1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ****************************************************************************** |
5 | * |
6 | * Copyright (C) 2000-2016, International Business Machines |
7 | * Corporation and others. All Rights Reserved. |
8 | * |
9 | ****************************************************************************** |
10 | * file name: ucnvscsu.c |
11 | * encoding: UTF-8 |
12 | * tab size: 8 (not used) |
13 | * indentation:4 |
14 | * |
15 | * created on: 2000nov18 |
16 | * created by: Markus W. Scherer |
17 | * |
18 | * This is an implementation of the Standard Compression Scheme for Unicode |
19 | * as defined in https://www.unicode.org/reports/tr6/ . |
20 | * Reserved commands and window settings are treated as illegal sequences and |
21 | * will result in callback calls. |
22 | */ |
23 | |
24 | #include "unicode/utypes.h" |
25 | |
26 | #if !UCONFIG_NO_CONVERSION && !UCONFIG_ONLY_HTML_CONVERSION |
27 | |
28 | #include "unicode/ucnv.h" |
29 | #include "unicode/ucnv_cb.h" |
30 | #include "unicode/utf16.h" |
31 | #include "ucnv_bld.h" |
32 | #include "ucnv_cnv.h" |
33 | #include "cmemory.h" |
34 | |
35 | /* SCSU definitions --------------------------------------------------------- */ |
36 | |
37 | /* SCSU command byte values */ |
38 | enum { |
39 | SQ0=0x01, /* Quote from window pair 0 */ |
40 | SQ7=0x08, /* Quote from window pair 7 */ |
41 | SDX=0x0B, /* Define a window as extended */ |
42 | Srs=0x0C, /* reserved */ |
43 | SQU=0x0E, /* Quote a single Unicode character */ |
44 | SCU=0x0F, /* Change to Unicode mode */ |
45 | SC0=0x10, /* Select window 0 */ |
46 | SC7=0x17, /* Select window 7 */ |
47 | SD0=0x18, /* Define and select window 0 */ |
48 | SD7=0x1F, /* Define and select window 7 */ |
49 | |
50 | UC0=0xE0, /* Select window 0 */ |
51 | UC7=0xE7, /* Select window 7 */ |
52 | UD0=0xE8, /* Define and select window 0 */ |
53 | UD7=0xEF, /* Define and select window 7 */ |
54 | UQU=0xF0, /* Quote a single Unicode character */ |
55 | UDX=0xF1, /* Define a Window as extended */ |
56 | Urs=0xF2 /* reserved */ |
57 | }; |
58 | |
59 | enum { |
60 | /* |
61 | * Unicode code points from 3400 to E000 are not adressible by |
62 | * dynamic window, since in these areas no short run alphabets are |
63 | * found. Therefore add gapOffset to all values from gapThreshold. |
64 | */ |
65 | gapThreshold=0x68, |
66 | gapOffset=0xAC00, |
67 | |
68 | /* values between reservedStart and fixedThreshold are reserved */ |
69 | reservedStart=0xA8, |
70 | |
71 | /* use table of predefined fixed offsets for values from fixedThreshold */ |
72 | fixedThreshold=0xF9 |
73 | }; |
74 | |
75 | /* constant offsets for the 8 static windows */ |
76 | static const uint32_t staticOffsets[8]={ |
77 | 0x0000, /* ASCII for quoted tags */ |
78 | 0x0080, /* Latin - 1 Supplement (for access to punctuation) */ |
79 | 0x0100, /* Latin Extended-A */ |
80 | 0x0300, /* Combining Diacritical Marks */ |
81 | 0x2000, /* General Punctuation */ |
82 | 0x2080, /* Currency Symbols */ |
83 | 0x2100, /* Letterlike Symbols and Number Forms */ |
84 | 0x3000 /* CJK Symbols and punctuation */ |
85 | }; |
86 | |
87 | /* initial offsets for the 8 dynamic (sliding) windows */ |
88 | static const uint32_t initialDynamicOffsets[8]={ |
89 | 0x0080, /* Latin-1 */ |
90 | 0x00C0, /* Latin Extended A */ |
91 | 0x0400, /* Cyrillic */ |
92 | 0x0600, /* Arabic */ |
93 | 0x0900, /* Devanagari */ |
94 | 0x3040, /* Hiragana */ |
95 | 0x30A0, /* Katakana */ |
96 | 0xFF00 /* Fullwidth ASCII */ |
97 | }; |
98 | |
99 | /* Table of fixed predefined Offsets */ |
100 | static const uint32_t fixedOffsets[]={ |
101 | /* 0xF9 */ 0x00C0, /* Latin-1 Letters + half of Latin Extended A */ |
102 | /* 0xFA */ 0x0250, /* IPA extensions */ |
103 | /* 0xFB */ 0x0370, /* Greek */ |
104 | /* 0xFC */ 0x0530, /* Armenian */ |
105 | /* 0xFD */ 0x3040, /* Hiragana */ |
106 | /* 0xFE */ 0x30A0, /* Katakana */ |
107 | /* 0xFF */ 0xFF60 /* Halfwidth Katakana */ |
108 | }; |
109 | |
110 | /* state values */ |
111 | enum { |
112 | readCommand, |
113 | quotePairOne, |
114 | quotePairTwo, |
115 | quoteOne, |
116 | definePairOne, |
117 | definePairTwo, |
118 | defineOne |
119 | }; |
120 | |
121 | typedef struct SCSUData { |
122 | /* dynamic window offsets, initialize to default values from initialDynamicOffsets */ |
123 | uint32_t toUDynamicOffsets[8]; |
124 | uint32_t fromUDynamicOffsets[8]; |
125 | |
126 | /* state machine state - toUnicode */ |
127 | UBool toUIsSingleByteMode; |
128 | uint8_t toUState; |
129 | int8_t toUQuoteWindow, toUDynamicWindow; |
130 | uint8_t toUByteOne; |
131 | uint8_t toUPadding[3]; |
132 | |
133 | /* state machine state - fromUnicode */ |
134 | UBool fromUIsSingleByteMode; |
135 | int8_t fromUDynamicWindow; |
136 | |
137 | /* |
138 | * windowUse[] keeps track of the use of the dynamic windows: |
139 | * At nextWindowUseIndex there is the least recently used window, |
140 | * and the following windows (in a wrapping manner) are more and more |
141 | * recently used. |
142 | * At nextWindowUseIndex-1 there is the most recently used window. |
143 | */ |
144 | uint8_t locale; |
145 | int8_t nextWindowUseIndex; |
146 | int8_t windowUse[8]; |
147 | } SCSUData; |
148 | |
149 | static const int8_t initialWindowUse[8]={ 7, 0, 3, 2, 4, 5, 6, 1 }; |
150 | static const int8_t initialWindowUse_ja[8]={ 3, 2, 4, 1, 0, 7, 5, 6 }; |
151 | |
152 | enum { |
153 | lGeneric, l_ja |
154 | }; |
155 | |
156 | /* SCSU setup functions ----------------------------------------------------- */ |
157 | U_CDECL_BEGIN |
158 | static void U_CALLCONV |
159 | _SCSUReset(UConverter *cnv, UConverterResetChoice choice) { |
160 | SCSUData *scsu=(SCSUData *)cnv->extraInfo; |
161 | |
162 | if(choice<=UCNV_RESET_TO_UNICODE) { |
163 | /* reset toUnicode */ |
164 | uprv_memcpy(scsu->toUDynamicOffsets, initialDynamicOffsets, 32); |
165 | |
166 | scsu->toUIsSingleByteMode=true; |
167 | scsu->toUState=readCommand; |
168 | scsu->toUQuoteWindow=scsu->toUDynamicWindow=0; |
169 | scsu->toUByteOne=0; |
170 | |
171 | cnv->toULength=0; |
172 | } |
173 | if(choice!=UCNV_RESET_TO_UNICODE) { |
174 | /* reset fromUnicode */ |
175 | uprv_memcpy(scsu->fromUDynamicOffsets, initialDynamicOffsets, 32); |
176 | |
177 | scsu->fromUIsSingleByteMode=true; |
178 | scsu->fromUDynamicWindow=0; |
179 | |
180 | scsu->nextWindowUseIndex=0; |
181 | switch(scsu->locale) { |
182 | case l_ja: |
183 | uprv_memcpy(scsu->windowUse, initialWindowUse_ja, 8); |
184 | break; |
185 | default: |
186 | uprv_memcpy(scsu->windowUse, initialWindowUse, 8); |
187 | break; |
188 | } |
189 | |
190 | cnv->fromUChar32=0; |
191 | } |
192 | } |
193 | |
194 | static void U_CALLCONV |
195 | _SCSUOpen(UConverter *cnv, |
196 | UConverterLoadArgs *pArgs, |
197 | UErrorCode *pErrorCode) { |
198 | const char *locale=pArgs->locale; |
199 | if(pArgs->onlyTestIsLoadable) { |
200 | return; |
201 | } |
202 | cnv->extraInfo=uprv_malloc(sizeof(SCSUData)); |
203 | if(cnv->extraInfo!=nullptr) { |
204 | if(locale!=nullptr && locale[0]=='j' && locale[1]=='a' && (locale[2]==0 || locale[2]=='_')) { |
205 | ((SCSUData *)cnv->extraInfo)->locale=l_ja; |
206 | } else { |
207 | ((SCSUData *)cnv->extraInfo)->locale=lGeneric; |
208 | } |
209 | _SCSUReset(cnv, UCNV_RESET_BOTH); |
210 | } else { |
211 | *pErrorCode=U_MEMORY_ALLOCATION_ERROR; |
212 | } |
213 | |
214 | /* Set the substitution character U+fffd as a Unicode string. */ |
215 | cnv->subUChars[0]=0xfffd; |
216 | cnv->subCharLen=-1; |
217 | } |
218 | |
219 | static void U_CALLCONV |
220 | _SCSUClose(UConverter *cnv) { |
221 | if(cnv->extraInfo!=nullptr) { |
222 | if(!cnv->isExtraLocal) { |
223 | uprv_free(cnv->extraInfo); |
224 | } |
225 | cnv->extraInfo=nullptr; |
226 | } |
227 | } |
228 | |
229 | /* SCSU-to-Unicode conversion functions ------------------------------------- */ |
230 | |
231 | static void U_CALLCONV |
232 | _SCSUToUnicodeWithOffsets(UConverterToUnicodeArgs *pArgs, |
233 | UErrorCode *pErrorCode) { |
234 | UConverter *cnv; |
235 | SCSUData *scsu; |
236 | const uint8_t *source, *sourceLimit; |
237 | char16_t *target; |
238 | const char16_t *targetLimit; |
239 | int32_t *offsets; |
240 | UBool isSingleByteMode; |
241 | uint8_t state, byteOne; |
242 | int8_t quoteWindow, dynamicWindow; |
243 | |
244 | int32_t sourceIndex, nextSourceIndex; |
245 | |
246 | uint8_t b; |
247 | |
248 | /* set up the local pointers */ |
249 | cnv=pArgs->converter; |
250 | scsu=(SCSUData *)cnv->extraInfo; |
251 | |
252 | source=(const uint8_t *)pArgs->source; |
253 | sourceLimit=(const uint8_t *)pArgs->sourceLimit; |
254 | target=pArgs->target; |
255 | targetLimit=pArgs->targetLimit; |
256 | offsets=pArgs->offsets; |
257 | |
258 | /* get the state machine state */ |
