• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2 *******************************************************************************
3 *
4 *   Copyright (C) 2004-2014, International Business Machines
5 *   Corporation and others.  All Rights Reserved.
6 *
7 *******************************************************************************
8 *   file name:  ucase.cpp
9 *   encoding:   US-ASCII
10 *   tab size:   8 (not used)
11 *   indentation:4
12 *
13 *   created on: 2004aug30
14 *   created by: Markus W. Scherer
15 *
16 *   Low-level Unicode character/string case mapping code.
17 *   Much code moved here (and modified) from uchar.c.
18 */
19 
20 #include "unicode/utypes.h"
21 #include "unicode/unistr.h"
22 #include "unicode/uset.h"
23 #include "unicode/udata.h" /* UDataInfo */
24 #include "unicode/utf16.h"
25 #include "ucmndata.h" /* DataHeader */
26 #include "udatamem.h"
27 #include "umutex.h"
28 #include "uassert.h"
29 #include "cmemory.h"
30 #include "utrie2.h"
31 #include "ucase.h"
32 
33 struct UCaseProps {
34     UDataMemory *mem;
35     const int32_t *indexes;
36     const uint16_t *exceptions;
37     const uint16_t *unfold;
38 
39     UTrie2 trie;
40     uint8_t formatVersion[4];
41 };
42 
43 /* ucase_props_data.h is machine-generated by gencase --csource */
44 #define INCLUDED_FROM_UCASE_CPP
45 #include "ucase_props_data.h"
46 
47 /* UCaseProps singleton ----------------------------------------------------- */
48 
49 U_CAPI const UCaseProps * U_EXPORT2
ucase_getSingleton()50 ucase_getSingleton() {
51     return &ucase_props_singleton;
52 }
53 
54 /* set of property starts for UnicodeSet ------------------------------------ */
55 
56 static UBool U_CALLCONV
_enumPropertyStartsRange(const void * context,UChar32 start,UChar32,uint32_t)57 _enumPropertyStartsRange(const void *context, UChar32 start, UChar32 /*end*/, uint32_t /*value*/) {
58     /* add the start code point to the USet */
59     const USetAdder *sa=(const USetAdder *)context;
60     sa->add(sa->set, start);
61     return TRUE;
62 }
63 
64 U_CFUNC void U_EXPORT2
ucase_addPropertyStarts(const UCaseProps * csp,const USetAdder * sa,UErrorCode * pErrorCode)65 ucase_addPropertyStarts(const UCaseProps *csp, const USetAdder *sa, UErrorCode *pErrorCode) {
66     if(U_FAILURE(*pErrorCode)) {
67         return;
68     }
69 
70     /* add the start code point of each same-value range of the trie */
71     utrie2_enum(&csp->trie, NULL, _enumPropertyStartsRange, sa);
72 
73     /* add code points with hardcoded properties, plus the ones following them */
74 
75     /* (none right now, see comment below) */
76 
77     /*
78      * Omit code points with hardcoded specialcasing properties
79      * because we do not build property UnicodeSets for them right now.
80      */
81 }
82 
83 /* data access primitives --------------------------------------------------- */
84 
85 #define GET_EXCEPTIONS(csp, props) ((csp)->exceptions+((props)>>UCASE_EXC_SHIFT))
86 
87 #define PROPS_HAS_EXCEPTION(props) ((props)&UCASE_EXCEPTION)
88 
89 /* number of bits in an 8-bit integer value */
90 static const uint8_t flagsOffset[256]={
91     0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4,
92     1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
93     1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
94     2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
95     1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
96     2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
97     2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
98     3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
99     1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
100     2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
101     2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
102     3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
103     2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
104     3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
105     3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
106     4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8
107 };
108 
109 #define HAS_SLOT(flags, idx) ((flags)&(1<<(idx)))
110 #define SLOT_OFFSET(flags, idx) flagsOffset[(flags)&((1<<(idx))-1)]
111 
112 /*
113  * Get the value of an optional-value slot where HAS_SLOT(excWord, idx).
114  *
115  * @param excWord (in) initial exceptions word
116  * @param idx (in) desired slot index
117  * @param pExc16 (in/out) const uint16_t * after excWord=*pExc16++;
118  *               moved to the last uint16_t of the value, use +1 for beginning of next slot
119  * @param value (out) int32_t or uint32_t output if hasSlot, otherwise not modified
120  */
121 #define GET_SLOT_VALUE(excWord, idx, pExc16, value) \
122     if(((excWord)&UCASE_EXC_DOUBLE_SLOTS)==0) { \
123         (pExc16)+=SLOT_OFFSET(excWord, idx); \
124         (value)=*pExc16; \
125     } else { \
126         (pExc16)+=2*SLOT_OFFSET(excWord, idx); \
127         (value)=*pExc16++; \
128         (value)=((value)<<16)|*pExc16; \
129     }
130 
131 /* simple case mappings ----------------------------------------------------- */
132 
133 U_CAPI UChar32 U_EXPORT2
ucase_tolower(const UCaseProps * csp,UChar32 c)134 ucase_tolower(const UCaseProps *csp, UChar32 c) {
135     uint16_t props=UTRIE2_GET16(&csp->trie, c);
136     if(!PROPS_HAS_EXCEPTION(props)) {
137         if(UCASE_GET_TYPE(props)>=UCASE_UPPER) {
138             c+=UCASE_GET_DELTA(props);
139         }
140     } else {
141         const uint16_t *pe=GET_EXCEPTIONS(csp, props);
142         uint16_t excWord=*pe++;
143         if(HAS_SLOT(excWord, UCASE_EXC_LOWER)) {
144             GET_SLOT_VALUE(excWord, UCASE_EXC_LOWER, pe, c);
145         }
146     }
147     return c;
148 }
149 
150 U_CAPI UChar32 U_EXPORT2
ucase_toupper(const UCaseProps * csp,UChar32 c)151 ucase_toupper(const UCaseProps *csp, UChar32 c) {
152     uint16_t props=UTRIE2_GET16(&csp->trie, c);
153     if(!PROPS_HAS_EXCEPTION(props)) {
154         if(UCASE_GET_TYPE(props)==UCASE_LOWER) {
155             c+=UCASE_GET_DELTA(props);
156         }
157     } else {
158         const uint16_t *pe=GET_EXCEPTIONS(csp, props);
159         uint16_t excWord=*pe++;
160         if(HAS_SLOT(excWord, UCASE_EXC_UPPER)) {
161             GET_SLOT_VALUE(excWord, UCASE_EXC_UPPER, pe, c);
162         }
163     }
164     return c;
165 }
166 
167 U_CAPI UChar32 U_EXPORT2
ucase_totitle(const UCaseProps * csp,UChar32 c)168 ucase_totitle(const UCaseProps *csp, UChar32 c) {
169     uint16_t props=UTRIE2_GET16(&csp->trie, c);
170     if(!PROPS_HAS_EXCEPTION(props)) {
171         if(UCASE_GET_TYPE(props)==UCASE_LOWER) {
172             c+=UCASE_GET_DELTA(props);
173         }
174     } else {
175         const uint16_t *pe=GET_EXCEPTIONS(csp, props);
176         uint16_t excWord=*pe++;
177         int32_t idx;
178         if(HAS_SLOT(excWord, UCASE_EXC_TITLE)) {
179             idx=UCASE_EXC_TITLE;
180         } else if(HAS_SLOT(excWord, UCASE_EXC_UPPER)) {
181             idx=UCASE_EXC_UPPER;
182         } else {
183             return c;
184         }
185         GET_SLOT_VALUE(excWord, idx, pe, c);
186     }
187     return c;
188 }
189 
190 static const UChar iDot[2] = { 0x69, 0x307 };
191 static const UChar jDot[2] = { 0x6a, 0x307 };
192 static const UChar iOgonekDot[3] = { 0x12f, 0x307 };
193 static const UChar iDotGrave[3] = { 0x69, 0x307, 0x300 };
194 static const UChar iDotAcute[3] = { 0x69, 0x307, 0x301 };
195 static const UChar iDotTilde[3] = { 0x69, 0x307, 0x303 };
196 
197 
198 U_CFUNC void U_EXPORT2
ucase_addCaseClosure(const UCaseProps * csp,UChar32 c,const USetAdder * sa)199 ucase_addCaseClosure(const UCaseProps *csp, UChar32 c, const USetAdder *sa) {
200     uint16_t props;
201 
202     /*
203      * Hardcode the case closure of i and its relatives and ignore the
204      * data file data for these characters.
