]>
Commit | Line | Data |
---|---|---|
4388f060 A |
1 | /* |
2 | ******************************************************************************* | |
3 | * Copyright (C) 2010-2011, International Business Machines | |
4 | * Corporation and others. All Rights Reserved. | |
5 | ******************************************************************************* | |
6 | * file name: bytestriebuilder.cpp | |
7 | * encoding: US-ASCII | |
8 | * tab size: 8 (not used) | |
9 | * indentation:4 | |
10 | * | |
11 | * created on: 2010sep25 | |
12 | * created by: Markus W. Scherer | |
13 | */ | |
14 | ||
15 | #include "unicode/utypes.h" | |
16 | #include "unicode/bytestrie.h" | |
17 | #include "unicode/bytestriebuilder.h" | |
18 | #include "unicode/stringpiece.h" | |
19 | #include "charstr.h" | |
20 | #include "cmemory.h" | |
21 | #include "uhash.h" | |
22 | #include "uarrsort.h" | |
23 | #include "uassert.h" | |
24 | #include "ustr_imp.h" | |
25 | ||
26 | U_NAMESPACE_BEGIN | |
27 | ||
28 | /* | |
29 | * Note: This builder implementation stores (bytes, value) pairs with full copies | |
30 | * of the byte sequences, until the BytesTrie is built. | |
31 | * It might(!) take less memory if we collected the data in a temporary, dynamic trie. | |
32 | */ | |
33 | ||
34 | class BytesTrieElement : public UMemory { | |
35 | public: | |
36 | // Use compiler's default constructor, initializes nothing. | |
37 | ||
38 | void setTo(const StringPiece &s, int32_t val, CharString &strings, UErrorCode &errorCode); | |
39 | ||
40 | StringPiece getString(const CharString &strings) const { | |
41 | int32_t offset=stringOffset; | |
42 | int32_t length; | |
43 | if(offset>=0) { | |
44 | length=(uint8_t)strings[offset++]; | |
45 | } else { | |
46 | offset=~offset; | |
47 | length=((int32_t)(uint8_t)strings[offset]<<8)|(uint8_t)strings[offset+1]; | |
48 | offset+=2; | |
49 | } | |
50 | return StringPiece(strings.data()+offset, length); | |
51 | } | |
52 | int32_t getStringLength(const CharString &strings) const { | |
53 | int32_t offset=stringOffset; | |
54 | if(offset>=0) { | |
55 | return (uint8_t)strings[offset]; | |
56 | } else { | |
57 | offset=~offset; | |
58 | return ((int32_t)(uint8_t)strings[offset]<<8)|(uint8_t)strings[offset+1]; | |
59 | } | |
60 | } | |
61 | ||
62 | char charAt(int32_t index, const CharString &strings) const { return data(strings)[index]; } | |
63 | ||
64 | int32_t getValue() const { return value; } | |
65 | ||
66 | int32_t compareStringTo(const BytesTrieElement &o, const CharString &strings) const; | |
67 | ||
68 | private: | |
69 | const char *data(const CharString &strings) const { | |
70 | int32_t offset=stringOffset; | |
71 | if(offset>=0) { | |
72 | ++offset; | |
73 | } else { | |
74 | offset=~offset+2; | |
75 | } | |
76 | return strings.data()+offset; | |
77 | } | |
78 | ||
79 | // If the stringOffset is non-negative, then the first strings byte contains | |
80 | // the string length. | |
81 | // If the stringOffset is negative, then the first two strings bytes contain | |
82 | // the string length (big-endian), and the offset needs to be bit-inverted. | |
83 | // (Compared with a stringLength field here, this saves 3 bytes per string for most strings.) | |
84 | int32_t stringOffset; | |
85 | int32_t value; | |
86 | }; | |
87 | ||
88 | void | |
89 | BytesTrieElement::setTo(const StringPiece &s, int32_t val, | |
90 | CharString &strings, UErrorCode &errorCode) { | |
91 | if(U_FAILURE(errorCode)) { | |
92 | return; | |
93 | } | |
94 | int32_t length=s.length(); | |
95 | if(length>0xffff) { | |
