]> git.saurik.com Git - apple/icu.git/blob - icuSources/common/rbbidata.cpp
ICU-551.51.4.tar.gz
[apple/icu.git] / icuSources / common / rbbidata.cpp
1 /*
2 ***************************************************************************
3 * Copyright (C) 1999-2014 International Business Machines Corporation *
4 * and others. All rights reserved. *
5 ***************************************************************************
6 */
7
8 #include "unicode/utypes.h"
9
10 #if !UCONFIG_NO_BREAK_ITERATION
11
12 #include "unicode/utypes.h"
13 #include "rbbidata.h"
14 #include "rbbirb.h"
15 #include "utrie.h"
16 #include "udatamem.h"
17 #include "cmemory.h"
18 #include "cstring.h"
19 #include "umutex.h"
20
21 #include "uassert.h"
22
23
24 //-----------------------------------------------------------------------------------
25 //
26 // Trie access folding function. Copied as-is from properties code in uchar.c
27 //
28 //-----------------------------------------------------------------------------------
29 U_CDECL_BEGIN
30 static int32_t U_CALLCONV
31 getFoldingOffset(uint32_t data) {
32 /* if bit 15 is set, then the folding offset is in bits 14..0 of the 16-bit trie result */
33 if(data&0x8000) {
34 return (int32_t)(data&0x7fff);
35 } else {
36 return 0;
37 }
38 }
39 U_CDECL_END
40
41 U_NAMESPACE_BEGIN
42
43 //-----------------------------------------------------------------------------
44 //
45 // Constructors.
46 //
47 //-----------------------------------------------------------------------------
48 RBBIDataWrapper::RBBIDataWrapper(const RBBIDataHeader *data, UErrorCode &status) {
49 init0();
50 init(data, status);
51 }
52
53 RBBIDataWrapper::RBBIDataWrapper(const RBBIDataHeader *data, enum EDontAdopt, UErrorCode &status) {
54 init0();
55 init(data, status);
56 fDontFreeData = TRUE;
57 }
58
59 RBBIDataWrapper::RBBIDataWrapper(UDataMemory* udm, UErrorCode &status) {
60 init0();
61 if (U_FAILURE(status)) {
62 return;
63 }
64 const DataHeader *dh = udm->pHeader;
65 int32_t headerSize = dh->dataHeader.headerSize;
66 if ( !(headerSize >= 20 &&
67 dh->info.isBigEndian == U_IS_BIG_ENDIAN &&
68 dh->info.charsetFamily == U_CHARSET_FAMILY &&
69 dh->info.dataFormat[0] == 0x42 && // dataFormat="Brk "
70 dh->info.dataFormat[1] == 0x72 &&
71 dh->info.dataFormat[2] == 0x6b &&
72 dh->info.dataFormat[3] == 0x20)
73 // Note: info.fFormatVersion is duplicated in the RBBIDataHeader, and is
74 // validated when checking that.
75 ) {
76 status = U_INVALID_FORMAT_ERROR;
77 return;
78 }
79 const char *dataAsBytes = reinterpret_cast<const char *>(dh);
80 const RBBIDataHeader *rbbidh = reinterpret_cast<const RBBIDataHeader *>(dataAsBytes + headerSize);
81 init(rbbidh, status);
82 fUDataMem = udm;
83 }
84
85 //-----------------------------------------------------------------------------
86 //
87 // init(). Does most of the work of construction, shared between the
88 // constructors.
89 //
90 //-----------------------------------------------------------------------------
91 void RBBIDataWrapper::init0() {
92 fHeader = NULL;
93 fForwardTable = NULL;
94 fReverseTable = NULL;
95 fSafeFwdTable = NULL;
96 fSafeRevTable = NULL;
97 fRuleSource = NULL;
98 fRuleStatusTable = NULL;
99 fUDataMem = NULL;
100 fRefCount = 0;
101 fDontFreeData = TRUE;
102 }
103
104 void RBBIDataWrapper::init(const RBBIDataHeader *data, UErrorCode &status) {
105 if (U_FAILURE(status)) {
106 return;
107 }
108 fHeader = data;
109 if (fHeader->fMagic != 0xb1a0 || fHeader->fFormatVersion[0] != 3)
110 {
111 status = U_INVALID_FORMAT_ERROR;
112 return;
113 }
114 // Note: in ICU version 3.2 and earlier, there was a formatVersion 1
115 // that is no longer supported. At that time fFormatVersion was
116 // an int32_t field, rather than an array of 4 bytes.
