]>
Commit | Line | Data |
---|---|---|
374ca955 A |
1 | /* |
2 | ******************************************************************************* | |
3 | * | |
4 | * Copyright (C) 2003-2004, International Business Machines | |
5 | * Corporation and others. All Rights Reserved. | |
6 | * | |
7 | ******************************************************************************* | |
8 | * file name: gencnvex.c | |
9 | * encoding: US-ASCII | |
10 | * tab size: 8 (not used) | |
11 | * indentation:4 | |
12 | * | |
13 | * created on: 2003oct12 | |
14 | * created by: Markus W. Scherer | |
15 | */ | |
16 | ||
17 | #include <stdio.h> | |
18 | #include "unicode/utypes.h" | |
19 | #include "unicode/ustring.h" | |
20 | #include "cstring.h" | |
21 | #include "cmemory.h" | |
22 | #include "ucnv_cnv.h" | |
23 | #include "ucnvmbcs.h" | |
24 | #include "toolutil.h" | |
25 | #include "unewdata.h" | |
26 | #include "ucm.h" | |
27 | #include "makeconv.h" | |
28 | #include "genmbcs.h" | |
29 | ||
30 | #define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0])) | |
31 | ||
32 | static void | |
33 | CnvExtClose(NewConverter *cnvData); | |
34 | ||
35 | static UBool | |
36 | CnvExtIsValid(NewConverter *cnvData, | |
37 | const uint8_t *bytes, int32_t length); | |
38 | ||
39 | static UBool | |
40 | CnvExtAddTable(NewConverter *cnvData, UCMTable *table, UConverterStaticData *staticData); | |
41 | ||
42 | static uint32_t | |
43 | CnvExtWrite(NewConverter *cnvData, const UConverterStaticData *staticData, | |
44 | UNewDataMemory *pData, int32_t tableType); | |
45 | ||
46 | typedef struct CnvExtData { | |
47 | NewConverter newConverter; | |
48 | ||
49 | UCMFile *ucm; | |
50 | ||
51 | /* toUnicode (state table in ucm->states) */ | |
52 | UToolMemory *toUTable, *toUUChars; | |
53 | ||
54 | /* fromUnicode */ | |
55 | UToolMemory *fromUTableUChars, *fromUTableValues, *fromUBytes; | |
56 | ||
57 | uint16_t stage1[MBCS_STAGE_1_SIZE]; | |
58 | uint16_t stage2[MBCS_STAGE_2_SIZE]; | |
59 | uint16_t stage3[0x10000<<UCNV_EXT_STAGE_2_LEFT_SHIFT]; /* 0x10000 because of 16-bit stage 2/3 indexes */ | |
60 | uint32_t stage3b[0x10000]; | |
61 | ||
62 | int32_t stage1Top, stage2Top, stage3Top, stage3bTop; | |
63 | ||
64 | /* for stage3 compaction of <subchar1> |2 mappings */ | |
65 | uint16_t stage3Sub1Block; | |
66 | ||
67 | /* statistics */ | |
68 | int32_t | |
69 | maxInBytes, maxOutBytes, maxBytesPerUChar, | |
70 | maxInUChars, maxOutUChars, maxUCharsPerByte; | |
71 | } CnvExtData; | |
72 | ||
73 | NewConverter * | |
74 | CnvExtOpen(UCMFile *ucm) { | |
75 | CnvExtData *extData; | |
76 | ||
77 | extData=(CnvExtData *)uprv_malloc(sizeof(CnvExtData)); | |
78 | if(extData!=NULL) { | |
79 | uprv_memset(extData, 0, sizeof(CnvExtData)); | |
80 | ||
81 | extData->ucm=ucm; /* aliased, not owned */ | |
82 | ||
83 | extData->newConverter.close=CnvExtClose; | |
84 | extData->newConverter.isValid=CnvExtIsValid; | |
85 | extData->newConverter.addTable=CnvExtAddTable; | |
86 | extData->newConverter.write=CnvExtWrite; | |
87 | } | |
88 | return &extData->newConverter; | |
89 | } | |
90 | ||
91 | static void | |
92 | CnvExtClose(NewConverter *cnvData) { | |
93 | CnvExtData *extData=(CnvExtData *)cnvData; | |
94 | if(extData!=NULL) { | |
95 | utm_close(extData->toUTable); | |
96 | utm_close(extData->toUUChars); | |
97 | utm_close(extData->fromUTableUChars); | |
98 | utm_close(extData->fromUTableValues); | |
99 | utm_close(extData->fromUBytes); | |
100 | } | |
101 | } | |
102 | ||
103 | /* we do not expect this to be called */ | |
104 | static UBool | |
105 | CnvExtIsValid(NewConverter *cnvData, | |
106 | const uint8_t *bytes, int32_t length) { | |
107 | return FALSE; | |
108 | } | |
109 | ||
110 | static uint32_t | |
111 | CnvExtWrite(NewConverter *cnvData, const UConverterStaticData *staticData, | |
112 | UNewDataMemory *pData, int32_t tableType) { | |
113 | CnvExtData *extData=(CnvExtData *)cnvData; | |
114 | int32_t length, top, headerSize; | |
115 | ||
116 | int32_t indexes[UCNV_EXT_INDEXES_MIN_LENGTH]={ 0 }; | |
117 | ||
118 | if(tableType&TABLE_BASE) { | |
119 | headerSize=0; | |
120 | } else { | |
121 | _MBCSHeader header={ { 0, 0, 0, 0 }, 0, 0, 0, 0, 0, 0, 0 }; | |
122 | ||
123 | /* write the header and base table name for an extension-only table */ | |
124 | length=(int32_t)uprv_strlen(extData->ucm->baseName)+1; | |
125 | while(length&3) { | |
126 | /* add padding */ | |
127 | extData->ucm->baseName[length++]=0; | |
128 | } | |
129 | ||
130 | headerSize=sizeof(header)+length; | |
131 | ||
132 | /* fill the header */ | |
133 | header.version[0]=4; | |
134 | header.version[1]=2; | |
135 | header.flags=(uint32_t)((headerSize<<8)|MBCS_OUTPUT_EXT_ONLY); | |
136 | ||
137 | /* write the header and the base table name */ | |
138 | udata_writeBlock(pData, &header, sizeof(header)); | |
