-
Notifications
You must be signed in to change notification settings - Fork 10.2k
/
StringGraphemeBreaking.swift
1131 lines (983 loc) · 36.4 KB
/
StringGraphemeBreaking.swift
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
//===----------------------------------------------------------------------===//
//
// This source file is part of the Swift.org open source project
//
// Copyright (c) 2014 - 2023 Apple Inc. and the Swift project authors
// Licensed under Apache License v2.0 with Runtime Library Exception
//
// See https://swift.org/LICENSE.txt for license information
// See https://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
//
//===----------------------------------------------------------------------===//
import SwiftShims
/// CR and LF are common special cases in grapheme breaking logic
private var _CR: UInt8 { return 0x0d }
private var _LF: UInt8 { return 0x0a }
internal func _hasGraphemeBreakBetween(
_ lhs: Unicode.Scalar, _ rhs: Unicode.Scalar
) -> Bool {
// CR-LF is a special case: no break between these
if lhs == Unicode.Scalar(_CR) && rhs == Unicode.Scalar(_LF) {
return false
}
// Whether the given scalar, when it appears paired with another scalar
// satisfying this property, has a grapheme break between it and the other
// scalar.
func hasBreakWhenPaired(_ x: Unicode.Scalar) -> Bool {
// TODO: This doesn't generate optimal code, tune/re-write at a lower
// level.
//
// NOTE: Order of case ranges affects codegen, and thus performance. All
// things being equal, keep existing order below.
switch x.value {
// Unified CJK Han ideographs, common and some supplemental, amongst
// others:
// U+3400 ~ U+A4CF
case 0x3400...0xa4cf: return true
// Repeat sub-300 check, this is beneficial for common cases of Latin
// characters embedded within non-Latin script (e.g. newlines, spaces,
// proper nouns and/or jargon, punctuation).
//
// NOTE: CR-LF special case has already been checked.
case 0x0000...0x02ff: return true
// Non-combining kana:
// U+3041 ~ U+3096
// U+30A1 ~ U+30FC
case 0x3041...0x3096: return true
case 0x30a1...0x30fc: return true
// Non-combining modern (and some archaic) Cyrillic:
// U+0400 ~ U+0482 (first half of Cyrillic block)
case 0x0400...0x0482: return true
// Modern Arabic, excluding extenders and prependers:
// U+061D ~ U+064A
case 0x061d...0x064a: return true
// Precomposed Hangul syllables:
// U+AC00 ~ U+D7AF
case 0xac00...0xd7af: return true
// Common general use punctuation, excluding extenders:
// U+2010 ~ U+2029
case 0x2010...0x2029: return true
// CJK punctuation characters, excluding extenders:
// U+3000 ~ U+3029
case 0x3000...0x3029: return true
// Full-width forms:
// U+FF01 ~ U+FF9D
case 0xFF01...0xFF9D: return true
default: return false
}
}
return hasBreakWhenPaired(lhs) && hasBreakWhenPaired(rhs)
}
extension _StringGuts {
@inline(__always)
internal func roundDownToNearestCharacter(
_ i: String.Index
) -> String.Index {
_internalInvariant(i._isScalarAligned)
_internalInvariant(hasMatchingEncoding(i))
_internalInvariant(i._encodedOffset <= count)
let offset = i._encodedOffset
if _fastPath(i._isCharacterAligned) { return i }
if offset == 0 || offset == count { return i._characterAligned }
return _slowRoundDownToNearestCharacter(i)
}
@inline(never)
internal func _slowRoundDownToNearestCharacter(
_ i: String.Index
) -> String.Index {
let offset = i._encodedOffset
let start = offset - _opaqueCharacterStride(endingAt: offset)
let stride = _opaqueCharacterStride(startingAt: start)
_internalInvariant(offset <= start + stride,
"Grapheme breaking inconsistency")
if offset >= start + stride {
// Already aligned, or grapheme breaking returned an unexpected result.
return i._characterAligned
}
let r = String.Index(encodedOffset: start, characterStride: stride)
return markEncoding(r._characterAligned)
}
@inline(__always)
internal func roundDownToNearestCharacter(
_ i: String.Index,
in bounds: Range<String.Index>
) -> String.Index {
_internalInvariant(
bounds.lowerBound._isScalarAligned && bounds.upperBound._isScalarAligned)
_internalInvariant(
hasMatchingEncoding(bounds.lowerBound)
&& hasMatchingEncoding(bounds.upperBound))
_internalInvariant(bounds.upperBound <= endIndex)
_internalInvariant(i._isScalarAligned)
_internalInvariant(hasMatchingEncoding(i))
_internalInvariant(i >= bounds.lowerBound && i <= bounds.upperBound)
// We can only use the `_isCharacterAligned` bit if the start index is also
// character-aligned.
if _fastPath(
bounds.lowerBound._isCharacterAligned && i._isCharacterAligned
) {
return i
}
if i == bounds.lowerBound || i == bounds.upperBound { return i }
return _slowRoundDownToNearestCharacter(i, in: bounds)
}
@inline(never)
internal func _slowRoundDownToNearestCharacter(
_ i: String.Index,
in bounds: Range<String.Index>
) -> String.Index {
let offset = i._encodedOffset
let offsetBounds = bounds._encodedOffsetRange
let prior =
offset - _opaqueCharacterStride(endingAt: offset, in: offsetBounds)
let stride = _opaqueCharacterStride(startingAt: prior)
_internalInvariant(offset <= prior + stride,
"Grapheme breaking inconsistency")
if offset >= prior + stride {
// Already aligned, or grapheme breaking returned an unexpected result.
return i
}
var r = String.Index(encodedOffset: prior, characterStride: stride)
if bounds.lowerBound._isCharacterAligned {
r = r._characterAligned
} else {
r = r._scalarAligned
}
return markEncoding(r)
}
}
extension _StringGuts {
@usableFromInline @inline(never)
@_effects(releasenone)
internal func isOnGraphemeClusterBoundary(_ i: String.Index) -> Bool {
if i._isCharacterAligned { return true }
guard i.transcodedOffset == 0 else { return false }
let offset = i._encodedOffset
if offset == 0 || offset == self.count { return true }
guard isOnUnicodeScalarBoundary(i) else { return false }
let nearest = roundDownToNearestCharacter(i._scalarAligned)
return i == nearest
}
}
extension _StringGuts {
/// Return the length of the extended grapheme cluster starting at offset `i`,
/// assuming it falls on a grapheme cluster boundary.
///
/// Note: This does not look behind at data preceding `i`, so if `i` is not on
/// a grapheme cluster boundary, then it may return results that are
/// inconsistent with `_opaqueCharacterStride(endingAt:)`. On the other hand,
/// this behavior makes this suitable for use in substrings whose start index
/// itself does not fall on a cluster boundary.
@usableFromInline @inline(__always)
@_effects(releasenone)
internal func _opaqueCharacterStride(startingAt i: Int) -> Int {
_internalInvariant(i < endIndex._encodedOffset)
if isFastUTF8 {
let fast = withFastUTF8 { utf8 in
if i &+ 1 == utf8.count { return true }
let pair = UnsafeRawPointer(
utf8.baseAddress.unsafelyUnwrapped
).loadUnaligned(fromByteOffset: i, as: UInt16.self)
//& 0x8080 == 0 is "both not ASCII", != 0x0A0D is "not CRLF"
return pair & 0x8080 == 0 && pair != 0x0A0D
}
if _fastPath(fast) {
_internalInvariant(_opaqueComplexCharacterStride(startingAt: i) == 1)
return 1
}
}
return _opaqueComplexCharacterStride(startingAt: i)
}
@_effects(releasenone) @inline(never)
internal func _opaqueComplexCharacterStride(startingAt i: Int) -> Int {
if _slowPath(isForeign) {
return _foreignOpaqueCharacterStride(startingAt: i)
}
let nextIdx = withFastUTF8 { utf8 in
nextBoundary(startingAt: i) { j in
_internalInvariant(j >= 0)
guard j < utf8.count else { return nil }
let (scalar, len) = _decodeScalar(utf8, startingAt: j)
return (scalar, j &+ len)
}
}
return nextIdx &- i
}
/// Return the length of the extended grapheme cluster ending at offset `i`,
/// or if `i` happens to be in the middle of a grapheme cluster, find and
/// return the distance to its start.
///
/// Note: unlike `_opaqueCharacterStride(startingAt:)`, this method always
/// finds a correct grapheme cluster boundary.
@usableFromInline @inline(__always)
@_effects(releasenone)
internal func _opaqueCharacterStride(endingAt i: Int) -> Int {
if i <= 1 {
return i
}
if isFastUTF8 {
let fast = withFastUTF8 { utf8 in
let pair = UnsafeRawPointer(
utf8.baseAddress.unsafelyUnwrapped
).loadUnaligned(fromByteOffset: i &- 2, as: UInt16.self)
//& 0x8080 == 0 is "both not ASCII", != 0x0A0D is "not CRLF"
return pair & 0x8080 == 0 && pair != 0x0A0D
}
if _fastPath(fast) {
_internalInvariant(_opaqueComplexCharacterStride(endingAt: i) == 1)
return 1
}
}
return _opaqueComplexCharacterStride(endingAt: i)
}
@_effects(releasenone) @inline(never)
internal func _opaqueComplexCharacterStride(endingAt i: Int) -> Int {
if _slowPath(isForeign) {
return _foreignOpaqueCharacterStride(endingAt: i)
}
let previousIdx = withFastUTF8 { utf8 in
previousBoundary(endingAt: i) { j in
_internalInvariant(j <= utf8.count)
guard j > 0 else { return nil }
let (scalar, len) = _decodeScalar(utf8, endingAt: j)
return (scalar, j &- len)
}
}
return i &- previousIdx
}
/// Return the length of the extended grapheme cluster ending at offset `i` in
/// bounds, or if `i` happens to be in the middle of a grapheme cluster, find
/// and return the distance to its start.
///
/// Note: unlike `_opaqueCharacterStride(startingAt:)`, this method always
/// finds a correct grapheme cluster boundary within the substring defined by
/// the specified bounds.
@_effects(releasenone)
internal func _opaqueCharacterStride(
endingAt i: Int,
in bounds: Range<Int>
) -> Int {
_internalInvariant(i > bounds.lowerBound && i <= bounds.upperBound)
if _slowPath(isForeign) {
return _foreignOpaqueCharacterStride(endingAt: i, in: bounds)
}
let previousIdx = withFastUTF8 { utf8 in
previousBoundary(endingAt: i) { j in
_internalInvariant(j <= bounds.upperBound)
guard j > bounds.lowerBound else { return nil }
let (scalar, len) = _decodeScalar(utf8, endingAt: j)
return (scalar, j &- len)
}
}
_internalInvariant(bounds.contains(previousIdx))
return i &- previousIdx
}
@inline(never)
@_effects(releasenone)
private func _foreignOpaqueCharacterStride(startingAt i: Int) -> Int {
#if _runtime(_ObjC)
_internalInvariant(isForeign)
let nextIdx = nextBoundary(startingAt: i) { j in
_internalInvariant(j >= 0)
guard j < count else { return nil }
let scalars = String.UnicodeScalarView(self)
let idx = String.Index(_encodedOffset: j)
let scalar = scalars[idx]
let nextIdx = scalars.index(after: idx)
return (scalar, nextIdx._encodedOffset)
}
return nextIdx &- i
#else
fatalError("No foreign strings on Linux in this version of Swift")
#endif
}
@inline(never)
@_effects(releasenone)
private func _foreignOpaqueCharacterStride(
startingAt i: Int,
in bounds: Range<Int>
) -> Int {
#if _runtime(_ObjC)
_internalInvariant(isForeign)
_internalInvariant(bounds.contains(i))
let nextIdx = nextBoundary(startingAt: i) { j in
_internalInvariant(j >= bounds.lowerBound)
guard j < bounds.upperBound else { return nil }
let scalars = String.UnicodeScalarView(self)
let idx = String.Index(_encodedOffset: j)
let scalar = scalars[idx]
let nextIdx = scalars.index(after: idx)
return (scalar, nextIdx._encodedOffset)
}
return nextIdx &- i
#else
fatalError("No foreign strings on Linux in this version of Swift")
#endif
}
@inline(never)
@_effects(releasenone)
private func _foreignOpaqueCharacterStride(endingAt i: Int) -> Int {
#if _runtime(_ObjC)
_internalInvariant(isForeign)
let previousIdx = previousBoundary(endingAt: i) { j in
_internalInvariant(j <= self.count)
guard j > 0 else { return nil }
let scalars = String.UnicodeScalarView(self)
let idx = String.Index(_encodedOffset: j)
let previousIdx = scalars.index(before: idx)
let scalar = scalars[previousIdx]
return (scalar, previousIdx._encodedOffset)
}
return i &- previousIdx
#else
fatalError("No foreign strings on Linux in this version of Swift")
#endif
}
@inline(never)
@_effects(releasenone)
private func _foreignOpaqueCharacterStride(
endingAt i: Int,
in bounds: Range<Int>
) -> Int {
#if _runtime(_ObjC)
_internalInvariant(isForeign)
_internalInvariant(i > bounds.lowerBound && i <= bounds.upperBound)
let previousIdx = previousBoundary(endingAt: i) { j in
_internalInvariant(j <= bounds.upperBound)
guard j > bounds.lowerBound else { return nil }
let scalars = String.UnicodeScalarView(self)
let idx = String.Index(_encodedOffset: j)
let previousIdx = scalars.index(before: idx)
let scalar = scalars[previousIdx]
return (scalar, previousIdx._encodedOffset)
}
return i &- previousIdx
#else
fatalError("No foreign strings on Linux in this version of Swift")
#endif
}
}
extension Unicode.Scalar {
fileprivate var _isLinkingConsonant: Bool {
_swift_stdlib_isLinkingConsonant(value)
}
fileprivate var _isVirama: Bool {
switch value {
// Devanagari
case 0x94D:
return true
// Bengali
case 0x9CD:
return true
// Gujarati
case 0xACD:
return true
// Oriya
case 0xB4D:
return true
// Telugu
case 0xC4D:
return true
// Malayalam
case 0xD4D:
return true
default:
return false
}
}
}
internal struct _GraphemeBreakingState: Sendable, Equatable {
// When we're looking through an indic sequence, one of the requirements is
// that there is at LEAST 1 Virama present between two linking consonants.
// This value helps ensure that when we ultimately need to decide whether or
// not to break that we've at least seen 1 when walking.
var hasSeenVirama = false
// When walking forwards in a string, we need to know whether or not we've
// entered an emoji sequence to be able to eventually break after all of the
// emoji's various extenders and zero width joiners. This bit allows us to
// keep track of whether or not we're still in an emoji sequence when deciding
// to break.
var isInEmojiSequence = false
// Similar to emoji sequences, we need to know not to break an Indic grapheme
// sequence. This sequence is (potentially) composed of many scalars and isn't
// as trivial as comparing two grapheme properties.
var isInIndicSequence = false
// When walking forward in a string, we need to not break on emoji flag
// sequences. Emoji flag sequences are composed of 2 regional indicators, so
// when we see our first (.regionalIndicator, .regionalIndicator) decision,
// we need to know to return false in this case. However, if the next scalar
// is another regional indicator, we reach the same decision rule, but in this
// case we actually need to break there's a boundary between emoji flag
// sequences.
var shouldBreakRI = false
}
extension _GraphemeBreakingState: CustomStringConvertible {
var description: String {
var r = "["
if hasSeenVirama { r += "V" }
if isInEmojiSequence { r += "E" }
if isInIndicSequence { r += "I" }
if shouldBreakRI { r += "R" }
r += "]"
return r
}
}
extension Unicode {
/// A state machine for recognizing character (i.e., extended grapheme
/// cluster) boundaries in an arbitrary series of Unicode scalars.
///
/// To detect grapheme breaks in a sequence of Unicode scalars, feed each of
/// them to the `hasBreak(before:)` method. The method returns true if the
/// sequence has a grapheme break preceding the given value.
///
/// The results produced by this state machine are guaranteed to match the way
/// `String` splits its contents into `Character` values.
@available(SwiftStdlib 5.8, *)
public // SPI(Foundation) FIXME: We need API for this
struct _CharacterRecognizer: Sendable {
internal var _previous: Unicode.Scalar
internal var _state: _GraphemeBreakingState
/// Returns a non-nil value if it can be determined whether there is a
/// grapheme break between `scalar1` and `scalar2` without knowing anything
/// about the scalars that precede `scalar1`. This can optionally be used as
/// a fast (but incomplete) test before spinning up a full state machine
/// session.
@_effects(releasenone)
public static func quickBreak(
between scalar1: Unicode.Scalar,
and scalar2: Unicode.Scalar
) -> Bool? {
if scalar1.value == 0xD, scalar2.value == 0xA {
return false
}
if _hasGraphemeBreakBetween(scalar1, scalar2) {
return true
}
return nil
}
/// Initialize a new character recognizer at the _start of text_ (sot)
/// position.
///
/// The resulting state machine will report a grapheme break on the
/// first scalar that is fed to it.
public init() {
_state = _GraphemeBreakingState()
// To avoid having to handle the empty case specially, we use NUL as the
// placeholder before the first scalar. NUL is a control character, so per
// rule GB5, it will induce an unconditional grapheme break before the
// first actual scalar, emulating GB1.
_previous = Unicode.Scalar(0 as UInt8)
}
/// Feeds the next scalar to the state machine, returning a Boolean value
/// indicating whether it starts a new extended grapheme cluster.
///
/// This method will always report a break the first time it is called
/// on a newly initialized recognizer.
///
/// The state machine does not carry information across character
/// boundaries. I.e., if this method returns true, then `self` after the
/// call is equivalent to feeding the same scalar to a newly initialized
/// recognizer instance.
@_effects(releasenone)
public mutating func hasBreak(
before next: Unicode.Scalar
) -> Bool {
let r = _state.shouldBreak(between: _previous, and: next)
if r {
_state = _GraphemeBreakingState()
}
_previous = next
return r
}
/// Decode the scalars in the given UTF-8 buffer and feed them to the
/// recognizer up to and including the scalar following the first grapheme
/// break. If the buffer contains a grapheme break, then this function
/// returns the index range of the scalar that follows the first one;
/// otherwise it returns `nil`.
///
/// On return, the state of the recognizer is updated to reflect the scalars
/// up to and including the returned one. You can detect additional grapheme
/// breaks by feeding the recognizer subsequent data.
///
/// - Parameter buffer: A buffer containing valid UTF-8 data, starting and
/// ending on Unicode scalar boundaries.
///
/// - Parameter start: A valid index into `buffer`, addressing the first
/// code unit of a UTF-8 scalar in the buffer, or the end.
///
/// - Returns: The index range of the scalar that follows the first grapheme
/// break in the buffer, if there is one. If the buffer contains no
/// grapheme breaks, then this function returns `nil`.
///
/// - Warning: This function does not validate that the buffer contains
/// valid UTF-8 data; its behavior is undefined if given invalid input.
@_effects(releasenone)
public mutating func _firstBreak(
inUncheckedUnsafeUTF8Buffer buffer: UnsafeBufferPointer<UInt8>,
startingAt start: Int = 0
) -> Range<Int>? {
var i = start
while i < buffer.endIndex {
let (next, n) = _decodeScalar(buffer, startingAt: i)
if hasBreak(before: next) {
return Range(_uncheckedBounds: (i, i &+ n))
}
i &+= n
}
return nil
}
}
}
@available(SwiftStdlib 5.9, *)
extension Unicode._CharacterRecognizer: Equatable {
public static func ==(left: Self, right: Self) -> Bool {
left._previous == right._previous && left._state == right._state
}
}
@available(SwiftStdlib 5.9, *)
extension Unicode._CharacterRecognizer: CustomStringConvertible {
public var description: String {
return "\(_state)U+\(String(_previous.value, radix: 16, uppercase: true))"
}
}
extension _StringGuts {
// Returns the stride of the grapheme cluster starting at offset `index`,
// assuming it is on a grapheme cluster boundary.
//
// This method never looks at data below `index`. If `index` isn't on a
// grapheme cluster boundary, then the result may not be consistent with the
// actual breaks in the string. `Substring` relies on this to generate the
// right breaks if its start index isn't aligned on one -- in this case, the
// substring's breaks may not match the ones in its base string.
internal func nextBoundary(
startingAt index: Int,
nextScalar: (Int) -> (scalar: Unicode.Scalar, end: Int)?
) -> Int {
_internalInvariant(index < endIndex._encodedOffset)
// Note: If `index` in't already on a boundary, then starting with an empty
// state here sometimes leads to this method returning results that diverge
// from the true breaks in the string.
var state = _GraphemeBreakingState()
var (scalar, index) = nextScalar(index)!
while true {
guard let (scalar2, nextIndex) = nextScalar(index) else { break }
if state.shouldBreak(between: scalar, and: scalar2) {
break
}
index = nextIndex
scalar = scalar2
}
return index
}
// Returns the stride of the grapheme cluster ending at offset `index`.
//
// This method uses `previousScalar` to looks back in the string as far as
// necessary to find a correct grapheme cluster boundary, whether or not
// `index` happens to be on a boundary itself.
internal func previousBoundary(
endingAt index: Int,
previousScalar: (Int) -> (scalar: Unicode.Scalar, start: Int)?
) -> Int {
// FIXME: This requires potentially arbitrary lookback in each iteration,
// leading to quadratic behavior in some edge cases. Ideally lookback should
// only be done once per cluster (or in the case of RI sequences, once per
// flag sequence). One way to avoid most quadratic behavior is to replace
// this implementation with a scheme that first searches backwards for a
// safe point then iterates forward using the regular `shouldBreak` until we
// reach `index`, as recommended in section 6.4 of TR#29.
//
// https://www.unicode.org/reports/tr29/#Random_Access
var (scalar2, index) = previousScalar(index)!
while true {
guard let (scalar1, previousIndex) = previousScalar(index) else { break }
if shouldBreakWithLookback(
between: scalar1, and: scalar2, at: index, with: previousScalar
) {
break
}
index = previousIndex
scalar2 = scalar1
}
return index
}
}
extension _GraphemeBreakingState {
// Return true if there is an extended grapheme cluster boundary between two
// scalars, based on state information previously collected about preceding
// scalars.
//
// This method never looks at scalars other than the two that are explicitly
// passed to it. The `state` parameter is assumed to hold all contextual
// information necessary to make a correct decision; it gets updated with more
// data as needed.
//
// This is based on the Unicode Annex #29 for [Grapheme Cluster Boundary
// Rules](https://unicode.org/reports/tr29/#Grapheme_Cluster_Boundary_Rules).
internal mutating func shouldBreak(
between scalar1: Unicode.Scalar,
and scalar2: Unicode.Scalar
) -> Bool {
// GB3
if scalar1.value == 0xD, scalar2.value == 0xA {
return false
}
if _hasGraphemeBreakBetween(scalar1, scalar2) {
return true
}
let x = Unicode._GraphemeBreakProperty(from: scalar1)
// GB4 handled here because we don't need to know `y` for this csae
if x == .control {
return true
}
// This variable and the defer statement help toggle the isInEmojiSequence
// state variable to false after every decision of 'shouldBreak'. If we
// happen to see a rhs .extend or .zwj, then it's a signal that we should
// continue treating the current grapheme cluster as an emoji sequence.
var enterEmojiSequence = false
// Very similar to emoji sequences, but for Indic grapheme sequences.
var enterIndicSequence = false
defer {
self.isInEmojiSequence = enterEmojiSequence
self.isInIndicSequence = enterIndicSequence
}
let y = Unicode._GraphemeBreakProperty(from: scalar2)
switch (x, y) {
// Fast path: If we know our scalars have no properties the decision is
// trivial and we don't need to crawl to the default statement.
case (.any, .any):
return true
// (GB4 is handled above)
// GB5
case (_, .control):
return true
// GB6
case (.l, .l),
(.l, .v),
(.l, .lv),
(.l, .lvt):
return false
// GB7
case (.lv, .v),
(.v, .v),
(.lv, .t),
(.v, .t):
return false
// GB8
case (.lvt, .t),
(.t, .t):
return false
// GB9 (partial GB11)
case (_, .extend),
(_, .zwj):
// Prepare for recognizing GB11, by remembering if we're in an emoji
// sequence.
//
// GB11: Extended_Pictographic Extend* ZWJ × Extended_Pictographic
//
// If our left-side scalar is a pictograph, then it starts a new emoji
// sequence; the sequence continues through subsequent extend/extend and
// extend/zwj pairs.
if (
x == .extendedPictographic || (self.isInEmojiSequence && x == .extend)
) {
enterEmojiSequence = true
}
// If we're currently in an indic sequence (or if our lhs is a linking
// consonant), then this check and everything underneath ensures that
// we continue being in one and may check if this extend is a Virama.
if self.isInIndicSequence || scalar1._isLinkingConsonant {
if y == .extend {
let extendNormData = Unicode._NormData(scalar2, fastUpperbound: 0x300)
// If our extend's CCC is 0, then this rule does not apply.
guard extendNormData.ccc != 0 else {
return false
}
}
enterIndicSequence = true
if scalar2._isVirama {
self.hasSeenVirama = true
}
}
return false
// GB9a
case (_, .spacingMark):
return false
// GB9b
case (.prepend, _):
return false
// GB11
case (.zwj, .extendedPictographic):
return !self.isInEmojiSequence
// GB12 & GB13
case (.regionalIndicator, .regionalIndicator):
defer {
self.shouldBreakRI.toggle()
}
return self.shouldBreakRI
// GB999
default:
// GB9c
if
self.isInIndicSequence,
self.hasSeenVirama,
scalar2._isLinkingConsonant
{
self.hasSeenVirama = false
return false
}
return true
}
}
}
extension _StringGuts {
// Return true if there is an extended grapheme cluster boundary between two
// scalars, with no previous knowledge about preceding scalars.
//
// This method looks back as far as it needs to determine the correct
// placement of boundaries.
//
// This is based off of the Unicode Annex #29 for [Grapheme Cluster Boundary
// Rules](https://unicode.org/reports/tr29/#Grapheme_Cluster_Boundary_Rules).
internal func shouldBreakWithLookback(
between scalar1: Unicode.Scalar,
and scalar2: Unicode.Scalar,
at index: Int,
with previousScalar: (Int) -> (scalar: Unicode.Scalar, start: Int)?
) -> Bool {
// GB3
if scalar1.value == 0xD, scalar2.value == 0xA {
return false
}
if _hasGraphemeBreakBetween(scalar1, scalar2) {
return true
}
let x = Unicode._GraphemeBreakProperty(from: scalar1)
let y = Unicode._GraphemeBreakProperty(from: scalar2)
switch (x, y) {
// Fast path: If we know our scalars have no properties the decision is
// trivial and we don't need to crawl to the default statement.
case (.any, .any):
return true
// GB4
case (.control, _):
return true
// GB5
case (_, .control):
return true
// GB6
case (.l, .l),
(.l, .v),
(.l, .lv),
(.l, .lvt):
return false
// GB7
case (.lv, .v),
(.v, .v),
(.lv, .t),
(.v, .t):
return false
// GB8
case (.lvt, .t),
(.t, .t):
return false
// GB9 (partial GB11)
case (_, .extend),
(_, .zwj):
return false
// GB9a
case (_, .spacingMark):
return false
// GB9b
case (.prepend, _):
return false
// GB11
case (.zwj, .extendedPictographic):
return !checkIfInEmojiSequence(at: index, with: previousScalar)
// GB12 & GB13
case (.regionalIndicator, .regionalIndicator):
return countRIs(at: index, with: previousScalar)
// GB999
default:
// GB9c
switch (x, scalar2._isLinkingConsonant) {
case (.extend, true):
let extendNormData = Unicode._NormData(scalar1, fastUpperbound: 0x300)
guard extendNormData.ccc != 0 else {
return true
}
return !checkIfInIndicSequence(at: index, with: previousScalar)
case (.zwj, true):
return !checkIfInIndicSequence(at: index, with: previousScalar)
default:
return true
}
}
}
// When walking backwards, it's impossible to know whether we were in an emoji
// sequence without walking further backwards. This walks the string backwards
// enough until we figure out whether or not to break our
// (.zwj, .extendedPictographic) question. For example:
//
// Scalar view #1:
//
// [.control, .zwj, .extendedPictographic]
// ^
// | = To determine whether or not we break here, we need
// to see the previous scalar's grapheme property.
// ^
// | = This is neither .extendedPictographic nor .extend, thus we
// were never in an emoji sequence, so break between the .zwj
// and .extendedPictographic.
//
// Scalar view #2:
//
// [.extendedPictographic, .zwj, .extendedPictographic]
// ^
// | = Same as above, move backwards one to
// view the previous scalar's property.
// ^
// | = This is an .extendedPictographic, so this indicates that
// we are in an emoji sequence, so we should NOT break
// between the .zwj and .extendedPictographic.
//
// Scalar view #3:
//
// [.extendedPictographic, .extend, .extend, .zwj, .extendedPictographic]
// ^
// | = Same as above
// ^
// | = This is an .extend which means
// there is a potential emoji
// sequence, walk further backwards
// to find an .extendedPictographic.
//
// <-- = Another extend, go backwards more.
// ^
// | = We found our starting .extendedPictographic letting us
// know that we are in an emoji sequence so our initial
// break question is answered as NO.
internal func checkIfInEmojiSequence(
at index: Int,
with previousScalar: (Int) -> (scalar: Unicode.Scalar, start: Int)?
) -> Bool {
guard var i = previousScalar(index)?.start else { return false }
while let prev = previousScalar(i) {
i = prev.start