205      * The Turkic dotless i and dotted I with their case mapping conditions
206      * and case folding option make the related characters behave specially.
207      * This code matches their closure behavior to their case folding behavior.
208      */
209 
210     switch(c) {
211     case 0x49:
212         /* regular i and I are in one equivalence class */
213         sa->add(sa->set, 0x69);
214         return;
215     case 0x69:
216         sa->add(sa->set, 0x49);
217         return;
218     case 0x130:
219         /* dotted I is in a class with <0069 0307> (for canonical equivalence with <0049 0307>) */
220         sa->addString(sa->set, iDot, 2);
221         return;
222     case 0x131:
223         /* dotless i is in a class by itself */
224         return;
225     default:
226         /* otherwise use the data file data */
227         break;
228     }
229 
230     props=UTRIE2_GET16(&csp->trie, c);
231     if(!PROPS_HAS_EXCEPTION(props)) {
232         if(UCASE_GET_TYPE(props)!=UCASE_NONE) {
233             /* add the one simple case mapping, no matter what type it is */
234             int32_t delta=UCASE_GET_DELTA(props);
235             if(delta!=0) {
236                 sa->add(sa->set, c+delta);
237             }
238         }
239     } else {
240         /*
241          * c has exceptions, so there may be multiple simple and/or
242          * full case mappings. Add them all.
243          */
244         const uint16_t *pe0, *pe=GET_EXCEPTIONS(csp, props);
245         const UChar *closure;
246         uint16_t excWord=*pe++;
247         int32_t idx, closureLength, fullLength, length;
248 
249         pe0=pe;
250 
251         /* add all simple case mappings */
252         for(idx=UCASE_EXC_LOWER; idx<=UCASE_EXC_TITLE; ++idx) {
253             if(HAS_SLOT(excWord, idx)) {
254                 pe=pe0;
255                 GET_SLOT_VALUE(excWord, idx, pe, c);
256                 sa->add(sa->set, c);
257             }
258         }
259 
260         /* get the closure string pointer & length */
261         if(HAS_SLOT(excWord, UCASE_EXC_CLOSURE)) {
262             pe=pe0;
263             GET_SLOT_VALUE(excWord, UCASE_EXC_CLOSURE, pe, closureLength);
264             closureLength&=UCASE_CLOSURE_MAX_LENGTH; /* higher bits are reserved */
265             closure=(const UChar *)pe+1; /* behind this slot, unless there are full case mappings */
266         } else {
267             closureLength=0;
268             closure=NULL;
269         }
270 
271         /* add the full case folding */
272         if(HAS_SLOT(excWord, UCASE_EXC_FULL_MAPPINGS)) {
273             pe=pe0;
274             GET_SLOT_VALUE(excWord, UCASE_EXC_FULL_MAPPINGS, pe, fullLength);
275 
276             /* start of full case mapping strings */
277             ++pe;
278 
279             fullLength&=0xffff; /* bits 16 and higher are reserved */
280 
281             /* skip the lowercase result string */
282             pe+=fullLength&UCASE_FULL_LOWER;
283             fullLength>>=4;
284 
285             /* add the full case folding string */
286             length=fullLength&0xf;
287             if(length!=0) {
288                 sa->addString(sa->set, (const UChar *)pe, length);
289                 pe+=length;
290             }
291 
292             /* skip the uppercase and titlecase strings */
293             fullLength>>=4;
294             pe+=fullLength&0xf;
295             fullLength>>=4;
296             pe+=fullLength;
297 
298             closure=(const UChar *)pe; /* behind full case mappings */
299         }
300 
301         /* add each code point in the closure string */
302         for(idx=0; idx<closureLength;) {
303             U16_NEXT_UNSAFE(closure, idx, c);
304             sa->add(sa->set, c);
305         }
306     }
307 }
308 
309 /*
310  * compare s, which has a length, with t, which has a maximum length or is NUL-terminated
311  * must be length>0 and max>0 and length<=max
312  */
313 static inline int32_t
strcmpMax(const UChar * s,int32_t length,const UChar * t,int32_t max)314 strcmpMax(const UChar *s, int32_t length, const UChar *t, int32_t max) {
315     int32_t c1, c2;
316 
317     max-=length; /* we require length<=max, so no need to decrement max in the loop */
318     do {
319         c1=*s++;
320         c2=*t++;
321         if(c2==0) {
322             return 1; /* reached the end of t but not of s */
323         }
324         c1-=c2;
325         if(c1!=0) {
326             return c1; /* return difference result */
327         }
328     } while(--length>0);
329     /* ends with length==0 */
330 
331     if(max==0 || *t==0) {
332         return 0; /* equal to length of both strings */
333     } else {
334         return -max; /* return lengh difference */
335     }
336 }
337 
338 U_CFUNC UBool U_EXPORT2
ucase_addStringCaseClosure(const UCaseProps * csp,const UChar * s,int32_t length,const USetAdder * sa)339 ucase_addStringCaseClosure(const UCaseProps *csp, const UChar *s, int32_t length, const USetAdder *sa) {
340     int32_t i, start, limit, result, unfoldRows, unfoldRowWidth, unfoldStringWidth;
341 
342     if(csp->unfold==NULL || s==NULL) {
343         return FALSE; /* no reverse case folding data, or no string */
344     }
345     if(length<=1) {
346         /* the string is too short to find any match */
347         /*
348          * more precise would be:
349          * if(!u_strHasMoreChar32Than(s, length, 1))
350          * but this does not make much practical difference because
351          * a single supplementary code point would just not be found
352          */
353         return FALSE;
354     }
355 
356     const uint16_t *unfold=csp->unfold;
357     unfoldRows=unfold[UCASE_UNFOLD_ROWS];
358     unfoldRowWidth=unfold[UCASE_UNFOLD_ROW_WIDTH];
359     unfoldStringWidth=unfold[UCASE_UNFOLD_STRING_WIDTH];
360     unfold+=unfoldRowWidth;
361 
362     if(length>unfoldStringWidth) {
363         /* the string is too long to find any match */
364         return FALSE;
365     }
366 
367     /* do a binary search for the string */
368     start=0;
369     limit=unfoldRows;
370     while(start<limit) {
371         i=(start+limit)/2;
372         const UChar *p=reinterpret_cast<const UChar *>(unfold+(i*unfoldRowWidth));
373         result=strcmpMax(s, length, p, unfoldStringWidth);
374 
375         if(result==0) {
376             /* found the string: add each code point, and its case closure */
377             UChar32 c;
378 
379             for(i=unfoldStringWidth; i<unfoldRowWidth && p[i]!=0;) {
380                 U16_NEXT_UNSAFE(p, i, c);
381                 sa->add(sa->set, c);
382                 ucase_addCaseClosure(csp, c, sa);
383             }
384             return TRUE;
385         } else if(result<0) {
386             limit=i;
387         } else /* result>0 */ {
388             start=i+1;
389         }
390     }
391 
392     return FALSE; /* string not found */
393 }
394 
395 U_NAMESPACE_BEGIN
396 
FullCaseFoldingIterator()397 FullCaseFoldingIterator::FullCaseFoldingIterator()
398         : unfold(reinterpret_cast<const UChar *>(ucase_props_singleton.unfold)),
399           unfoldRows(unfold[UCASE_UNFOLD_ROWS]),
400           unfoldRowWidth(unfold[UCASE_UNFOLD_ROW_WIDTH]),
401           unfoldStringWidth(unfold[UCASE_UNFOLD_STRING_WIDTH]),
402           currentRow(0),
403           rowCpIndex(unfoldStringWidth) {
404     unfold+=unfoldRowWidth;
405 }
406 
407 UChar32
next(UnicodeString & full)408 FullCaseFoldingIterator::next(UnicodeString &full) {
409     // Advance past the last-delivered code point.
