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) 2002-2016, International Business Machines
7 * Corporation and others. All Rights Reserved.
8 *
9 *******************************************************************************
10 * file name: uprops.h
11 * encoding: UTF-8
12 * tab size: 8 (not used)
13 * indentation:4
14 *
15 * created on: 2002feb24
16 * created by: Markus W. Scherer
17 *
18 * Constants for mostly non-core Unicode character properties
19 * stored in uprops.icu.
20 */
21
22 #ifndef __UPROPS_H__
23 #define __UPROPS_H__
24
25 #include "unicode/utypes.h"
26 #include "unicode/uset.h"
27 #include "uset_imp.h"
28 #include "udataswp.h"
29
30 /* indexes[] entries */
31 enum {
32 UPROPS_PROPS32_INDEX,
33 UPROPS_EXCEPTIONS_INDEX,
34 UPROPS_EXCEPTIONS_TOP_INDEX,
35
36 UPROPS_ADDITIONAL_TRIE_INDEX,
37 UPROPS_ADDITIONAL_VECTORS_INDEX,
38 UPROPS_ADDITIONAL_VECTORS_COLUMNS_INDEX,
39
40 UPROPS_SCRIPT_EXTENSIONS_INDEX,
41
42 UPROPS_RESERVED_INDEX_7,
43 UPROPS_RESERVED_INDEX_8,
44
45 /* size of the data file (number of 32-bit units after the header) */
46 UPROPS_DATA_TOP_INDEX,
47
48 /* maximum values for code values in vector word 0 */
49 UPROPS_MAX_VALUES_INDEX=10,
50 /* maximum values for code values in vector word 2 */
51 UPROPS_MAX_VALUES_2_INDEX,
52
53 UPROPS_INDEX_COUNT=16
54 };
55
56 /* definitions for the main properties words */
57 enum {
58 /* general category shift==0 0 (5 bits) */
59 /* reserved 5 (1 bit) */
60 UPROPS_NUMERIC_TYPE_VALUE_SHIFT=6 /* 6 (10 bits) */
61 };
62
63 #define GET_CATEGORY(props) ((props)&0x1f)
64 #define CAT_MASK(props) U_MASK(GET_CATEGORY(props))
65
66 #define GET_NUMERIC_TYPE_VALUE(props) ((props)>>UPROPS_NUMERIC_TYPE_VALUE_SHIFT)
67
68 /* constants for the storage form of numeric types and values */
69 enum {
70 /** No numeric value. */
71 UPROPS_NTV_NONE=0,
72 /** Decimal digits: nv=0..9 */
73 UPROPS_NTV_DECIMAL_START=1,
74 /** Other digits: nv=0..9 */
75 UPROPS_NTV_DIGIT_START=11,
76 /** Small integers: nv=0..154 */
77 UPROPS_NTV_NUMERIC_START=21,
78 /** Fractions: ((ntv>>4)-12) / ((ntv&0xf)+1) = -1..17 / 1..16 */
79 UPROPS_NTV_FRACTION_START=0xb0,
80 /**
81 * Large integers:
82 * ((ntv>>5)-14) * 10^((ntv&0x1f)+2) = (1..9)*(10^2..10^33)
83 * (only one significant decimal digit)
84 */
85 UPROPS_NTV_LARGE_START=0x1e0,
86 /**
87 * Sexagesimal numbers:
88 * ((ntv>>2)-0xbf) * 60^((ntv&3)+1) = (1..9)*(60^1..60^4)
89 */
90 UPROPS_NTV_BASE60_START=0x300,
91 /**
92 * Fraction-20 values:
93 * frac20 = ntv-0x324 = 0..0x17 -> 1|3|5|7 / 20|40|80|160|320|640
94 * numerator: num = 2*(frac20&3)+1
95 * denominator: den = 20<<(frac20>>2)
96 */
97 UPROPS_NTV_FRACTION20_START=UPROPS_NTV_BASE60_START+36, // 0x300+9*4=0x324
98 /**
99 * Fraction-32 values:
100 * frac32 = ntv-0x34c = 0..15 -> 1|3|5|7 / 32|64|128|256
101 * numerator: num = 2*(frac32&3)+1
102 * denominator: den = 32<<(frac32>>2)
103 */