259 | isSingleByteMode=scsu->toUIsSingleByteMode; |
260 | state=scsu->toUState; |
261 | quoteWindow=scsu->toUQuoteWindow; |
262 | dynamicWindow=scsu->toUDynamicWindow; |
263 | byteOne=scsu->toUByteOne; |
264 | |
265 | /* sourceIndex=-1 if the current character began in the previous buffer */ |
266 | sourceIndex=state==readCommand ? 0 : -1; |
267 | nextSourceIndex=0; |
268 | |
269 | /* |
270 | * conversion "loop" |
271 | * |
272 | * For performance, this is not a normal C loop. |
273 | * Instead, there are two code blocks for the two SCSU modes. |
274 | * The function branches to either one, and a change of the mode is done with a goto to |
275 | * the other branch. |
276 | * |
277 | * Each branch has two conventional loops: |
278 | * - a fast-path loop for the most common codes in the mode |
279 | * - a loop for all other codes in the mode |
280 | * When the fast-path runs into a code that it cannot handle, its loop ends and it |
281 | * runs into the following loop to handle the other codes. |
282 | * The end of the input or output buffer is also handled by the slower loop. |
283 | * The slow loop jumps (goto) to the fast-path loop again as soon as possible. |
284 | * |
285 | * The callback handling is done by returning with an error code. |
286 | * The conversion framework actually calls the callback function. |
287 | */ |
288 | if(isSingleByteMode) { |
289 | /* fast path for single-byte mode */ |
290 | if(state==readCommand) { |
291 | fastSingle: |
292 | while(source<sourceLimit && target<targetLimit && (b=*source)>=0x20) { |
293 | ++source; |
294 | ++nextSourceIndex; |
295 | if(b<=0x7f) { |
296 | /* write US-ASCII graphic character or DEL */ |
297 | *target++=(char16_t)b; |
298 | if(offsets!=nullptr) { |
299 | *offsets++=sourceIndex; |
300 | } |
301 | } else { |
302 | /* write from dynamic window */ |
303 | uint32_t c=scsu->toUDynamicOffsets[dynamicWindow]+(b&0x7f); |
304 | if(c<=0xffff) { |
305 | *target++=(char16_t)c; |
306 | if(offsets!=nullptr) { |
307 | *offsets++=sourceIndex; |
308 | } |
309 | } else { |
310 | /* output surrogate pair */ |
311 | *target++=(char16_t)(0xd7c0+(c>>10)); |
312 | if(target<targetLimit) { |
313 | *target++=(char16_t)(0xdc00|(c&0x3ff)); |
314 | if(offsets!=nullptr) { |
315 | *offsets++=sourceIndex; |
316 | *offsets++=sourceIndex; |
317 | } |
318 | } else { |
319 | /* target overflow */ |
320 | if(offsets!=nullptr) { |
321 | *offsets++=sourceIndex; |
322 | } |
323 | cnv->UCharErrorBuffer[0]=(char16_t)(0xdc00|(c&0x3ff)); |
324 | cnv->UCharErrorBufferLength=1; |
325 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
326 | goto endloop; |
327 | } |
328 | } |
329 | } |
330 | sourceIndex=nextSourceIndex; |
331 | } |
332 | } |
333 | |
334 | /* normal state machine for single-byte mode, minus handling for what fastSingle covers */ |
335 | singleByteMode: |
336 | while(source<sourceLimit) { |
337 | if(target>=targetLimit) { |
338 | /* target is full */ |
339 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
340 | break; |
341 | } |
342 | b=*source++; |
343 | ++nextSourceIndex; |
344 | switch(state) { |
345 | case readCommand: |
346 | /* redundant conditions are commented out */ |
347 | /* here: b<0x20 because otherwise we would be in fastSingle */ |
348 | if((1UL<<b)&0x2601 /* binary 0010 0110 0000 0001, check for b==0xd || b==0xa || b==9 || b==0 */) { |
349 | /* CR/LF/TAB/NUL */ |
350 | *target++=(char16_t)b; |
351 | if(offsets!=nullptr) { |
352 | *offsets++=sourceIndex; |
353 | } |
354 | sourceIndex=nextSourceIndex; |
355 | goto fastSingle; |
356 | } else if(SC0<=b) { |
357 | if(b<=SC7) { |
358 | dynamicWindow=(int8_t)(b-SC0); |
359 | sourceIndex=nextSourceIndex; |
360 | goto fastSingle; |
361 | } else /* if(SD0<=b && b<=SD7) */ { |
362 | dynamicWindow=(int8_t)(b-SD0); |
363 | state=defineOne; |
364 | } |
365 | } else if(/* SQ0<=b && */ b<=SQ7) { |
366 | quoteWindow=(int8_t)(b-SQ0); |
367 | state=quoteOne; |
368 | } else if(b==SDX) { |
369 | state=definePairOne; |
370 | } else if(b==SQU) { |
371 | state=quotePairOne; |
372 | } else if(b==SCU) { |
373 | sourceIndex=nextSourceIndex; |
374 | isSingleByteMode=false; |
375 | goto fastUnicode; |
376 | } else /* Srs */ { |
377 | /* callback(illegal) */ |
378 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
379 | cnv->toUBytes[0]=b; |
380 | cnv->toULength=1; |
381 | goto endloop; |
382 | } |
383 | |
384 | /* store the first byte of a multibyte sequence in toUBytes[] */ |
385 | cnv->toUBytes[0]=b; |
386 | cnv->toULength=1; |
387 | break; |
388 | case quotePairOne: |
389 | byteOne=b; |
390 | cnv->toUBytes[1]=b; |
391 | cnv->toULength=2; |
392 | state=quotePairTwo; |
393 | break; |
394 | case quotePairTwo: |
395 | *target++=(char16_t)((byteOne<<8)|b); |
396 | if(offsets!=nullptr) { |
397 | *offsets++=sourceIndex; |
398 | } |
399 | sourceIndex=nextSourceIndex; |
400 | state=readCommand; |
401 | goto fastSingle; |
402 | case quoteOne: |
403 | if(b<0x80) { |
404 | /* all static offsets are in the BMP */ |
405 | *target++=(char16_t)(staticOffsets[quoteWindow]+b); |
406 | if(offsets!=nullptr) { |
407 | *offsets++=sourceIndex; |
408 | } |
409 | } else { |
410 | /* write from dynamic window */ |
411 | uint32_t c=scsu->toUDynamicOffsets[quoteWindow]+(b&0x7f); |
412 | if(c<=0xffff) { |
413 | *target++=(char16_t)c; |
414 | if(offsets!=nullptr) { |
415 | *offsets++=sourceIndex; |
416 | } |
417 | } else { |
418 | /* output surrogate pair */ |
419 | *target++=(char16_t)(0xd7c0+(c>>10)); |
420 | if(target<targetLimit) { |
421 | *target++=(char16_t)(0xdc00|(c&0x3ff)); |
422 | if(offsets!=nullptr) { |
423 | *offsets++=sourceIndex; |
424 | *offsets++=sourceIndex; |
425 | } |
426 | } else { |
427 | /* target overflow */ |
428 | if(offsets!=nullptr) { |
429 | *offsets++=sourceIndex; |
430 | } |
431 | cnv->UCharErrorBuffer[0]=(char16_t)(0xdc00|(c&0x3ff)); |
432 | cnv->UCharErrorBufferLength=1; |
433 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
434 | goto endloop; |
435 | } |
436 | } |
437 | } |
438 | sourceIndex=nextSourceIndex; |
439 | state=readCommand; |
440 | goto fastSingle; |
441 | case definePairOne: |
442 | dynamicWindow=(int8_t)((b>>5)&7); |
443 | byteOne=(uint8_t)(b&0x1f); |
444 | cnv->toUBytes[1]=b; |
445 | cnv->toULength=2; |
446 | state=definePairTwo; |
447 | break; |
448 | case definePairTwo: |
449 | scsu->toUDynamicOffsets[dynamicWindow]=0x10000+(byteOne<<15UL | b<<7UL); |
450 | sourceIndex=nextSourceIndex; |
451 | state=readCommand; |
452 | goto fastSingle; |
453 | case defineOne: |
454 | if(b==0) { |
455 | /* callback(illegal): Reserved window offset value 0 */ |
456 | cnv->toUBytes[1]=b; |
457 | cnv->toULength=2; |
458 | goto endloop; |
459 | } else if(b<gapThreshold) { |
460 | scsu->toUDynamicOffsets[dynamicWindow]=b<<7UL; |
461 | } else if((uint8_t)(b-gapThreshold)<(reservedStart-gapThreshold)) { |
462 | scsu->toUDynamicOffsets[dynamicWindow]=(b<<7UL)+gapOffset; |
463 | } else if(b>=fixedThreshold) { |
464 | scsu->toUDynamicOffsets[dynamicWindow]=fixedOffsets[b-fixedThreshold]; |
465 | } else { |
466 | /* callback(illegal): Reserved window offset value 0xa8..0xf8 */ |
467 | cnv->toUBytes[1]=b; |
468 | cnv->toULength=2; |
469 | goto endloop; |
470 | } |
471 | sourceIndex=nextSourceIndex; |
472 | state=readCommand; |
473 | goto fastSingle; |
474 | } |
475 | } |
476 | } else { |
477 | /* fast path for Unicode mode */ |
478 | if(state==readCommand) { |
479 | fastUnicode: |
480 | while(source+1<sourceLimit && target<targetLimit && (uint8_t)((b=*source)-UC0)>(Urs-UC0)) { |
481 | *target++=(char16_t)((b<<8)|source[1]); |
482 | if(offsets!=nullptr) { |
483 | *offsets++=sourceIndex; |
484 | } |
485 | sourceIndex=nextSourceIndex; |
486 | nextSourceIndex+=2; |
487 | source+=2; |
488 | } |
489 | } |
490 | |
491 | /* normal state machine for Unicode mode */ |
492 | /* unicodeByteMode: */ |
493 | while(source<sourceLimit) { |
494 | if(target>=targetLimit) { |
495 | /* target is full */ |
496 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
497 | break; |
498 | } |
499 | b=*source++; |
500 | ++nextSourceIndex; |
501 | switch(state) { |
502 | case readCommand: |
503 | if((uint8_t)(b-UC0)>(Urs-UC0)) { |
504 | byteOne=b; |
505 | cnv->toUBytes[0]=b; |
506 | cnv->toULength=1; |
507 | state=quotePairTwo; |
508 | } else if(/* UC0<=b && */ b<=UC7) { |
509 | dynamicWindow=(int8_t)(b-UC0); |
510 | sourceIndex=nextSourceIndex; |
511 | isSingleByteMode=true; |
512 | goto fastSingle; |
513 | } else if(/* UD0<=b && */ b<=UD7) { |
514 | dynamicWindow=(int8_t)(b-UD0); |
515 | isSingleByteMode=true; |
516 | cnv->toUBytes[0]=b; |
517 | cnv->toULength=1; |
518 | state=defineOne; |
519 | goto singleByteMode; |
520 | } else if(b==UDX) { |
521 | isSingleByteMode=true; |