410     const UChar *p=unfold+(currentRow*unfoldRowWidth);
411     if(rowCpIndex>=unfoldRowWidth || p[rowCpIndex]==0) {
412         ++currentRow;
413         p+=unfoldRowWidth;
414         rowCpIndex=unfoldStringWidth;
415     }
416     if(currentRow>=unfoldRows) { return U_SENTINEL; }
417     // Set "full" to the NUL-terminated string in the first unfold column.
418     int32_t length=unfoldStringWidth;
419     while(length>0 && p[length-1]==0) { --length; }
420     full.setTo(FALSE, p, length);
421     // Return the code point.
422     UChar32 c;
423     U16_NEXT_UNSAFE(p, rowCpIndex, c);
424     return c;
425 }
426 
427 U_NAMESPACE_END
428 
429 /** @return UCASE_NONE, UCASE_LOWER, UCASE_UPPER, UCASE_TITLE */
430 U_CAPI int32_t U_EXPORT2
ucase_getType(const UCaseProps * csp,UChar32 c)431 ucase_getType(const UCaseProps *csp, UChar32 c) {
432     uint16_t props=UTRIE2_GET16(&csp->trie, c);
433     return UCASE_GET_TYPE(props);
434 }
435 
436 /** @return same as ucase_getType() and set bit 2 if c is case-ignorable */
437 U_CAPI int32_t U_EXPORT2
ucase_getTypeOrIgnorable(const UCaseProps * csp,UChar32 c)438 ucase_getTypeOrIgnorable(const UCaseProps *csp, UChar32 c) {
439     uint16_t props=UTRIE2_GET16(&csp->trie, c);
440     return UCASE_GET_TYPE_AND_IGNORABLE(props);
441 }
442 
443 /** @return UCASE_NO_DOT, UCASE_SOFT_DOTTED, UCASE_ABOVE, UCASE_OTHER_ACCENT */
444 static inline int32_t
getDotType(const UCaseProps * csp,UChar32 c)445 getDotType(const UCaseProps *csp, UChar32 c) {
446     uint16_t props=UTRIE2_GET16(&csp->trie, c);
447     if(!PROPS_HAS_EXCEPTION(props)) {
448         return props&UCASE_DOT_MASK;
449     } else {
450         const uint16_t *pe=GET_EXCEPTIONS(csp, props);
451         return (*pe>>UCASE_EXC_DOT_SHIFT)&UCASE_DOT_MASK;
452     }
453 }
454 
455 U_CAPI UBool U_EXPORT2
ucase_isSoftDotted(const UCaseProps * csp,UChar32 c)456 ucase_isSoftDotted(const UCaseProps *csp, UChar32 c) {
457     return (UBool)(getDotType(csp, c)==UCASE_SOFT_DOTTED);
458 }
459 
460 U_CAPI UBool U_EXPORT2
ucase_isCaseSensitive(const UCaseProps * csp,UChar32 c)461 ucase_isCaseSensitive(const UCaseProps *csp, UChar32 c) {
462     uint16_t props=UTRIE2_GET16(&csp->trie, c);
463     return (UBool)((props&UCASE_SENSITIVE)!=0);
464 }
465 
466 /* string casing ------------------------------------------------------------ */
467 
468 /*
469  * These internal functions form the core of string case mappings.
470  * They map single code points to result code points or strings and take
471  * all necessary conditions (context, locale ID, options) into account.
472  *
473  * They do not iterate over the source or write to the destination
474  * so that the same functions are useful for non-standard string storage,
475  * such as in a Replaceable (for Transliterator) or UTF-8/32 strings etc.
476  * For the same reason, the "surrounding text" context is passed in as a
477  * UCaseContextIterator which does not make any assumptions about
478  * the underlying storage.
479  *
480  * This section contains helper functions that check for conditions
481  * in the input text surrounding the current code point
482  * according to SpecialCasing.txt.
483  *
484  * Each helper function gets the index
485  * - after the current code point if it looks at following text
486  * - before the current code point if it looks at preceding text
487  *
488  * Unicode 3.2 UAX 21 "Case Mappings" defines the conditions as follows:
489  *
490  * Final_Sigma
491  *   C is preceded by a sequence consisting of
492  *     a cased letter and a case-ignorable sequence,
493  *   and C is not followed by a sequence consisting of
494  *     an ignorable sequence and then a cased letter.
495  *
496  * More_Above
497  *   C is followed by one or more characters of combining class 230 (ABOVE)
498  *   in the combining character sequence.
499  *
500  * After_Soft_Dotted
501  *   The last preceding character with combining class of zero before C
502  *   was Soft_Dotted,
503  *   and there is no intervening combining character class 230 (ABOVE).
504  *
505  * Before_Dot
506  *   C is followed by combining dot above (U+0307).
507  *   Any sequence of characters with a combining class that is neither 0 nor 230
508  *   may intervene between the current character and the combining dot above.
509  *
510  * The erratum from 2002-10-31 adds the condition
511  *
512  * After_I
513  *   The last preceding base character was an uppercase I, and there is no
514  *   intervening combining character class 230 (ABOVE).
515  *
516  *   (See Jitterbug 2344 and the comments on After_I below.)