104 UPROPS_NTV_FRACTION32_START=UPROPS_NTV_FRACTION20_START+24, // 0x324+6*4=0x34c
105 /** No numeric value (yet). */
106 UPROPS_NTV_RESERVED_START=UPROPS_NTV_FRACTION32_START+16, // 0x34c+4*4=0x35c
107
108 UPROPS_NTV_MAX_SMALL_INT=UPROPS_NTV_FRACTION_START-UPROPS_NTV_NUMERIC_START-1
109 };
110
111 #define UPROPS_NTV_GET_TYPE(ntv) \
112 ((ntv==UPROPS_NTV_NONE) ? U_NT_NONE : \
113 (ntv<UPROPS_NTV_DIGIT_START) ? U_NT_DECIMAL : \
114 (ntv<UPROPS_NTV_NUMERIC_START) ? U_NT_DIGIT : \
115 U_NT_NUMERIC)
116
117 /* number of properties vector words */
118 #define UPROPS_VECTOR_WORDS 3
119
120 /*
121 * Properties in vector word 0
122 * Bits
123 * 31..24 DerivedAge version major/minor one nibble each
124 * 23..22 3..1: Bits 21..20 & 7..0 = Script_Extensions index
125 * 3: Script value from Script_Extensions
126 * 2: Script=Inherited
127 * 1: Script=Common
128 * 0: Script=bits 21..20 & 7..0
129 * 21..20 Bits 9..8 of the UScriptCode, or index to Script_Extensions
130 * 19..17 East Asian Width
131 * 16.. 8 UBlockCode
132 * 7.. 0 UScriptCode, or index to Script_Extensions
133 */
134
135 /* derived age: one nibble each for major and minor version numbers */
136 #define UPROPS_AGE_MASK 0xff000000
137 #define UPROPS_AGE_SHIFT 24
138
139 /* Script_Extensions: mask includes Script */
140 #define UPROPS_SCRIPT_X_MASK 0x00f000ff
141 #define UPROPS_SCRIPT_X_SHIFT 22
142
143 // The UScriptCode or Script_Extensions index is split across two bit fields.
144 // (Starting with Unicode 13/ICU 66/2019 due to more varied Script_Extensions.)
145 // Shift the high bits right by 12 to assemble the full value.
146 #define UPROPS_SCRIPT_HIGH_MASK 0x00300000
147 #define UPROPS_SCRIPT_HIGH_SHIFT 12
148 #define UPROPS_MAX_SCRIPT 0x3ff
149
150 #define UPROPS_EA_MASK 0x000e0000
151 #define UPROPS_EA_SHIFT 17
152
153 #define UPROPS_BLOCK_MASK 0x0001ff00
154 #define UPROPS_BLOCK_SHIFT 8
155
156 #define UPROPS_SCRIPT_LOW_MASK 0x000000ff
157
158 /* UPROPS_SCRIPT_X_WITH_COMMON must be the lowest value that involves Script_Extensions. */
159 #define UPROPS_SCRIPT_X_WITH_COMMON 0x400000
160 #define UPROPS_SCRIPT_X_WITH_INHERITED 0x800000
161 #define UPROPS_SCRIPT_X_WITH_OTHER 0xc00000
162
163 #ifdef __cplusplus
164
165 namespace {
166
uprops_mergeScriptCodeOrIndex(uint32_t scriptX)167 inline uint32_t uprops_mergeScriptCodeOrIndex(uint32_t scriptX) {
168 return
169 ((scriptX & UPROPS_SCRIPT_HIGH_MASK) >> UPROPS_SCRIPT_HIGH_SHIFT) |
170 (scriptX & UPROPS_SCRIPT_LOW_MASK);
171 }
172
173 } // namespace
174
175 #endif // __cplusplus
176
177 /*
178 * Properties in vector word 1
179 * Each bit encodes one binary property.
180 * The following constants represent the bit number, use 1<<UPROPS_XYZ.
181 * UPROPS_BINARY_1_TOP<=32!
182 *
183 * Keep this list of property enums in sync with
184 * propListNames[] in icu/source/tools/genprops/props2.c!
185 *
186 * ICU 2.6/uprops format version 3.2 stores full properties instead of "Other_".