522 | cnv->toUBytes[0]=b; |
523 | cnv->toULength=1; |
524 | state=definePairOne; |
525 | goto singleByteMode; |
526 | } else if(b==UQU) { |
527 | cnv->toUBytes[0]=b; |
528 | cnv->toULength=1; |
529 | state=quotePairOne; |
530 | } else /* Urs */ { |
531 | /* callback(illegal) */ |
532 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
533 | cnv->toUBytes[0]=b; |
534 | cnv->toULength=1; |
535 | goto endloop; |
536 | } |
537 | break; |
538 | case quotePairOne: |
539 | byteOne=b; |
540 | cnv->toUBytes[1]=b; |
541 | cnv->toULength=2; |
542 | state=quotePairTwo; |
543 | break; |
544 | case quotePairTwo: |
545 | *target++=(char16_t)((byteOne<<8)|b); |
546 | if(offsets!=nullptr) { |
547 | *offsets++=sourceIndex; |
548 | } |
549 | sourceIndex=nextSourceIndex; |
550 | state=readCommand; |
551 | goto fastUnicode; |
552 | } |
553 | } |
554 | } |
555 | endloop: |
556 | |
557 | /* set the converter state back into UConverter */ |
558 | if(U_FAILURE(*pErrorCode) && *pErrorCode!=U_BUFFER_OVERFLOW_ERROR) { |
559 | /* reset to deal with the next character */ |
560 | state=readCommand; |
561 | } else if(state==readCommand) { |
562 | /* not in a multi-byte sequence, reset toULength */ |
563 | cnv->toULength=0; |
564 | } |
565 | scsu->toUIsSingleByteMode=isSingleByteMode; |
566 | scsu->toUState=state; |
567 | scsu->toUQuoteWindow=quoteWindow; |
568 | scsu->toUDynamicWindow=dynamicWindow; |
569 | scsu->toUByteOne=byteOne; |
570 | |
571 | /* write back the updated pointers */ |
572 | pArgs->source=(const char *)source; |
573 | pArgs->target=target; |
574 | pArgs->offsets=offsets; |
575 | return; |
576 | } |
577 | |
578 | /* |
579 | * Identical to _SCSUToUnicodeWithOffsets but without offset handling. |
580 | * If a change is made in the original function, then either |
581 | * change this function the same way or |
582 | * re-copy the original function and remove the variables |
583 | * offsets, sourceIndex, and nextSourceIndex. |
584 | */ |
585 | static void U_CALLCONV |
586 | _SCSUToUnicode(UConverterToUnicodeArgs *pArgs, |
587 | UErrorCode *pErrorCode) { |
588 | UConverter *cnv; |
589 | SCSUData *scsu; |
590 | const uint8_t *source, *sourceLimit; |
591 | char16_t *target; |
592 | const char16_t *targetLimit; |
593 | UBool isSingleByteMode; |
594 | uint8_t state, byteOne; |
595 | int8_t quoteWindow, dynamicWindow; |
596 | |
597 | uint8_t b; |
598 | |
599 | /* set up the local pointers */ |
600 | cnv=pArgs->converter; |
601 | scsu=(SCSUData *)cnv->extraInfo; |
602 | |
603 | source=(const uint8_t *)pArgs->source; |
604 | sourceLimit=(const uint8_t *)pArgs->sourceLimit; |
605 | target=pArgs->target; |
606 | targetLimit=pArgs->targetLimit; |
607 | |
608 | /* get the state machine state */ |
609 | isSingleByteMode=scsu->toUIsSingleByteMode; |
610 | state=scsu->toUState; |
611 | quoteWindow=scsu->toUQuoteWindow; |
612 | dynamicWindow=scsu->toUDynamicWindow; |
613 | byteOne=scsu->toUByteOne; |
614 | |
615 | /* |
616 | * conversion "loop" |
617 | * |
618 | * For performance, this is not a normal C loop. |
619 | * Instead, there are two code blocks for the two SCSU modes. |
620 | * The function branches to either one, and a change of the mode is done with a goto to |
621 | * the other branch. |
622 | * |
623 | * Each branch has two conventional loops: |
624 | * - a fast-path loop for the most common codes in the mode |
625 | * - a loop for all other codes in the mode |
626 | * When the fast-path runs into a code that it cannot handle, its loop ends and it |
627 | * runs into the following loop to handle the other codes. |
628 | * The end of the input or output buffer is also handled by the slower loop. |
629 | * The slow loop jumps (goto) to the fast-path loop again as soon as possible. |
630 | * |
631 | * The callback handling is done by returning with an error code. |
632 | * The conversion framework actually calls the callback function. |
633 | */ |
634 | if(isSingleByteMode) { |
635 | /* fast path for single-byte mode */ |
636 | if(state==readCommand) { |
637 | fastSingle: |
638 | while(source<sourceLimit && target<targetLimit && (b=*source)>=0x20) { |
639 | ++source; |
640 | if(b<=0x7f) { |
641 | /* write US-ASCII graphic character or DEL */ |
642 | *target++=(char16_t)b; |
643 | } else { |
644 | /* write from dynamic window */ |
645 | uint32_t c=scsu->toUDynamicOffsets[dynamicWindow]+(b&0x7f); |
646 | if(c<=0xffff) { |
647 | *target++=(char16_t)c; |
648 | } else { |
649 | /* output surrogate pair */ |
650 | *target++=(char16_t)(0xd7c0+(c>>10)); |
651 | if(target<targetLimit) { |
652 | *target++=(char16_t)(0xdc00|(c&0x3ff)); |
653 | } else { |
654 | /* target overflow */ |
655 | cnv->UCharErrorBuffer[0]=(char16_t)(0xdc00|(c&0x3ff)); |
656 | cnv->UCharErrorBufferLength=1; |
657 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
658 | goto endloop; |
659 | } |
660 | } |
661 | } |
662 | } |
663 | } |
664 | |
665 | /* normal state machine for single-byte mode, minus handling for what fastSingle covers */ |
666 | singleByteMode: |
667 | while(source<sourceLimit) { |
668 | if(target>=targetLimit) { |
669 | /* target is full */ |
670 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
671 | break; |
672 | } |
673 | b=*source++; |
674 | switch(state) { |
675 | case readCommand: |
676 | /* redundant conditions are commented out */ |
677 | /* here: b<0x20 because otherwise we would be in fastSingle */ |
678 | if((1UL<<b)&0x2601 /* binary 0010 0110 0000 0001, check for b==0xd || b==0xa || b==9 || b==0 */) { |
679 | /* CR/LF/TAB/NUL */ |
680 | *target++=(char16_t)b; |
681 | goto fastSingle; |
682 | } else if(SC0<=b) { |
683 | if(b<=SC7) { |
684 | dynamicWindow=(int8_t)(b-SC0); |
685 | goto fastSingle; |
686 | } else /* if(SD0<=b && b<=SD7) */ { |
687 | dynamicWindow=(int8_t)(b-SD0); |
688 | state=defineOne; |
689 | } |
690 | } else if(/* SQ0<=b && */ b<=SQ7) { |
691 | quoteWindow=(int8_t)(b-SQ0); |
692 | state=quoteOne; |
693 | } else if(b==SDX) { |
694 | state=definePairOne; |
695 | } else if(b==SQU) { |
696 | state=quotePairOne; |
697 | } else if(b==SCU) { |
698 | isSingleByteMode=false; |
699 | goto fastUnicode; |
700 | } else /* Srs */ { |
701 | /* callback(illegal) */ |
702 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
703 | cnv->toUBytes[0]=b; |
704 | cnv->toULength=1; |
705 | goto endloop; |
706 | } |
707 | |
708 | /* store the first byte of a multibyte sequence in toUBytes[] */ |
709 | cnv->toUBytes[0]=b; |
710 | cnv->toULength=1; |
711 | break; |
712 | case quotePairOne: |
713 | byteOne=b; |
714 | cnv->toUBytes[1]=b; |
715 | cnv->toULength=2; |
716 | state=quotePairTwo; |
717 | break; |
718 | case quotePairTwo: |
719 | *target++=(char16_t)((byteOne<<8)|b); |
720 | state=readCommand; |
721 | goto fastSingle; |
722 | case quoteOne: |
723 | if(b<0x80) { |
724 | /* all static offsets are in the BMP */ |
725 | *target++=(char16_t)(staticOffsets[quoteWindow]+b); |
726 | } else { |
727 | /* write from dynamic window */ |
728 | uint32_t c=scsu->toUDynamicOffsets[quoteWindow]+(b&0x7f); |
729 | if(c<=0xffff) { |
730 | *target++=(char16_t)c; |
731 | } else { |
732 | /* output surrogate pair */ |
733 | *target++=(char16_t)(0xd7c0+(c>>10)); |
734 | if(target<targetLimit) { |
735 | *target++=(char16_t)(0xdc00|(c&0x3ff)); |
736 | } else { |
737 | /* target overflow */ |
738 | cnv->UCharErrorBuffer[0]=(char16_t)(0xdc00|(c&0x3ff)); |
739 | cnv->UCharErrorBufferLength=1; |
740 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
741 | goto endloop; |
742 | } |
743 | } |
744 | } |
745 | state=readCommand; |
746 | goto fastSingle; |
747 | case definePairOne: |
748 | dynamicWindow=(int8_t)((b>>5)&7); |
749 | byteOne=(uint8_t)(b&0x1f); |
750 | cnv->toUBytes[1]=b; |
751 | cnv->toULength=2; |
752 | state=definePairTwo; |
753 | break; |
754 | case definePairTwo: |
755 | scsu->toUDynamicOffsets[dynamicWindow]=0x10000+(byteOne<<15UL | b<<7UL); |
756 | state=readCommand; |
757 | goto fastSingle; |
758 | case defineOne: |
759 | if(b==0) { |
760 | /* callback(illegal): Reserved window offset value 0 */ |
761 | cnv->toUBytes[1]=b; |
762 | cnv->toULength=2; |
763 | goto endloop; |
764 | } else if(b<gapThreshold) { |
765 | scsu->toUDynamicOffsets[dynamicWindow]=b<<7UL; |
766 | } else if((uint8_t)(b-gapThreshold)<(reservedStart-gapThreshold)) { |
767 | scsu->toUDynamicOffsets[dynamicWindow]=(b<<7UL)+gapOffset; |
768 | } else if(b>=fixedThreshold) { |
769 | scsu->toUDynamicOffsets[dynamicWindow]=fixedOffsets[b-fixedThreshold]; |
770 | } else { |
771 | /* callback(illegal): Reserved window offset value 0xa8..0xf8 */ |
772 | cnv->toUBytes[1]=b; |
773 | cnv->toULength=2; |
774 | goto endloop; |
775 | } |
776 | state=readCommand; |
777 | goto fastSingle; |
778 | } |
779 | } |
780 | } else { |