517  *
518  * Helper definitions in Unicode 3.2 UAX 21:
519  *
520  * D1. A character C is defined to be cased
521  *     if it meets any of the following criteria:
522  *
523  *   - The general category of C is Titlecase Letter (Lt)
524  *   - In [CoreProps], C has one of the properties Uppercase, or Lowercase
525  *   - Given D = NFD(C), then it is not the case that:
526  *     D = UCD_lower(D) = UCD_upper(D) = UCD_title(D)
527  *     (This third criterium does not add any characters to the list
528  *      for Unicode 3.2. Ignored.)
529  *
530  * D2. A character C is defined to be case-ignorable
531  *     if it meets either of the following criteria:
532  *
533  *   - The general category of C is
534  *     Nonspacing Mark (Mn), or Enclosing Mark (Me), or Format Control (Cf), or
535  *     Letter Modifier (Lm), or Symbol Modifier (Sk)
536  *   - C is one of the following characters
537  *     U+0027 APOSTROPHE
538  *     U+00AD SOFT HYPHEN (SHY)
539  *     U+2019 RIGHT SINGLE QUOTATION MARK
540  *            (the preferred character for apostrophe)
541  *
542  * D3. A case-ignorable sequence is a sequence of
543  *     zero or more case-ignorable characters.
544  */
545 
546 #define is_a(c) ((c)=='a' || (c)=='A')
547 #define is_d(c) ((c)=='d' || (c)=='D')
548 #define is_e(c) ((c)=='e' || (c)=='E')
549 #define is_i(c) ((c)=='i' || (c)=='I')
550 #define is_l(c) ((c)=='l' || (c)=='L')
551 #define is_n(c) ((c)=='n' || (c)=='N')
552 #define is_r(c) ((c)=='r' || (c)=='R')
553 #define is_t(c) ((c)=='t' || (c)=='T')
554 #define is_u(c) ((c)=='u' || (c)=='U')
555 #define is_z(c) ((c)=='z' || (c)=='Z')
556 
557 /* separator? */
558 #define is_sep(c) ((c)=='_' || (c)=='-' || (c)==0)
559 
560 /**
561  * Requires non-NULL locale ID but otherwise does the equivalent of
562  * checking for language codes as if uloc_getLanguage() were called:
563  * Accepts both 2- and 3-letter codes and accepts case variants.
564  */
565 U_CFUNC int32_t
ucase_getCaseLocale(const char * locale,int32_t * locCache)566 ucase_getCaseLocale(const char *locale, int32_t *locCache) {
567     int32_t result;
568     char c;
569 
570     if(locCache!=NULL && (result=*locCache)!=UCASE_LOC_UNKNOWN) {
571         return result;
572     }
573 
574     result=UCASE_LOC_ROOT;
575 
576     /*
577      * This function used to use uloc_getLanguage(), but the current code
578      * removes the dependency of this low-level code on uloc implementation code
579      * and is faster because not the whole locale ID has to be
580      * examined and copied/transformed.
581      *
582      * Because this code does not want to depend on uloc, the caller must
583      * pass in a non-NULL locale, i.e., may need to call uloc_getDefault().
584      */
585     c=*locale++;
586     if(is_t(c)) {
587         /* tr or tur? */
588         c=*locale++;
589         if(is_u(c)) {
590             c=*locale++;
591         }
592         if(is_r(c)) {
593             c=*locale;
594             if(is_sep(c)) {
595                 result=UCASE_LOC_TURKISH;
596             }
597         }
598     } else if(is_a(c)) {
599         /* az or aze? */
600         c=*locale++;
601         if(is_z(c)) {
602             c=*locale++;
603             if(is_e(c)) {
604                 c=*locale;
605             }
606             if(is_sep(c)) {
607                 result=UCASE_LOC_TURKISH;
608             }
609         }
610     } else if(is_l(c)) {
611         /* lt or lit? */
612         c=*locale++;
613         if(is_i(c)) {
614             c=*locale++;
615         }
616         if(is_t(c)) {
617             c=*locale;
618             if(is_sep(c)) {
619                 result=UCASE_LOC_LITHUANIAN;
620             }
621         }
622     } else if(is_n(c)) {
623         /* nl or nld? */
624         c=*locale++;
625         if(is_l(c)) {
626             c=*locale++;
627             if(is_d(c)) {
628                 c=*locale;
629             }
630             if(is_sep(c)) {
631                 result=UCASE_LOC_DUTCH;
632             }
633         }
634     }
635 
636     if(locCache!=NULL) {
637         *locCache=result;
638     }
639     return result;
640 }
641 
642 /*
643  * Is followed by
644  *   {case-ignorable}* cased
645  * ?
646  * (dir determines looking forward/backward)
647  * If a character is case-ignorable, it is skipped regardless of whether
648  * it is also cased or not.
649  */
650 static UBool
isFollowedByCasedLetter(const UCaseProps * csp,UCaseContextIterator * iter,void * context,int8_t dir)651 isFollowedByCasedLetter(const UCaseProps *csp, UCaseContextIterator *iter, void *context, int8_t dir) {
652     UChar32 c;
653 
654     if(iter==NULL) {
655         return FALSE;
656     }
657 
658     for(/* dir!=0 sets direction */; (c=iter(context, dir))>=0; dir=0) {
659         int32_t type=ucase_getTypeOrIgnorable(csp, c);
660         if(type&4) {
661             /* case-ignorable, continue with the loop */
662         } else if(type!=UCASE_NONE) {
663             return TRUE; /* followed by cased letter */
664         } else {
665             return FALSE; /* uncased and not case-ignorable */
666         }
667     }
668 
669     return FALSE; /* not followed by cased letter */
670 }
671 
672 /* Is preceded by Soft_Dotted character with no intervening cc=230 ? */
673 static UBool
isPrecededBySoftDotted(const UCaseProps * csp,UCaseContextIterator * iter,void * context)674 isPrecededBySoftDotted(const UCaseProps *csp, UCaseContextIterator *iter, void *context) {
675     UChar32 c;
676     int32_t dotType;
677     int8_t dir;
678 
679     if(iter==NULL) {
680         return FALSE;
681     }
682 
683     for(dir=-1; (c=iter(context, dir))>=0; dir=0) {
684         dotType=getDotType(csp, c);
685         if(dotType==UCASE_SOFT_DOTTED) {
686             return TRUE; /* preceded by TYPE_i */
687         } else if(dotType!=UCASE_OTHER_ACCENT) {
688             return FALSE; /* preceded by different base character (not TYPE_i), or intervening cc==230 */
689         }
690     }
691 
692     return FALSE; /* not preceded by TYPE_i */
693 }
694 
695 /*
696  * See Jitterbug 2344:
697  * The condition After_I for Turkic-lowercasing of U+0307 combining dot above
698  * is checked in ICU 2.0, 2.1, 2.6 but was not in 2.2 & 2.4 because
699  * we made those releases compatible with Unicode 3.2 which had not fixed
700  * a related bug in SpecialCasing.txt.
701  *
702  * From the Jitterbug 2344 text:
703  * ... this bug is listed as a Unicode erratum
704  * from 2002-10-31 at http://www.unicode.org/uni2errata/UnicodeErrata.html
705  * <quote>
706  * There are two errors in SpecialCasing.txt.
