-
Notifications
You must be signed in to change notification settings - Fork 4
/
deepdiff.go
709 lines (630 loc) · 19.3 KB
/
deepdiff.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
package deepdiff
import (
"context"
"encoding/hex"
"hash"
"hash/fnv"
"reflect"
"sort"
)
// Config are any possible configuration parameters for calculating diffs
type Config struct {
// Setting CalcChanges to true will have diff represent in-place value shifts
// as changes instead of add-delete pairs
CalcChanges bool
}
// DiffOption is a function that adjust a config, zero or more DiffOptions
// can be passed to the Diff function
type DiffOption func(cfg *Config)
// DeepDiff is a configuration for performing diffs
type DeepDiff struct {
changes bool
}
// New creates a deepdiff struct
func New(opts ...DiffOption) *DeepDiff {
cfg := &Config{}
for _, opt := range opts {
opt(cfg)
}
return &DeepDiff{
changes: cfg.CalcChanges,
}
}
// Diff computes a slice of deltas that define an edit script for turning a
// into b.
// currently Diff will never return an error, error returns are reserved for
// future use. specifically: bailing before delta calculation based on a
// configurable threshold
func (dd *DeepDiff) Diff(ctx context.Context, a, b interface{}) (Deltas, error) {
deepdiff := &diff{changes: dd.changes, d1: a, d2: b}
return deepdiff.diff(ctx), nil
}
// StatDiff calculates a diff script and diff stats
func (dd *DeepDiff) StatDiff(ctx context.Context, a, b interface{}) (Deltas, *Stats, error) {
deepdiff := &diff{changes: dd.changes, d1: a, d2: b, stats: &Stats{}}
return deepdiff.diff(ctx), deepdiff.stats, nil
}
// Stat calculates the DiffStats between two documents
func (dd *DeepDiff) Stat(ctx context.Context, a, b interface{}) (*Stats, error) {
deepdiff := &diff{changes: dd.changes, d1: a, d2: b, stats: &Stats{}}
deepdiff.diff(ctx)
return deepdiff.stats, nil
}
// diff is a state machine for calculating an edit script that transitions
// between two state trees
type diff struct {
changes bool // calculate changes flag
stats *Stats
d1, d2 interface{}
t1, t2 node
t1Nodes map[string][]node
}
// diff calculates a structl diff for two given tree states
// generating an edit script as a list of Delta changes:
//
// 1. prepTrees - register in a map a unique signature (hash value) for every
// subtree of the d1 (old) document
// 2. queueMatch - consider every subtree in d2 document, starting from the
// largest. check if it is identitical to some the subtrees in
// d1, if so match both subtrees.
// 3. attempt to match the parents of two matched subtrees
// by checking labels (in our case, types of parent object or array)
// controlling for bad matches based on length of path to the
// ancestor and the weight of the matching subtrees. eg: a large
// subtree may force the matching of its ancestors up to the root
// a small subtree may not even force matching of its parent
// 4. Consider the largest subtrees of d2 in order. If one candidate
// has it's parent already matched to the parent of the considered
// node, it is certianly the best candidate.
// 5. At this point we might have matched all of d2. A node may not
// match b/c its been inserted, or we missed matching it. We can now
// do peephole optimization pass to retry some of the rejected nodes
// once no more matchings can be obtained, unmatched nodes in d2
// correspond to inserted nodes.
// 6. consider each matching node and decide if the node is at its right
// place, or whether it has been moved.
func (d *diff) diff(ctx context.Context) Deltas {
d.t1, d.t2, d.t1Nodes = d.prepTrees(ctx)
d.queueMatch(d.t1Nodes, d.t2)
d.optimize(d.t1, d.t2)
// TODO (b5): a second optimize pass seems to help on larger diffs, which
// to me seems we should propagating matches more aggressively in the optimize pass,
// removing the need for this second call (which is effectively doing the same
// thing as recursive/aggressive match propagation)
d.optimize(d.t1, d.t2)
d.optimize(d.t1, d.t2)
return d.calcDeltas(d.t1, d.t2)
}
// NewHash returns a new hash interface, wrapped in a function for easy
// hash algorithm switching, package consumers can override NewHash
// with their own desired hash.Hash implementation if the value space is
// particularly large. default is 64-bit FNV 1 for fast, cheap,
// (non-cryptographic) hashing
var NewHash = func() hash.Hash {
return fnv.New64()
}
// hashString converts a hash sum to a string using hex encoding
// localized here for easy ecnsoding swapping
func hashStr(sum []byte) string {
return hex.EncodeToString(sum)
}
func (d *diff) queueMatch(t1Nodes map[string][]node, t2 node) {
queue := make(chan node)
done := make(chan struct{})
considering := 1
t2Weight := t2.Weight()
go func() {
var candidates []node
for n2 := range queue {
key := hashStr(n2.Hash())
candidates = t1Nodes[key]
switch len(candidates) {
case 0:
// no candidates. check if node has children. If so, add them.