96 | // Too long: We store the length in 1 or 2 bytes. | |
97 | errorCode=U_INDEX_OUTOFBOUNDS_ERROR; | |
98 | return; | |
99 | } | |
100 | int32_t offset=strings.length(); | |
101 | if(length>0xff) { | |
102 | offset=~offset; | |
103 | strings.append((char)(length>>8), errorCode); | |
104 | } | |
105 | strings.append((char)length, errorCode); | |
106 | stringOffset=offset; | |
107 | value=val; | |
108 | strings.append(s, errorCode); | |
109 | } | |
110 | ||
111 | int32_t | |
112 | BytesTrieElement::compareStringTo(const BytesTrieElement &other, const CharString &strings) const { | |
113 | // TODO: add StringPiece::compare(), see ticket #8187 | |
114 | StringPiece thisString=getString(strings); | |
115 | StringPiece otherString=other.getString(strings); | |
116 | int32_t lengthDiff=thisString.length()-otherString.length(); | |
117 | int32_t commonLength; | |
118 | if(lengthDiff<=0) { | |
119 | commonLength=thisString.length(); | |
120 | } else { | |
121 | commonLength=otherString.length(); | |
122 | } | |
123 | int32_t diff=uprv_memcmp(thisString.data(), otherString.data(), commonLength); | |
124 | return diff!=0 ? diff : lengthDiff; | |
125 | } | |
126 | ||
127 | BytesTrieBuilder::BytesTrieBuilder(UErrorCode &errorCode) | |
128 | : strings(NULL), elements(NULL), elementsCapacity(0), elementsLength(0), | |
129 | bytes(NULL), bytesCapacity(0), bytesLength(0) { | |
130 | if(U_FAILURE(errorCode)) { | |
131 | return; | |
132 | } | |
133 | strings=new CharString(); | |
134 | if(strings==NULL) { | |
135 | errorCode=U_MEMORY_ALLOCATION_ERROR; | |
136 | } | |
137 | } | |
138 | ||
139 | BytesTrieBuilder::~BytesTrieBuilder() { | |
140 | delete strings; | |
141 | delete[] elements; | |
142 | uprv_free(bytes); | |
143 | } | |
144 | ||
145 | BytesTrieBuilder & | |
146 | BytesTrieBuilder::add(const StringPiece &s, int32_t value, UErrorCode &errorCode) { | |
147 | if(U_FAILURE(errorCode)) { | |
148 | return *this; | |
149 | } | |
150 | if(bytesLength>0) { | |
151 | // Cannot add elements after building. | |
152 | errorCode=U_NO_WRITE_PERMISSION; | |
153 | return *this; | |
154 | } | |
155 | if(elementsLength==elementsCapacity) { | |
156 | int32_t newCapacity; | |
157 | if(elementsCapacity==0) { | |
158 | newCapacity=1024; | |
159 | } else { | |
160 | newCapacity=4*elementsCapacity; | |
161 | } | |
162 | BytesTrieElement *newElements=new BytesTrieElement[newCapacity]; | |
163 | if(newElements==NULL) { | |
164 | errorCode=U_MEMORY_ALLOCATION_ERROR; | |
165 | return *this; // error instead of dereferencing null | |
166 | } | |
167 | if(elementsLength>0) { | |
168 | uprv_memcpy(newElements, elements, elementsLength*sizeof(BytesTrieElement)); | |
169 | } | |
170 | delete[] elements; | |
171 | elements=newElements; | |
172 | elementsCapacity=newCapacity; | |
173 | } | |
174 | elements[elementsLength++].setTo(s, value, *strings, errorCode); | |
175 | return *this; | |
176 | } | |
177 | ||
178 | U_CDECL_BEGIN | |
179 | ||
180 | static int32_t U_CALLCONV | |
181 | compareElementStrings(const void *context, const void *left, const void *right) { | |
182 | const CharString *strings=reinterpret_cast<const CharString *>(context); | |
183 | const BytesTrieElement *leftElement=reinterpret_cast<const BytesTrieElement *>(left); | |
184 | const BytesTrieElement *rightElement=reinterpret_cast<const BytesTrieElement *>(right); | |
185 | return leftElement->compareStringTo(*rightElement, *strings); | |
186 | } | |
187 | ||
188 | U_CDECL_END | |
189 | ||
190 | BytesTrie * | |
191 | BytesTrieBuilder::build(UStringTrieBuildOption buildOption, UErrorCode &errorCode) { | |