139 | udata_writeBlock(pData, extData->ucm->baseName, length); | |
140 | } | |
141 | ||
142 | /* fill indexes[] - offsets/indexes are in units of the target array */ | |
143 | top=0; | |
144 | ||
145 | indexes[UCNV_EXT_INDEXES_LENGTH]=length=UCNV_EXT_INDEXES_MIN_LENGTH; | |
146 | top+=length*4; | |
147 | ||
148 | indexes[UCNV_EXT_TO_U_INDEX]=top; | |
149 | indexes[UCNV_EXT_TO_U_LENGTH]=length=utm_countItems(extData->toUTable); | |
150 | top+=length*4; | |
151 | ||
152 | indexes[UCNV_EXT_TO_U_UCHARS_INDEX]=top; | |
153 | indexes[UCNV_EXT_TO_U_UCHARS_LENGTH]=length=utm_countItems(extData->toUUChars); | |
154 | top+=length*2; | |
155 | ||
156 | indexes[UCNV_EXT_FROM_U_UCHARS_INDEX]=top; | |
157 | length=utm_countItems(extData->fromUTableUChars); | |
158 | top+=length*2; | |
159 | ||
160 | if(top&3) { | |
161 | /* add padding */ | |
162 | *((UChar *)utm_alloc(extData->fromUTableUChars))=0; | |
163 | *((uint32_t *)utm_alloc(extData->fromUTableValues))=0; | |
164 | ++length; | |
165 | top+=2; | |
166 | } | |
167 | indexes[UCNV_EXT_FROM_U_LENGTH]=length; | |
168 | ||
169 | indexes[UCNV_EXT_FROM_U_VALUES_INDEX]=top; | |
170 | top+=length*4; | |
171 | ||
172 | indexes[UCNV_EXT_FROM_U_BYTES_INDEX]=top; | |
173 | length=utm_countItems(extData->fromUBytes); | |
174 | top+=length; | |
175 | ||
176 | if(top&1) { | |
177 | /* add padding */ | |
178 | *((uint8_t *)utm_alloc(extData->fromUBytes))=0; | |
179 | ++length; | |
180 | ++top; | |
181 | } | |
182 | indexes[UCNV_EXT_FROM_U_BYTES_LENGTH]=length; | |
183 | ||
184 | indexes[UCNV_EXT_FROM_U_STAGE_12_INDEX]=top; | |
185 | indexes[UCNV_EXT_FROM_U_STAGE_1_LENGTH]=length=extData->stage1Top; | |
186 | indexes[UCNV_EXT_FROM_U_STAGE_12_LENGTH]=length+=extData->stage2Top; | |
187 | top+=length*2; | |
188 | ||
189 | indexes[UCNV_EXT_FROM_U_STAGE_3_INDEX]=top; | |
190 | length=extData->stage3Top; | |
191 | top+=length*2; | |
192 | ||
193 | if(top&3) { | |
194 | /* add padding */ | |
195 | extData->stage3[extData->stage3Top++]=0; | |
196 | ++length; | |
197 | top+=2; | |
198 | } | |
199 | indexes[UCNV_EXT_FROM_U_STAGE_3_LENGTH]=length; | |
200 | ||
201 | indexes[UCNV_EXT_FROM_U_STAGE_3B_INDEX]=top; | |
202 | indexes[UCNV_EXT_FROM_U_STAGE_3B_LENGTH]=length=extData->stage3bTop; | |
203 | top+=length*4; | |
204 | ||
205 | indexes[UCNV_EXT_SIZE]=top; | |
206 | ||
207 | /* statistics */ | |
208 | indexes[UCNV_EXT_COUNT_BYTES]= | |
209 | (extData->maxInBytes<<16)| | |
210 | (extData->maxOutBytes<<8)| | |
211 | extData->maxBytesPerUChar; | |
212 | indexes[UCNV_EXT_COUNT_UCHARS]= | |
213 | (extData->maxInUChars<<16)| | |
214 | (extData->maxOutUChars<<8)| | |
215 | extData->maxUCharsPerByte; | |
216 | ||
217 | indexes[UCNV_EXT_FLAGS]=extData->ucm->ext->unicodeMask; | |
218 | ||
219 | /* write the extension data */ | |
220 | udata_writeBlock(pData, indexes, sizeof(indexes)); | |
221 | udata_writeBlock(pData, utm_getStart(extData->toUTable), indexes[UCNV_EXT_TO_U_LENGTH]*4); | |
222 | udata_writeBlock(pData, utm_getStart(extData->toUUChars), indexes[UCNV_EXT_TO_U_UCHARS_LENGTH]*2); | |
223 | ||
224 | udata_writeBlock(pData, utm_getStart(extData->fromUTableUChars), indexes[UCNV_EXT_FROM_U_LENGTH]*2); | |
225 | udata_writeBlock(pData, utm_getStart(extData->fromUTableValues), indexes[UCNV_EXT_FROM_U_LENGTH]*4); | |
226 | udata_writeBlock(pData, utm_getStart(extData->fromUBytes), indexes[UCNV_EXT_FROM_U_BYTES_LENGTH]); | |
227 | ||
228 | udata_writeBlock(pData, extData->stage1, extData->stage1Top*2); | |
229 | udata_writeBlock(pData, extData->stage2, extData->stage2Top*2); | |
230 | udata_writeBlock(pData, extData->stage3, extData->stage3Top*2); | |
231 | udata_writeBlock(pData, extData->stage3b, extData->stage3bTop*4); | |
232 | ||
233 | #if 0 | |
234 | { | |
235 | int32_t i, j; | |
236 | ||
237 | length=extData->stage1Top; | |
238 | printf("\nstage1[%x]:\n", length); | |
239 | ||
240 | for(i=0; i<length; ++i) { | |
241 | if(extData->stage1[i]!=length) { | |
242 | printf("stage1[%04x]=%04x\n", i, extData->stage1[i]); | |
243 | } | |
244 | } | |
245 | ||
246 | j=length; | |
247 | length=extData->stage2Top; | |
248 | printf("\nstage2[%x]:\n", length); | |
249 | ||
250 | for(i=0; i<length; ++j, ++i) { | |
251 | if(extData->stage2[i]!=0) { | |
252 | printf("stage12[%04x]=%04x\n", j, extData->stage2[i]); | |
253 | } | |
254 | } | |
255 | ||
256 | length=extData->stage3Top; | |
257 | printf("\nstage3[%x]:\n", length); | |
258 | ||
259 | for(i=0; i<length; ++i) { | |
260 | if(extData->stage3[i]!=0) { | |
261 | printf("stage3[%04x]=%04x\n", i, extData->stage3[i]); | |
262 | } | |
263 | } | |
264 | ||
265 | length=extData->stage3bTop; | |
266 | printf("\nstage3b[%x]:\n", length); | |
267 | ||
268 | for(i=0; i<length; ++i) { | |
269 | if(extData->stage3b[i]!=0) { | |
270 | printf("stage3b[%04x]=%08x\n", i, extData->stage3b[i]); | |
271 | } | |
272 | } | |
273 | } | |