707  * 1. Missing semicolons on two lines. ... [irrelevant for ICU]
708  * 2. An incorrect context definition. Correct as follows:
709  * < 0307; ; 0307; 0307; tr After_Soft_Dotted; # COMBINING DOT ABOVE
710  * < 0307; ; 0307; 0307; az After_Soft_Dotted; # COMBINING DOT ABOVE
711  * ---
712  * > 0307; ; 0307; 0307; tr After_I; # COMBINING DOT ABOVE
713  * > 0307; ; 0307; 0307; az After_I; # COMBINING DOT ABOVE
714  * where the context After_I is defined as:
715  * The last preceding base character was an uppercase I, and there is no
716  * intervening combining character class 230 (ABOVE).
717  * </quote>
718  *
719  * Note that SpecialCasing.txt even in Unicode 3.2 described the condition as:
720  *
721  * # When lowercasing, remove dot_above in the sequence I + dot_above, which will turn into i.
722  * # This matches the behavior of the canonically equivalent I-dot_above
723  *
724  * See also the description in this place in older versions of uchar.c (revision 1.100).
725  *
726  * Markus W. Scherer 2003-feb-15
727  */
728 
729 /* Is preceded by base character 'I' with no intervening cc=230 ? */
730 static UBool
isPrecededBy_I(const UCaseProps * csp,UCaseContextIterator * iter,void * context)731 isPrecededBy_I(const UCaseProps *csp, UCaseContextIterator *iter, void *context) {
732     UChar32 c;
733     int32_t dotType;
734     int8_t dir;
735 
736     if(iter==NULL) {
737         return FALSE;
738     }
739 
740     for(dir=-1; (c=iter(context, dir))>=0; dir=0) {
741         if(c==0x49) {
742             return TRUE; /* preceded by I */
743         }
744         dotType=getDotType(csp, c);
745         if(dotType!=UCASE_OTHER_ACCENT) {
746             return FALSE; /* preceded by different base character (not I), or intervening cc==230 */
747         }
748     }
749 
750     return FALSE; /* not preceded by I */
751 }
752 
753 /* Is followed by one or more cc==230 ? */
754 static UBool
isFollowedByMoreAbove(const UCaseProps * csp,UCaseContextIterator * iter,void * context)755 isFollowedByMoreAbove(const UCaseProps *csp, UCaseContextIterator *iter, void *context) {
756     UChar32 c;
757     int32_t dotType;
758     int8_t dir;
759 
760     if(iter==NULL) {
761         return FALSE;
762     }
763 
764     for(dir=1; (c=iter(context, dir))>=0; dir=0) {
765         dotType=getDotType(csp, c);
766         if(dotType==UCASE_ABOVE) {
767             return TRUE; /* at least one cc==230 following */
768         } else if(dotType!=UCASE_OTHER_ACCENT) {
769             return FALSE; /* next base character, no more cc==230 following */
770         }
771     }
772 
773     return FALSE; /* no more cc==230 following */
774 }
775 
776 /* Is followed by a dot above (without cc==230 in between) ? */
777 static UBool
isFollowedByDotAbove(const UCaseProps * csp,UCaseContextIterator * iter,void * context)778 isFollowedByDotAbove(const UCaseProps *csp, UCaseContextIterator *iter, void *context) {
779     UChar32 c;
780     int32_t dotType;
781     int8_t dir;
782 
783     if(iter==NULL) {
784         return FALSE;
785     }
786 
787     for(dir=1; (c=iter(context, dir))>=0; dir=0) {
788         if(c==0x307) {
789             return TRUE;
790         }
791         dotType=getDotType(csp, c);
792         if(dotType!=UCASE_OTHER_ACCENT) {
793             return FALSE; /* next base character or cc==230 in between */
794         }
795     }
796 
797     return FALSE; /* no dot above following */
798 }
799 
800 U_CAPI int32_t U_EXPORT2
ucase_toFullLower(const UCaseProps * csp,UChar32 c,UCaseContextIterator * iter,void * context,const UChar ** pString,const char * locale,int32_t * locCache)801 ucase_toFullLower(const UCaseProps *csp, UChar32 c,
802                   UCaseContextIterator *iter, void *context,
803                   const UChar **pString,
804                   const char *locale, int32_t *locCache)
805 {
806     UChar32 result=c;
807     uint16_t props=UTRIE2_GET16(&csp->trie, c);
808     if(!PROPS_HAS_EXCEPTION(props)) {
809         if(UCASE_GET_TYPE(props)>=UCASE_UPPER) {
810             result=c+UCASE_GET_DELTA(props);
811         }
812     } else {
813         const uint16_t *pe=GET_EXCEPTIONS(csp, props), *pe2;
814         uint16_t excWord=*pe++;
815         int32_t full;
816 
817         pe2=pe;
818 
819         if(excWord&UCASE_EXC_CONDITIONAL_SPECIAL) {
820             /* use hardcoded conditions and mappings */
821             int32_t loc=ucase_getCaseLocale(locale, locCache);
822 
823             /*
824              * Test for conditional mappings first
825              *   (otherwise the unconditional default mappings are always taken),
826              * then test for characters that have unconditional mappings in SpecialCasing.txt,
827              * then get the UnicodeData.txt mappings.
828              */
829             if( loc==UCASE_LOC_LITHUANIAN &&
830                     /* base characters, find accents above */
831                     (((c==0x49 || c==0x4a || c==0x12e) &&
832                         isFollowedByMoreAbove(csp, iter, context)) ||
833                     /* precomposed with accent above, no need to find one */
834                     (c==0xcc || c==0xcd || c==0x128))
835             ) {
836                 /*
837                     # Lithuanian
838 
839                     # Lithuanian retains the dot in a lowercase i when followed by accents.
840 
841                     # Introduce an explicit dot above when lowercasing capital I's and J's
842                     # whenever there are more accents above.
843                     # (of the accents used in Lithuanian: grave, acute, tilde above, and ogonek)
844 
845                     0049; 0069 0307; 0049; 0049; lt More_Above; # LATIN CAPITAL LETTER I
846                     004A; 006A 0307; 004A; 004A; lt More_Above; # LATIN CAPITAL LETTER J
847                     012E; 012F 0307; 012E; 012E; lt More_Above; # LATIN CAPITAL LETTER I WITH OGONEK
848                     00CC; 0069 0307 0300; 00CC; 00CC; lt; # LATIN CAPITAL LETTER I WITH GRAVE
849                     00CD; 0069 0307 0301; 00CD; 00CD; lt; # LATIN CAPITAL LETTER I WITH ACUTE
850                     0128; 0069 0307 0303; 0128; 0128; lt; # LATIN CAPITAL LETTER I WITH TILDE
851                  */
852                 switch(c) {
853                 case 0x49:  /* LATIN CAPITAL LETTER I */
854                     *pString=iDot;
855                     return 2;
856                 case 0x4a:  /* LATIN CAPITAL LETTER J */
857                     *pString=jDot;
858                     return 2;
859                 case 0x12e: /* LATIN CAPITAL LETTER I WITH OGONEK */
860                     *pString=iOgonekDot;
861                     return 2;
862                 case 0xcc:  /* LATIN CAPITAL LETTER I WITH GRAVE */
863                     *pString=iDotGrave;
864                     return 3;
865                 case 0xcd:  /* LATIN CAPITAL LETTER I WITH ACUTE */
866                     *pString=iDotAcute;
867                     return 3;
868                 case 0x128: /* LATIN CAPITAL LETTER I WITH TILDE */
869                     *pString=iDotTilde;
870                     return 3;
871                 default:
872                     return 0; /* will not occur */
873                 }
874             /* # Turkish and Azeri */
875             } else if(loc==UCASE_LOC_TURKISH && c==0x130) {
876                 /*
877                     # I and i-dotless; I-dot and i are case pairs in Turkish and Azeri
878                     # The following rules handle those cases.