192 | buildBytes(buildOption, errorCode); | |
193 | BytesTrie *newTrie=NULL; | |
194 | if(U_SUCCESS(errorCode)) { | |
195 | newTrie=new BytesTrie(bytes, bytes+(bytesCapacity-bytesLength)); | |
196 | if(newTrie==NULL) { | |
197 | errorCode=U_MEMORY_ALLOCATION_ERROR; | |
198 | } else { | |
199 | bytes=NULL; // The new trie now owns the array. | |
200 | bytesCapacity=0; | |
201 | } | |
202 | } | |
203 | return newTrie; | |
204 | } | |
205 | ||
206 | StringPiece | |
207 | BytesTrieBuilder::buildStringPiece(UStringTrieBuildOption buildOption, UErrorCode &errorCode) { | |
208 | buildBytes(buildOption, errorCode); | |
209 | StringPiece result; | |
210 | if(U_SUCCESS(errorCode)) { | |
211 | result.set(bytes+(bytesCapacity-bytesLength), bytesLength); | |
212 | } | |
213 | return result; | |
214 | } | |
215 | ||
216 | void | |
217 | BytesTrieBuilder::buildBytes(UStringTrieBuildOption buildOption, UErrorCode &errorCode) { | |
218 | if(U_FAILURE(errorCode)) { | |
219 | return; | |
220 | } | |
221 | if(bytes!=NULL && bytesLength>0) { | |
222 | // Already built. | |
223 | return; | |
224 | } | |
225 | if(bytesLength==0) { | |
226 | if(elementsLength==0) { | |
227 | errorCode=U_INDEX_OUTOFBOUNDS_ERROR; | |
228 | return; | |
229 | } | |
230 | uprv_sortArray(elements, elementsLength, (int32_t)sizeof(BytesTrieElement), | |
231 | compareElementStrings, strings, | |
232 | FALSE, // need not be a stable sort | |
233 | &errorCode); | |
234 | if(U_FAILURE(errorCode)) { | |
235 | return; | |
236 | } | |
237 | // Duplicate strings are not allowed. | |
238 | StringPiece prev=elements[0].getString(*strings); | |
239 | for(int32_t i=1; i<elementsLength; ++i) { | |
240 | StringPiece current=elements[i].getString(*strings); | |
241 | if(prev==current) { | |
242 | errorCode=U_ILLEGAL_ARGUMENT_ERROR; | |
243 | return; | |
244 | } | |
245 | prev=current; | |
246 | } | |
247 | } | |
248 | // Create and byte-serialize the trie for the elements. | |
249 | bytesLength=0; | |
250 | int32_t capacity=strings->length(); | |
251 | if(capacity<1024) { | |
252 | capacity=1024; | |
253 | } | |
254 | if(bytesCapacity<capacity) { | |
255 | uprv_free(bytes); | |
256 | bytes=reinterpret_cast<char *>(uprv_malloc(capacity)); | |
257 | if(bytes==NULL) { | |
258 | errorCode=U_MEMORY_ALLOCATION_ERROR; | |
259 | bytesCapacity=0; | |
260 | return; | |
261 | } | |
262 | bytesCapacity=capacity; | |
263 | } | |
264 | StringTrieBuilder::build(buildOption, elementsLength, errorCode); | |
265 | if(bytes==NULL) { | |
266 | errorCode=U_MEMORY_ALLOCATION_ERROR; | |
267 | } | |
268 | } | |
269 | ||
270 | BytesTrieBuilder & | |
271 | BytesTrieBuilder::clear() { | |
272 | strings->clear(); | |
273 | elementsLength=0; | |
274 | bytesLength=0; | |
275 | return *this; | |
276 | } | |
277 | ||
278 | int32_t | |
279 | BytesTrieBuilder::getElementStringLength(int32_t i) const { | |
280 | return elements[i].getStringLength(*strings); | |
281 | } | |
282 | ||
283 | UChar | |
284 | BytesTrieBuilder::getElementUnit(int32_t i, int32_t byteIndex) const { | |
285 | return (uint8_t)elements[i].charAt(byteIndex, *strings); | |
286 | } | |
287 | ||
288 | int32_t | |
289 | BytesTrieBuilder::getElementValue(int32_t i) const { | |
290 | return elements[i].getValue(); | |
291 | } | |
292 | ||
293 | int32_t | |
294 | BytesTrieBuilder::getLimitOfLinearMatch(int32_t first, int32_t last, int32_t byteIndex) const { | |
295 | const BytesTrieElement &firstElement=elements[first]; | |
296 | const BytesTrieElement &lastElement=elements[last]; | |