187 */
188 enum {
189 UPROPS_WHITE_SPACE,
190 UPROPS_DASH,
191 UPROPS_HYPHEN,
192 UPROPS_QUOTATION_MARK,
193 UPROPS_TERMINAL_PUNCTUATION,
194 UPROPS_MATH,
195 UPROPS_HEX_DIGIT,
196 UPROPS_ASCII_HEX_DIGIT,
197 UPROPS_ALPHABETIC,
198 UPROPS_IDEOGRAPHIC,
199 UPROPS_DIACRITIC,
200 UPROPS_EXTENDER,
201 UPROPS_NONCHARACTER_CODE_POINT,
202 UPROPS_GRAPHEME_EXTEND,
203 UPROPS_GRAPHEME_LINK,
204 UPROPS_IDS_BINARY_OPERATOR,
205 UPROPS_IDS_TRINARY_OPERATOR,
206 UPROPS_RADICAL,
207 UPROPS_UNIFIED_IDEOGRAPH,
208 UPROPS_DEFAULT_IGNORABLE_CODE_POINT,
209 UPROPS_DEPRECATED,
210 UPROPS_LOGICAL_ORDER_EXCEPTION,
211 UPROPS_XID_START,
212 UPROPS_XID_CONTINUE,
213 UPROPS_ID_START, /* ICU 2.6, uprops format version 3.2 */
214 UPROPS_ID_CONTINUE,
215 UPROPS_GRAPHEME_BASE,
216 UPROPS_S_TERM, /* new in ICU 3.0 and Unicode 4.0.1 */
217 UPROPS_VARIATION_SELECTOR,
218 UPROPS_PATTERN_SYNTAX, /* new in ICU 3.4 and Unicode 4.1 */
219 UPROPS_PATTERN_WHITE_SPACE,
220 UPROPS_PREPENDED_CONCATENATION_MARK, // new in ICU 60 and Unicode 10
221 UPROPS_BINARY_1_TOP /* ==32 - full! */
222 };
223
224 /*
225 * Properties in vector word 2
226 * Bits
227 * 31..26 ICU 75: Identifier_Type bit set
228 * ICU 70..74: unused
229 * ICU 57..69: emoji properties; moved to uemoji.icu in ICU 70
230 * 25..20 Line Break
231 * 19..15 Sentence Break
232 * 14..10 Word Break
233 * 9.. 5 Grapheme Cluster Break
234 * 4.. 0 Decomposition Type
235 */
236
237 #ifdef __cplusplus
238
239 // https://www.unicode.org/reports/tr39/#Identifier_Status_and_Type
240 // The Identifier_Type maps each code point to a *set* of one or more values.
241 // Some can be combined with others, some can only occur alone.
242 // Exclusion & Limited_Use are combinable bits, but cannot occur together.
243 // We use this forbidden combination for enumerated values.
244 // We use 6 bits for all possible combinations.
245 // If more combinable values are added, then we need to use more bits.
246 //
247 // We do not store separate data for Identifier_Status:
248 // We can derive that from the encoded Identifier_Type via a simple range check.
249
250 inline constexpr uint32_t UPROPS_2_ID_TYPE_MASK = 0xfc000000;
251 inline constexpr int32_t UPROPS_2_ID_TYPE_SHIFT = 26;
252
253 enum {
254 // A high bit for use in idTypeToEncoded[] but not used in the data
255 UPROPS_ID_TYPE_BIT = 0x80,
256
257 // Combinable bits
258 UPROPS_ID_TYPE_EXCLUSION = 0x20,
259 UPROPS_ID_TYPE_LIMITED_USE = 0x10,
260 UPROPS_ID_TYPE_UNCOMMON_USE = 8,
261 UPROPS_ID_TYPE_TECHNICAL = 4,
262 UPROPS_ID_TYPE_OBSOLETE = 2,
263 UPROPS_ID_TYPE_NOT_XID = 1,
264
265 // Exclusive values
266 UPROPS_ID_TYPE_NOT_CHARACTER = 0,
267
268 // Forbidden bit combination used for enumerating other exclusive values
269 UPROPS_ID_TYPE_FORBIDDEN = UPROPS_ID_TYPE_EXCLUSION | UPROPS_ID_TYPE_LIMITED_USE, // 0x30
270 UPROPS_ID_TYPE_DEPRECATED = UPROPS_ID_TYPE_FORBIDDEN, // 0x30
271 UPROPS_ID_TYPE_DEFAULT_IGNORABLE, // 0x31
272 UPROPS_ID_TYPE_NOT_NFKC, // 0x32
273
274 UPROPS_ID_TYPE_ALLOWED_MIN = UPROPS_ID_TYPE_FORBIDDEN + 0xc, // 0x3c
275 UPROPS_ID_TYPE_INCLUSION = UPROPS_ID_TYPE_FORBIDDEN + 0xe, // 0x3e
276 UPROPS_ID_TYPE_RECOMMENDED = UPROPS_ID_TYPE_FORBIDDEN + 0xf, // 0x3f
277 };
278
279 /**
280 * Maps UIdentifierType to encoded bits.
281 * When UPROPS_ID_TYPE_BIT is set, then use "&" to test whether the value bit is set.
282 * When UPROPS_ID_TYPE_BIT is not set, then compare ("==") the array value with the data value.