274 | #endif | |
275 | ||
276 | if(VERBOSE) { | |
277 | printf("size of extension data: %ld\n", (long)top); | |
278 | } | |
279 | ||
280 | /* return the number of bytes that should have been written */ | |
281 | return (uint32_t)(headerSize+top); | |
282 | } | |
283 | ||
284 | /* to Unicode --------------------------------------------------------------- */ | |
285 | ||
286 | /* | |
287 | * Remove fromUnicode fallbacks and SUB mappings which are irrelevant for | |
288 | * the toUnicode table. | |
289 | * The table must be sorted. | |
290 | * Destroys previous data in the reverseMap. | |
291 | */ | |
292 | static int32_t | |
293 | reduceToUMappings(UCMTable *table) { | |
294 | UCMapping *mappings; | |
295 | int32_t *map; | |
296 | int32_t i, j, count; | |
297 | int8_t flag; | |
298 | ||
299 | mappings=table->mappings; | |
300 | map=table->reverseMap; | |
301 | count=table->mappingsLength; | |
302 | ||
303 | /* leave the map alone for the initial mappings with desired flags */ | |
304 | for(i=j=0; i<count; ++i) { | |
305 | flag=mappings[map[i]].f; | |
306 | if(flag!=0 && flag!=3) { | |
307 | break; | |
308 | } | |
309 | } | |
310 | ||
311 | /* reduce from here to the rest */ | |
312 | for(j=i; i<count; ++i) { | |
313 | flag=mappings[map[i]].f; | |
314 | if(flag==0 || flag==3) { | |
315 | map[j++]=map[i]; | |
316 | } | |
317 | } | |
318 | ||
319 | return j; | |
320 | } | |
321 | ||
322 | static uint32_t | |
323 | getToUnicodeValue(CnvExtData *extData, UCMTable *table, UCMapping *m) { | |
324 | UChar32 *u32; | |
325 | UChar *u; | |
326 | uint32_t value; | |
327 | int32_t u16Length, ratio; | |
328 | UErrorCode errorCode; | |
329 | ||
330 | /* write the Unicode result code point or string index */ | |
331 | if(m->uLen==1) { | |
332 | u16Length=U16_LENGTH(m->u); | |
333 | value=(uint32_t)(UCNV_EXT_TO_U_MIN_CODE_POINT+m->u); | |
334 | } else { | |
335 | /* the parser enforces m->uLen<=UCNV_EXT_MAX_UCHARS */ | |
336 | ||
337 | /* get the result code point string and its 16-bit string length */ | |
338 | u32=UCM_GET_CODE_POINTS(table, m); | |
339 | errorCode=U_ZERO_ERROR; | |
340 | u_strFromUTF32(NULL, 0, &u16Length, u32, m->uLen, &errorCode); | |
341 | if(U_FAILURE(errorCode) && errorCode!=U_BUFFER_OVERFLOW_ERROR) { | |
342 | exit(errorCode); | |
343 | } | |
344 | ||
345 | /* allocate it and put its length and index into the value */ | |
346 | value= | |
347 | (((uint32_t)m->uLen+UCNV_EXT_TO_U_LENGTH_OFFSET)<<UCNV_EXT_TO_U_LENGTH_SHIFT)| | |
348 | ((uint32_t)utm_countItems(extData->toUUChars)); | |
349 | u=utm_allocN(extData->toUUChars, u16Length); | |
350 | ||
351 | /* write the result 16-bit string */ | |
352 | errorCode=U_ZERO_ERROR; | |
353 | u_strFromUTF32(u, u16Length, NULL, u32, m->uLen, &errorCode); | |
354 | if(U_FAILURE(errorCode) && errorCode!=U_BUFFER_OVERFLOW_ERROR) { | |
355 | exit(errorCode); | |
356 | } | |
357 | } | |
358 | if(m->f==0) { | |
359 | value|=UCNV_EXT_TO_U_ROUNDTRIP_FLAG; | |
360 | } | |
361 | ||
362 | /* update statistics */ | |
363 | if(m->bLen>extData->maxInBytes) { | |
364 | extData->maxInBytes=m->bLen; | |
365 | } | |
366 | if(u16Length>extData->maxOutUChars) { | |
367 | extData->maxOutUChars=u16Length; | |
368 | } | |
369 | ||
370 | ratio=(u16Length+(m->bLen-1))/m->bLen; | |
371 | if(ratio>extData->maxUCharsPerByte) { | |
372 | extData->maxUCharsPerByte=ratio; | |
373 | } | |
374 | ||
375 | return value; | |
376 | } | |
377 | ||
378 | /* | |
379 | * Recursive toUTable generator core function. | |
380 | * Preconditions: | |
381 | * - start<limit (There is at least one mapping.) | |
382 | * - The mappings are sorted lexically. (Access is through the reverseMap.) | |
383 | * - All mappings between start and limit have input sequences that share | |
384 | * the same prefix of unitIndex length, and therefore all of these sequences | |
385 | * are at least unitIndex+1 long. | |
386 | * - There are only relevant mappings available through the reverseMap, | |
387 | * see reduceToUMappings(). | |
388 | * | |
389 | * One function invocation generates one section table. | |
390 | * | |
391 | * Steps: | |
392 | * 1. Count the number of unique unit values and get the low/high unit values | |
393 | * that occur at unitIndex. | |
394 | * 2. Allocate the section table with possible optimization for linear access. | |
395 | * 3. Write temporary version of the section table with start indexes of | |
396 | * subsections, each corresponding to one unit value at unitIndex. | |
397 | * 4. Iterate through the table once more, and depending on the subsection length: | |
398 | * 0: write 0 as a result value (unused byte in linear-access section table) | |
399 | * >0: if there is one mapping with an input unit sequence of unitIndex+1 | |
400 | * then defaultValue=compute the mapping result for this whole sequence | |
401 | * else defaultValue=0 | |