297 | int32_t minStringLength=firstElement.getStringLength(*strings); | |
298 | while(++byteIndex<minStringLength && | |
299 | firstElement.charAt(byteIndex, *strings)== | |
300 | lastElement.charAt(byteIndex, *strings)) {} | |
301 | return byteIndex; | |
302 | } | |
303 | ||
304 | int32_t | |
305 | BytesTrieBuilder::countElementUnits(int32_t start, int32_t limit, int32_t byteIndex) const { | |
306 | int32_t length=0; // Number of different bytes at byteIndex. | |
307 | int32_t i=start; | |
308 | do { | |
309 | char byte=elements[i++].charAt(byteIndex, *strings); | |
310 | while(i<limit && byte==elements[i].charAt(byteIndex, *strings)) { | |
311 | ++i; | |
312 | } | |
313 | ++length; | |
314 | } while(i<limit); | |
315 | return length; | |
316 | } | |
317 | ||
318 | int32_t | |
319 | BytesTrieBuilder::skipElementsBySomeUnits(int32_t i, int32_t byteIndex, int32_t count) const { | |
320 | do { | |
321 | char byte=elements[i++].charAt(byteIndex, *strings); | |
322 | while(byte==elements[i].charAt(byteIndex, *strings)) { | |
323 | ++i; | |
324 | } | |
325 | } while(--count>0); | |
326 | return i; | |
327 | } | |
328 | ||
329 | int32_t | |
330 | BytesTrieBuilder::indexOfElementWithNextUnit(int32_t i, int32_t byteIndex, UChar byte) const { | |
331 | char b=(char)byte; | |
332 | while(b==elements[i].charAt(byteIndex, *strings)) { | |
333 | ++i; | |
334 | } | |
335 | return i; | |
336 | } | |
337 | ||
338 | BytesTrieBuilder::BTLinearMatchNode::BTLinearMatchNode(const char *bytes, int32_t len, Node *nextNode) | |
339 | : LinearMatchNode(len, nextNode), s(bytes) { | |
340 | hash=hash*37+ustr_hashCharsN(bytes, len); | |
341 | } | |
342 | ||
343 | UBool | |
344 | BytesTrieBuilder::BTLinearMatchNode::operator==(const Node &other) const { | |
345 | if(this==&other) { | |
346 | return TRUE; | |
347 | } | |
348 | if(!LinearMatchNode::operator==(other)) { | |
349 | return FALSE; | |
350 | } | |
351 | const BTLinearMatchNode &o=(const BTLinearMatchNode &)other; | |
352 | return 0==uprv_memcmp(s, o.s, length); | |
353 | } | |
354 | ||
355 | void | |
356 | BytesTrieBuilder::BTLinearMatchNode::write(StringTrieBuilder &builder) { | |
357 | BytesTrieBuilder &b=(BytesTrieBuilder &)builder; | |
358 | next->write(builder); | |
359 | b.write(s, length); | |
360 | offset=b.write(b.getMinLinearMatch()+length-1); | |
361 | } | |
362 | ||
363 | StringTrieBuilder::Node * | |
364 | BytesTrieBuilder::createLinearMatchNode(int32_t i, int32_t byteIndex, int32_t length, | |
365 | Node *nextNode) const { | |
366 | return new BTLinearMatchNode( | |
367 | elements[i].getString(*strings).data()+byteIndex, | |
368 | length, | |
369 | nextNode); | |
370 | } | |
371 | ||
372 | UBool | |
373 | BytesTrieBuilder::ensureCapacity(int32_t length) { | |
374 | if(bytes==NULL) { | |
375 | return FALSE; // previous memory allocation had failed | |
376 | } | |
377 | if(length>bytesCapacity) { | |
378 | int32_t newCapacity=bytesCapacity; | |
379 | do { | |
380 | newCapacity*=2; | |
381 | } while(newCapacity<=length); | |
382 | char *newBytes=reinterpret_cast<char *>(uprv_malloc(newCapacity)); | |
383 | if(newBytes==NULL) { | |
384 | // unable to allocate memory | |
385 | uprv_free(bytes); | |
386 | bytes=NULL; | |
387 | bytesCapacity=0; | |
388 | return FALSE; | |
389 | } | |
390 | uprv_memcpy(newBytes+(newCapacity-bytesLength), | |
391 | bytes+(bytesCapacity-bytesLength), bytesLength); | |
392 | uprv_free(bytes); | |
393 | bytes=newBytes; | |
394 | bytesCapacity=newCapacity; | |
395 | } | |
396 | return TRUE; | |
397 | } | |
