1 // © 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
3 /*
4 ***************************************************************************
5 * Copyright (C) 1999-2014 International Business Machines Corporation *
6 * and others. All rights reserved. *
7 ***************************************************************************
8 */
9
10 #include "unicode/utypes.h"
11
12 #if !UCONFIG_NO_BREAK_ITERATION
13
14 #include "unicode/utypes.h"
15 #include "rbbidata.h"
16 #include "rbbirb.h"
17 #include "utrie2.h"
18 #include "udatamem.h"
19 #include "cmemory.h"
20 #include "cstring.h"
21 #include "umutex.h"
22
23 #include "uassert.h"
24
25
26 U_NAMESPACE_BEGIN
27
28 //-----------------------------------------------------------------------------
29 //
30 // Constructors.
31 //
32 //-----------------------------------------------------------------------------
RBBIDataWrapper(const RBBIDataHeader * data,UErrorCode & status)33 RBBIDataWrapper::RBBIDataWrapper(const RBBIDataHeader *data, UErrorCode &status) {
34 init0();
35 init(data, status);
36 }
37
RBBIDataWrapper(const RBBIDataHeader * data,enum EDontAdopt,UErrorCode & status)38 RBBIDataWrapper::RBBIDataWrapper(const RBBIDataHeader *data, enum EDontAdopt, UErrorCode &status) {
39 init0();
40 init(data, status);
41 fDontFreeData = TRUE;
42 }
43
RBBIDataWrapper(UDataMemory * udm,UErrorCode & status)44 RBBIDataWrapper::RBBIDataWrapper(UDataMemory* udm, UErrorCode &status) {
45 init0();
46 if (U_FAILURE(status)) {
47 return;
48 }
49 const DataHeader *dh = udm->pHeader;
50 int32_t headerSize = dh->dataHeader.headerSize;
51 if ( !(headerSize >= 20 &&
52 dh->info.isBigEndian == U_IS_BIG_ENDIAN &&
53 dh->info.charsetFamily == U_CHARSET_FAMILY &&
54 dh->info.dataFormat[0] == 0x42 && // dataFormat="Brk "
55 dh->info.dataFormat[1] == 0x72 &&
56 dh->info.dataFormat[2] == 0x6b &&
57 dh->info.dataFormat[3] == 0x20 &&
58 isDataVersionAcceptable(dh->info.formatVersion))
59 ) {
60 status = U_INVALID_FORMAT_ERROR;
61 return;
62 }
63 const char *dataAsBytes = reinterpret_cast<const char *>(dh);
64 const RBBIDataHeader *rbbidh = reinterpret_cast<const RBBIDataHeader *>(dataAsBytes + headerSize);
65 init(rbbidh, status);
66 fUDataMem = udm;
67 }
68
isDataVersionAcceptable(const UVersionInfo version)69 UBool RBBIDataWrapper::isDataVersionAcceptable(const UVersionInfo version) {
70 return RBBI_DATA_FORMAT_VERSION[0] == version[0];
71 }
72
73
74 //-----------------------------------------------------------------------------
75 //
76 // init(). Does most of the work of construction, shared between the
77 // constructors.
78 //
79 //-----------------------------------------------------------------------------
init0()80 void RBBIDataWrapper::init0() {
81 fHeader = NULL;
82 fForwardTable = NULL;
83 fReverseTable = NULL;
84 fSafeFwdTable = NULL;
85 fSafeRevTable = NULL;
86 fRuleSource = NULL;
87 fRuleStatusTable = NULL;
88 fTrie = NULL;
89 fUDataMem = NULL;
90 fRefCount = 0;
91 fDontFreeData = TRUE;
92 }
93
init(const RBBIDataHeader * data,UErrorCode & status)94 void RBBIDataWrapper::init(const RBBIDataHeader *data, UErrorCode &status) {
95 if (U_FAILURE(status)) {
96 return;
97 }
98 fHeader = data;
99 if (fHeader->fMagic != 0xb1a0 || !isDataVersionAcceptable(fHeader->fFormatVersion)) {
100 status = U_INVALID_FORMAT_ERROR;
101 return;
102 }
103 // Note: in ICU version 3.2 and earlier, there was a formatVersion 1
104 // that is no longer supported. At that time fFormatVersion was
105 // an int32_t field, rather than an array of 4 bytes.
106
107 fDontFreeData = FALSE;
108 if (data->fFTableLen != 0) {
109 fForwardTable = (RBBIStateTable *)((char *)data + fHeader->fFTable);
110 }
111 if (data->fRTableLen != 0) {
112 fReverseTable = (RBBIStateTable *)((char *)data + fHeader->fRTable);
113 }
114 if (data->fSFTableLen != 0) {
115 fSafeFwdTable = (RBBIStateTable *)((char *)data + fHeader->fSFTable);
116 }
117 if (data->fSRTableLen != 0) {
118 fSafeRevTable = (RBBIStateTable *)((char *)data + fHeader->fSRTable);
119 }
120
121 // Rule Compatibility Hacks
122 // If a rule set includes reverse rules but does not explicitly include safe reverse rules,
123 // the reverse rules are to be treated as safe reverse rules.