879 
880                     0130; 0069; 0130; 0130; tr # LATIN CAPITAL LETTER I WITH DOT ABOVE
881                     0130; 0069; 0130; 0130; az # LATIN CAPITAL LETTER I WITH DOT ABOVE
882                  */
883                 return 0x69;
884             } else if(loc==UCASE_LOC_TURKISH && c==0x307 && isPrecededBy_I(csp, iter, context)) {
885                 /*
886                     # When lowercasing, remove dot_above in the sequence I + dot_above, which will turn into i.
887                     # This matches the behavior of the canonically equivalent I-dot_above
888 
889                     0307; ; 0307; 0307; tr After_I; # COMBINING DOT ABOVE
890                     0307; ; 0307; 0307; az After_I; # COMBINING DOT ABOVE
891                  */
892                 return 0; /* remove the dot (continue without output) */
893             } else if(loc==UCASE_LOC_TURKISH && c==0x49 && !isFollowedByDotAbove(csp, iter, context)) {
894                 /*
895                     # When lowercasing, unless an I is before a dot_above, it turns into a dotless i.
896 
897                     0049; 0131; 0049; 0049; tr Not_Before_Dot; # LATIN CAPITAL LETTER I
898                     0049; 0131; 0049; 0049; az Not_Before_Dot; # LATIN CAPITAL LETTER I
899                  */
900                 return 0x131;
901             } else if(c==0x130) {
902                 /*
903                     # Preserve canonical equivalence for I with dot. Turkic is handled below.
904 
905                     0130; 0069 0307; 0130; 0130; # LATIN CAPITAL LETTER I WITH DOT ABOVE
906                  */
907                 *pString=iDot;
908                 return 2;
909             } else if(  c==0x3a3 &&
910                         !isFollowedByCasedLetter(csp, iter, context, 1) &&
911                         isFollowedByCasedLetter(csp, iter, context, -1) /* -1=preceded */
912             ) {
913                 /* greek capital sigma maps depending on surrounding cased letters (see SpecialCasing.txt) */
914                 /*
915                     # Special case for final form of sigma
916 
917                     03A3; 03C2; 03A3; 03A3; Final_Sigma; # GREEK CAPITAL LETTER SIGMA
918                  */
919                 return 0x3c2; /* greek small final sigma */
920             } else {
921                 /* no known conditional special case mapping, use a normal mapping */
922             }
923         } else if(HAS_SLOT(excWord, UCASE_EXC_FULL_MAPPINGS)) {
924             GET_SLOT_VALUE(excWord, UCASE_EXC_FULL_MAPPINGS, pe, full);
925             full&=UCASE_FULL_LOWER;
926             if(full!=0) {
927                 /* set the output pointer to the lowercase mapping */
928                 *pString=reinterpret_cast<const UChar *>(pe+1);
929 
930                 /* return the string length */
931                 return full;
932             }
933         }
934 
935         if(HAS_SLOT(excWord, UCASE_EXC_LOWER)) {
936             GET_SLOT_VALUE(excWord, UCASE_EXC_LOWER, pe2, result);
937         }
938     }
939 
940     return (result==c) ? ~result : result;
941 }
942 
943 /* internal */
944 static int32_t
toUpperOrTitle(const UCaseProps * csp,UChar32 c,UCaseContextIterator * iter,void * context,const UChar ** pString,const char * locale,int32_t * locCache,UBool upperNotTitle)945 toUpperOrTitle(const UCaseProps *csp, UChar32 c,
946                UCaseContextIterator *iter, void *context,
947                const UChar **pString,
948                const char *locale, int32_t *locCache,
949                UBool upperNotTitle) {
950     UChar32 result=c;
951     uint16_t props=UTRIE2_GET16(&csp->trie, c);
952     if(!PROPS_HAS_EXCEPTION(props)) {
953         if(UCASE_GET_TYPE(props)==UCASE_LOWER) {
954             result=c+UCASE_GET_DELTA(props);
955         }
956     } else {
957         const uint16_t *pe=GET_EXCEPTIONS(csp, props), *pe2;
958         uint16_t excWord=*pe++;
959         int32_t full, idx;
960 
961         pe2=pe;
962 
963         if(excWord&UCASE_EXC_CONDITIONAL_SPECIAL) {
964             /* use hardcoded conditions and mappings */
965             int32_t loc=ucase_getCaseLocale(locale, locCache);
966 
967             if(loc==UCASE_LOC_TURKISH && c==0x69) {
968                 /*
969                     # Turkish and Azeri
970 
971                     # I and i-dotless; I-dot and i are case pairs in Turkish and Azeri
972                     # The following rules handle those cases.
973 
974                     # When uppercasing, i turns into a dotted capital I
975 
976                     0069; 0069; 0130; 0130; tr; # LATIN SMALL LETTER I
977                     0069; 0069; 0130; 0130; az; # LATIN SMALL LETTER I
978                 */
979                 return 0x130;
980             } else if(loc==UCASE_LOC_LITHUANIAN && c==0x307 && isPrecededBySoftDotted(csp, iter, context)) {
981                 /*
982                     # Lithuanian
983 
984                     # Lithuanian retains the dot in a lowercase i when followed by accents.
985 
986                     # Remove DOT ABOVE after "i" with upper or titlecase
987 
988                     0307; 0307; ; ; lt After_Soft_Dotted; # COMBINING DOT ABOVE
989                  */
990                 return 0; /* remove the dot (continue without output) */
991             } else {
992                 /* no known conditional special case mapping, use a normal mapping */
993             }
994         } else if(HAS_SLOT(excWord, UCASE_EXC_FULL_MAPPINGS)) {
995             GET_SLOT_VALUE(excWord, UCASE_EXC_FULL_MAPPINGS, pe, full);
996 
997             /* start of full case mapping strings */
998             ++pe;
999 
1000             /* skip the lowercase and case-folding result strings */
1001             pe+=full&UCASE_FULL_LOWER;
1002             full>>=4;
1003             pe+=full&0xf;
1004             full>>=4;
1005 
1006             if(upperNotTitle) {
1007                 full&=0xf;
1008             } else {
1009                 /* skip the uppercase result string */
1010                 pe+=full&0xf;
1011                 full=(full>>4)&0xf;
1012             }
1013 
1014             if(full!=0) {
1015                 /* set the output pointer to the result string */
1016                 *pString=reinterpret_cast<const UChar *>(pe);
1017 
1018                 /* return the string length */
1019                 return full;
1020             }
1021         }
1022 
1023         if(!upperNotTitle && HAS_SLOT(excWord, UCASE_EXC_TITLE)) {
1024             idx=UCASE_EXC_TITLE;
1025         } else if(HAS_SLOT(excWord, UCASE_EXC_UPPER)) {
1026             /* here, titlecase is same as uppercase */
1027             idx=UCASE_EXC_UPPER;
1028         } else {
1029             return ~c;
1030         }
1031         GET_SLOT_VALUE(excWord, idx, pe2, result);
1032     }
1033 
1034     return (result==c) ? ~result : result;
1035 }
1036 
1037 U_CAPI int32_t U_EXPORT2
ucase_toFullUpper(const UCaseProps * csp,UChar32 c,UCaseContextIterator * iter,void * context,const UChar ** pString,const char * locale,int32_t * locCache)1038 ucase_toFullUpper(const UCaseProps *csp, UChar32 c,
1039                   UCaseContextIterator *iter, void *context,
1040                   const UChar **pString,
1041                   const char *locale, int32_t *locCache) {
1042     return toUpperOrTitle(csp, c, iter, context, pString, locale, locCache, TRUE);
1043 }
1044 
1045 U_CAPI int32_t U_EXPORT2
ucase_toFullTitle(const UCaseProps * csp,UChar32 c,UCaseContextIterator * iter,void * context,const UChar ** pString,const char * locale,int32_t * locCache)1046 ucase_toFullTitle(const UCaseProps *csp, UChar32 c,
1047                   UCaseContextIterator *iter, void *context,
1048                   const UChar **pString,
1049                   const char *locale, int32_t *locCache) {
1050     return toUpperOrTitle(csp, c, iter, context, pString, locale, locCache, FALSE);
1051 }
1052 
1053 /* case folding ------------------------------------------------------------- */
1054 
1055 /*
1056  * Case folding is similar to lowercasing.