398 | ||
399 | int32_t | |
400 | BytesTrieBuilder::write(int32_t byte) { | |
401 | int32_t newLength=bytesLength+1; | |
402 | if(ensureCapacity(newLength)) { | |
403 | bytesLength=newLength; | |
404 | bytes[bytesCapacity-bytesLength]=(char)byte; | |
405 | } | |
406 | return bytesLength; | |
407 | } | |
408 | ||
409 | int32_t | |
410 | BytesTrieBuilder::write(const char *b, int32_t length) { | |
411 | int32_t newLength=bytesLength+length; | |
412 | if(ensureCapacity(newLength)) { | |
413 | bytesLength=newLength; | |
414 | uprv_memcpy(bytes+(bytesCapacity-bytesLength), b, length); | |
415 | } | |
416 | return bytesLength; | |
417 | } | |
418 | ||
419 | int32_t | |
420 | BytesTrieBuilder::writeElementUnits(int32_t i, int32_t byteIndex, int32_t length) { | |
421 | return write(elements[i].getString(*strings).data()+byteIndex, length); | |
422 | } | |
423 | ||
424 | int32_t | |
425 | BytesTrieBuilder::writeValueAndFinal(int32_t i, UBool isFinal) { | |
426 | if(0<=i && i<=BytesTrie::kMaxOneByteValue) { | |
427 | return write(((BytesTrie::kMinOneByteValueLead+i)<<1)|isFinal); | |
428 | } | |
429 | char intBytes[5]; | |
430 | int32_t length=1; | |
431 | if(i<0 || i>0xffffff) { | |
432 | intBytes[0]=(char)BytesTrie::kFiveByteValueLead; | |
433 | intBytes[1]=(char)(i>>24); | |
434 | intBytes[2]=(char)(i>>16); | |
435 | intBytes[3]=(char)(i>>8); | |
436 | intBytes[4]=(char)i; | |
437 | length=5; | |
438 | // } else if(i<=BytesTrie::kMaxOneByteValue) { | |
439 | // intBytes[0]=(char)(BytesTrie::kMinOneByteValueLead+i); | |
440 | } else { | |
441 | if(i<=BytesTrie::kMaxTwoByteValue) { | |
442 | intBytes[0]=(char)(BytesTrie::kMinTwoByteValueLead+(i>>8)); | |
443 | } else { | |
444 | if(i<=BytesTrie::kMaxThreeByteValue) { | |
445 | intBytes[0]=(char)(BytesTrie::kMinThreeByteValueLead+(i>>16)); | |
446 | } else { | |
447 | intBytes[0]=(char)BytesTrie::kFourByteValueLead; | |
448 | intBytes[1]=(char)(i>>16); | |
449 | length=2; | |
450 | } | |
451 | intBytes[length++]=(char)(i>>8); | |
452 | } | |
453 | intBytes[length++]=(char)i; | |
454 | } | |
455 | intBytes[0]=(char)((intBytes[0]<<1)|isFinal); | |
456 | return write(intBytes, length); | |
457 | } | |
458 | ||
459 | int32_t | |
460 | BytesTrieBuilder::writeValueAndType(UBool hasValue, int32_t value, int32_t node) { | |
461 | int32_t offset=write(node); | |
462 | if(hasValue) { | |
463 | offset=writeValueAndFinal(value, FALSE); | |
464 | } | |
465 | return offset; | |
466 | } | |
467 | ||
468 | int32_t | |
469 | BytesTrieBuilder::writeDeltaTo(int32_t jumpTarget) { | |
470 | int32_t i=bytesLength-jumpTarget; | |
471 | U_ASSERT(i>=0); | |
472 | if(i<=BytesTrie::kMaxOneByteDelta) { | |
473 | return write(i); | |
474 | } | |
475 | char intBytes[5]; | |
476 | int32_t length; | |
477 | if(i<=BytesTrie::kMaxTwoByteDelta) { | |
478 | intBytes[0]=(char)(BytesTrie::kMinTwoByteDeltaLead+(i>>8)); | |
479 | length=1; | |
480 | } else { | |
481 | if(i<=BytesTrie::kMaxThreeByteDelta) { | |
482 | intBytes[0]=(char)(BytesTrie::kMinThreeByteDeltaLead+(i>>16)); | |
483 | length=2; | |
484 | } else { | |
485 | if(i<=0xffffff) { | |
486 | intBytes[0]=(char)BytesTrie::kFourByteDeltaLead; | |
487 | length=3; | |
488 | } else { | |
489 | intBytes[0]=(char)BytesTrie::kFiveByteDeltaLead; | |
490 | intBytes[1]=(char)(i>>24); | |
491 | length=4; | |
492 | } | |
493 | intBytes[1]=(char)(i>>16); | |
494 | } | |
495 | intBytes[1]=(char)(i>>8); | |
496 | } | |
497 | intBytes[length++]=(char)i; | |
498 | return write(intBytes, length); | |
499 | } | |
500 | ||
501 | U_NAMESPACE_END |