402 | * | |
403 | * recurse into the subsection | |
404 | */ | |
405 | static UBool | |
406 | generateToUTable(CnvExtData *extData, UCMTable *table, | |
407 | int32_t start, int32_t limit, int32_t unitIndex, | |
408 | uint32_t defaultValue) { | |
409 | UCMapping *mappings, *m; | |
410 | int32_t *map; | |
411 | int32_t i, j, uniqueCount, count, subStart, subLimit; | |
412 | ||
413 | uint8_t *bytes; | |
414 | int32_t low, high, prev; | |
415 | ||
416 | uint32_t *section; | |
417 | ||
418 | mappings=table->mappings; | |
419 | map=table->reverseMap; | |
420 | ||
421 | /* step 1: examine the input units; set low, high, uniqueCount */ | |
422 | m=mappings+map[start]; | |
423 | bytes=UCM_GET_BYTES(table, m); | |
424 | low=bytes[unitIndex]; | |
425 | uniqueCount=1; | |
426 | ||
427 | prev=high=low; | |
428 | for(i=start+1; i<limit; ++i) { | |
429 | m=mappings+map[i]; | |
430 | bytes=UCM_GET_BYTES(table, m); | |
431 | high=bytes[unitIndex]; | |
432 | ||
433 | if(high!=prev) { | |
434 | prev=high; | |
435 | ++uniqueCount; | |
436 | } | |
437 | } | |
438 | ||
439 | /* step 2: allocate the section; set count, section */ | |
440 | count=(high-low)+1; | |
441 | if(unitIndex==0 || uniqueCount>=(3*count)/4) { | |
442 | /* | |
443 | * for the root table and for fairly full tables: | |
444 | * allocate for direct, linear array access | |
445 | * by keeping count, to write an entry for each unit value | |
446 | * from low to high | |
447 | */ | |
448 | } else { | |
449 | count=uniqueCount; | |
450 | } | |
451 | ||
452 | /* allocate the section: 1 entry for the header + count for the items */ | |
453 | section=(uint32_t *)utm_allocN(extData->toUTable, 1+count); | |
454 | ||
455 | /* write the section header */ | |
456 | *section++=((uint32_t)count<<UCNV_EXT_TO_U_BYTE_SHIFT)|defaultValue; | |
457 | ||
458 | /* step 3: write temporary section table with subsection starts */ | |
459 | prev=low-1; /* just before low to prevent empty subsections before low */ | |
460 | j=0; /* section table index */ | |
461 | for(i=start; i<limit; ++i) { | |
462 | m=mappings+map[i]; | |
463 | bytes=UCM_GET_BYTES(table, m); | |
464 | high=bytes[unitIndex]; | |
465 | ||
466 | if(high!=prev) { | |
467 | /* start of a new subsection for unit high */ | |
468 | if(count>uniqueCount) { | |
469 | /* write empty subsections for unused units in a linear table */ | |
470 | while(++prev<high) { | |
471 | section[j++]=((uint32_t)prev<<UCNV_EXT_TO_U_BYTE_SHIFT)|(uint32_t)i; | |
472 | } | |
473 | } else { | |
474 | prev=high; | |
475 | } | |
476 | ||
477 | /* write the entry with the subsection start */ | |
478 | section[j++]=((uint32_t)high<<UCNV_EXT_TO_U_BYTE_SHIFT)|(uint32_t)i; | |
479 | } | |
480 | } | |
481 | /* assert(j==count) */ | |
482 | ||
483 | /* step 4: recurse and write results */ | |
484 | subLimit=UCNV_EXT_TO_U_GET_VALUE(section[0]); | |
485 | for(j=0; j<count; ++j) { | |
486 | subStart=subLimit; | |
487 | subLimit= (j+1)<count ? UCNV_EXT_TO_U_GET_VALUE(section[j+1]) : limit; | |
488 | ||
489 | /* remove the subStart temporary value */ | |
490 | section[j]&=~UCNV_EXT_TO_U_VALUE_MASK; | |
491 | ||
492 | if(subStart==subLimit) { | |
493 | /* leave the value zero: empty subsection for unused unit in a linear table */ | |
494 | continue; | |
495 | } | |
496 | ||
497 | /* see if there is exactly one input unit sequence of length unitIndex+1 */ | |
498 | defaultValue=0; | |
499 | m=mappings+map[subStart]; | |
500 | if(m->bLen==unitIndex+1) { | |
501 | /* do not include this in generateToUTable() */ | |
502 | ++subStart; | |
503 | ||
504 | if(subStart<subLimit && mappings[map[subStart]].bLen==unitIndex+1) { | |
505 | /* print error for multiple same-input-sequence mappings */ | |
506 | fprintf(stderr, "error: multiple mappings from same bytes\n"); | |
507 | ucm_printMapping(table, m, stderr); | |
508 | ucm_printMapping(table, mappings+map[subStart], stderr); | |
509 | return FALSE; | |
510 | } | |
511 | ||
512 | defaultValue=getToUnicodeValue(extData, table, m); | |
513 | } | |
514 | ||
515 | if(subStart==subLimit) { | |
516 | /* write the result for the input sequence ending here */ | |
517 | section[j]|=defaultValue; | |
518 | } else { | |
519 | /* write the index to the subsection table */ | |
520 | section[j]|=(uint32_t)utm_countItems(extData->toUTable); | |
521 | ||
522 | /* recurse */ | |
523 | if(!generateToUTable(extData, table, subStart, subLimit, unitIndex+1, defaultValue)) { | |
524 | return FALSE; | |
525 | } | |
526 | } | |
527 | } | |
528 | return TRUE; | |
529 | } | |
530 | ||
531 | /* | |
532 | * Generate the toUTable and toUUChars from the input table. | |
533 | * The input table must be sorted, and all precision flags must be 0..3. | |
534 | * This function will modify the table's reverseMap. | |
535 | */ | |
536 | static UBool | |
537 | makeToUTable(CnvExtData *extData, UCMTable *table) { | |
