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 fRuleSource = NULL;
85 fRuleStatusTable = NULL;
86 fTrie = NULL;
87 fUDataMem = NULL;
88 fRefCount = 0;
89 fDontFreeData = TRUE;
90 }
91
init(const RBBIDataHeader * data,UErrorCode & status)92 void RBBIDataWrapper::init(const RBBIDataHeader *data, UErrorCode &status) {
93 if (U_FAILURE(status)) {
94 return;
95 }
96 fHeader = data;
97 if (fHeader->fMagic != 0xb1a0 || !isDataVersionAcceptable(fHeader->fFormatVersion)) {
98 status = U_INVALID_FORMAT_ERROR;
99 return;
100 }
101 // Note: in ICU version 3.2 and earlier, there was a formatVersion 1
102 // that is no longer supported. At that time fFormatVersion was
103 // an int32_t field, rather than an array of 4 bytes.
104
105 fDontFreeData = FALSE;
106 if (data->fFTableLen != 0) {
107 fForwardTable = (RBBIStateTable *)((char *)data + fHeader->fFTable);
108 }
109 if (data->fRTableLen != 0) {
110 fReverseTable = (RBBIStateTable *)((char *)data + fHeader->fRTable);
111 }
112
113 fTrie = utrie2_openFromSerialized(UTRIE2_16_VALUE_BITS,
114 (uint8_t *)data + fHeader->fTrie,
115 fHeader->fTrieLen,
116 NULL, // *actual length
117 &status);
118 if (U_FAILURE(status)) {
119 return;
120 }
121
122 fRuleSource = (UChar *)((char *)data + fHeader->fRuleSource);
123 fRuleString.setTo(TRUE, fRuleSource, -1);
124 U_ASSERT(data->fRuleSourceLen > 0);
125
126 fRuleStatusTable = (int32_t *)((char *)data + fHeader->fStatusTable);
127 fStatusMaxIdx = data->fStatusTableLen / sizeof(int32_t);
128
129 fRefCount = 1;
130
131 #ifdef RBBI_DEBUG
132 char *debugEnv = getenv("U_RBBIDEBUG");
133 if (debugEnv && uprv_strstr(debugEnv, "data")) {this->printData();}
134 #endif
135 }
136
137
138 //-----------------------------------------------------------------------------
139 //
140 // Destructor. Don't call this - use removeReference() instead.
141 //
142 //-----------------------------------------------------------------------------
~RBBIDataWrapper()143 RBBIDataWrapper::~RBBIDataWrapper() {
144 U_ASSERT(fRefCount == 0);
145 utrie2_close(fTrie);
146 fTrie = NULL;
147 if (fUDataMem) {
148 udata_close(fUDataMem);
149 } else if (!fDontFreeData) {
150 uprv_free((void *)fHeader);
151 }
152 }
153
154
155
156 //-----------------------------------------------------------------------------
157 //
158 // Operator == Consider two RBBIDataWrappers to be equal if they
159 // refer to the same underlying data. Although
160 // the data wrappers are normally shared between
161 // iterator instances, it's possible to independently
162 // open the same data twice, and get two instances, which
163 // should still be ==.
164 //
165 //-----------------------------------------------------------------------------
operator ==(const RBBIDataWrapper & other) const166 UBool RBBIDataWrapper::operator ==(const RBBIDataWrapper &other) const {
167 if (fHeader == other.fHeader) {
168 return TRUE;
169 }
170 if (fHeader->fLength != other.fHeader->fLength) {
171 return FALSE;
172 }
173 if (uprv_memcmp(fHeader, other.fHeader, fHeader->fLength) == 0) {
174 return TRUE;
175 }
176 return FALSE;
177 }
178
hashCode()179 int32_t RBBIDataWrapper::hashCode() {
180 return fHeader->fFTableLen;
181 }
182
183
184
185 //-----------------------------------------------------------------------------
186 //
187 // Reference Counting. A single RBBIDataWrapper object is shared among
188 // however many RulesBasedBreakIterator instances are
189 // referencing the same data.
190 //
191 //-----------------------------------------------------------------------------
removeReference()192 void RBBIDataWrapper::removeReference() {
193 if (umtx_atomic_dec(&fRefCount) == 0) {
194 delete this;
195 }
196 }
197
198
addReference()199 RBBIDataWrapper *RBBIDataWrapper::addReference() {
200 umtx_atomic_inc(&fRefCount);
201 return this;
202 }
203
204
205
206 //-----------------------------------------------------------------------------
207 //
208 // getRuleSourceString
209 //
210 //-----------------------------------------------------------------------------
getRuleSourceString() const211 const UnicodeString &RBBIDataWrapper::getRuleSourceString() const {
212 return fRuleString;
213 }
214
215
216 //-----------------------------------------------------------------------------
217 //
218 // print - debugging function to dump the runtime data tables.