781 | /* fast path for Unicode mode */ |
782 | if(state==readCommand) { |
783 | fastUnicode: |
784 | while(source+1<sourceLimit && target<targetLimit && (uint8_t)((b=*source)-UC0)>(Urs-UC0)) { |
785 | *target++=(char16_t)((b<<8)|source[1]); |
786 | source+=2; |
787 | } |
788 | } |
789 | |
790 | /* normal state machine for Unicode mode */ |
791 | /* unicodeByteMode: */ |
792 | while(source<sourceLimit) { |
793 | if(target>=targetLimit) { |
794 | /* target is full */ |
795 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
796 | break; |
797 | } |
798 | b=*source++; |
799 | switch(state) { |
800 | case readCommand: |
801 | if((uint8_t)(b-UC0)>(Urs-UC0)) { |
802 | byteOne=b; |
803 | cnv->toUBytes[0]=b; |
804 | cnv->toULength=1; |
805 | state=quotePairTwo; |
806 | } else if(/* UC0<=b && */ b<=UC7) { |
807 | dynamicWindow=(int8_t)(b-UC0); |
808 | isSingleByteMode=true; |
809 | goto fastSingle; |
810 | } else if(/* UD0<=b && */ b<=UD7) { |
811 | dynamicWindow=(int8_t)(b-UD0); |
812 | isSingleByteMode=true; |
813 | cnv->toUBytes[0]=b; |
814 | cnv->toULength=1; |
815 | state=defineOne; |
816 | goto singleByteMode; |
817 | } else if(b==UDX) { |
818 | isSingleByteMode=true; |
819 | cnv->toUBytes[0]=b; |
820 | cnv->toULength=1; |
821 | state=definePairOne; |
822 | goto singleByteMode; |
823 | } else if(b==UQU) { |
824 | cnv->toUBytes[0]=b; |
825 | cnv->toULength=1; |
826 | state=quotePairOne; |
827 | } else /* Urs */ { |
828 | /* callback(illegal) */ |
829 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
830 | cnv->toUBytes[0]=b; |
831 | cnv->toULength=1; |
832 | goto endloop; |
833 | } |
834 | break; |
835 | case quotePairOne: |
836 | byteOne=b; |
837 | cnv->toUBytes[1]=b; |
838 | cnv->toULength=2; |
839 | state=quotePairTwo; |
840 | break; |
841 | case quotePairTwo: |
842 | *target++=(char16_t)((byteOne<<8)|b); |
843 | state=readCommand; |
844 | goto fastUnicode; |
845 | } |
846 | } |
847 | } |
848 | endloop: |
849 | |
850 | /* set the converter state back into UConverter */ |
851 | if(U_FAILURE(*pErrorCode) && *pErrorCode!=U_BUFFER_OVERFLOW_ERROR) { |
852 | /* reset to deal with the next character */ |
853 | state=readCommand; |
854 | } else if(state==readCommand) { |
855 | /* not in a multi-byte sequence, reset toULength */ |
856 | cnv->toULength=0; |
857 | } |
858 | scsu->toUIsSingleByteMode=isSingleByteMode; |
859 | scsu->toUState=state; |
860 | scsu->toUQuoteWindow=quoteWindow; |
861 | scsu->toUDynamicWindow=dynamicWindow; |
862 | scsu->toUByteOne=byteOne; |
863 | |
864 | /* write back the updated pointers */ |
865 | pArgs->source=(const char *)source; |
866 | pArgs->target=target; |
867 | return; |
868 | } |
869 | U_CDECL_END |
870 | /* SCSU-from-Unicode conversion functions ----------------------------------- */ |
871 | |
872 | /* |
873 | * This SCSU Encoder is fairly simple but uses all SCSU commands to achieve |
874 | * reasonable results. The lookahead is minimal. |
875 | * Many cases are simple: |
876 | * A character fits directly into the current mode, a dynamic or static window, |
877 | * or is not compressible. These cases are tested first. |
878 | * Real compression heuristics are applied to the rest, in code branches for |
879 | * single/Unicode mode and BMP/supplementary code points. |
880 | * The heuristics used here are extremely simple. |
881 | */ |
882 | |
883 | /* get the number of the window that this character is in, or -1 */ |
884 | static int8_t |
885 | getWindow(const uint32_t offsets[8], uint32_t c) { |
886 | int i; |
887 | for(i=0; i<8; ++i) { |
888 | if((uint32_t)(c-offsets[i])<=0x7f) { |
889 | return (int8_t)(i); |
890 | } |
891 | } |
892 | return -1; |
893 | } |
894 | |
895 | /* is the character in the dynamic window starting at the offset, or in the direct-encoded range? */ |
896 | static UBool |
897 | isInOffsetWindowOrDirect(uint32_t offset, uint32_t c) { |
898 | return (UBool)(c<=offset+0x7f && |
899 | (c>=offset || (c<=0x7f && |
900 | (c>=0x20 || (1UL<<c)&0x2601)))); |
901 | /* binary 0010 0110 0000 0001, |
902 | check for b==0xd || b==0xa || b==9 || b==0 */ |
903 | } |
904 | |
905 | /* |
906 | * getNextDynamicWindow returns the next dynamic window to be redefined |
907 | */ |
908 | static int8_t |
909 | getNextDynamicWindow(SCSUData *scsu) { |
910 | int8_t window=scsu->windowUse[scsu->nextWindowUseIndex]; |
911 | if(++scsu->nextWindowUseIndex==8) { |
912 | scsu->nextWindowUseIndex=0; |
913 | } |
914 | return window; |
915 | } |
916 | |
917 | /* |
918 | * useDynamicWindow() adjusts |
919 | * windowUse[] and nextWindowUseIndex for the algorithm to choose |
920 | * the next dynamic window to be defined; |
921 | * a subclass may override it and provide its own algorithm. |
922 | */ |
923 | static void |
924 | useDynamicWindow(SCSUData *scsu, int8_t window) { |
925 | /* |
926 | * move the existing window, which just became the most recently used one, |
927 | * up in windowUse[] to nextWindowUseIndex-1 |
928 | */ |
929 | |
930 | /* first, find the index of the window - backwards to favor the more recently used windows */ |
931 | int i, j; |
932 | |
933 | i=scsu->nextWindowUseIndex; |
934 | do { |
935 | if(--i<0) { |
936 | i=7; |
937 | } |
938 | } while(scsu->windowUse[i]!=window); |
939 | |
940 | /* now copy each windowUse[i+1] to [i] */ |
941 | j=i+1; |
942 | if(j==8) { |
943 | j=0; |
944 | } |
945 | while(j!=scsu->nextWindowUseIndex) { |
946 | scsu->windowUse[i]=scsu->windowUse[j]; |
947 | i=j; |
948 | if(++j==8) { j=0; } |
949 | } |
950 | |
951 | /* finally, set the window into the most recently used index */ |
952 | scsu->windowUse[i]=window; |
953 | } |
954 | |
955 | /* |
956 | * calculate the offset and the code for a dynamic window that contains the character |
957 | * takes fixed offsets into account |
958 | * the offset of the window is stored in the offset variable, |
959 | * the code is returned |
960 | * |
961 | * return offset code: -1 none <=0xff code for SDn/UDn else code for SDX/UDX, subtract 0x200 to get the true code |
962 | */ |
963 | static int |
964 | getDynamicOffset(uint32_t c, uint32_t *pOffset) { |
965 | int i; |
966 | |
967 | for(i=0; i<7; ++i) { |
968 | if((uint32_t)(c-fixedOffsets[i])<=0x7f) { |
969 | *pOffset=fixedOffsets[i]; |
970 | return 0xf9+i; |
971 | } |
972 | } |
973 | |
974 | if(c<0x80) { |
975 | /* No dynamic window for US-ASCII. */ |
976 | return -1; |
977 | } else if(c<0x3400 || |
978 | (uint32_t)(c-0x10000)<(0x14000-0x10000) || |
979 | (uint32_t)(c-0x1d000)<=(0x1ffff-0x1d000) |
980 | ) { |
981 | /* This character is in a code range for a "small", i.e., reasonably windowable, script. */ |
982 | *pOffset=c&0x7fffff80; |
983 | return (int)(c>>7); |
984 | } else if(0xe000<=c && c!=0xfeff && c<0xfff0) { |
985 | /* For these characters we need to take the gapOffset into account. */ |
986 | *pOffset=c&0x7fffff80; |
987 | return (int)((c-gapOffset)>>7); |
988 | } else { |
989 | return -1; |
990 | } |
991 | } |
992 | U_CDECL_BEGIN |
993 | /* |
994 | * Idea for compression: |
995 | * - save SCSUData and other state before really starting work |
996 | * - at endloop, see if compression could be better with just unicode mode |
997 | * - don't do this if a callback has been called |
998 | * - if unicode mode would be smaller, then override the results with it - may need SCU at the beginning |
999 | * - different buffer handling! |
1000 | * |
1001 | * Drawback or need for corrective handling: |
1002 | * it is desirable to encode U+feff as SQU fe ff for the SCSU signature, and |
1003 | * it is desirable to start a document in US-ASCII/Latin-1 for as long as possible |
1004 | * not only for compression but also for HTML/XML documents with following charset/encoding announcers. |
1005 | * |
1006 | * How to achieve both? |
1007 | * - Only replace the result after an SDX or SCU? |
1008 | */ |
1009 | |
1010 | static void U_CALLCONV |
1011 | _SCSUFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, |
1012 | UErrorCode *pErrorCode) { |
1013 | UConverter *cnv; |
1014 | SCSUData *scsu; |
1015 | const char16_t *source, *sourceLimit; |
1016 | uint8_t *target; |
1017 | int32_t targetCapacity; |
1018 | int32_t *offsets; |
1019 | |
1020 | UBool isSingleByteMode; |
1021 | uint8_t dynamicWindow; |
1022 | uint32_t currentOffset; |
1023 | |
1024 | uint32_t c, delta; |
1025 | |
1026 | int32_t sourceIndex, nextSourceIndex; |
1027 | |
1028 | int32_t length; |
1029 | |
1030 | /* variables for compression heuristics */ |
1031 | uint32_t offset; |
1032 | char16_t lead, trail; |
1033 | int code; |
1034 | int8_t window; |
1035 | |
1036 | /* set up the local pointers */ |
1037 | cnv=pArgs->converter; |
1038 | scsu=(SCSUData *)cnv->extraInfo; |
1039 | |
1040 | /* set up the local pointers */ |