1057  * The result may be a simple mapping, i.e., a single code point, or
1058  * a full mapping, i.e., a string.
1059  * If the case folding for a code point is the same as its simple (1:1) lowercase mapping,
1060  * then only the lowercase mapping is stored.
1061  *
1062  * Some special cases are hardcoded because their conditions cannot be
1063  * parsed and processed from CaseFolding.txt.
1064  *
1065  * Unicode 3.2 CaseFolding.txt specifies for its status field:
1066 
1067 # C: common case folding, common mappings shared by both simple and full mappings.
1068 # F: full case folding, mappings that cause strings to grow in length. Multiple characters are separated by spaces.
1069 # S: simple case folding, mappings to single characters where different from F.
1070 # T: special case for uppercase I and dotted uppercase I
1071 #    - For non-Turkic languages, this mapping is normally not used.
1072 #    - For Turkic languages (tr, az), this mapping can be used instead of the normal mapping for these characters.
1073 #
1074 # Usage:
1075 #  A. To do a simple case folding, use the mappings with status C + S.
1076 #  B. To do a full case folding, use the mappings with status C + F.
1077 #
1078 #    The mappings with status T can be used or omitted depending on the desired case-folding
1079 #    behavior. (The default option is to exclude them.)
1080 
1081  * Unicode 3.2 has 'T' mappings as follows:
1082 
1083 0049; T; 0131; # LATIN CAPITAL LETTER I
1084 0130; T; 0069; # LATIN CAPITAL LETTER I WITH DOT ABOVE
1085 
1086  * while the default mappings for these code points are:
1087 
1088 0049; C; 0069; # LATIN CAPITAL LETTER I
1089 0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE
1090 
1091  * U+0130 has no simple case folding (simple-case-folds to itself).
1092  */
1093 
1094 /* return the simple case folding mapping for c */
1095 U_CAPI UChar32 U_EXPORT2
ucase_fold(const UCaseProps * csp,UChar32 c,uint32_t options)1096 ucase_fold(const UCaseProps *csp, UChar32 c, uint32_t options) {
1097     uint16_t props=UTRIE2_GET16(&csp->trie, c);
1098     if(!PROPS_HAS_EXCEPTION(props)) {
1099         if(UCASE_GET_TYPE(props)>=UCASE_UPPER) {
1100             c+=UCASE_GET_DELTA(props);
1101         }
1102     } else {
1103         const uint16_t *pe=GET_EXCEPTIONS(csp, props);
1104         uint16_t excWord=*pe++;
1105         int32_t idx;
1106         if(excWord&UCASE_EXC_CONDITIONAL_FOLD) {
1107             /* special case folding mappings, hardcoded */
1108             if((options&_FOLD_CASE_OPTIONS_MASK)==U_FOLD_CASE_DEFAULT) {
1109                 /* default mappings */
1110                 if(c==0x49) {
1111                     /* 0049; C; 0069; # LATIN CAPITAL LETTER I */
1112                     return 0x69;
1113                 } else if(c==0x130) {
1114                     /* no simple case folding for U+0130 */
1115                     return c;
1116                 }
1117             } else {
1118                 /* Turkic mappings */
1119                 if(c==0x49) {
1120                     /* 0049; T; 0131; # LATIN CAPITAL LETTER I */
1121                     return 0x131;
1122                 } else if(c==0x130) {
1123                     /* 0130; T; 0069; # LATIN CAPITAL LETTER I WITH DOT ABOVE */
1124                     return 0x69;
1125                 }
1126             }
1127         }
1128         if(HAS_SLOT(excWord, UCASE_EXC_FOLD)) {
1129             idx=UCASE_EXC_FOLD;
1130         } else if(HAS_SLOT(excWord, UCASE_EXC_LOWER)) {
1131             idx=UCASE_EXC_LOWER;
1132         } else {
1133             return c;
1134         }
1135         GET_SLOT_VALUE(excWord, idx, pe, c);
1136     }
1137     return c;
1138 }
1139 
1140 /*
1141  * Issue for canonical caseless match (UAX #21):
1142  * Turkic casefolding (using "T" mappings in CaseFolding.txt) does not preserve
1143  * canonical equivalence, unlike default-option casefolding.
1144  * For example, I-grave and I + grave fold to strings that are not canonically
1145  * equivalent.
1146  * For more details, see the comment in unorm_compare() in unorm.cpp
1147  * and the intermediate prototype changes for Jitterbug 2021.
1148  * (For example, revision 1.104 of uchar.c and 1.4 of CaseFolding.txt.)
1149  *
1150  * This did not get fixed because it appears that it is not possible to fix
1151  * it for uppercase and lowercase characters (I-grave vs. i-grave)
1152  * together in a way that they still fold to common result strings.