219 //
220 //-----------------------------------------------------------------------------
221 #ifdef RBBI_DEBUG
printTable(const char * heading,const RBBIStateTable * table)222 void RBBIDataWrapper::printTable(const char *heading, const RBBIStateTable *table) {
223 uint32_t c;
224 uint32_t s;
225
226 RBBIDebugPrintf(" %s\n", heading);
227
228 RBBIDebugPrintf("State | Acc LA TagIx");
229 for (c=0; c<fHeader->fCatCount; c++) {RBBIDebugPrintf("%3d ", c);}
230 RBBIDebugPrintf("\n------|---------------"); for (c=0;c<fHeader->fCatCount; c++) {
231 RBBIDebugPrintf("----");
232 }
233 RBBIDebugPrintf("\n");
234
235 if (table == NULL) {
236 RBBIDebugPrintf(" N U L L T A B L E\n\n");
237 return;
238 }
239 for (s=0; s<table->fNumStates; s++) {
240 RBBIStateTableRow *row = (RBBIStateTableRow *)
241 (table->fTableData + (table->fRowLen * s));
242 RBBIDebugPrintf("%4d | %3d %3d %3d ", s, row->fAccepting, row->fLookAhead, row->fTagIdx);
243 for (c=0; c<fHeader->fCatCount; c++) {
244 RBBIDebugPrintf("%3d ", row->fNextState[c]);
245 }
246 RBBIDebugPrintf("\n");
247 }
248 RBBIDebugPrintf("\n");
249 }
250 #endif
251
252
printData()253 void RBBIDataWrapper::printData() {
254 #ifdef RBBI_DEBUG
255 RBBIDebugPrintf("RBBI Data at %p\n", (void *)fHeader);
256 RBBIDebugPrintf(" Version = {%d %d %d %d}\n", fHeader->fFormatVersion[0], fHeader->fFormatVersion[1],
257 fHeader->fFormatVersion[2], fHeader->fFormatVersion[3]);
258 RBBIDebugPrintf(" total length of data = %d\n", fHeader->fLength);
259 RBBIDebugPrintf(" number of character categories = %d\n\n", fHeader->fCatCount);
260
261 printTable("Forward State Transition Table", fForwardTable);
262 printTable("Reverse State Transition Table", fReverseTable);
263
264 RBBIDebugPrintf("\nOrignal Rules source:\n");
265 for (int32_t c=0; fRuleSource[c] != 0; c++) {
266 RBBIDebugPrintf("%c", fRuleSource[c]);
267 }
268 RBBIDebugPrintf("\n\n");
269 #endif
270 }
271
272
273 U_NAMESPACE_END
274 U_NAMESPACE_USE
275
276 //-----------------------------------------------------------------------------
277 //
278 // ubrk_swap - byte swap and char encoding swap of RBBI data
279 //
280 //-----------------------------------------------------------------------------
281
282 U_CAPI int32_t U_EXPORT2
ubrk_swap(const UDataSwapper * ds,const void * inData,int32_t length,void * outData,UErrorCode * status)283 ubrk_swap(const UDataSwapper *ds, const void *inData, int32_t length, void *outData,
284 UErrorCode *status) {
285
286 if (status == NULL || U_FAILURE(*status)) {
287 return 0;
288 }
289 if(ds==NULL || inData==NULL || length<-1 || (length>0 && outData==NULL)) {
290 *status=U_ILLEGAL_ARGUMENT_ERROR;
291 return 0;
292 }
293
294 //
295 // Check that the data header is for for break data.
296 // (Header contents are defined in genbrk.cpp)
297 //
298 const UDataInfo *pInfo = (const UDataInfo *)((const char *)inData+4);
299 if(!( pInfo->dataFormat[0]==0x42 && /* dataFormat="Brk " */
300 pInfo->dataFormat[1]==0x72 &&
301 pInfo->dataFormat[2]==0x6b &&
302 pInfo->dataFormat[3]==0x20 &&
303 RBBIDataWrapper::isDataVersionAcceptable(pInfo->formatVersion) )) {
304 udata_printError(ds, "ubrk_swap(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized\n",
305 pInfo->dataFormat[0], pInfo->dataFormat[1],
306 pInfo->dataFormat[2], pInfo->dataFormat[3],
307 pInfo->formatVersion[0]);
308 *status=U_UNSUPPORTED_ERROR;
309 return 0;
310 }
311
312 //
313 // Swap the data header. (This is the generic ICU Data Header, not the RBBI Specific
314 // RBBIDataHeader). This swap also conveniently gets us
315 // the size of the ICU d.h., which lets us locate the start
316 // of the RBBI specific data.