1041 | source=pArgs->source; |
1042 | sourceLimit=pArgs->sourceLimit; |
1043 | target=(uint8_t *)pArgs->target; |
1044 | targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); |
1045 | offsets=pArgs->offsets; |
1046 | |
1047 | /* get the state machine state */ |
1048 | isSingleByteMode=scsu->fromUIsSingleByteMode; |
1049 | dynamicWindow=scsu->fromUDynamicWindow; |
1050 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1051 | |
1052 | c=cnv->fromUChar32; |
1053 | |
1054 | /* sourceIndex=-1 if the current character began in the previous buffer */ |
1055 | sourceIndex= c==0 ? 0 : -1; |
1056 | nextSourceIndex=0; |
1057 | |
1058 | /* similar conversion "loop" as in toUnicode */ |
1059 | loop: |
1060 | if(isSingleByteMode) { |
1061 | if(c!=0 && targetCapacity>0) { |
1062 | goto getTrailSingle; |
1063 | } |
1064 | |
1065 | /* state machine for single-byte mode */ |
1066 | /* singleByteMode: */ |
1067 | while(source<sourceLimit) { |
1068 | if(targetCapacity<=0) { |
1069 | /* target is full */ |
1070 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1071 | break; |
1072 | } |
1073 | c=*source++; |
1074 | ++nextSourceIndex; |
1075 | |
1076 | if((c-0x20)<=0x5f) { |
1077 | /* pass US-ASCII graphic character through */ |
1078 | *target++=(uint8_t)c; |
1079 | if(offsets!=nullptr) { |
1080 | *offsets++=sourceIndex; |
1081 | } |
1082 | --targetCapacity; |
1083 | } else if(c<0x20) { |
1084 | if((1UL<<c)&0x2601 /* binary 0010 0110 0000 0001, check for b==0xd || b==0xa || b==9 || b==0 */) { |
1085 | /* CR/LF/TAB/NUL */ |
1086 | *target++=(uint8_t)c; |
1087 | if(offsets!=nullptr) { |
1088 | *offsets++=sourceIndex; |
1089 | } |
1090 | --targetCapacity; |
1091 | } else { |
1092 | /* quote C0 control character */ |
1093 | c|=SQ0<<8; |
1094 | length=2; |
1095 | goto outputBytes; |
1096 | } |
1097 | } else if((delta=c-currentOffset)<=0x7f) { |
1098 | /* use the current dynamic window */ |
1099 | *target++=(uint8_t)(delta|0x80); |
1100 | if(offsets!=nullptr) { |
1101 | *offsets++=sourceIndex; |
1102 | } |
1103 | --targetCapacity; |
1104 | } else if(U16_IS_SURROGATE(c)) { |
1105 | if(U16_IS_SURROGATE_LEAD(c)) { |
1106 | getTrailSingle: |
1107 | lead=(char16_t)c; |
1108 | if(source<sourceLimit) { |
1109 | /* test the following code unit */ |
1110 | trail=*source; |
1111 | if(U16_IS_TRAIL(trail)) { |
1112 | ++source; |
1113 | ++nextSourceIndex; |
1114 | c=U16_GET_SUPPLEMENTARY(c, trail); |
1115 | /* convert this surrogate code point */ |
1116 | /* exit this condition tree */ |
1117 | } else { |
1118 | /* this is an unmatched lead code unit (1st surrogate) */ |
1119 | /* callback(illegal) */ |
1120 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1121 | goto endloop; |
1122 | } |
1123 | } else { |
1124 | /* no more input */ |
1125 | break; |
1126 | } |
1127 | } else { |
1128 | /* this is an unmatched trail code unit (2nd surrogate) */ |
1129 | /* callback(illegal) */ |
1130 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1131 | goto endloop; |
1132 | } |
1133 | |
1134 | /* compress supplementary character U+10000..U+10ffff */ |
1135 | if((delta=c-currentOffset)<=0x7f) { |
1136 | /* use the current dynamic window */ |
1137 | *target++=(uint8_t)(delta|0x80); |
1138 | if(offsets!=nullptr) { |
1139 | *offsets++=sourceIndex; |
1140 | } |
1141 | --targetCapacity; |
1142 | } else if((window=getWindow(scsu->fromUDynamicOffsets, c))>=0) { |
1143 | /* there is a dynamic window that contains this character, change to it */ |
1144 | dynamicWindow=window; |
1145 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1146 | useDynamicWindow(scsu, dynamicWindow); |
1147 | c=((uint32_t)(SC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1148 | length=2; |
1149 | goto outputBytes; |
1150 | } else if((code=getDynamicOffset(c, &offset))>=0) { |
1151 | /* might check if there are more characters in this window to come */ |
1152 | /* define an extended window with this character */ |
1153 | code-=0x200; |
1154 | dynamicWindow=getNextDynamicWindow(scsu); |
1155 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1156 | useDynamicWindow(scsu, dynamicWindow); |
1157 | c=((uint32_t)SDX<<24)|((uint32_t)dynamicWindow<<21)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1158 | length=4; |
1159 | goto outputBytes; |
1160 | } else { |
1161 | /* change to Unicode mode and output this (lead, trail) pair */ |
1162 | isSingleByteMode=false; |
1163 | *target++=(uint8_t)SCU; |
1164 | if(offsets!=nullptr) { |
1165 | *offsets++=sourceIndex; |
1166 | } |
1167 | --targetCapacity; |
1168 | c=((uint32_t)lead<<16)|trail; |
1169 | length=4; |
1170 | goto outputBytes; |
1171 | } |
1172 | } else if(c<0xa0) { |
1173 | /* quote C1 control character */ |
1174 | c=(c&0x7f)|(SQ0+1)<<8; /* SQ0+1==SQ1 */ |
1175 | length=2; |
1176 | goto outputBytes; |
1177 | } else if(c==0xfeff || c>=0xfff0) { |
1178 | /* quote signature character=byte order mark and specials */ |
1179 | c|=SQU<<16; |
1180 | length=3; |
1181 | goto outputBytes; |
1182 | } else { |
1183 | /* compress all other BMP characters */ |
1184 | if((window=getWindow(scsu->fromUDynamicOffsets, c))>=0) { |
1185 | /* there is a window defined that contains this character - switch to it or quote from it? */ |
1186 | if(source>=sourceLimit || isInOffsetWindowOrDirect(scsu->fromUDynamicOffsets[window], *source)) { |
1187 | /* change to dynamic window */ |
1188 | dynamicWindow=window; |
1189 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1190 | useDynamicWindow(scsu, dynamicWindow); |
1191 | c=((uint32_t)(SC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1192 | length=2; |
1193 | goto outputBytes; |
1194 | } else { |
1195 | /* quote from dynamic window */ |
1196 | c=((uint32_t)(SQ0+window)<<8)|(c-scsu->fromUDynamicOffsets[window])|0x80; |
1197 | length=2; |
1198 | goto outputBytes; |
1199 | } |
1200 | } else if((window=getWindow(staticOffsets, c))>=0) { |
1201 | /* quote from static window */ |
1202 | c=((uint32_t)(SQ0+window)<<8)|(c-staticOffsets[window]); |
1203 | length=2; |
1204 | goto outputBytes; |
1205 | } else if((code=getDynamicOffset(c, &offset))>=0) { |
1206 | /* define a dynamic window with this character */ |
1207 | dynamicWindow=getNextDynamicWindow(scsu); |
1208 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1209 | useDynamicWindow(scsu, dynamicWindow); |
1210 | c=((uint32_t)(SD0+dynamicWindow)<<16)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1211 | length=3; |
1212 | goto outputBytes; |
1213 | } else if((uint32_t)(c-0x3400)<(0xd800-0x3400) && |
1214 | (source>=sourceLimit || (uint32_t)(*source-0x3400)<(0xd800-0x3400)) |
1215 | ) { |
1216 | /* |
1217 | * this character is not compressible (a BMP ideograph or similar); |
1218 | * switch to Unicode mode if this is the last character in the block |
1219 | * or there is at least one more ideograph following immediately |
1220 | */ |
1221 | isSingleByteMode=false; |
1222 | c|=SCU<<16; |
1223 | length=3; |
1224 | goto outputBytes; |
1225 | } else { |
1226 | /* quote Unicode */ |
1227 | c|=SQU<<16; |
1228 | length=3; |
1229 | goto outputBytes; |
1230 | } |
1231 | } |
1232 | |
1233 | /* normal end of conversion: prepare for a new character */ |
1234 | c=0; |
1235 | sourceIndex=nextSourceIndex; |
1236 | } |
1237 | } else { |
1238 | if(c!=0 && targetCapacity>0) { |
1239 | goto getTrailUnicode; |
1240 | } |
1241 | |
1242 | /* state machine for Unicode mode */ |
1243 | /* unicodeByteMode: */ |
1244 | while(source<sourceLimit) { |
1245 | if(targetCapacity<=0) { |
1246 | /* target is full */ |
1247 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1248 | break; |
1249 | } |
1250 | c=*source++; |
1251 | ++nextSourceIndex; |
1252 | |
1253 | if((uint32_t)(c-0x3400)<(0xd800-0x3400)) { |
1254 | /* not compressible, write character directly */ |
1255 | if(targetCapacity>=2) { |
1256 | *target++=(uint8_t)(c>>8); |
1257 | *target++=(uint8_t)c; |
1258 | if(offsets!=nullptr) { |
1259 | *offsets++=sourceIndex; |
1260 | *offsets++=sourceIndex; |
1261 | } |
1262 | targetCapacity-=2; |
1263 | } else { |
1264 | length=2; |
1265 | goto outputBytes; |
1266 | } |
1267 | } else if((uint32_t)(c-0x3400)>=(0xf300-0x3400) /* c<0x3400 || c>=0xf300 */) { |
1268 | /* compress BMP character if the following one is not an uncompressible ideograph */ |
1269 | if(!(source<sourceLimit && (uint32_t)(*source-0x3400)<(0xd800-0x3400))) { |
1270 | if(((uint32_t)(c-0x30)<10 || (uint32_t)(c-0x61)<26 || (uint32_t)(c-0x41)<26)) { |
1271 | /* ASCII digit or letter */ |
1272 | isSingleByteMode=true; |
1273 | c|=((uint32_t)(UC0+dynamicWindow)<<8)|c; |
1274 | length=2; |
1275 | goto outputBytes; |
1276 | } else if((window=getWindow(scsu->fromUDynamicOffsets, c))>=0) { |