124
125 if (fSafeRevTable == NULL && fReverseTable != NULL) {
126 fSafeRevTable = fReverseTable;
127 fReverseTable = NULL;
128 }
129
130 fTrie = utrie2_openFromSerialized(UTRIE2_16_VALUE_BITS,
131 (uint8_t *)data + fHeader->fTrie,
132 fHeader->fTrieLen,
133 NULL, // *actual length
134 &status);
135 if (U_FAILURE(status)) {
136 return;
137 }
138
139 fRuleSource = (UChar *)((char *)data + fHeader->fRuleSource);
140 fRuleString.setTo(TRUE, fRuleSource, -1);
141 U_ASSERT(data->fRuleSourceLen > 0);
142
143 fRuleStatusTable = (int32_t *)((char *)data + fHeader->fStatusTable);
144 fStatusMaxIdx = data->fStatusTableLen / sizeof(int32_t);
145
146 fRefCount = 1;
147
148 #ifdef RBBI_DEBUG
149 char *debugEnv = getenv("U_RBBIDEBUG");
150 if (debugEnv && uprv_strstr(debugEnv, "data")) {this->printData();}
151 #endif
152 }
153
154
155 //-----------------------------------------------------------------------------
156 //
157 // Destructor. Don't call this - use removeReference() instead.
158 //
159 //-----------------------------------------------------------------------------
~RBBIDataWrapper()160 RBBIDataWrapper::~RBBIDataWrapper() {
161 U_ASSERT(fRefCount == 0);
162 utrie2_close(fTrie);
163 fTrie = NULL;
164 if (fUDataMem) {
165 udata_close(fUDataMem);
166 } else if (!fDontFreeData) {
167 uprv_free((void *)fHeader);
168 }
169 }
170
171
172
173 //-----------------------------------------------------------------------------
174 //
175 // Operator == Consider two RBBIDataWrappers to be equal if they
176 // refer to the same underlying data. Although
177 // the data wrappers are normally shared between
178 // iterator instances, it's possible to independently
179 // open the same data twice, and get two instances, which
180 // should still be ==.
181 //
182 //-----------------------------------------------------------------------------
operator ==(const RBBIDataWrapper & other) const183 UBool RBBIDataWrapper::operator ==(const RBBIDataWrapper &other) const {
184 if (fHeader == other.fHeader) {
185 return TRUE;
186 }
187 if (fHeader->fLength != other.fHeader->fLength) {
188 return FALSE;
189 }
190 if (uprv_memcmp(fHeader, other.fHeader, fHeader->fLength) == 0) {
191 return TRUE;
192 }
193 return FALSE;
194 }
195
hashCode()196 int32_t RBBIDataWrapper::hashCode() {
197 return fHeader->fFTableLen;
198 }
199
200
201
202 //-----------------------------------------------------------------------------
203 //
204 // Reference Counting. A single RBBIDataWrapper object is shared among
205 // however many RulesBasedBreakIterator instances are
206 // referencing the same data.
207 //
208 //-----------------------------------------------------------------------------
removeReference()209 void RBBIDataWrapper::removeReference() {
210 if (umtx_atomic_dec(&fRefCount) == 0) {
211 delete this;
212 }
213 }
214
215
addReference()216 RBBIDataWrapper *RBBIDataWrapper::addReference() {
217 umtx_atomic_inc(&fRefCount);
218 return this;
219 }
220
221
222
223 //-----------------------------------------------------------------------------
224 //
225 // getRuleSourceString
226 //
227 //-----------------------------------------------------------------------------
getRuleSourceString() const228 const UnicodeString &RBBIDataWrapper::getRuleSourceString() const {
229 return fRuleString;
230 }
231
232
233 //-----------------------------------------------------------------------------
234 //
235 // print - debugging function to dump the runtime data tables.