317 //
318 int32_t headerSize=udata_swapDataHeader(ds, inData, length, outData, status);
319
320
321 //
322 // Get the RRBI Data Header, and check that it appears to be OK.
323 //
324 const uint8_t *inBytes =(const uint8_t *)inData+headerSize;
325 RBBIDataHeader *rbbiDH = (RBBIDataHeader *)inBytes;
326 if (ds->readUInt32(rbbiDH->fMagic) != 0xb1a0 ||
327 !RBBIDataWrapper::isDataVersionAcceptable(rbbiDH->fFormatVersion) ||
328 ds->readUInt32(rbbiDH->fLength) < sizeof(RBBIDataHeader)) {
329 udata_printError(ds, "ubrk_swap(): RBBI Data header is invalid.\n");
330 *status=U_UNSUPPORTED_ERROR;
331 return 0;
332 }
333
334 //
335 // Prefight operation? Just return the size
336 //
337 int32_t breakDataLength = ds->readUInt32(rbbiDH->fLength);
338 int32_t totalSize = headerSize + breakDataLength;
339 if (length < 0) {
340 return totalSize;
341 }
342
343 //
344 // Check that length passed in is consistent with length from RBBI data header.
345 //
346 if (length < totalSize) {
347 udata_printError(ds, "ubrk_swap(): too few bytes (%d after ICU Data header) for break data.\n",
348 breakDataLength);
349 *status=U_INDEX_OUTOFBOUNDS_ERROR;
350 return 0;
351 }
352
353
354 //
355 // Swap the Data. Do the data itself first, then the RBBI Data Header, because
356 // we need to reference the header to locate the data, and an
357 // inplace swap of the header leaves it unusable.
358 //
359 uint8_t *outBytes = (uint8_t *)outData + headerSize;
360 RBBIDataHeader *outputDH = (RBBIDataHeader *)outBytes;
361
362 int32_t tableStartOffset;
363 int32_t tableLength;
364
365 //
366 // If not swapping in place, zero out the output buffer before starting.
367 // Individual tables and other data items within are aligned to 8 byte boundaries
368 // when originally created. Any unused space between items needs to be zero.
369 //
370 if (inBytes != outBytes) {
371 uprv_memset(outBytes, 0, breakDataLength);
372 }
373
374 //
375 // Each state table begins with several 32 bit fields. Calculate the size
376 // in bytes of these.
377 //
378 int32_t topSize = offsetof(RBBIStateTable, fTableData);
379
380 // Forward state table.
381 tableStartOffset = ds->readUInt32(rbbiDH->fFTable);
382 tableLength = ds->readUInt32(rbbiDH->fFTableLen);
383
384 if (tableLength > 0) {
385 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
386 outBytes+tableStartOffset, status);
387 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
388 outBytes+tableStartOffset+topSize, status);
389 }
390
391 // Reverse state table. Same layout as forward table, above.
392 tableStartOffset = ds->readUInt32(rbbiDH->fRTable);
393 tableLength = ds->readUInt32(rbbiDH->fRTableLen);
394
395 if (tableLength > 0) {
396 ds->swapArray32(ds, inBytes+tableStartOffset, topSize,
397 outBytes+tableStartOffset, status);
398 ds->swapArray16(ds, inBytes+tableStartOffset+topSize, tableLength-topSize,
399 outBytes+tableStartOffset+topSize, status);
400 }
401
402 // Trie table for character categories
403 utrie2_swap(ds, inBytes+ds->readUInt32(rbbiDH->fTrie), ds->readUInt32(rbbiDH->fTrieLen),
404 outBytes+ds->readUInt32(rbbiDH->fTrie), status);
405
406 // Source Rules Text. It's UChar data
407 ds->swapArray16(ds, inBytes+ds->readUInt32(rbbiDH->fRuleSource), ds->readUInt32(rbbiDH->fRuleSourceLen),
408 outBytes+ds->readUInt32(rbbiDH->fRuleSource), status);
409
410 // Table of rule status values. It's all int_32 values
411 ds->swapArray32(ds, inBytes+ds->readUInt32(rbbiDH->fStatusTable), ds->readUInt32(rbbiDH->fStatusTableLen),
412 outBytes+ds->readUInt32(rbbiDH->fStatusTable), status);
413
414 // And, last, the header.
415 // It is all int32_t values except for fFormataVersion, which is an array of four bytes.
416 // Swap the whole thing as int32_t, then re-swap the one field.
417 //
418 ds->swapArray32(ds, inBytes, sizeof(RBBIDataHeader), outBytes, status);
419 ds->swapArray32(ds, outputDH->fFormatVersion, 4, outputDH->fFormatVersion, status);
420
421 return totalSize;
422 }
423
424
425 #endif /* #if !UCONFIG_NO_BREAK_ITERATION */
426