117
118 fDontFreeData = FALSE;
119 if (data->fFTableLen != 0) {
120 fForwardTable = (RBBIStateTable *)((char *)data + fHeader->fFTable);
121 }
122 if (data->fRTableLen != 0) {
123 fReverseTable = (RBBIStateTable *)((char *)data + fHeader->fRTable);
124 }
125 if (data->fSFTableLen != 0) {
126 fSafeFwdTable = (RBBIStateTable *)((char *)data + fHeader->fSFTable);
127 }
128 if (data->fSRTableLen != 0) {
129 fSafeRevTable = (RBBIStateTable *)((char *)data + fHeader->fSRTable);
130 }
131
132
133 utrie_unserialize(&fTrie,
134 (uint8_t *)data + fHeader->fTrie,
135 fHeader->fTrieLen,
136 &status);
137 if (U_FAILURE(status)) {
138 return;
139 }
140 fTrie.getFoldingOffset=getFoldingOffset;
141
142
143 fRuleSource = (UChar *)((char *)data + fHeader->fRuleSource);
144 fRuleString.setTo(TRUE, fRuleSource, -1);
145 U_ASSERT(data->fRuleSourceLen > 0);
146
147 fRuleStatusTable = (int32_t *)((char *)data + fHeader->fStatusTable);
148 fStatusMaxIdx = data->fStatusTableLen / sizeof(int32_t);
149
150 fRefCount = 1;
151
152 #ifdef RBBI_DEBUG
153 char *debugEnv = getenv("U_RBBIDEBUG");
154 if (debugEnv && uprv_strstr(debugEnv, "data")) {this->printData();}
155 #endif
156 }
157
158
159 //-----------------------------------------------------------------------------
160 //
161 // Destructor. Don't call this - use removeReference() instead.
162 //
163 //-----------------------------------------------------------------------------
164 RBBIDataWrapper::~RBBIDataWrapper() {
165 U_ASSERT(fRefCount == 0);
166 if (fUDataMem) {
167 udata_close(fUDataMem);
168 } else if (!fDontFreeData) {
169 uprv_free((void *)fHeader);
170 }
171 }
172
173
174
175 //-----------------------------------------------------------------------------
176 //
177 // Operator == Consider two RBBIDataWrappers to be equal if they
178 // refer to the same underlying data. Although
179 // the data wrappers are normally shared between
180 // iterator instances, it's possible to independently
181 // open the same data twice, and get two instances, which
182 // should still be ==.
183 //
184 //-----------------------------------------------------------------------------
185 UBool RBBIDataWrapper::operator ==(const RBBIDataWrapper &other) const {
186 if (fHeader == other.fHeader) {
187 return TRUE;
188 }
189 if (fHeader->fLength != other.fHeader->fLength) {
190 return FALSE;
191 }
192 if (uprv_memcmp(fHeader, other.fHeader, fHeader->fLength) == 0) {
193 return TRUE;
194 }
195 return FALSE;
196 }
197
198 int32_t RBBIDataWrapper::hashCode() {
199 return fHeader->fFTableLen;
200 }
201
202
203
204 //-----------------------------------------------------------------------------
205 //
206 // Reference Counting. A single RBBIDataWrapper object is shared among
207 // however many RulesBasedBreakIterator instances are
208 // referencing the same data.
209 //
210 //-----------------------------------------------------------------------------
211 void RBBIDataWrapper::removeReference() {
212 if (umtx_atomic_dec(&fRefCount) == 0) {
213 delete this;
214 }
215 }
216
217
218 RBBIDataWrapper *RBBIDataWrapper::addReference() {
219 umtx_atomic_inc(&fRefCount);
220 return this;
221 }
222
223
224
225 //-----------------------------------------------------------------------------
226 //
227 // getRuleSourceString
228 //
229 //-----------------------------------------------------------------------------
230 const UnicodeString &RBBIDataWrapper::getRuleSourceString() const {
231 return fRuleString;
232 }
233
234
235 //-----------------------------------------------------------------------------
236 //
237 // print - debugging function to dump the runtime data tables.
