | } |
| 1353 | } |
| 1354 | |
| 1355 | |
| 1356 | |
| 1357 | U_CAPI UDataMemory * U_EXPORT2 |
| 1358 | udata_openChoice(const char *path, const char *type, const char *name, |
| 1359 | UDataMemoryIsAcceptable *isAcceptable, void *context, |
| 1360 | UErrorCode *pErrorCode) { |
| 1361 | #ifdef UDATA_DEBUG |
| 1362 | fprintf(stderr, "udata_openChoice(): Opening: %s : %s . %s\n", (path?path:"NULL"), name, type); |
| 1363 | #endif |
| 1364 | |
| 1365 | if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) { |
| 1366 | return NULL; |
| 1367 | } else if(name==NULL || *name==0 || isAcceptable==NULL) { |
| 1368 | *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; |
| 1369 | return NULL; |
| 1370 | } else { |
| 1371 | return doOpenChoice(path, type, name, isAcceptable, context, pErrorCode); |
| 1372 | } |
| 1373 | } |
| 1374 | |
| 1375 | |
| 1376 | |
| 1377 | U_CAPI void U_EXPORT2 |
| 1378 | udata_getInfo(UDataMemory *pData, UDataInfo *pInfo) { |
| 1379 | if(pInfo!=NULL) { |
| 1380 | if(pData!=NULL && pData->pHeader!=NULL) { |
| 1381 | const UDataInfo *info=&pData->pHeader->info; |
| 1382 | uint16_t dataInfoSize=udata_getInfoSize(info); |
| 1383 | if(pInfo->size>dataInfoSize) { |
| 1384 | pInfo->size=dataInfoSize; |
| 1385 | } |
| 1386 | uprv_memcpy((uint16_t *)pInfo+1, (const uint16_t *)info+1, pInfo->size-2); |
| 1387 | if(info->isBigEndian!=U_IS_BIG_ENDIAN) { |
| 1388 | /* opposite endianness */ |
| 1389 | uint16_t x=info->reservedWord; |
| 1390 | pInfo->reservedWord=(uint16_t)((x<<8)|(x>>8)); |
| 1391 | } |
| 1392 | } else { |
| 1393 | pInfo->size=0; |
| 1394 | } |
| 1395 | } |
| 1396 | } |
| 1397 | |
| 1398 | |
| 1399 | U_CAPI void U_EXPORT2 udata_setFileAccess(UDataFileAccess access, UErrorCode * /*status*/) |
| 1400 | { |
| 1401 | gDataFileAccess = access; |
| 1402 | } |