538 | int32_t toUCount; | |
539 | ||
540 | toUCount=reduceToUMappings(table); | |
541 | ||
542 | extData->toUTable=utm_open("cnv extension toUTable", 0x10000, UCNV_EXT_TO_U_MIN_CODE_POINT, 4); | |
543 | extData->toUUChars=utm_open("cnv extension toUUChars", 0x10000, UCNV_EXT_TO_U_INDEX_MASK+1, 2); | |
544 | ||
545 | return generateToUTable(extData, table, 0, toUCount, 0, 0); | |
546 | } | |
547 | ||
548 | /* from Unicode ------------------------------------------------------------- */ | |
549 | ||
550 | /* | |
551 | * preprocessing: | |
552 | * rebuild reverseMap with mapping indexes for mappings relevant for from Unicode | |
553 | * change each Unicode string to encode all but the first code point in 16-bit form | |
554 | * | |
555 | * generation: | |
556 | * for each unique code point | |
557 | * write an entry in the 3-stage trie | |
558 | * check that there is only one single-code point sequence | |
559 | * start recursion for following 16-bit input units | |
560 | */ | |
561 | ||
562 | /* | |
563 | * Remove toUnicode fallbacks and non-<subchar1> SUB mappings | |
564 | * which are irrelevant for the fromUnicode extension table. | |
565 | * Overwrite the reverseMap with an index array to the relevant mappings. | |
566 | * Modify the code point sequences to a generator-friendly format where | |
567 | * the first code points remains unchanged but the following are recoded | |
568 | * into 16-bit Unicode string form. | |
569 | * The table must be sorted. | |
570 | * Destroys previous data in the reverseMap. | |
571 | */ | |
572 | static int32_t | |
573 | prepareFromUMappings(UCMTable *table) { | |
574 | UCMapping *mappings, *m; | |
575 | int32_t *map; | |
576 | int32_t i, j, count; | |
577 | int8_t flag; | |
578 | ||
579 | mappings=table->mappings; | |
580 | map=table->reverseMap; | |
581 | count=table->mappingsLength; | |
582 | ||
583 | /* | |
584 | * we do not go through the map on input because the mappings are | |
585 | * sorted lexically | |
586 | */ | |
587 | m=mappings; | |
588 | ||
589 | for(i=j=0; i<count; ++m, ++i) { | |
590 | flag=m->f; | |
591 | if(flag==0 || flag==1 || (flag==2 && m->bLen==1)) { | |
592 | map[j++]=i; | |
593 | ||
594 | if(m->uLen>1) { | |
595 | /* recode all but the first code point to 16-bit Unicode */ | |
596 | UChar32 *u32; | |
597 | UChar *u; | |
598 | UChar32 c; | |
599 | int32_t q, r; | |
600 | ||
601 | u32=UCM_GET_CODE_POINTS(table, m); | |
602 | u=(UChar *)u32; /* destructive in-place recoding */ | |
603 | for(r=2, q=1; q<m->uLen; ++q) { | |
604 | c=u32[q]; | |
605 | U16_APPEND_UNSAFE(u, r, c); | |
606 | } | |
607 | ||
608 | /* counts the first code point always at 2 - the first 16-bit unit is at 16-bit index 2 */ | |
609 | m->uLen=(int8_t)r; | |
610 | } | |
611 | } | |
612 | } | |
613 | ||
614 | return j; | |
615 | } | |
616 | ||
617 | static uint32_t | |
618 | getFromUBytesValue(CnvExtData *extData, UCMTable *table, UCMapping *m) { | |
619 | uint8_t *bytes, *resultBytes; | |
620 | uint32_t value; | |
621 | int32_t u16Length, ratio; | |
622 | ||
623 | if(m->f==2) { | |
624 | /* | |
625 | * no mapping, <subchar1> preferred | |
626 | * | |
627 | * no need to count in statistics because the subchars are already | |
628 | * counted for maxOutBytes and maxBytesPerUChar in UConverterStaticData, | |
629 | * and this non-mapping does not count for maxInUChars which are always | |
630 | * trivially at least two if counting unmappable supplementary code points | |
631 | */ | |
632 | return UCNV_EXT_FROM_U_SUBCHAR1; | |
633 | } | |
634 | ||
635 | bytes=UCM_GET_BYTES(table, m); | |
636 | value=0; | |
637 | switch(m->bLen) { | |
638 | /* 1..3: store the bytes in the value word */ | |
639 | case 3: | |
640 | value=((uint32_t)*bytes++)<<16; | |
641 | case 2: | |
642 | value|=((uint32_t)*bytes++)<<8; | |
643 | case 1: | |
644 | value|=*bytes; | |
645 | break; | |
646 | default: | |
647 | /* the parser enforces m->bLen<=UCNV_EXT_MAX_BYTES */ | |
648 | /* store the bytes in fromUBytes[] and the index in the value word */ | |
649 | value=(uint32_t)utm_countItems(extData->fromUBytes); | |
650 | resultBytes=utm_allocN(extData->fromUBytes, m->bLen); | |
651 | uprv_memcpy(resultBytes, bytes, m->bLen); | |
652 | break; | |
653 | } | |
654 | value|=(uint32_t)m->bLen<<UCNV_EXT_FROM_U_LENGTH_SHIFT; | |
655 | if(m->f==0) { | |
656 | value|=UCNV_EXT_FROM_U_ROUNDTRIP_FLAG; | |
657 | } | |
658 | ||
659 | /* calculate the real UTF-16 length (see recoding in prepareFromUMappings()) */ | |
660 | if(m->uLen==1) { | |
661 | u16Length=U16_LENGTH(m->u); | |
662 | } else { | |
663 | u16Length=U16_LENGTH(UCM_GET_CODE_POINTS(table, m)[0])+(m->uLen-2); | |
664 | } | |
665 | ||
666 | /* update statistics */ | |
667 | if(u16Length>extData->maxInUChars) { | |
668 | extData->maxInUChars=u16Length; | |
669 | } | |
670 | if(m->bLen>extData->maxOutBytes) { | |