283 */
284 inline constexpr uint8_t uprops_idTypeToEncoded[] = {
285 UPROPS_ID_TYPE_NOT_CHARACTER,
286 UPROPS_ID_TYPE_DEPRECATED,
287 UPROPS_ID_TYPE_DEFAULT_IGNORABLE,
288 UPROPS_ID_TYPE_NOT_NFKC,
289 UPROPS_ID_TYPE_BIT | UPROPS_ID_TYPE_NOT_XID,
290 UPROPS_ID_TYPE_BIT | UPROPS_ID_TYPE_EXCLUSION,
291 UPROPS_ID_TYPE_BIT | UPROPS_ID_TYPE_OBSOLETE,
292 UPROPS_ID_TYPE_BIT | UPROPS_ID_TYPE_TECHNICAL,
293 UPROPS_ID_TYPE_BIT | UPROPS_ID_TYPE_UNCOMMON_USE,
294 UPROPS_ID_TYPE_BIT | UPROPS_ID_TYPE_LIMITED_USE,
295 UPROPS_ID_TYPE_INCLUSION,
296 UPROPS_ID_TYPE_RECOMMENDED
297 };
298
299 #endif // __cplusplus
300
301 #define UPROPS_LB_MASK 0x03f00000
302 #define UPROPS_LB_SHIFT 20
303
304 #define UPROPS_SB_MASK 0x000f8000
305 #define UPROPS_SB_SHIFT 15
306
307 #define UPROPS_WB_MASK 0x00007c00
308 #define UPROPS_WB_SHIFT 10
309
310 #define UPROPS_GCB_MASK 0x000003e0
311 #define UPROPS_GCB_SHIFT 5
312
313 #define UPROPS_DT_MASK 0x0000001f
314
315 /**
316 * Gets the main properties value for a code point.
317 * Implemented in uchar.c for uprops.cpp.
318 */
319 U_CFUNC uint32_t
320 u_getMainProperties(UChar32 c);
321
322 /**
323 * Get a properties vector word for a code point.
324 * Implemented in uchar.c for uprops.cpp.
325 * @return 0 if no data or illegal argument
326 */
327 U_CFUNC uint32_t
328 u_getUnicodeProperties(UChar32 c, int32_t column);
329
330 /**
331 * Get the the maximum values for some enum/int properties.
332 * Use the same column numbers as for u_getUnicodeProperties().
333 * The returned value will contain maximum values stored in the same bit fields
334 * as where the enum values are stored in the u_getUnicodeProperties()
335 * return values for the same columns.
336 *
337 * Valid columns are those for properties words that contain enumerated values.
338 * (ICU 2.6: columns 0 and 2)
339 * For other column numbers, this function will return 0.
340 *
341 * @internal
342 */
343 U_CFUNC int32_t
344 uprv_getMaxValues(int32_t column);
345
346 /**
347 * Checks if c is alphabetic, or a decimal digit; implements UCHAR_POSIX_ALNUM.
348 * @internal
349 */
350 U_CFUNC UBool
351 u_isalnumPOSIX(UChar32 c);
352
353 /**
354 * Checks if c is in
355 * [^\p{space}\p{gc=Control}\p{gc=Surrogate}\p{gc=Unassigned}]
356 * with space=\p{Whitespace} and Control=Cc.
357 * Implements UCHAR_POSIX_GRAPH.
358 * @internal
359 */
360 U_CFUNC UBool
361 u_isgraphPOSIX(UChar32 c);
362
363 /**
364 * Checks if c is in \p{graph}\p{blank} - \p{cntrl}.
365 * Implements UCHAR_POSIX_PRINT.
366 * @internal
367 */
368 U_CFUNC UBool
369 u_isprintPOSIX(UChar32 c);
370
371 /** Some code points. @internal */
372 enum {
373 TAB =0x0009,
374 LF =0x000a,
375 FF =0x000c,
376 CR =0x000d,
377 NBSP =0x00a0,
378 CGJ =0x034f,
379 FIGURESP=0x2007,
380 HAIRSP =0x200a,
381 ZWNJ =0x200c,
382 ZWJ =0x200d,
383 RLM =0x200f,
384 NNBSP =0x202f,
385 ZWNBSP =0xfeff
386 };
387
388 /**
389 * Get the maximum length of a (regular/1.0/extended) character name.
390 * @return 0 if no character names available.
391 */
392 U_CAPI int32_t U_EXPORT2
393 uprv_getMaxCharNameLength(void);
394
395 /**
396 * Fills set with characters that are used in Unicode character names.