236 //
237 //-----------------------------------------------------------------------------
238 #ifdef RBBI_DEBUG
printTable(const char * heading,const RBBIStateTable * table)239 void RBBIDataWrapper::printTable(const char *heading, const RBBIStateTable *table) {
240 uint32_t c;
241 uint32_t s;
242
243 RBBIDebugPrintf(" %s\n", heading);
244
245 RBBIDebugPrintf("State | Acc LA TagIx");
246 for (c=0; c<fHeader->fCatCount; c++) {RBBIDebugPrintf("%3d ", c);}
247 RBBIDebugPrintf("\n------|---------------"); for (c=0;c<fHeader->fCatCount; c++) {
248 RBBIDebugPrintf("----");
249 }
250 RBBIDebugPrintf("\n");
251
252 if (table == NULL) {
253 RBBIDebugPrintf(" N U L L T A B L E\n\n");
254 return;
255 }
256 for (s=0; s<table->fNumStates; s++) {
257 RBBIStateTableRow *row = (RBBIStateTableRow *)
258 (table->fTableData + (table->fRowLen * s));
259 RBBIDebugPrintf("%4d | %3d %3d %3d ", s, row->fAccepting, row->fLookAhead, row->fTagIdx);
260 for (c=0; c<fHeader->fCatCount; c++) {
261 RBBIDebugPrintf("%3d ", row->fNextState[c]);
262 }
263 RBBIDebugPrintf("\n");
264 }
265 RBBIDebugPrintf("\n");
266 }
267 #endif
268
269
270 #ifdef RBBI_DEBUG
printData()271 void RBBIDataWrapper::printData() {
272 RBBIDebugPrintf("RBBI Data at %p\n", (void *)fHeader);
273 RBBIDebugPrintf(" Version = {%d %d %d %d}\n", fHeader->fFormatVersion[0], fHeader->fFormatVersion[1],
274 fHeader->fFormatVersion[2], fHeader->fFormatVersion[3]);
275 RBBIDebugPrintf(" total length of data = %d\n", fHeader->fLength);
276 RBBIDebugPrintf(" number of character categories = %d\n\n", fHeader->fCatCount);
277
278 printTable("Forward State Transition Table", fForwardTable);
279 printTable("Reverse State Transition Table", fReverseTable);
280 printTable("Safe Forward State Transition Table", fSafeFwdTable);
281 printTable("Safe Reverse State Transition Table", fSafeRevTable);
282
283 RBBIDebugPrintf("\nOrignal Rules source:\n");
284 for (int32_t c=0; fRuleSource[c] != 0; c++) {
285 RBBIDebugPrintf("%c", fRuleSource[c]);
286 }
287 RBBIDebugPrintf("\n\n");
288 }
289 #endif
290
291
292 U_NAMESPACE_END
293 U_NAMESPACE_USE
294
295 //-----------------------------------------------------------------------------
296 //
297 // ubrk_swap - byte swap and char encoding swap of RBBI data
298 //
299 //-----------------------------------------------------------------------------
300
301 U_CAPI int32_t U_EXPORT2
ubrk_swap(const UDataSwapper * ds,const void * inData,int32_t length,void * outData,UErrorCode * status)302 ubrk_swap(const UDataSwapper *ds, const void *inData, int32_t length, void *outData,
303 UErrorCode *status) {
304
305 if (status == NULL || U_FAILURE(*status)) {
306 return 0;
307 }
308 if(ds==NULL || inData==NULL || length<-1 || (length>0 && outData==NULL)) {
309 *status=U_ILLEGAL_ARGUMENT_ERROR;
310 return 0;
311 }
312
313 //
314 // Check that the data header is for for break data.
315 // (Header contents are defined in genbrk.cpp)
316 //
317 const UDataInfo *pInfo = (const UDataInfo *)((const char *)inData+4);
318 if(!( pInfo->dataFormat[0]==0x42 && /* dataFormat="Brk " */
319 pInfo->dataFormat[1]==0x72 &&
320 pInfo->dataFormat[2]==0x6b &&
321 pInfo->dataFormat[3]==0x20 &&
322 RBBIDataWrapper::isDataVersionAcceptable(pInfo->formatVersion) )) {
323 udata_printError(ds, "ubrk_swap(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized\n",
324 pInfo->dataFormat[0], pInfo->dataFormat[1],
325 pInfo->dataFormat[2], pInfo->dataFormat[3],
326 pInfo->formatVersion[0]);
327 *status=U_UNSUPPORTED_ERROR;
328 return 0;
329 }
330
331 //
332 // Swap the data header. (This is the generic ICU Data Header, not the RBBI Specific
333 // RBBIDataHeader). This swap also conveniently gets us
334 // the size of the ICU d.h., which lets us locate the start
335 // of the RBBI specific data.
336 //
337 int32_t headerSize=udata_swapDataHeader(ds, inData, length, outData, status);
338
339
340 //
341 // Get the RRBI Data Header, and check that it appears to be OK.