671 | extData->maxOutBytes=m->bLen; | |
672 | } | |
673 | ||
674 | ratio=(m->bLen+(u16Length-1))/u16Length; | |
675 | if(ratio>extData->maxBytesPerUChar) { | |
676 | extData->maxBytesPerUChar=ratio; | |
677 | } | |
678 | ||
679 | return value; | |
680 | } | |
681 | ||
682 | /* | |
683 | * works like generateToUTable(), except that the | |
684 | * output section consists of two arrays, one for input UChars and one | |
685 | * for result values | |
686 | * | |
687 | * also, fromUTable sections are always stored in a compact form for | |
688 | * access via binary search | |
689 | */ | |
690 | static UBool | |
691 | generateFromUTable(CnvExtData *extData, UCMTable *table, | |
692 | int32_t start, int32_t limit, int32_t unitIndex, | |
693 | uint32_t defaultValue) { | |
694 | UCMapping *mappings, *m; | |
695 | int32_t *map; | |
696 | int32_t i, j, uniqueCount, count, subStart, subLimit; | |
697 | ||
698 | UChar *uchars; | |
699 | UChar32 low, high, prev; | |
700 | ||
701 | UChar *sectionUChars; | |
702 | uint32_t *sectionValues; | |
703 | ||
704 | mappings=table->mappings; | |
705 | map=table->reverseMap; | |
706 | ||
707 | /* step 1: examine the input units; set low, high, uniqueCount */ | |
708 | m=mappings+map[start]; | |
709 | uchars=(UChar *)UCM_GET_CODE_POINTS(table, m); | |
710 | low=uchars[unitIndex]; | |
711 | uniqueCount=1; | |
712 | ||
713 | prev=high=low; | |
714 | for(i=start+1; i<limit; ++i) { | |
715 | m=mappings+map[i]; | |
716 | uchars=(UChar *)UCM_GET_CODE_POINTS(table, m); | |
717 | high=uchars[unitIndex]; | |
718 | ||
719 | if(high!=prev) { | |
720 | prev=high; | |
721 | ++uniqueCount; | |
722 | } | |
723 | } | |
724 | ||
725 | /* step 2: allocate the section; set count, section */ | |
726 | /* the fromUTable always stores for access via binary search */ | |
727 | count=uniqueCount; | |
728 | ||
729 | /* allocate the section: 1 entry for the header + count for the items */ | |
730 | sectionUChars=(UChar *)utm_allocN(extData->fromUTableUChars, 1+count); | |
731 | sectionValues=(uint32_t *)utm_allocN(extData->fromUTableValues, 1+count); | |
732 | ||
733 | /* write the section header */ | |
734 | *sectionUChars++=(UChar)count; | |
735 | *sectionValues++=defaultValue; | |
736 | ||
737 | /* step 3: write temporary section table with subsection starts */ | |
738 | prev=low-1; /* just before low to prevent empty subsections before low */ | |
739 | j=0; /* section table index */ | |
740 | for(i=start; i<limit; ++i) { | |
741 | m=mappings+map[i]; | |
742 | uchars=(UChar *)UCM_GET_CODE_POINTS(table, m); | |
743 | high=uchars[unitIndex]; | |
744 | ||
745 | if(high!=prev) { | |
746 | /* start of a new subsection for unit high */ | |
747 | prev=high; | |
748 | ||
749 | /* write the entry with the subsection start */ | |
750 | sectionUChars[j]=(UChar)high; | |
751 | sectionValues[j]=(uint32_t)i; | |
752 | ++j; | |
753 | } | |
754 | } | |
755 | /* assert(j==count) */ | |
756 | ||
757 | /* step 4: recurse and write results */ | |
758 | subLimit=(int32_t)(sectionValues[0]); | |
759 | for(j=0; j<count; ++j) { | |
760 | subStart=subLimit; | |
761 | subLimit= (j+1)<count ? (int32_t)(sectionValues[j+1]) : limit; | |
762 | ||
763 | /* see if there is exactly one input unit sequence of length unitIndex+1 */ | |
764 | defaultValue=0; | |
765 | m=mappings+map[subStart]; | |
766 | if(m->uLen==unitIndex+1) { | |
767 | /* do not include this in generateToUTable() */ | |
768 | ++subStart; | |
769 | ||
770 | if(subStart<subLimit && mappings[map[subStart]].uLen==unitIndex+1) { | |
771 | /* print error for multiple same-input-sequence mappings */ | |
772 | fprintf(stderr, "error: multiple mappings from same Unicode code points\n"); | |
773 | ucm_printMapping(table, m, stderr); | |
774 | ucm_printMapping(table, mappings+map[subStart], stderr); | |
775 | return FALSE; | |
776 | } | |
777 | ||
778 | defaultValue=getFromUBytesValue(extData, table, m); | |
779 | } | |
780 | ||
781 | if(subStart==subLimit) { | |
782 | /* write the result for the input sequence ending here */ | |
783 | sectionValues[j]=defaultValue; | |
784 | } else { | |
785 | /* write the index to the subsection table */ | |
786 | sectionValues[j]=(uint32_t)utm_countItems(extData->fromUTableValues); | |
787 | ||
788 | /* recurse */ | |
789 | if(!generateFromUTable(extData, table, subStart, subLimit, unitIndex+1, defaultValue)) { | |
790 | return FALSE; | |
791 | } | |
792 | } | |
793 | } | |
794 | return TRUE; | |
795 | } | |
796 | ||
797 | /* | |
798 | * add entries to the fromUnicode trie, | |
799 | * assume to be called with code points in ascending order | |
800 | * and use that to build the trie in precompacted form | |
801 | */ | |
802 | static void | |
803 | addFromUTrieEntry(CnvExtData *extData, UChar32 c, uint32_t value) { | |
804 | int32_t i1, i2, i3, i3b, nextOffset, min, newBlock; | |