1277 | /* there is a dynamic window that contains this character, change to it */ |
1278 | isSingleByteMode=true; |
1279 | dynamicWindow=window; |
1280 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1281 | useDynamicWindow(scsu, dynamicWindow); |
1282 | c=((uint32_t)(UC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1283 | length=2; |
1284 | goto outputBytes; |
1285 | } else if((code=getDynamicOffset(c, &offset))>=0) { |
1286 | /* define a dynamic window with this character */ |
1287 | isSingleByteMode=true; |
1288 | dynamicWindow=getNextDynamicWindow(scsu); |
1289 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1290 | useDynamicWindow(scsu, dynamicWindow); |
1291 | c=((uint32_t)(UD0+dynamicWindow)<<16)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1292 | length=3; |
1293 | goto outputBytes; |
1294 | } |
1295 | } |
1296 | |
1297 | /* don't know how to compress this character, just write it directly */ |
1298 | length=2; |
1299 | goto outputBytes; |
1300 | } else if(c<0xe000) { |
1301 | /* c is a surrogate */ |
1302 | if(U16_IS_SURROGATE_LEAD(c)) { |
1303 | getTrailUnicode: |
1304 | lead=(char16_t)c; |
1305 | if(source<sourceLimit) { |
1306 | /* test the following code unit */ |
1307 | trail=*source; |
1308 | if(U16_IS_TRAIL(trail)) { |
1309 | ++source; |
1310 | ++nextSourceIndex; |
1311 | c=U16_GET_SUPPLEMENTARY(c, trail); |
1312 | /* convert this surrogate code point */ |
1313 | /* exit this condition tree */ |
1314 | } else { |
1315 | /* this is an unmatched lead code unit (1st surrogate) */ |
1316 | /* callback(illegal) */ |
1317 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1318 | goto endloop; |
1319 | } |
1320 | } else { |
1321 | /* no more input */ |
1322 | break; |
1323 | } |
1324 | } else { |
1325 | /* this is an unmatched trail code unit (2nd surrogate) */ |
1326 | /* callback(illegal) */ |
1327 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1328 | goto endloop; |
1329 | } |
1330 | |
1331 | /* compress supplementary character */ |
1332 | if( (window=getWindow(scsu->fromUDynamicOffsets, c))>=0 && |
1333 | !(source<sourceLimit && (uint32_t)(*source-0x3400)<(0xd800-0x3400)) |
1334 | ) { |
1335 | /* |
1336 | * there is a dynamic window that contains this character and |
1337 | * the following character is not uncompressible, |
1338 | * change to the window |
1339 | */ |
1340 | isSingleByteMode=true; |
1341 | dynamicWindow=window; |
1342 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1343 | useDynamicWindow(scsu, dynamicWindow); |
1344 | c=((uint32_t)(UC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1345 | length=2; |
1346 | goto outputBytes; |
1347 | } else if(source<sourceLimit && lead==*source && /* too lazy to check trail in same window as source[1] */ |
1348 | (code=getDynamicOffset(c, &offset))>=0 |
1349 | ) { |
1350 | /* two supplementary characters in (probably) the same window - define an extended one */ |
1351 | isSingleByteMode=true; |
1352 | code-=0x200; |
1353 | dynamicWindow=getNextDynamicWindow(scsu); |
1354 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1355 | useDynamicWindow(scsu, dynamicWindow); |
1356 | c=((uint32_t)UDX<<24)|((uint32_t)dynamicWindow<<21)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1357 | length=4; |
1358 | goto outputBytes; |
1359 | } else { |
1360 | /* don't know how to compress this character, just write it directly */ |
1361 | c=((uint32_t)lead<<16)|trail; |
1362 | length=4; |
1363 | goto outputBytes; |
1364 | } |
1365 | } else /* 0xe000<=c<0xf300 */ { |
1366 | /* quote to avoid SCSU tags */ |
1367 | c|=UQU<<16; |
1368 | length=3; |
1369 | goto outputBytes; |
1370 | } |
1371 | |
1372 | /* normal end of conversion: prepare for a new character */ |
1373 | c=0; |
1374 | sourceIndex=nextSourceIndex; |
1375 | } |
1376 | } |
1377 | endloop: |
1378 | |
1379 | /* set the converter state back into UConverter */ |
1380 | scsu->fromUIsSingleByteMode=isSingleByteMode; |
1381 | scsu->fromUDynamicWindow=dynamicWindow; |
1382 | |
1383 | cnv->fromUChar32=c; |
1384 | |
1385 | /* write back the updated pointers */ |
1386 | pArgs->source=source; |
1387 | pArgs->target=(char *)target; |
1388 | pArgs->offsets=offsets; |
1389 | return; |
1390 | |
1391 | outputBytes: |
1392 | /* write the output character bytes from c and length [code copied from ucnvmbcs.c] */ |
1393 | /* from the first if in the loop we know that targetCapacity>0 */ |
1394 | if(length<=targetCapacity) { |
1395 | if(offsets==nullptr) { |
1396 | switch(length) { |
1397 | /* each branch falls through to the next one */ |
1398 | case 4: |
1399 | *target++=(uint8_t)(c>>24); |
1400 | U_FALLTHROUGH; |
1401 | case 3: |
1402 | *target++=(uint8_t)(c>>16); |
1403 | U_FALLTHROUGH; |
1404 | case 2: |
1405 | *target++=(uint8_t)(c>>8); |
1406 | U_FALLTHROUGH; |
1407 | case 1: |
1408 | *target++=(uint8_t)c; |
1409 | U_FALLTHROUGH; |
1410 | default: |
1411 | /* will never occur */ |
1412 | break; |
1413 | } |
1414 | } else { |
1415 | switch(length) { |
1416 | /* each branch falls through to the next one */ |
1417 | case 4: |
1418 | *target++=(uint8_t)(c>>24); |
1419 | *offsets++=sourceIndex; |
1420 | U_FALLTHROUGH; |
1421 | case 3: |
1422 | *target++=(uint8_t)(c>>16); |
1423 | *offsets++=sourceIndex; |
1424 | U_FALLTHROUGH; |
1425 | case 2: |
1426 | *target++=(uint8_t)(c>>8); |
1427 | *offsets++=sourceIndex; |
1428 | U_FALLTHROUGH; |
1429 | case 1: |
1430 | *target++=(uint8_t)c; |
1431 | *offsets++=sourceIndex; |
1432 | U_FALLTHROUGH; |
1433 | default: |
1434 | /* will never occur */ |
1435 | break; |
1436 | } |
1437 | } |
1438 | targetCapacity-=length; |
1439 | |
1440 | /* normal end of conversion: prepare for a new character */ |
1441 | c=0; |
1442 | sourceIndex=nextSourceIndex; |
1443 | goto loop; |
1444 | } else { |
1445 | uint8_t *p; |
1446 | |
1447 | /* |
1448 | * We actually do this backwards here: |
1449 | * In order to save an intermediate variable, we output |
1450 | * first to the overflow buffer what does not fit into the |
1451 | * regular target. |
1452 | */ |
1453 | /* we know that 0<=targetCapacity<length<=4 */ |
1454 | /* targetCapacity==0 when SCU+supplementary where SCU used up targetCapacity==1 */ |
1455 | length-=targetCapacity; |
1456 | p=(uint8_t *)cnv->charErrorBuffer; |
1457 | switch(length) { |
1458 | /* each branch falls through to the next one */ |
1459 | case 4: |
1460 | *p++=(uint8_t)(c>>24); |
1461 | U_FALLTHROUGH; |
1462 | case 3: |
1463 | *p++=(uint8_t)(c>>16); |
1464 | U_FALLTHROUGH; |
1465 | case 2: |
1466 | *p++=(uint8_t)(c>>8); |
1467 | U_FALLTHROUGH; |
1468 | case 1: |
1469 | *p=(uint8_t)c; |
1470 | U_FALLTHROUGH; |
1471 | default: |
1472 | /* will never occur */ |
1473 | break; |
1474 | } |
1475 | cnv->charErrorBufferLength=(int8_t)length; |
1476 | |
1477 | /* now output what fits into the regular target */ |
1478 | c>>=8*length; /* length was reduced by targetCapacity */ |
1479 | switch(targetCapacity) { |
1480 | /* each branch falls through to the next one */ |
1481 | case 3: |
1482 | *target++=(uint8_t)(c>>16); |
1483 | if(offsets!=nullptr) { |
1484 | *offsets++=sourceIndex; |
1485 | } |
1486 | U_FALLTHROUGH; |
1487 | case 2: |
1488 | *target++=(uint8_t)(c>>8); |
1489 | if(offsets!=nullptr) { |
1490 | *offsets++=sourceIndex; |
1491 | } |
1492 | U_FALLTHROUGH; |
1493 | case 1: |
1494 | *target++=(uint8_t)c; |
1495 | if(offsets!=nullptr) { |
1496 | *offsets++=sourceIndex; |
1497 | } |
1498 | U_FALLTHROUGH; |
1499 | default: |
1500 | break; |
1501 | } |
1502 | |
1503 | /* target overflow */ |
1504 | targetCapacity=0; |
1505 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1506 | c=0; |
1507 | goto endloop; |
1508 | } |
1509 | } |
1510 | |
1511 | /* |
1512 | * Identical to _SCSUFromUnicodeWithOffsets but without offset handling. |
1513 | * If a change is made in the original function, then either |
1514 | * change this function the same way or |
1515 | * re-copy the original function and remove the variables |
1516 | * offsets, sourceIndex, and nextSourceIndex. |
1517 | */ |
1518 | static void U_CALLCONV |
1519 | _SCSUFromUnicode(UConverterFromUnicodeArgs *pArgs, |
1520 | UErrorCode *pErrorCode) { |
1521 | UConverter *cnv; |
1522 | SCSUData *scsu; |
1523 | const char16_t *source, *sourceLimit; |
1524 | uint8_t *target; |
1525 | int32_t targetCapacity; |
1526 | |
1527 | UBool isSingleByteMode; |
1528 | uint8_t dynamicWindow; |
1529 | uint32_t currentOffset; |
1530 | |
1531 | uint32_t c, delta; |
1532 | |
1533 | int32_t length; |
1534 | |
1535 | /* variables for compression heuristics */ |
1536 | uint32_t offset; |
1537 | char16_t lead, trail; |
1538 | int code; |
1539 | int8_t window; |
1540 | |
1541 | /* set up the local pointers */ |
1542 | cnv=pArgs->converter; |
1543 | scsu=(SCSUData *)cnv->extraInfo; |