342 //
343 const uint8_t *inBytes =(const uint8_t *)inData+headerSize;
344 RBBIDataHeader *rbbiDH = (RBBIDataHeader *)inBytes;
345 if (ds->readUInt32(rbbiDH->fMagic) != 0xb1a0 ||
346 !RBBIDataWrapper::isDataVersionAcceptable(rbbiDH->fFormatVersion) ||
347 ds->readUInt32(rbbiDH->fLength) < sizeof(RBBIDataHeader)) {
348 udata_printError(ds, "ubrk_swap(): RBBI Data header is invalid.\n");
349 *status=U_UNSUPPORTED_ERROR;
350 return 0;
351 }
352
353 //
354 // Prefight operation? Just return the size
355 //
356 int32_t breakDataLength = ds->readUInt32(rbbiDH->fLength);
357 int32_t totalSize = headerSize + breakDataLength;
358 if (length < 0) {
359 return totalSize;
360 }
361
362 //
363 // Check that length passed in is consistent with length from RBBI data header.
364 //
365 if (length < totalSize) {
366 udata_printError(ds, "ubrk_swap(): too few bytes (%d after ICU Data header) for break data.\n",
367 breakDataLength);
368 *status=U_INDEX_OUTOFBOUNDS_ERROR;
369 return 0;
370 }
371
372
373 //
374 // Swap the Data. Do the data itself first, then the RBBI Data Header, because
375 // we need to reference the header to locate the data, and an
376 // inplace swap of the header leaves it unusable.
377 //
378 uint8_t *outBytes = (uint8_t *)outData + headerSize;
379 RBBIDataHeader *outputDH = (RBBIDataHeader *)outBytes;
380
381 int32_t tableStartOffset;
382 int32_t tableLength;
383
384 //
385 // If not swapping in place, zero out the output buffer before starting.
386 // Individual tables and other data items within are aligned to 8 byte boundaries
387 // when originally created. Any unused space between items needs to be zero.
388 //
389 if (inBytes != outBytes) {
390 uprv_memset(outBytes, 0, breakDataLength);
391 }
392
393 //
394 // Each state table begins with several 32 bit fields. Calculate the size
395 // in bytes of these.
396 //
397 int32_t topSize = offsetof(RBBIStateTable, fTableData);
398
399 // Forward state table.
400 tableStartOffset = ds->readUInt32(rbbiDH->fFTable);
401 tableLength = ds->readUInt32(rbbiDH->fFTableLen);
402
403 if (tableLength > 0) {
404 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
405 outBytes+tableStartOffset, status);
406 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
407 outBytes+tableStartOffset+topSize, status);
408 }
409
410 // Reverse state table. Same layout as forward table, above.
411 tableStartOffset = ds->readUInt32(rbbiDH->fRTable);
412 tableLength = ds->readUInt32(rbbiDH->fRTableLen);
413
414 if (tableLength > 0) {
415 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
416 outBytes+tableStartOffset, status);
417 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
418 outBytes+tableStartOffset+topSize, status);
419 }
420
421 // Safe Forward state table. Same layout as forward table, above.
422 tableStartOffset = ds->readUInt32(rbbiDH->fSFTable);
423 tableLength = ds->readUInt32(rbbiDH->fSFTableLen);
424
425 if (tableLength > 0) {
426 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
427 outBytes+tableStartOffset, status);
428 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
429 outBytes+tableStartOffset+topSize, status);
430 }
431
432 // Safe Reverse state table. Same layout as forward table, above.
433 tableStartOffset = ds->readUInt32(rbbiDH->fSRTable);
434 tableLength = ds->readUInt32(rbbiDH->fSRTableLen);
435
436 if (tableLength > 0) {
437 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
438 outBytes+tableStartOffset, status);
439 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
440 outBytes+tableStartOffset+topSize, status);
441 }
442
443 // Trie table for character categories
444 utrie2_swap(ds, inBytes+ds->readUInt32(rbbiDH->fTrie), ds->readUInt32(rbbiDH->fTrieLen),
445 outBytes+ds->readUInt32(rbbiDH->fTrie), status);
446
447 // Source Rules Text. It's UChar data
448 ds->swapArray16(ds, inBytes+ds->readUInt32(rbbiDH->fRuleSource), ds->readUInt32(rbbiDH->fRuleSourceLen),
449 outBytes+ds->readUInt32(rbbiDH->fRuleSource), status);
450
451 // Table of rule status values. It's all int_32 values
452 ds->swapArray32(ds, inBytes+ds->readUInt32(rbbiDH->fStatusTable), ds->readUInt32(rbbiDH->fStatusTableLen),
453 outBytes+ds->readUInt32(rbbiDH->fStatusTable), status);
454
455 // And, last, the header.
456 // It is all int32_t values except for fFormataVersion, which is an array of four bytes.
457 // Swap the whole thing as int32_t, then re-swap the one field.
458 //
459 ds->swapArray32(ds, inBytes, sizeof(RBBIDataHeader), outBytes, status);
460 ds->swapArray32(ds, outputDH->fFormatVersion, 4, outputDH->fFormatVersion, status);
461
462 return totalSize;
463 }
464
465
466 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
467