397 * Includes all characters that are used in regular/Unicode 1.0/extended names.
398 * Just empties the set if no character names are available.
399 * @param sa USetAdder to receive characters.
400 */
401 U_CAPI void U_EXPORT2
402 uprv_getCharNameCharacters(const USetAdder *sa);
403
404 /**
405 * Constants for which data and implementation files provide which properties.
406 * Used by UnicodeSet for service-specific property enumeration.
407 * @internal
408 */
409 enum UPropertySource {
410 /** No source, not a supported property. */
411 UPROPS_SRC_NONE,
412 /** From uchar.c/uprops.icu main trie */
413 UPROPS_SRC_CHAR,
414 /** From uchar.c/uprops.icu properties vectors trie */
415 UPROPS_SRC_PROPSVEC,
416 /** From unames.c/unames.icu */
417 UPROPS_SRC_NAMES,
418 /** From ucase.c/ucase.icu */
419 UPROPS_SRC_CASE,
420 /** From ubidi_props.c/ubidi.icu */
421 UPROPS_SRC_BIDI,
422 /** From uchar.c/uprops.icu main trie as well as properties vectors trie */
423 UPROPS_SRC_CHAR_AND_PROPSVEC,
424 /** From ucase.c/ucase.icu as well as unorm.cpp/unorm.icu */
425 UPROPS_SRC_CASE_AND_NORM,
426 /** From normalizer2impl.cpp/nfc.nrm */
427 UPROPS_SRC_NFC,
428 /** From normalizer2impl.cpp/nfkc.nrm */
429 UPROPS_SRC_NFKC,
430 /** From normalizer2impl.cpp/nfkc_cf.nrm */
431 UPROPS_SRC_NFKC_CF,
432 /** From normalizer2impl.cpp/nfc.nrm canonical iterator data */
433 UPROPS_SRC_NFC_CANON_ITER,
434 // Text layout properties.
435 UPROPS_SRC_INPC,
436 UPROPS_SRC_INSC,
437 UPROPS_SRC_VO,
438 UPROPS_SRC_EMOJI,
439 UPROPS_SRC_IDSU,
440 UPROPS_SRC_ID_COMPAT_MATH,
441 /** One more than the highest UPropertySource (UPROPS_SRC_) constant. */
442 UPROPS_SRC_COUNT
443 };
444 typedef enum UPropertySource UPropertySource;
445
446 /**
447 * @see UPropertySource
448 * @internal
449 */
450 U_CFUNC UPropertySource U_EXPORT2
451 uprops_getSource(UProperty which);
452
453 /**
454 * Enumerate uprops.icu's main data trie and add the
455 * start of each range of same properties to the set.
456 * @internal
457 */
458 U_CFUNC void U_EXPORT2
459 uchar_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
460
461 /**
462 * Enumerate uprops.icu's properties vectors trie and add the
463 * start of each range of same properties to the set.
464 * @internal
465 */
466 U_CFUNC void U_EXPORT2
467 upropsvec_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
468
469 U_CFUNC void U_EXPORT2
470 uprops_addPropertyStarts(UPropertySource src, const USetAdder *sa, UErrorCode *pErrorCode);
471
472 /**
473 * Return a set of characters for property enumeration.
474 * For each two consecutive characters (start, limit) in the set,
475 * all of the properties for start..limit-1 are all the same.
476 *
477 * @param sa USetAdder to receive result. Existing contents are lost.
478 * @internal
479 */
480 /*U_CFUNC void U_EXPORT2
481 uprv_getInclusions(const USetAdder *sa, UErrorCode *pErrorCode);
482 */
483
484 /**
485 * Swap the ICU Unicode character names file. See uchar.c.
486 * @internal
487 */
488 U_CAPI int32_t U_EXPORT2
489 uchar_swapNames(const UDataSwapper *ds,
490 const void *inData, int32_t length, void *outData,
491 UErrorCode *pErrorCode);
492
493 #ifdef __cplusplus
494
495 U_NAMESPACE_BEGIN
496
497 class UnicodeSet;
498
499 class CharacterProperties {
500 public:
501 CharacterProperties() = delete;
502 static const UnicodeSet *getInclusionsForProperty(UProperty prop, UErrorCode &errorCode);
503 static const UnicodeSet *getBinaryPropertySet(UProperty property, UErrorCode &errorCode);
504 };
505
506 // implemented in uniset_props.cpp
507 U_CFUNC UnicodeSet *
508 uniset_getUnicode32Instance(UErrorCode &errorCode);
509
510 U_NAMESPACE_END
511
512 #endif
513
514 #endif
515