805 | ||
806 | if(value==0) { | |
807 | return; | |
808 | } | |
809 | ||
810 | /* | |
811 | * compute the index for each stage, | |
812 | * allocate a stage block if necessary, | |
813 | * and write the stage value | |
814 | */ | |
815 | i1=c>>10; | |
816 | if(i1>=extData->stage1Top) { | |
817 | extData->stage1Top=i1+1; | |
818 | } | |
819 | ||
820 | nextOffset=(c>>4)&0x3f; | |
821 | ||
822 | if(extData->stage1[i1]==0) { | |
823 | /* allocate another block in stage 2; overlap with the previous block */ | |
824 | newBlock=extData->stage2Top; | |
825 | min=newBlock-nextOffset; /* minimum block start with overlap */ | |
826 | while(min<newBlock && extData->stage2[newBlock-1]==0) { | |
827 | --newBlock; | |
828 | } | |
829 | ||
830 | extData->stage1[i1]=(uint16_t)newBlock; | |
831 | extData->stage2Top=newBlock+MBCS_STAGE_2_BLOCK_SIZE; | |
832 | if(extData->stage2Top>LENGTHOF(extData->stage2)) { | |
833 | fprintf(stderr, "error: too many stage 2 entries at U+%04x\n", (int)c); | |
834 | exit(U_MEMORY_ALLOCATION_ERROR); | |
835 | } | |
836 | } | |
837 | ||
838 | i2=extData->stage1[i1]+nextOffset; | |
839 | nextOffset=c&0xf; | |
840 | ||
841 | if(extData->stage2[i2]==0) { | |
842 | /* allocate another block in stage 3; overlap with the previous block */ | |
843 | newBlock=extData->stage3Top; | |
844 | min=newBlock-nextOffset; /* minimum block start with overlap */ | |
845 | while(min<newBlock && extData->stage3[newBlock-1]==0) { | |
846 | --newBlock; | |
847 | } | |
848 | ||
849 | /* round up to a multiple of stage 3 granularity >1 (similar to utrie.c) */ | |
850 | newBlock=(newBlock+(UCNV_EXT_STAGE_3_GRANULARITY-1))&~(UCNV_EXT_STAGE_3_GRANULARITY-1); | |
851 | extData->stage2[i2]=(uint16_t)(newBlock>>UCNV_EXT_STAGE_2_LEFT_SHIFT); | |
852 | ||
853 | extData->stage3Top=newBlock+MBCS_STAGE_3_BLOCK_SIZE; | |
854 | if(extData->stage3Top>LENGTHOF(extData->stage3)) { | |
855 | fprintf(stderr, "error: too many stage 3 entries at U+%04x\n", (int)c); | |
856 | exit(U_MEMORY_ALLOCATION_ERROR); | |
857 | } | |
858 | } | |
859 | ||
860 | i3=((int32_t)extData->stage2[i2]<<UCNV_EXT_STAGE_2_LEFT_SHIFT)+nextOffset; | |
861 | /* | |
862 | * assume extData->stage3[i3]==0 because we get | |
863 | * code points in strictly ascending order | |
864 | */ | |
865 | ||
866 | if(value==UCNV_EXT_FROM_U_SUBCHAR1) { | |
867 | /* <subchar1> SUB mapping, see getFromUBytesValue() and prepareFromUMappings() */ | |
868 | extData->stage3[i3]=1; | |
869 | ||
870 | /* | |
871 | * precompaction is not optimal for <subchar1> |2 mappings because | |
872 | * stage3 values for them are all the same, unlike for other mappings | |
873 | * which all have unique values; | |
874 | * use a simple compaction of reusing a whole block filled with these | |
875 | * mappings | |
876 | */ | |
877 | ||
878 | /* is the entire block filled with <subchar1> |2 mappings? */ | |
879 | if(nextOffset==MBCS_STAGE_3_BLOCK_SIZE-1) { | |
880 | for(min=i3-nextOffset; | |
881 | min<i3 && extData->stage3[min]==1; | |
882 | ++min) {} | |
883 | ||
884 | if(min==i3) { | |
885 | /* the entire block is filled with these mappings */ | |
886 | if(extData->stage3Sub1Block!=0) { | |
887 | /* point to the previous such block and remove this block from stage3 */ | |
888 | extData->stage2[i2]=extData->stage3Sub1Block; | |
889 | extData->stage3Top-=MBCS_STAGE_3_BLOCK_SIZE; | |
890 | uprv_memset(extData->stage3+extData->stage3Top, 0, MBCS_STAGE_3_BLOCK_SIZE*2); | |
891 | } else { | |
892 | /* remember this block's stage2 entry */ | |
893 | extData->stage3Sub1Block=extData->stage2[i2]; | |
894 | } | |
895 | } | |
896 | } | |
897 | } else { | |
898 | if((i3b=extData->stage3bTop++)>=LENGTHOF(extData->stage3b)) { | |
899 | fprintf(stderr, "error: too many stage 3b entries at U+%04x\n", (int)c); | |
900 | exit(U_MEMORY_ALLOCATION_ERROR); | |
901 | } | |
902 | ||
903 | /* roundtrip or fallback mapping */ | |
904 | extData->stage3[i3]=(uint16_t)i3b; | |
905 | extData->stage3b[i3b]=value; | |
906 | } | |
907 | } | |
908 | ||
909 | static UBool | |
910 | generateFromUTrie(CnvExtData *extData, UCMTable *table, int32_t mapLength) { | |
911 | UCMapping *mappings, *m; | |
912 | int32_t *map; | |
913 | uint32_t value; | |
914 | int32_t subStart, subLimit; | |
915 | ||
916 | UChar32 *codePoints; | |
917 | UChar32 c, next; | |
918 | ||
919 | if(mapLength==0) { | |
920 | return TRUE; | |
921 | } | |
922 | ||
923 | mappings=table->mappings; | |
924 | map=table->reverseMap; | |
925 | ||
926 | /* | |
927 | * iterate over same-initial-code point mappings, | |
928 | * enter the initial code point into the trie, | |
929 | * and start a recursion on the corresponding mappings section | |
930 | * with generateFromUTable() | |
931 | */ | |
932 | m=mappings+map[0]; | |
933 | codePoints=UCM_GET_CODE_POINTS(table, m); | |
934 | next=codePoints[0]; | |