1544 | |
1545 | /* set up the local pointers */ |
1546 | source=pArgs->source; |
1547 | sourceLimit=pArgs->sourceLimit; |
1548 | target=(uint8_t *)pArgs->target; |
1549 | targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); |
1550 | |
1551 | /* get the state machine state */ |
1552 | isSingleByteMode=scsu->fromUIsSingleByteMode; |
1553 | dynamicWindow=scsu->fromUDynamicWindow; |
1554 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1555 | |
1556 | c=cnv->fromUChar32; |
1557 | |
1558 | /* similar conversion "loop" as in toUnicode */ |
1559 | loop: |
1560 | if(isSingleByteMode) { |
1561 | if(c!=0 && targetCapacity>0) { |
1562 | goto getTrailSingle; |
1563 | } |
1564 | |
1565 | /* state machine for single-byte mode */ |
1566 | /* singleByteMode: */ |
1567 | while(source<sourceLimit) { |
1568 | if(targetCapacity<=0) { |
1569 | /* target is full */ |
1570 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1571 | break; |
1572 | } |
1573 | c=*source++; |
1574 | |
1575 | if((c-0x20)<=0x5f) { |
1576 | /* pass US-ASCII graphic character through */ |
1577 | *target++=(uint8_t)c; |
1578 | --targetCapacity; |
1579 | } else if(c<0x20) { |
1580 | if((1UL<<c)&0x2601 /* binary 0010 0110 0000 0001, check for b==0xd || b==0xa || b==9 || b==0 */) { |
1581 | /* CR/LF/TAB/NUL */ |
1582 | *target++=(uint8_t)c; |
1583 | --targetCapacity; |
1584 | } else { |
1585 | /* quote C0 control character */ |
1586 | c|=SQ0<<8; |
1587 | length=2; |
1588 | goto outputBytes; |
1589 | } |
1590 | } else if((delta=c-currentOffset)<=0x7f) { |
1591 | /* use the current dynamic window */ |
1592 | *target++=(uint8_t)(delta|0x80); |
1593 | --targetCapacity; |
1594 | } else if(U16_IS_SURROGATE(c)) { |
1595 | if(U16_IS_SURROGATE_LEAD(c)) { |
1596 | getTrailSingle: |
1597 | lead=(char16_t)c; |
1598 | if(source<sourceLimit) { |
1599 | /* test the following code unit */ |
1600 | trail=*source; |
1601 | if(U16_IS_TRAIL(trail)) { |
1602 | ++source; |
1603 | c=U16_GET_SUPPLEMENTARY(c, trail); |
1604 | /* convert this surrogate code point */ |
1605 | /* exit this condition tree */ |
1606 | } else { |
1607 | /* this is an unmatched lead code unit (1st surrogate) */ |
1608 | /* callback(illegal) */ |
1609 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1610 | goto endloop; |
1611 | } |
1612 | } else { |
1613 | /* no more input */ |
1614 | break; |
1615 | } |
1616 | } else { |
1617 | /* this is an unmatched trail code unit (2nd surrogate) */ |
1618 | /* callback(illegal) */ |
1619 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1620 | goto endloop; |
1621 | } |
1622 | |
1623 | /* compress supplementary character U+10000..U+10ffff */ |
1624 | if((delta=c-currentOffset)<=0x7f) { |
1625 | /* use the current dynamic window */ |
1626 | *target++=(uint8_t)(delta|0x80); |
1627 | --targetCapacity; |
1628 | } else if((window=getWindow(scsu->fromUDynamicOffsets, c))>=0) { |
1629 | /* there is a dynamic window that contains this character, change to it */ |
1630 | dynamicWindow=window; |
1631 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1632 | useDynamicWindow(scsu, dynamicWindow); |
1633 | c=((uint32_t)(SC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1634 | length=2; |
1635 | goto outputBytes; |
1636 | } else if((code=getDynamicOffset(c, &offset))>=0) { |
1637 | /* might check if there are more characters in this window to come */ |
1638 | /* define an extended window with this character */ |
1639 | code-=0x200; |
1640 | dynamicWindow=getNextDynamicWindow(scsu); |
1641 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1642 | useDynamicWindow(scsu, dynamicWindow); |
1643 | c=((uint32_t)SDX<<24)|((uint32_t)dynamicWindow<<21)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1644 | length=4; |
1645 | goto outputBytes; |
1646 | } else { |
1647 | /* change to Unicode mode and output this (lead, trail) pair */ |
1648 | isSingleByteMode=false; |
1649 | *target++=(uint8_t)SCU; |
1650 | --targetCapacity; |
1651 | c=((uint32_t)lead<<16)|trail; |
1652 | length=4; |
1653 | goto outputBytes; |
1654 | } |
1655 | } else if(c<0xa0) { |
1656 | /* quote C1 control character */ |
1657 | c=(c&0x7f)|(SQ0+1)<<8; /* SQ0+1==SQ1 */ |
1658 | length=2; |
1659 | goto outputBytes; |
1660 | } else if(c==0xfeff || c>=0xfff0) { |
1661 | /* quote signature character=byte order mark and specials */ |
1662 | c|=SQU<<16; |
1663 | length=3; |
1664 | goto outputBytes; |
1665 | } else { |
1666 | /* compress all other BMP characters */ |
1667 | if((window=getWindow(scsu->fromUDynamicOffsets, c))>=0) { |
1668 | /* there is a window defined that contains this character - switch to it or quote from it? */ |
1669 | if(source>=sourceLimit || isInOffsetWindowOrDirect(scsu->fromUDynamicOffsets[window], *source)) { |
1670 | /* change to dynamic window */ |
1671 | dynamicWindow=window; |
1672 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1673 | useDynamicWindow(scsu, dynamicWindow); |
1674 | c=((uint32_t)(SC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1675 | length=2; |
1676 | goto outputBytes; |
1677 | } else { |
1678 | /* quote from dynamic window */ |
1679 | c=((uint32_t)(SQ0+window)<<8)|(c-scsu->fromUDynamicOffsets[window])|0x80; |
1680 | length=2; |
1681 | goto outputBytes; |
1682 | } |
1683 | } else if((window=getWindow(staticOffsets, c))>=0) { |
1684 | /* quote from static window */ |
1685 | c=((uint32_t)(SQ0+window)<<8)|(c-staticOffsets[window]); |
1686 | length=2; |
1687 | goto outputBytes; |
1688 | } else if((code=getDynamicOffset(c, &offset))>=0) { |
1689 | /* define a dynamic window with this character */ |
1690 | dynamicWindow=getNextDynamicWindow(scsu); |
1691 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1692 | useDynamicWindow(scsu, dynamicWindow); |
1693 | c=((uint32_t)(SD0+dynamicWindow)<<16)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1694 | length=3; |
1695 | goto outputBytes; |
1696 | } else if((uint32_t)(c-0x3400)<(0xd800-0x3400) && |
1697 | (source>=sourceLimit || (uint32_t)(*source-0x3400)<(0xd800-0x3400)) |
1698 | ) { |
1699 | /* |
1700 | * this character is not compressible (a BMP ideograph or similar); |
1701 | * switch to Unicode mode if this is the last character in the block |
1702 | * or there is at least one more ideograph following immediately |
1703 | */ |
1704 | isSingleByteMode=false; |
1705 | c|=SCU<<16; |
1706 | length=3; |
1707 | goto outputBytes; |
1708 | } else { |
1709 | /* quote Unicode */ |
1710 | c|=SQU<<16; |
1711 | length=3; |
1712 | goto outputBytes; |
1713 | } |
1714 | } |
1715 | |
1716 | /* normal end of conversion: prepare for a new character */ |
1717 | c=0; |
1718 | } |
1719 | } else { |
1720 | if(c!=0 && targetCapacity>0) { |
1721 | goto getTrailUnicode; |
1722 | } |
1723 | |
1724 | /* state machine for Unicode mode */ |
1725 | /* unicodeByteMode: */ |
1726 | while(source<sourceLimit) { |
1727 | if(targetCapacity<=0) { |
1728 | /* target is full */ |
1729 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1730 | break; |
1731 | } |
1732 | c=*source++; |
1733 | |
1734 | if((uint32_t)(c-0x3400)<(0xd800-0x3400)) { |
1735 | /* not compressible, write character directly */ |
1736 | if(targetCapacity>=2) { |
1737 | *target++=(uint8_t)(c>>8); |
1738 | *target++=(uint8_t)c; |
1739 | targetCapacity-=2; |
1740 | } else { |
1741 | length=2; |
1742 | goto outputBytes; |
1743 | } |
1744 | } else if((uint32_t)(c-0x3400)>=(0xf300-0x3400) /* c<0x3400 || c>=0xf300 */) { |
1745 | /* compress BMP character if the following one is not an uncompressible ideograph */ |
1746 | if(!(source<sourceLimit && (uint32_t)(*source-0x3400)<(0xd800-0x3400))) { |
1747 | if(((uint32_t)(c-0x30)<10 || (uint32_t)(c-0x61)<26 || (uint32_t)(c-0x41)<26)) { |
1748 | /* ASCII digit or letter */ |
1749 | isSingleByteMode=true; |
1750 | c|=((uint32_t)(UC0+dynamicWindow)<<8)|c; |
1751 | length=2; |
1752 | goto outputBytes; |
1753 | } else if((window=getWindow(scsu->fromUDynamicOffsets, c))>=0) { |
1754 | /* there is a dynamic window that contains this character, change to it */ |
1755 | isSingleByteMode=true; |
1756 | dynamicWindow=window; |
1757 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1758 | useDynamicWindow(scsu, dynamicWindow); |
1759 | c=((uint32_t)(UC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1760 | length=2; |
1761 | goto outputBytes; |
1762 | } else if((code=getDynamicOffset(c, &offset))>=0) { |
1763 | /* define a dynamic window with this character */ |
1764 | isSingleByteMode=true; |
1765 | dynamicWindow=getNextDynamicWindow(scsu); |
1766 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1767 | useDynamicWindow(scsu, dynamicWindow); |
1768 | c=((uint32_t)(UD0+dynamicWindow)<<16)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1769 | length=3; |