238 //
239 //-----------------------------------------------------------------------------
240 #ifdef RBBI_DEBUG
241 void RBBIDataWrapper::printTable(const char *heading, const RBBIStateTable *table) {
242 uint32_t c;
243 uint32_t s;
244
245 RBBIDebugPrintf(" %s\n", heading);
246
247 RBBIDebugPrintf("State | Acc LA TagIx");
248 for (c=0; c<fHeader->fCatCount; c++) {RBBIDebugPrintf("%3d ", c);}
249 RBBIDebugPrintf("\n------|---------------"); for (c=0;c<fHeader->fCatCount; c++) {
250 RBBIDebugPrintf("----");
251 }
252 RBBIDebugPrintf("\n");
253
254 if (table == NULL) {
255 RBBIDebugPrintf(" N U L L T A B L E\n\n");
256 return;
257 }
258 for (s=0; s<table->fNumStates; s++) {
259 RBBIStateTableRow *row = (RBBIStateTableRow *)
260 (table->fTableData + (table->fRowLen * s));
261 RBBIDebugPrintf("%4d | %3d %3d %3d ", s, row->fAccepting, row->fLookAhead, row->fTagIdx);
262 for (c=0; c<fHeader->fCatCount; c++) {
263 RBBIDebugPrintf("%3d ", row->fNextState[c]);
264 }
265 RBBIDebugPrintf("\n");
266 }
267 RBBIDebugPrintf("\n");
268 }
269 #endif
270
271
272 #ifdef RBBI_DEBUG
273 void RBBIDataWrapper::printData() {
274 RBBIDebugPrintf("RBBI Data at %p\n", (void *)fHeader);
275 RBBIDebugPrintf(" Version = {%d %d %d %d}\n", fHeader->fFormatVersion[0], fHeader->fFormatVersion[1],
276 fHeader->fFormatVersion[2], fHeader->fFormatVersion[3]);
277 RBBIDebugPrintf(" total length of data = %d\n", fHeader->fLength);
278 RBBIDebugPrintf(" number of character categories = %d\n\n", fHeader->fCatCount);
279
280 printTable("Forward State Transition Table", fForwardTable);
281 printTable("Reverse State Transition Table", fReverseTable);
282 printTable("Safe Forward State Transition Table", fSafeFwdTable);
283 printTable("Safe Reverse State Transition Table", fSafeRevTable);
284
285 RBBIDebugPrintf("\nOrignal Rules source:\n");
286 for (int32_t c=0; fRuleSource[c] != 0; c++) {
287 RBBIDebugPrintf("%c", fRuleSource[c]);
288 }
289 RBBIDebugPrintf("\n\n");
290 }
291 #endif
292
293
294 U_NAMESPACE_END
295 U_NAMESPACE_USE
296
297 //-----------------------------------------------------------------------------
298 //
299 // ubrk_swap - byte swap and char encoding swap of RBBI data
300 //
301 //-----------------------------------------------------------------------------
302
303 U_CAPI int32_t U_EXPORT2
304 ubrk_swap(const UDataSwapper *ds, const void *inData, int32_t length, void *outData,
305 UErrorCode *status) {
306
307 if (status == NULL || U_FAILURE(*status)) {
308 return 0;
309 }
310 if(ds==NULL || inData==NULL || length<-1 || (length>0 && outData==NULL)) {
311 *status=U_ILLEGAL_ARGUMENT_ERROR;
312 return 0;
313 }
314
315 //
316 // Check that the data header is for for break data.
317 // (Header contents are defined in genbrk.cpp)
318 //
319 const UDataInfo *pInfo = (const UDataInfo *)((const char *)inData+4);
320 if(!( pInfo->dataFormat[0]==0x42 && /* dataFormat="Brk " */
321 pInfo->dataFormat[1]==0x72 &&
322 pInfo->dataFormat[2]==0x6b &&
323 pInfo->dataFormat[3]==0x20 &&
324 pInfo->formatVersion[0]==3 )) {
325 udata_printError(ds, "ubrk_swap(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized\n",
326 pInfo->dataFormat[0], pInfo->dataFormat[1],
327 pInfo->dataFormat[2], pInfo->dataFormat[3],
328 pInfo->formatVersion[0]);
329 *status=U_UNSUPPORTED_ERROR;
330 return 0;
331 }
332
333 //
334 // Swap the data header. (This is the generic ICU Data Header, not the RBBI Specific
335 // RBBIDataHeader). This swap also conveniently gets us
336 // the size of the ICU d.h., which lets us locate the start
337 // of the RBBI specific data.
338 //
339 int32_t headerSize=udata_swapDataHeader(ds, inData, length, outData, status);
340
341
342 //
343 // Get the RRBI Data Header, and check that it appears to be OK.
344 //
345 // Note: ICU 3.2 and earlier, RBBIDataHeader::fDataFormat was actually
346 // an int32_t with a value of 1. Starting with ICU 3.4,
347 // RBBI's fDataFormat matches the dataFormat field from the
348 // UDataInfo header, four int8_t bytes. The value is {3,1,0,0}
349 //
350 const uint8_t *inBytes =(const uint8_t *)inData+headerSize;
351 RBBIDataHeader *rbbiDH = (RBBIDataHeader *)inBytes;
352 if (ds->readUInt32(rbbiDH->fMagic) != 0xb1a0 ||
353 rbbiDH->fFormatVersion[0] != 3 ||
354 ds->readUInt32(rbbiDH->fLength) < sizeof(RBBIDataHeader))
355 {
356 udata_printError(ds, "ubrk_swap(): RBBI Data header is invalid.\n");
357 *status=U_UNSUPPORTED_ERROR;
358 return 0;
359 }
360
361 //
362 // Prefight operation? Just return the size
363 //
364 int32_t breakDataLength = ds->readUInt32(rbbiDH->fLength);
365 int32_t totalSize = headerSize + breakDataLength;
366 if (length < 0) {
367 return totalSize;
368 }
369
370 //
371 // Check that length passed in is consistent with length from RBBI data header.