1153  */
1154 
1155 U_CAPI int32_t U_EXPORT2
ucase_toFullFolding(const UCaseProps * csp,UChar32 c,const UChar ** pString,uint32_t options)1156 ucase_toFullFolding(const UCaseProps *csp, UChar32 c,
1157                     const UChar **pString,
1158                     uint32_t options)
1159 {
1160     UChar32 result=c;
1161     uint16_t props=UTRIE2_GET16(&csp->trie, c);
1162     if(!PROPS_HAS_EXCEPTION(props)) {
1163         if(UCASE_GET_TYPE(props)>=UCASE_UPPER) {
1164             result=c+UCASE_GET_DELTA(props);
1165         }
1166     } else {
1167         const uint16_t *pe=GET_EXCEPTIONS(csp, props), *pe2;
1168         uint16_t excWord=*pe++;
1169         int32_t full, idx;
1170 
1171         pe2=pe;
1172 
1173         if(excWord&UCASE_EXC_CONDITIONAL_FOLD) {
1174             /* use hardcoded conditions and mappings */
1175             if((options&_FOLD_CASE_OPTIONS_MASK)==U_FOLD_CASE_DEFAULT) {
1176                 /* default mappings */
1177                 if(c==0x49) {
1178                     /* 0049; C; 0069; # LATIN CAPITAL LETTER I */
1179                     return 0x69;
1180                 } else if(c==0x130) {
1181                     /* 0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE */
1182                     *pString=iDot;
1183                     return 2;
1184                 }
1185             } else {
1186                 /* Turkic mappings */
1187                 if(c==0x49) {
1188                     /* 0049; T; 0131; # LATIN CAPITAL LETTER I */
1189                     return 0x131;
1190                 } else if(c==0x130) {
1191                     /* 0130; T; 0069; # LATIN CAPITAL LETTER I WITH DOT ABOVE */
1192                     return 0x69;
1193                 }
1194             }
1195         } else if(HAS_SLOT(excWord, UCASE_EXC_FULL_MAPPINGS)) {
1196             GET_SLOT_VALUE(excWord, UCASE_EXC_FULL_MAPPINGS, pe, full);
1197 
1198             /* start of full case mapping strings */
1199             ++pe;
1200 
1201             /* skip the lowercase result string */
1202             pe+=full&UCASE_FULL_LOWER;
1203             full=(full>>4)&0xf;
1204 
1205             if(full!=0) {
1206                 /* set the output pointer to the result string */
1207                 *pString=reinterpret_cast<const UChar *>(pe);
1208 
1209                 /* return the string length */
1210                 return full;
1211             }
1212         }
1213 
1214         if(HAS_SLOT(excWord, UCASE_EXC_FOLD)) {
1215             idx=UCASE_EXC_FOLD;
1216         } else if(HAS_SLOT(excWord, UCASE_EXC_LOWER)) {
1217             idx=UCASE_EXC_LOWER;
1218         } else {
1219             return ~c;
1220         }
1221         GET_SLOT_VALUE(excWord, idx, pe2, result);
1222     }
1223 
1224     return (result==c) ? ~result : result;
1225 }
1226 
1227 /* case mapping properties API ---------------------------------------------- */
1228 
1229 #define GET_CASE_PROPS() &ucase_props_singleton
1230 
1231 /* public API (see uchar.h) */
1232 
1233 U_CAPI UBool U_EXPORT2
u_isULowercase(UChar32 c)1234 u_isULowercase(UChar32 c) {
1235     return (UBool)(UCASE_LOWER==ucase_getType(GET_CASE_PROPS(), c));
1236 }
1237 
1238 U_CAPI UBool U_EXPORT2
u_isUUppercase(UChar32 c)1239 u_isUUppercase(UChar32 c) {
1240     return (UBool)(UCASE_UPPER==ucase_getType(GET_CASE_PROPS(), c));
1241 }
1242 
1243 /* Transforms the Unicode character to its lower case equivalent.*/
1244 U_CAPI UChar32 U_EXPORT2
u_tolower(UChar32 c)1245 u_tolower(UChar32 c) {
1246     return ucase_tolower(GET_CASE_PROPS(), c);
1247 }
1248 
1249 /* Transforms the Unicode character to its upper case equivalent.*/
1250 U_CAPI UChar32 U_EXPORT2
u_toupper(UChar32 c)1251 u_toupper(UChar32 c) {
1252     return ucase_toupper(GET_CASE_PROPS(), c);
1253 }
1254 
1255 /* Transforms the Unicode character to its title case equivalent.*/
1256 U_CAPI UChar32 U_EXPORT2
u_totitle(UChar32 c)1257 u_totitle(UChar32 c) {
1258     return ucase_totitle(GET_CASE_PROPS(), c);
1259 }
1260 
1261 /* return the simple case folding mapping for c */
1262 U_CAPI UChar32 U_EXPORT2
u_foldCase(UChar32 c,uint32_t options)1263 u_foldCase(UChar32 c, uint32_t options) {
1264     return ucase_fold(GET_CASE_PROPS(), c, options);
1265 }
1266 
1267 U_CFUNC int32_t U_EXPORT2
ucase_hasBinaryProperty(UChar32 c,UProperty which)1268 ucase_hasBinaryProperty(UChar32 c, UProperty which) {
1269     /* case mapping properties */
1270     const UChar *resultString;
1271     int32_t locCache;
1272     const UCaseProps *csp=GET_CASE_PROPS();
1273     if(csp==NULL) {
1274         return FALSE;
1275     }
1276     switch(which) {
1277     case UCHAR_LOWERCASE:
1278         return (UBool)(UCASE_LOWER==ucase_getType(csp, c));
1279     case UCHAR_UPPERCASE:
1280         return (UBool)(UCASE_UPPER==ucase_getType(csp, c));
1281     case UCHAR_SOFT_DOTTED:
1282         return ucase_isSoftDotted(csp, c);
1283     case UCHAR_CASE_SENSITIVE:
1284         return ucase_isCaseSensitive(csp, c);
1285     case UCHAR_CASED:
1286         return (UBool)(UCASE_NONE!=ucase_getType(csp, c));
1287     case UCHAR_CASE_IGNORABLE:
1288         return (UBool)(ucase_getTypeOrIgnorable(csp, c)>>2);
1289     /*
1290      * Note: The following Changes_When_Xyz are defined as testing whether
1291      * the NFD form of the input changes when Xyz-case-mapped.
1292      * However, this simpler implementation of these properties,
1293      * ignoring NFD, passes the tests.
1294      * The implementation needs to be changed if the tests start failing.
1295      * When that happens, optimizations should be used to work with the
1296      * per-single-code point ucase_toFullXyz() functions unless
1297      * the NFD form has more than one code point,
1298      * and the property starts set needs to be the union of the
1299      * start sets for normalization and case mappings.
1300      */
1301     case UCHAR_CHANGES_WHEN_LOWERCASED:
1302         locCache=UCASE_LOC_ROOT;
1303         return (UBool)(ucase_toFullLower(csp, c, NULL, NULL, &resultString, "", &locCache)>=0);
1304     case UCHAR_CHANGES_WHEN_UPPERCASED:
1305         locCache=UCASE_LOC_ROOT;
1306         return (UBool)(ucase_toFullUpper(csp, c, NULL, NULL, &resultString, "", &locCache)>=0);
1307     case UCHAR_CHANGES_WHEN_TITLECASED:
1308         locCache=UCASE_LOC_ROOT;
1309         return (UBool)(ucase_toFullTitle(csp, c, NULL, NULL, &resultString, "", &locCache)>=0);
1310     /* case UCHAR_CHANGES_WHEN_CASEFOLDED: -- in uprops.c */
1311     case UCHAR_CHANGES_WHEN_CASEMAPPED:
1312         locCache=UCASE_LOC_ROOT;
1313         return (UBool)(
1314             ucase_toFullLower(csp, c, NULL, NULL, &resultString, "", &locCache)>=0 ||
1315             ucase_toFullUpper(csp, c, NULL, NULL, &resultString, "", &locCache)>=0 ||
1316             ucase_toFullTitle(csp, c, NULL, NULL, &resultString, "", &locCache)>=0);
1317     default:
1318         return FALSE;
1319     }
1320 }
1321