1770 | goto outputBytes; |
1771 | } |
1772 | } |
1773 | |
1774 | /* don't know how to compress this character, just write it directly */ |
1775 | length=2; |
1776 | goto outputBytes; |
1777 | } else if(c<0xe000) { |
1778 | /* c is a surrogate */ |
1779 | if(U16_IS_SURROGATE_LEAD(c)) { |
1780 | getTrailUnicode: |
1781 | lead=(char16_t)c; |
1782 | if(source<sourceLimit) { |
1783 | /* test the following code unit */ |
1784 | trail=*source; |
1785 | if(U16_IS_TRAIL(trail)) { |
1786 | ++source; |
1787 | c=U16_GET_SUPPLEMENTARY(c, trail); |
1788 | /* convert this surrogate code point */ |
1789 | /* exit this condition tree */ |
1790 | } else { |
1791 | /* this is an unmatched lead code unit (1st surrogate) */ |
1792 | /* callback(illegal) */ |
1793 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1794 | goto endloop; |
1795 | } |
1796 | } else { |
1797 | /* no more input */ |
1798 | break; |
1799 | } |
1800 | } else { |
1801 | /* this is an unmatched trail code unit (2nd surrogate) */ |
1802 | /* callback(illegal) */ |
1803 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1804 | goto endloop; |
1805 | } |
1806 | |
1807 | /* compress supplementary character */ |
1808 | if( (window=getWindow(scsu->fromUDynamicOffsets, c))>=0 && |
1809 | !(source<sourceLimit && (uint32_t)(*source-0x3400)<(0xd800-0x3400)) |
1810 | ) { |
1811 | /* |
1812 | * there is a dynamic window that contains this character and |
1813 | * the following character is not uncompressible, |
1814 | * change to the window |
1815 | */ |
1816 | isSingleByteMode=true; |
1817 | dynamicWindow=window; |
1818 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]; |
1819 | useDynamicWindow(scsu, dynamicWindow); |
1820 | c=((uint32_t)(UC0+dynamicWindow)<<8)|(c-currentOffset)|0x80; |
1821 | length=2; |
1822 | goto outputBytes; |
1823 | } else if(source<sourceLimit && lead==*source && /* too lazy to check trail in same window as source[1] */ |
1824 | (code=getDynamicOffset(c, &offset))>=0 |
1825 | ) { |
1826 | /* two supplementary characters in (probably) the same window - define an extended one */ |
1827 | isSingleByteMode=true; |
1828 | code-=0x200; |
1829 | dynamicWindow=getNextDynamicWindow(scsu); |
1830 | currentOffset=scsu->fromUDynamicOffsets[dynamicWindow]=offset; |
1831 | useDynamicWindow(scsu, dynamicWindow); |
1832 | c=((uint32_t)UDX<<24)|((uint32_t)dynamicWindow<<21)|((uint32_t)code<<8)|(c-currentOffset)|0x80; |
1833 | length=4; |
1834 | goto outputBytes; |
1835 | } else { |
1836 | /* don't know how to compress this character, just write it directly */ |
1837 | c=((uint32_t)lead<<16)|trail; |
1838 | length=4; |
1839 | goto outputBytes; |
1840 | } |
1841 | } else /* 0xe000<=c<0xf300 */ { |
1842 | /* quote to avoid SCSU tags */ |
1843 | c|=UQU<<16; |
1844 | length=3; |
1845 | goto outputBytes; |
1846 | } |
1847 | |
1848 | /* normal end of conversion: prepare for a new character */ |
1849 | c=0; |
1850 | } |
1851 | } |
1852 | endloop: |
1853 | |
1854 | /* set the converter state back into UConverter */ |
1855 | scsu->fromUIsSingleByteMode=isSingleByteMode; |
1856 | scsu->fromUDynamicWindow=dynamicWindow; |
1857 | |
1858 | cnv->fromUChar32=c; |
1859 | |
1860 | /* write back the updated pointers */ |
1861 | pArgs->source=source; |
1862 | pArgs->target=(char *)target; |
1863 | return; |
1864 | |
1865 | outputBytes: |
1866 | /* write the output character bytes from c and length [code copied from ucnvmbcs.c] */ |
1867 | /* from the first if in the loop we know that targetCapacity>0 */ |
1868 | if(length<=targetCapacity) { |
1869 | switch(length) { |
1870 | /* each branch falls through to the next one */ |
1871 | case 4: |
1872 | *target++=(uint8_t)(c>>24); |
1873 | U_FALLTHROUGH; |
1874 | case 3: |
1875 | *target++=(uint8_t)(c>>16); |
1876 | U_FALLTHROUGH; |
1877 | case 2: |
1878 | *target++=(uint8_t)(c>>8); |
1879 | U_FALLTHROUGH; |
1880 | case 1: |
1881 | *target++=(uint8_t)c; |
1882 | U_FALLTHROUGH; |
1883 | default: |
1884 | /* will never occur */ |
1885 | break; |
1886 | } |
1887 | targetCapacity-=length; |
1888 | |
1889 | /* normal end of conversion: prepare for a new character */ |
1890 | c=0; |
1891 | goto loop; |
1892 | } else { |
1893 | uint8_t *p; |
1894 | |
1895 | /* |
1896 | * We actually do this backwards here: |
1897 | * In order to save an intermediate variable, we output |
1898 | * first to the overflow buffer what does not fit into the |
1899 | * regular target. |
1900 | */ |
1901 | /* we know that 0<=targetCapacity<length<=4 */ |
1902 | /* targetCapacity==0 when SCU+supplementary where SCU used up targetCapacity==1 */ |
1903 | length-=targetCapacity; |
1904 | p=(uint8_t *)cnv->charErrorBuffer; |
1905 | switch(length) { |
1906 | /* each branch falls through to the next one */ |
1907 | case 4: |
1908 | *p++=(uint8_t)(c>>24); |
1909 | U_FALLTHROUGH; |
1910 | case 3: |
1911 | *p++=(uint8_t)(c>>16); |
1912 | U_FALLTHROUGH; |
1913 | case 2: |
1914 | *p++=(uint8_t)(c>>8); |
1915 | U_FALLTHROUGH; |
1916 | case 1: |
1917 | *p=(uint8_t)c; |
1918 | U_FALLTHROUGH; |
1919 | default: |
1920 | /* will never occur */ |
1921 | break; |
1922 | } |
1923 | cnv->charErrorBufferLength=(int8_t)length; |
1924 | |
1925 | /* now output what fits into the regular target */ |
1926 | c = (length == 4) ? 0 : c >> 8*length; /* length was reduced by targetCapacity */ |
1927 | switch(targetCapacity) { |
1928 | /* each branch falls through to the next one */ |
1929 | case 3: |
1930 | *target++=(uint8_t)(c>>16); |
1931 | U_FALLTHROUGH; |
1932 | case 2: |
1933 | *target++=(uint8_t)(c>>8); |
1934 | U_FALLTHROUGH; |
1935 | case 1: |
1936 | *target++=(uint8_t)c; |
1937 | U_FALLTHROUGH; |
1938 | default: |
1939 | break; |
1940 | } |
1941 | |
1942 | /* target overflow */ |
1943 | targetCapacity=0; |
1944 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1945 | c=0; |
1946 | goto endloop; |
1947 | } |
1948 | } |
1949 | |
1950 | /* miscellaneous ------------------------------------------------------------ */ |
1951 | |
1952 | static const char * U_CALLCONV |
1953 | _SCSUGetName(const UConverter *cnv) { |
1954 | SCSUData *scsu=(SCSUData *)cnv->extraInfo; |
1955 | |
1956 | switch(scsu->locale) { |
1957 | case l_ja: |
1958 | return "SCSU,locale=ja" ; |
1959 | default: |
1960 | return "SCSU" ; |
1961 | } |
1962 | } |
1963 | |
1964 | /* structure for SafeClone calculations */ |
1965 | struct cloneSCSUStruct |
1966 | { |
1967 | UConverter cnv; |
1968 | SCSUData mydata; |
1969 | }; |
1970 | |
1971 | static UConverter * U_CALLCONV |
1972 | _SCSUSafeClone(const UConverter *cnv, |
1973 | void *stackBuffer, |
1974 | int32_t *pBufferSize, |
1975 | UErrorCode *status) |
1976 | { |
1977 | struct cloneSCSUStruct * localClone; |
1978 | int32_t bufferSizeNeeded = sizeof(struct cloneSCSUStruct); |
1979 | |
1980 | if (U_FAILURE(*status)){ |
1981 | return 0; |
1982 | } |
1983 | |
1984 | if (*pBufferSize == 0){ /* 'preflighting' request - set needed size into *pBufferSize */ |
1985 | *pBufferSize = bufferSizeNeeded; |
1986 | return 0; |
1987 | } |
1988 | |
1989 | localClone = (struct cloneSCSUStruct *)stackBuffer; |
1990 | /* ucnv.c/ucnv_safeClone() copied the main UConverter already */ |
1991 | |
1992 | uprv_memcpy(&localClone->mydata, cnv->extraInfo, sizeof(SCSUData)); |
1993 | localClone->cnv.extraInfo = &localClone->mydata; |
1994 | localClone->cnv.isExtraLocal = true; |
1995 | |
1996 | return &localClone->cnv; |
1997 | } |
1998 | U_CDECL_END |
1999 | |
2000 | static const UConverterImpl _SCSUImpl={ |
2001 | UCNV_SCSU, |
2002 | |
2003 | nullptr, |
2004 | nullptr, |
2005 | |
2006 | _SCSUOpen, |
2007 | _SCSUClose, |
2008 | _SCSUReset, |
2009 | |
2010 | _SCSUToUnicode, |
2011 | _SCSUToUnicodeWithOffsets, |
2012 | _SCSUFromUnicode, |
2013 | _SCSUFromUnicodeWithOffsets, |
2014 | nullptr, |
2015 | |
2016 | nullptr, |
2017 | _SCSUGetName, |
2018 | nullptr, |
2019 | _SCSUSafeClone, |
2020 | ucnv_getCompleteUnicodeSet, |
2021 | nullptr, |
2022 | nullptr |
2023 | }; |
2024 | |
2025 | static const UConverterStaticData _SCSUStaticData={ |
2026 | sizeof(UConverterStaticData), |
2027 | "SCSU" , |
2028 | 1212, /* CCSID for SCSU */ |
2029 | UCNV_IBM, UCNV_SCSU, |
2030 | 1, 3, /* one char16_t generates at least 1 byte and at most 3 bytes */ |
2031 | /* |
2032 | * The subchar here is ignored because _SCSUOpen() sets U+fffd as a Unicode |
2033 | * substitution string. |
2034 | */ |
2035 | { 0x0e, 0xff, 0xfd, 0 }, 3, |
2036 | false, false, |
2037 | 0, |
2038 | 0, |
2039 | { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } /* reserved */ |
2040 | }; |
2041 | |
2042 | const UConverterSharedData _SCSUData= |
2043 | UCNV_IMMUTABLE_SHARED_DATA_INITIALIZER(&_SCSUStaticData, &_SCSUImpl); |
2044 | |
2045 | #endif |
2046 | |