372 //
373 if (length < totalSize) {
374 udata_printError(ds, "ubrk_swap(): too few bytes (%d after ICU Data header) for break data.\n",
375 breakDataLength);
376 *status=U_INDEX_OUTOFBOUNDS_ERROR;
377 return 0;
378 }
379
380
381 //
382 // Swap the Data. Do the data itself first, then the RBBI Data Header, because
383 // we need to reference the header to locate the data, and an
384 // inplace swap of the header leaves it unusable.
385 //
386 uint8_t *outBytes = (uint8_t *)outData + headerSize;
387 RBBIDataHeader *outputDH = (RBBIDataHeader *)outBytes;
388
389 int32_t tableStartOffset;
390 int32_t tableLength;
391
392 //
393 // If not swapping in place, zero out the output buffer before starting.
394 // Individual tables and other data items within are aligned to 8 byte boundaries
395 // when originally created. Any unused space between items needs to be zero.
396 //
397 if (inBytes != outBytes) {
398 uprv_memset(outBytes, 0, breakDataLength);
399 }
400
401 //
402 // Each state table begins with several 32 bit fields. Calculate the size
403 // in bytes of these.
404 //
405 int32_t topSize = offsetof(RBBIStateTable, fTableData);
406
407 // Forward state table.
408 tableStartOffset = ds->readUInt32(rbbiDH->fFTable);
409 tableLength = ds->readUInt32(rbbiDH->fFTableLen);
410
411 if (tableLength > 0) {
412 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
413 outBytes+tableStartOffset, status);
414 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
415 outBytes+tableStartOffset+topSize, status);
416 }
417
418 // Reverse state table. Same layout as forward table, above.
419 tableStartOffset = ds->readUInt32(rbbiDH->fRTable);
420 tableLength = ds->readUInt32(rbbiDH->fRTableLen);
421
422 if (tableLength > 0) {
423 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
424 outBytes+tableStartOffset, status);
425 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
426 outBytes+tableStartOffset+topSize, status);
427 }
428
429 // Safe Forward state table. Same layout as forward table, above.
430 tableStartOffset = ds->readUInt32(rbbiDH->fSFTable);
431 tableLength = ds->readUInt32(rbbiDH->fSFTableLen);
432
433 if (tableLength > 0) {
434 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
435 outBytes+tableStartOffset, status);
436 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
437 outBytes+tableStartOffset+topSize, status);
438 }
439
440 // Safe Reverse state table. Same layout as forward table, above.
441 tableStartOffset = ds->readUInt32(rbbiDH->fSRTable);
442 tableLength = ds->readUInt32(rbbiDH->fSRTableLen);
443
444 if (tableLength > 0) {
445 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
446 outBytes+tableStartOffset, status);
447 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
448 outBytes+tableStartOffset+topSize, status);
449 }
450
451 // Trie table for character categories
452 utrie_swap(ds, inBytes+ds->readUInt32(rbbiDH->fTrie), ds->readUInt32(rbbiDH->fTrieLen),
453 outBytes+ds->readUInt32(rbbiDH->fTrie), status);
454
455 // Source Rules Text. It's UChar data
456 ds->swapArray16(ds, inBytes+ds->readUInt32(rbbiDH->fRuleSource), ds->readUInt32(rbbiDH->fRuleSourceLen),
457 outBytes+ds->readUInt32(rbbiDH->fRuleSource), status);
458
459 // Table of rule status values. It's all int_32 values
460 ds->swapArray32(ds, inBytes+ds->readUInt32(rbbiDH->fStatusTable), ds->readUInt32(rbbiDH->fStatusTableLen),
461 outBytes+ds->readUInt32(rbbiDH->fStatusTable), status);
462
463 // And, last, the header.
464 // It is all int32_t values except for fFormataVersion, which is an array of four bytes.
465 // Swap the whole thing as int32_t, then re-swap the one field.
466 //
467 ds->swapArray32(ds, inBytes, sizeof(RBBIDataHeader), outBytes, status);
468 ds->swapArray32(ds, outputDH->fFormatVersion, 4, outputDH->fFormatVersion, status);
469
470 return totalSize;
471 }
472
473
474 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */