-
Notifications
You must be signed in to change notification settings - Fork 3.4k
/
replacer.go
859 lines (747 loc) · 26 KB
/
replacer.go
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
package tokenization
import (
"bytes"
"slices"
)
var (
placeholderNumber = []byte("<NUM>")
placeholderHex = []byte("<HEX>")
placeholderUUID = []byte("<UUID>")
placeholderTimestamp = []byte("<TIMESTAMP>")
placeholderDuration = []byte("<DURATION>")
placeholderBytesize = []byte("<BYTESIZE>")
placeholderIP = []byte("<IP>")
)
// Integer numbers after these words won't be replaced by `<NUM>`.
var numPreservingKeys = [][]byte{[]byte("status"), []byte("status_code"), []byte("httpStatus")}
const (
maxYear = 2100 // anything above that is unlikely to be a timestamp...
minHexLen = 12 // 48 bits min for free-standing hex strings (e.g "0123456789ab") or 42 bits for "0xABCDEF1234" strings
)
var boundaryChars = [256]bool{}
func init() {
for i := 0; i < 128; i++ { // for now, we don't consider non-ASCII chars as boundary
if i < '0' || (i > '9' && i < 'A') || (i > 'Z' && i < 'a') || i > 'z' {
boundaryChars[i] = true
}
}
// We need these keys sorted in an ascending length to efficiently compare them.
slices.SortStableFunc(numPreservingKeys, func(a, b []byte) int {
return len(a) - len(b)
})
}
type replacer struct {
source, dest []byte
// This is the last position that was copied to the destination buffer.
tail int
// This is the current position we are working with
cur int
// This is essentially used for lookahed operations
head int
// Here's some ASCII art to visualize that (the long string of dashes
// visualizes the log line:
//
// 0 <already in dest, copied or replaced > `tail` <about to be copied or replaced> `cur` <lookahead> `head` <remaining unprocessed log> `len(source)`
// |---------------------------------------------|------------------------------------------|---------------------|------------------------------------------|
// A somewhat hacky solution that allows us to preserve specific numeric
// values we care about, like status and status_code.
preserveNextNumbers bool
}
// commit advances the current position marker to the lookahead position, i.e.
// we are committing to consume everything we've looked ahead so far.
func (r *replacer) commit() {
r.cur = r.head
}
func (r *replacer) resetHead() {
r.head = r.cur
}
func (r *replacer) resetHeadExpr() bool {
r.head = r.cur
return true // useful when included in boolean expressions with advance()
}
func (r *replacer) backtrack() {
r.head--
}
func (r *replacer) consumeUpToCurrent() {
r.tail = r.cur
}
// advanceUintRet returns the actual value of the number we read, as well as its
// length. The value might overflow for large numbers, so it's also important to
// check the length!
func (r *replacer) advanceUintRet() (val uint, length uint) {
var c byte
for ; r.head < len(r.source); r.head++ {
c = r.source[r.head]
if c < '0' || '9' < c {
break
}
length++
val = val*10 + uint(c-'0') // TODO: use bitwise trick?
}
return val, length
}
func (r *replacer) advanceUintOrRangeRet(lc, hc byte) (length uint) {
var c byte
for ; r.head < len(r.source); r.head++ {
c = r.source[r.head]
if !(('0' <= c && c <= '9') || (lc <= c && c <= hc)) {
break
}
length++
}
return length
}
func (r *replacer) advanceUint() bool {
_, l := r.advanceUintRet()
return l > 0
}
func (r *replacer) advanceUintUpTo(val uint, length uint) bool {
foundVal, foundLength := r.advanceUintRet()
return foundLength > 0 && foundVal <= val && foundLength <= length
}
func (r *replacer) advanceYear() bool {
return r.advanceUintUpTo(maxYear, 4)
}
func (r *replacer) advanceChar(c byte) bool {
if r.head < len(r.source) && r.source[r.head] == c {
r.head++
return true
}
return false
}
func (r *replacer) peekNextIsBoundary() bool {
if r.head >= len(r.source) {
return true // we are at the end of the line
}
return boundaryChars[r.source[r.head]]
}
func (r *replacer) peekFirstNonInt() (c byte) {
for i := r.head; i < len(r.source); i++ {
c = r.source[i]
if c < '0' || '9' < c {
return c
}
}
return 0 // we can return the 0 value here!
}
func (r *replacer) peekNext4() (result [4]byte) {
overhead := len(r.source) - r.head
switch {
case overhead > 3:
result[3] = r.source[r.head+3]
fallthrough
case overhead > 2:
result[2] = r.source[r.head+2]
fallthrough
case overhead > 1:
result[1] = r.source[r.head+1]
fallthrough
case overhead > 0:
result[0] = r.source[r.head+0]
}
return result
}
func (r *replacer) advanceTimeZoneLetters() bool {
UCLetters := 0
for {
nc, ok := r.advance()
if !ok {
break
}
if nc < 'A' || nc > 'Z' {
r.backtrack()
break
}
UCLetters++
}
return UCLetters >= 2 && UCLetters <= 5
}
func (r *replacer) advanceNumericTimeZone() bool {
// See https://en.wikipedia.org/wiki/List_of_tz_database_time_zones
return r.advanceOneOf('+', '-') && r.advanceUintUpTo(14000, 5)
}
// helper for handleWeirdTimestamp()
func (r *replacer) advanceStringOrNumericTimeZone(isOptional bool) bool {
curHead := r.head
if r.advanceChar(' ') && r.advanceNumericTimeZone() {
return true
}
r.head = curHead
if r.advanceChar(' ') && r.advanceTimeZoneLetters() {
return true
}
r.head = curHead
return isOptional
}
func (r *replacer) advanceOneOf(chars ...byte) bool {
if r.head >= len(r.source) {
return false
}
c := r.source[r.head]
for _, ec := range chars {
if c == ec {
r.head++
return true
}
}
return false
}
func (r *replacer) advanceTime(secondsOptional bool) bool {
return r.advanceUintUpTo(24, 2) && r.advanceChar(':') && r.advanceUintUpTo(60, 2) && (secondsOptional || (r.advanceChar(':') && r.advanceUintUpTo(60, 2)))
}
// Mon|Tue|Wed|Thu|Fri|Sat|Sun
func (r *replacer) advanceDayOfTheWeek() bool {
c1, ok1 := r.advance()
c2, ok2 := r.advance()
c3, ok3 := r.advance()
if !ok1 || !ok2 || !ok3 {
return false
}
return (c1 == 'S' && ((c2 == 'a' && c3 == 't') || (c2 == 'u' && c3 == 'n'))) || // Sat, Sun
(c1 == 'M' && c2 == 'o' && c3 == 'n') ||
(c1 == 'T' && c2 == 'u' && c3 == 'e') ||
(c1 == 'W' && c2 == 'e' && c3 == 'd') ||
(c1 == 'T' && c2 == 'h' && c3 == 'u') ||
(c1 == 'F' && c2 == 'r' && c3 == 'i')
}
// Jan|Feb|Mar|Apr|May|Jul|Jun|Aug|Sep|Oct|Nov|Dec
func (r *replacer) advanceMonthName() bool {
c1, ok1 := r.advance()
c2, ok2 := r.advance()
c3, ok3 := r.advance()
if !ok1 || !ok2 || !ok3 {
return false
}
return (c1 == 'J' && ((c2 == 'u' && (c3 == 'n' || c3 == 'l')) || // Jun, Jul
(c2 == 'a' && c3 == 'n'))) || // Jan
(c1 == 'M' && c2 == 'a' && (c3 == 'r' || c3 == 'y')) || // Mar, May
(c2 == 'e' && ((c1 == 'F' && c3 == 'b') || (c1 == 'S' && c3 == 'p') || (c1 == 'D' && c3 == 'c'))) || // Feb, Sep, Dec
(c1 == 'A' && ((c2 == 'p' && c3 == 'r') || // Apr
(c2 == 'u' && c3 == 'g'))) || // Aug
(c1 == 'O' && c2 == 'c' && c3 == 't') || // Oct
(c1 == 'N' && c2 == 'o' && c3 == 'v') // Nov
}
// Check if we in the middle of an UUID, exactly after the first 8 characters
// and the dash after them, e.g after "abcd0123-":
func (r *replacer) advanceUUIDAfterFirstDash(lc, hc byte) (endsWithBoundary bool) {
return (r.advanceUintOrRangeRet(lc, hc) == 4) && r.advanceChar('-') &&
(r.advanceUintOrRangeRet(lc, hc) == 4) && r.advanceChar('-') &&
(r.advanceUintOrRangeRet(lc, hc) == 4) && r.advanceChar('-') &&
(r.advanceUintOrRangeRet(lc, hc) == 12) && r.peekNextIsBoundary()
}
// Only moves the head forward if it successfully matches a duration
func (r *replacer) advanceDuration() (matched bool) {
curHead := r.head
var secsLen int
n := r.peekNext4()
if n[0] == 'h' {
r.head++
if boundaryChars[n[1]] {
return true // e.g. "1h", "123h"
}
if !r.advanceUintUpTo(60, 2) {
goto restore
}
n = r.peekNext4()
}
if n[0] == 'm' && (boundaryChars[n[1]] || n[1] != 's') { // we don't want to match 'ms' here
r.head++
if boundaryChars[n[1]] {
return true // e.g. "2h21m", "121m"
}
if !(r.advanceUintUpTo(60, 2) && ((r.advanceChar('.') && r.advanceUint()) || true)) {
goto restore
}
n = r.peekNext4()
}
if n[0] == 's' && boundaryChars[n[1]] {
secsLen = 1
} else if n[1] == 's' && (n[0] == 'm' || n[0] == 'n' || n[0] == 'u') && boundaryChars[n[2]] {
secsLen = 2
} else if n[2] == 's' && ((n[0] == 0xC2 && n[1] == 0xB5) || (n[0] == 0xCE && n[1] == 0xBC)) && boundaryChars[n[3]] {
// This checks for the unicode "µs" (U+00B5 = micro symbol) and "μs" (U+03BC = Greek letter mu)
secsLen = 3
} else {
goto restore
}
r.head += secsLen
return true
restore: // should be faster than a defer
r.head = curHead
return false
}
var byteSizes = [256]bool{'k': true, 'K': true, 'm': true, 'M': true, 'g': true, 'G': true, 't': true, 'T': true, 'p': true, 'P': true}
// Only moves the head forward if it successfully matches a duration
func (r *replacer) advanceBytesize(c1 byte) (matched bool) {
if !byteSizes[c1] {
return false
}
n := r.peekNext4()
var unitLen int // not counting the first character c1, which is already advanced to
if (n[0] == 'b' || n[0] == 'B') && boundaryChars[n[1]] {
unitLen = 1
} else if n[0] == 'i' && (n[1] == 'b' || n[1] == 'B') && boundaryChars[n[2]] {
unitLen = 2
} else if ((n[0] == 'b' && n[1] == 'p' && n[2] == 's') || (n[0] == 'b' && n[1] == 'i' && n[2] == 't')) && boundaryChars[n[3]] {
unitLen = 3
}
if unitLen > 0 {
r.head += unitLen
return true
}
return false
}
func (r *replacer) advance() (c byte, advanced bool) {
if r.head >= len(r.source) {
return 0, false
}
c = r.source[r.head]
r.head++
return c, true
}
func (r *replacer) emitNumber() {
r.commit()
r.dest = append(r.dest, placeholderNumber...)
r.consumeUpToCurrent()
}
func (r *replacer) emitNumberOrCopyText(hasMinusPrefix bool) {
r.commit()
if !r.preserveNextNumbers {
r.dest = append(r.dest, placeholderNumber...)
r.consumeUpToCurrent()
} else {
r.maybeEmitDash(hasMinusPrefix)
r.copyUpToCurrent()
}
}
func (r *replacer) emit(hasMinusPrefix bool, placeholder []byte) {
r.commit()
r.maybeEmitDash(hasMinusPrefix)
r.dest = append(r.dest, placeholder...)
r.consumeUpToCurrent()
}
func (r *replacer) maybeEmitDash(hasMinusPrefix bool) {
// This minus was actually a dash, so we just copy it to the result
if hasMinusPrefix {
r.dest = append(r.dest, '-')
}
}
func (r *replacer) copyUpToCurrent() {
r.dest = append(r.dest, r.source[r.tail:r.cur]...)
r.consumeUpToCurrent()
}
func (r *replacer) handleHexOrUnit(hasMinusPrefix bool, n1, l1 uint, c1 byte) (endsWithBoundary bool) {
// Special case that is likely a hex string of the format "0x", but we don't
// know whether the rest is upper case or lower case yet.
zeroHex := false
if n1 == 0 && l1 == 1 && c1 == 'x' {
zeroHex = true // these cannot be the start of an UUID
c1 = r.peekFirstNonInt()
}
// Maybe we are at the start of a hex string, either something like
// "[0-9]+[a-f]", "[0-9]+[A-F]", or "0x". We support both lower and upper
// case letters, but to avoid false positives, we want hex replacements to
// happen only if the strings are fully lower case or fully upper case.
if 'a' <= c1 && c1 <= 'f' {
return r.handleHex(hasMinusPrefix, l1+1, 'a', 'f', !zeroHex)
} else if 'A' <= c1 && c1 <= 'F' {
return r.handleHex(hasMinusPrefix, l1+1, 'A', 'F', !zeroHex)
} else if zeroHex {
// Well, it probably wasn't a zero-hex after all, or it contained only
// digits, so try to handle that or emit what we absorbed
_, l2 := r.advanceUintRet()
if l2+2 >= minHexLen { // We consider "0x" to be part of the hex string when comparing with minHexLen
r.emit(hasMinusPrefix, placeholderHex)
} else {
r.maybeEmitDash(hasMinusPrefix)
r.commit()
r.copyUpToCurrent()
}
return r.peekNextIsBoundary()
}
return r.handlePotentialUnitWithDecimal(hasMinusPrefix, c1)
}
func (r *replacer) handleHex(hasMinusPrefix bool, l1 uint, lc, hc byte, canBeUUID bool) (endsWithBoundary bool) {
totalLen := l1 + r.advanceUintOrRangeRet(lc, hc)
r.commit()
if totalLen == 8 && canBeUUID {
// We might be at the first part of a UUID, right before the first dash
if r.advanceChar('-') && r.advanceUUIDAfterFirstDash(lc, hc) {
r.emit(hasMinusPrefix, placeholderUUID)
return true
}
r.resetHead()
}
if totalLen >= minHexLen && r.peekNextIsBoundary() {
r.emit(hasMinusPrefix, placeholderHex)
return true
}
r.copyUpToCurrent()
return r.peekNextIsBoundary()
}
func (r *replacer) handlePotentialUnitWithDecimal(hasMinusPrefix bool, c1 byte) (endsWithBoundary bool) {
if r.advanceBytesize(c1) {
// We do not subsume a minus sign - byte sizes are unlikely to be
// negative, it's more likely this is a dash as a part of a range
r.emit(hasMinusPrefix, placeholderBytesize)
return true
}
r.backtrack()
if r.advanceDuration() {
// We subsume hasMinusPrefix, since durations can be negative
r.emit(false, placeholderDuration)
return true
}
// We couldn't match anything, so just copy what existed.
r.maybeEmitDash(hasMinusPrefix)
r.copyUpToCurrent()
return false
}
func (r *replacer) handleNumberWithDecimal(hasMinusPrefix bool, n1 uint, l1 uint) (endsWithBoundary bool) {
n2, l2 := r.advanceUintRet()
if l2 == 0 {
// The dot wasn't followed by another number, so emit everything before it.
r.backtrack()
r.emitNumberOrCopyText(hasMinusPrefix)
return false
}
// See if the number after the decimal is also followed by a boundary
b2, hasNext := r.advance()
// We are at the end of the string, which is a boundary, replace evertything
// up to now with a number
if !hasNext {
r.emitNumber() // this also subsumes any minus sign we had
return true
}
// The decimal number isn't followed by a boundary char (which include
// things like '.', ':', '/', etc.), so the part after the decimal is either
// not a real number, or it's some sort of a unit that can support decimals
// like size (e.g. 12KB, 3MiB) or durations (e.g. 3.5124s), etc.
if !boundaryChars[b2] {
return r.handlePotentialUnitWithDecimal(hasMinusPrefix, b2)
}
// We have a decimal number followed by a non-dot boundary, so this is not
// an IP or a version number or anything like that.
if b2 != '.' {
r.backtrack()
r.emitNumber()
return true
}
n3, l3 := r.advanceUintRet()
if l3 == 0 || !r.peekNextIsBoundary() {
// The second dot wasn't followed by another number and a boundary, so
// emit just the first number.
r.resetHead()
r.emitNumber()
return true
}
// We have "<NUM>.<NUM>.<NUM>" at this point, so we either have something
// like a version number, or an IP address, but certainly not a simple
// decimal number we can just emit.
r.commit()
// Check if this is an IP address...
if n1 <= 255 && l1 <= 3 && n2 <= 255 && l2 <= 3 && n3 <= 255 && l3 <= 3 && r.advanceChar('.') && r.advanceUintUpTo(255, 3) && r.peekNextIsBoundary() {
r.emit(hasMinusPrefix, placeholderIP)
return true
}
// This wasn't an IP after all, so just emit "<NUM>.<NUM>.<NUM>". We don't
// want to assume this is a simple decimal number because of the second dot,
// e.g. this might be something like a version number.
r.resetHead()
r.maybeEmitDash(hasMinusPrefix) // we preserve the dashes here, since it's likely not a minus
r.dest = append(r.dest, placeholderNumber...)
r.dest = append(r.dest, '.')
r.dest = append(r.dest, placeholderNumber...)
r.dest = append(r.dest, '.')
r.dest = append(r.dest, placeholderNumber...)
r.consumeUpToCurrent()
return true
}
func (r *replacer) handleSaneTimestamp(hasMinusPrefix bool, n1 uint, delim byte) (endsWithBoundary bool) {
if r.advanceUintUpTo(12, 2) && r.advanceChar(delim) && r.advanceUintUpTo(31, 2) && r.advanceOneOf('T', ' ') && r.advanceTime(false) {
r.commit()
// continue down to parsing sub-second and timezone values
} else if r.resetHeadExpr() && n1 <= 31 && r.advanceChar(delim) && r.advanceMonthName() && r.advanceChar(delim) && r.advanceYear() && r.advanceChar(':') && r.advanceTime(false) && r.advanceChar(' ') && r.advanceNumericTimeZone() && r.peekNextIsBoundary() {
// We might not have a sane timestamp, but apparently a not-so-sane
// timestamp format first, e.g. "27/Mar/2024:14:34:37 +0000"...
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
} else {
// Apparently that wasn't it either, we probably just have a dash or
// forward slash after a number, so we backtrack and emit the number.
r.resetHead()
r.emitNumberOrCopyText(hasMinusPrefix)
return true
}
// We have a date that looks like `YY[YY]-MM-DD hh:mm:ss` or
// `YY[YY]/MM/DDZhh:mm:ss`
r.commit() // we want to keep this
// Now we need to also check for sub-second precision and time zones:
c, canAdvance := r.advance()
if !canAdvance {
// End of the string
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
}
if c == '.' {
_, intl := r.advanceUintRet()
if intl == 0 {
// No sub-second precision, the dot was not part of the timestamp
r.backtrack()
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
}
// We had sub-second precision, capture that too
r.commit()
// Get the next char to see if we have time zone
c, canAdvance = r.advance()
if !canAdvance {
// We are at the end of the sting after we captured the
// sub-second value.
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
}
}
if c == 'Z' || c == 'z' {
//UTC string, nothing to do after that
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
}
// See https://en.wikipedia.org/wiki/List_of_tz_database_time_zones
if (c == '+' || c == '-') && r.advanceUintUpTo(14, 2) && r.advanceChar(':') && r.advanceUintUpTo(60, 2) {
// e.g. "2020-09-30T00:00:59.9999+03:00"
r.commit()
} else if r.resetHeadExpr() && r.advanceChar(' ') && r.advanceNumericTimeZone() && r.advanceChar(' ') && r.advanceTimeZoneLetters() && r.peekNextIsBoundary() {
// e.g. "2023-09-05 23:20:28.030285153 +0000 UTC"
r.commit()
} else {
r.resetHead()
}
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
}
func (r *replacer) handleNumberStart(hasMinusPrefix bool) (endsWithBoundary bool) {
r.resetHead() // keep the head pos in sync with the current pos
// We know we were at a digit due to how handleNumber() is called
n1, l1 := r.advanceUintRet()
r.commit() // we will consume this one way or another for sure
// See if the number is followed by a boundary
b1, hasNext := r.advance()
switch {
// We are at the end of the string, which is a boundary, replace evertything
// up to now with a number
case !hasNext:
r.emitNumberOrCopyText(hasMinusPrefix) // this also may subsume any minus sign we had
return true
// The number isn't followed by a boundary char (which include things like
// '.', ' ', '/', etc.), so it's either not a real number or date, or it's
// some sort of a unit like a duration (e.g. 3h5m2s), size, (e.g. 12KB,
// 3MiB), etc.
case !boundaryChars[b1]:
return r.handleHexOrUnit(hasMinusPrefix, n1, l1, b1)
// We have a decimal point, so this can either be a plain number, a unit
// that can have a float value, or an IP address.
case b1 == '.':
return r.handleNumberWithDecimal(hasMinusPrefix, n1, l1)
// This might be a timestamp that looks like '2024-04-01...' or
// `2024/03/27...`; timestamps in the far future are not supported :)
case n1 <= maxYear && l1 <= 4 && (b1 == '-' || b1 == '/'):
return r.handleSaneTimestamp(hasMinusPrefix, n1, b1)
// Weird RFC822 dates like "02 Jan 06 15:04 MST"
case n1 <= 31 && l1 <= 2 && b1 == ' ':
if r.advanceMonthName() && r.advanceChar(' ') && r.advanceYear() && r.advanceChar(' ') && r.advanceTime(true) && r.advanceStringOrNumericTimeZone(false) {
r.commit()
r.emit(hasMinusPrefix, placeholderTimestamp)
return true
}
// if not, go to default handler after switch statement
// It could be a UUID that starts with 8 digits
case l1 == 8 && b1 == '-':
if r.advanceUUIDAfterFirstDash('a', 'f') || (r.resetHeadExpr() && r.advanceChar('-') && r.advanceUUIDAfterFirstDash('A', 'F')) {
r.emit(hasMinusPrefix, placeholderUUID)
return true
}
// if not, go to default handler after switch statement
}
// Number with an unknown boundary - emit the number and leave the boundary
// for the following passes.
r.resetHead()
r.emitNumberOrCopyText(hasMinusPrefix)
return true
}
var longDayNames = [...][]byte{
[]byte("Sunday"),
[]byte("Monday"),
[]byte("Tuesday"),
[]byte("Wednesday"),
[]byte("Thursday"),
[]byte("Friday"),
[]byte("Saturday"),
}
func (r *replacer) handleWeirdTimestamp() (endsWithBoundary bool) {
r.resetHead()
if r.advanceDayOfTheWeek() {
r.commit() // we will always consume this
// RFC1123 and RFC1123Z, e.g.:
// - "Mon, 02 Jan 2006 15:04:05 MST"
// - "Mon, 02 Jan 2006 15:04:05 -0700"
if r.advanceChar(',') && r.advanceChar(' ') && r.advanceUintUpTo(31, 2) && r.advanceChar(' ') && r.advanceMonthName() && r.advanceChar(' ') && r.advanceYear() && r.advanceChar(' ') && r.advanceTime(false) && r.advanceStringOrNumericTimeZone(false) {
r.emit(false, placeholderTimestamp)
return true
}
r.resetHead()
// ANSIC, UnixDatem, RubyDate e.g
// - "Mon Jan 2 15:04:05 2006"
// - "Mon Jan 2 15:04:05 MST 2006"
// - "Mon Jan 02 15:04:05 -0700 2006"
if r.advanceChar(' ') && r.advanceMonthName() && r.advanceChar(' ') && (r.advanceChar(' ') || true) && r.advanceUintUpTo(31, 2) && r.advanceChar(' ') && r.advanceTime(false) && r.advanceStringOrNumericTimeZone(true) && r.advanceChar(' ') && r.advanceYear() {
r.emit(false, placeholderTimestamp)
return true
}
r.resetHead()
// Linux, e.g.
// - "Mon 2 Jan 15:04:05 MST 2006"
// - "Tue 23 Jan 15:04:05 -0700 2023"
if r.advanceChar(' ') && (r.advanceChar(' ') || true) && r.advanceUintUpTo(31, 2) && r.advanceChar(' ') && r.advanceMonthName() && r.advanceChar(' ') && r.advanceTime(false) && r.advanceStringOrNumericTimeZone(false) && r.advanceChar(' ') && r.advanceYear() {
r.emit(false, placeholderTimestamp)
return true
}
r.resetHead()
// RFC850, e.g.
// - "Monday, 02-Jan-06 15:04:05 MST"
backtrackedSlice := r.source[r.head-3:]
var matchedDay []byte
for _, dw := range longDayNames {
if bytes.HasPrefix(backtrackedSlice, dw) {
matchedDay = dw
break
}
}
if matchedDay != nil {
r.head += len(matchedDay) - 3
if r.advanceChar(',') && r.advanceChar(' ') && r.advanceUintUpTo(31, 2) && r.advanceChar('-') && r.advanceMonthName() && r.advanceChar('-') && r.advanceUintUpTo(99, 2) && r.advanceChar(' ') && r.advanceTime(false) && r.advanceStringOrNumericTimeZone(true) {
r.emit(false, placeholderTimestamp)
return true
}
}
r.cur -= 3 // unconsume
r.resetHead()
return false
}
r.resetHead()
if r.advanceMonthName() {
r.commit() // provisionally consume this
// Linux journald logs and others similar like this:
// - Feb 29 23:00:14
// - Apr-10 23:43:46.807
// - Jul 1 00:21:28
if (r.advanceChar('-') || (r.advanceChar(' ') && (r.advanceChar(' ') || true))) && r.advanceUintUpTo(31, 2) && r.advanceChar(' ') && r.advanceTime(false) {
r.commit()
// This is already a timestamp, but let's try to match subseconds as well
if r.advanceChar('.') && r.advanceUint() {
r.commit()
} else {
r.resetHead()
}
r.emit(false, placeholderTimestamp)
return true
}
r.cur -= 3 // unconsume
r.resetHead()
return false
}
r.resetHead()
return false
}
func (r *replacer) wasNumPreservingKeyword() bool {
for _, key := range numPreservingKeys {
pPos := r.cur - 1 - len(key)
if pPos < -1 {
return false // all subsequent keys are longer
}
if pPos != -1 && !boundaryChars[r.source[pPos]] {
continue
}
if bytes.HasPrefix(r.source[pPos+1:], key) {
return true
}
}
return false
}
func (r *replacer) replaceWithPlaceholders() {
lineLen := len(r.source)
var c byte
onBoundary := true
for r.cur = 0; r.cur < lineLen; r.cur++ {
c = r.source[r.cur]
switch {
// If we're currently not at a boundary, the only thing we need to check
// is whether the current char would create a boundary in the next iteration.
case !onBoundary:
onBoundary = boundaryChars[c]
// We weren't at a boundary and now we are, so check if we match one
// of the special keywords that will preserve numbers
if onBoundary {
r.preserveNextNumbers = r.wasNumPreservingKeyword()
}
// If we've reached this far, it means we're currently at a boundary!
// A lot of very complex logic if we encounter a number at a boundary,
// so we move that to a helper function.
case '0' <= c && c <= '9':
r.copyUpToCurrent()
onBoundary = r.handleNumberStart(false)
// Handle negative numbers, potentially
case c == '-':
next := r.cur + 1
// This might be a number, a date, an IP address, etc. So we don't
// know if this is a minus sign to replace or a dash to copy yet.
if next < lineLen && '0' <= r.source[next] && r.source[next] <= '9' {
// Copy everything before the dash, but mark it as consumed.
r.copyUpToCurrent()
r.cur++
r.consumeUpToCurrent()
onBoundary = r.handleNumberStart(true)
} else {
onBoundary = true
}
// Try to match weird timestamps. They require a lot of remaining
// length, generally start with a capitalized day of the week or month
// name (1 upper case letter followed by 2 lower case letters).
//
// We are basically looking for something that may match this here:
// Mon|Tue|Wed|Thu|Fri|Sat|Sun|Jan|Feb|Mar|Apr|May|Jul|Jun|Aug|Sep|Oct|Nov|Dec
//
// The detailed check would be performed by the actual handler:
case 'A' <= c && c <= 'W' && lineLen-r.cur >= 14 &&
'a' <= r.source[r.cur+1] && r.source[r.cur+1] <= 'u' &&
'b' <= r.source[r.cur+2] && r.source[r.cur+2] <= 'y':
r.copyUpToCurrent()
onBoundary = r.handleWeirdTimestamp()
// This could be the start of an lower case hex string:
case 'a' <= c && c <= 'f':
r.copyUpToCurrent()
r.resetHead()
onBoundary = r.handleHex(false, 0, 'a', 'f', true)
// This could be the start of an upper case hex string:
case 'A' <= c && c <= 'F':
r.copyUpToCurrent()
r.resetHead()
onBoundary = r.handleHex(false, 0, 'A', 'F', true)
// If we haven't actually matched anything, update whether we're still
// on a boundary character and continue onto the next one.
default:
onBoundary = boundaryChars[c]
}
}
if r.cur > r.tail {
r.dest = append(r.dest, r.source[r.tail:]...)
r.consumeUpToCurrent()
}
}
func Preprocess(content []byte) []byte {
// ~floor(120%), to allow for some expansion from replacements, hopefully
// without needing to allocate more memory
r := replacer{source: content, dest: make([]byte, 0, len(content)*120/100)}
r.replaceWithPlaceholders()
return r.dest
}