935 | subLimit=0; | |
936 | while(subLimit<mapLength) { | |
937 | /* get a new subsection of mappings starting with the same code point */ | |
938 | subStart=subLimit; | |
939 | c=next; | |
940 | while(next==c && ++subLimit<mapLength) { | |
941 | m=mappings+map[subLimit]; | |
942 | codePoints=UCM_GET_CODE_POINTS(table, m); | |
943 | next=codePoints[0]; | |
944 | } | |
945 | ||
946 | /* | |
947 | * compute the value for this code point; | |
948 | * if there is a mapping for this code point alone, it is at subStart | |
949 | * because the table is sorted lexically | |
950 | */ | |
951 | value=0; | |
952 | m=mappings+map[subStart]; | |
953 | codePoints=UCM_GET_CODE_POINTS(table, m); | |
954 | if(m->uLen==1) { | |
955 | /* do not include this in generateFromUTable() */ | |
956 | ++subStart; | |
957 | ||
958 | if(subStart<subLimit && mappings[map[subStart]].uLen==1) { | |
959 | /* print error for multiple same-input-sequence mappings */ | |
960 | fprintf(stderr, "error: multiple mappings from same Unicode code points\n"); | |
961 | ucm_printMapping(table, m, stderr); | |
962 | ucm_printMapping(table, mappings+map[subStart], stderr); | |
963 | return FALSE; | |
964 | } | |
965 | ||
966 | value=getFromUBytesValue(extData, table, m); | |
967 | } | |
968 | ||
969 | if(subStart==subLimit) { | |
970 | /* write the result for this one code point */ | |
971 | addFromUTrieEntry(extData, c, value); | |
972 | } else { | |
973 | /* write the index to the subsection table */ | |
974 | addFromUTrieEntry(extData, c, (uint32_t)utm_countItems(extData->fromUTableValues)); | |
975 | ||
976 | /* recurse, starting from 16-bit-unit index 2, the first 16-bit unit after c */ | |
977 | if(!generateFromUTable(extData, table, subStart, subLimit, 2, value)) { | |
978 | return FALSE; | |
979 | } | |
980 | } | |
981 | } | |
982 | return TRUE; | |
983 | } | |
984 | ||
985 | /* | |
986 | * Generate the fromU data structures from the input table. | |
987 | * The input table must be sorted, and all precision flags must be 0..3. | |
988 | * This function will modify the table's reverseMap. | |
989 | */ | |
990 | static UBool | |
991 | makeFromUTable(CnvExtData *extData, UCMTable *table) { | |
992 | uint16_t *stage1; | |
993 | int32_t i, stage1Top, fromUCount; | |
994 | ||
995 | fromUCount=prepareFromUMappings(table); | |
996 | ||
997 | extData->fromUTableUChars=utm_open("cnv extension fromUTableUChars", 0x10000, UCNV_EXT_FROM_U_DATA_MASK+1, 2); | |
998 | extData->fromUTableValues=utm_open("cnv extension fromUTableValues", 0x10000, UCNV_EXT_FROM_U_DATA_MASK+1, 4); | |
999 | extData->fromUBytes=utm_open("cnv extension fromUBytes", 0x10000, UCNV_EXT_FROM_U_DATA_MASK+1, 1); | |
1000 | ||
1001 | /* allocate all-unassigned stage blocks */ | |
1002 | extData->stage2Top=MBCS_STAGE_2_FIRST_ASSIGNED; | |
1003 | extData->stage3Top=MBCS_STAGE_3_FIRST_ASSIGNED; | |
1004 | ||
1005 | /* | |
1006 | * stage 3b stores only unique values, and in | |
1007 | * index 0: 0 for "no mapping" | |
1008 | * index 1: "no mapping" with preference for <subchar1> rather than <subchar> | |
1009 | */ | |
1010 | extData->stage3b[1]=UCNV_EXT_FROM_U_SUBCHAR1; | |
1011 | extData->stage3bTop=2; | |
1012 | ||
1013 | /* allocate the first entry in the fromUTable because index 0 means "no result" */ | |
1014 | utm_alloc(extData->fromUTableUChars); | |
1015 | utm_alloc(extData->fromUTableValues); | |
1016 | ||
1017 | if(!generateFromUTrie(extData, table, fromUCount)) { | |
1018 | return FALSE; | |
1019 | } | |
1020 | ||
1021 | /* | |
1022 | * offset the stage 1 trie entries by stage1Top because they will | |
1023 | * be stored in a single array | |
1024 | */ | |
1025 | stage1=extData->stage1; | |
1026 | stage1Top=extData->stage1Top; | |
1027 | for(i=0; i<stage1Top; ++i) { | |
1028 | stage1[i]=(uint16_t)(stage1[i]+stage1Top); | |
1029 | } | |
1030 | ||
1031 | return TRUE; | |
1032 | } | |
1033 | ||
1034 | /* -------------------------------------------------------------------------- */ | |
1035 | ||
1036 | static UBool | |
1037 | CnvExtAddTable(NewConverter *cnvData, UCMTable *table, UConverterStaticData *staticData) { | |
1038 | CnvExtData *extData; | |
1039 | ||
1040 | if(table->unicodeMask&UCNV_HAS_SURROGATES) { | |
1041 | fprintf(stderr, "error: contains mappings for surrogate code points\n"); | |
1042 | return FALSE; | |
1043 | } | |
1044 | ||
1045 | staticData->conversionType=UCNV_MBCS; | |
1046 | ||
1047 | extData=(CnvExtData *)cnvData; | |
1048 | ||
1049 | /* | |
1050 | * assume that the table is sorted | |
1051 | * | |
1052 | * call the functions in this order because | |
1053 | * makeToUTable() modifies the original reverseMap, | |
1054 | * makeFromUTable() writes a whole new mapping into reverseMap | |
1055 | */ | |
1056 | return | |
1057 | makeToUTable(extData, table) && | |
1058 | makeFromUTable(extData, table); | |
1059 | } |