if n2c, ok := n2.(compound); ok {
for _, ch := range n2c.Children() {
considering++
go func(n node) {
queue <- n
}(ch)
}
}
case 1:
// connect an exact match. yay!
n1 := candidates[0]
matchNodes(n1, n2)
default:
// choose a best candidate. let the sketchiness begin.
bestCandidate(candidates, n2, t2Weight)
}
considering--
if considering == 0 {
done <- struct{}{}
break
}
}
}()
// start queue with t2 (root of tree)
queue <- t2
<-done
return
}
// matchNodes connects two nodes & tries to propagate that match upward to
// ancestors so long as labels match
func matchNodes(n1, n2 node) {
n1.SetMatch(n2)
n2.SetMatch(n1)
n1p := n1.Parent()
n2p := n2.Parent()
for n1p != nil && n2p != nil {
if n1p.Addr().Eq(n2p.Addr()) {
n1p.SetMatch(n2p)
n2p.SetMatch(n1p)
n1p = n1p.Parent()
n2p = n2p.Parent()
}
break
}
}
func bestCandidate(t1Candidates []node, n2 node, t2Weight int) {
if n2.Parent() == nil {
return
}
// Copy the candidate list so that this slice can be modified
nodeList := make([]node, len(t1Candidates))
copy(nodeList, t1Candidates)
maxDist := 1 + float32(n2.Weight())/float32(t2Weight)
dist := 1 + float32(n2.Parent().Weight()-n2.Weight())/float32(t2Weight)
n2 = n2.Parent()
for dist < maxDist {
for i, can := range nodeList {
// Some nodes are replaced with their parents, which may result in
// a nil pointer once we reach the root
if can == nil {
continue
}
if cp := can.Parent(); cp != nil {
if n2.Addr().Eq(cp.Addr()) {
matchNodes(cp, n2)
return
}
}
// Move to the candidates parent, which may result in a nil pointer
nodeList[i] = can.Parent()
}
if n2.Parent() == nil {
break
}
dist = 1 + float32(n2.Parent().Weight()-n2.Weight())/float32(t2Weight)
n2 = n2.Parent()
}
}
func (d *diff) optimize(t1, t2 node) {
walkPostfix(t1, nil, func(_ []Addr, n node) {
propagateMatchToParent(n)
})
walkPostfix(t2, nil, func(_ []Addr, n node) {
propagateMatchToParent(n)
})
walk(t1, nil, func(_ []Addr, n node) bool {
propagateMatchToChildren(n)
return true
})
walk(t2, nil, func(_ []Addr, n node) bool {
propagateMatchToChildren(n)
return true
})
}
func propagateMatchToParent(n node) {
// if n is a compound type that isn't matched
if cmp, ok := n.(compound); ok && n.Match() == nil {
var match node
// iterate each child
for _, ch := range cmp.Children() {
// if this child has a match, and the matches parent doesn't have a match,
// match the parents
if m := ch.Match(); m != nil && m.Parent() != nil && m.Parent().Match() == nil {
p := m.Parent()
if match == nil {
match = p
} else if p.Weight() > m.Weight() {
// if a match already exists, keep the heavier match
match = p
}
}
}
if match != nil {
matchNodes(match, n)
}
}
}
func propagateMatchToChildren(n node) {
// if a node is matched & a compound type,
if n1, ok := n.(compound); ok && n.Match() != nil {
if n2, ok := n.Match().(compound); ok {
if n1.Type() == ntObject && n2.Type() == ntObject {
// match any key names
for _, n1ch := range n1.Children() {
if n2ch := n2.Child(n1ch.Addr()); n2ch != nil {
n2ch.SetMatch(n1ch)
n1ch.SetMatch(n2ch)
}
}
}
if n1.Type() == ntArray && n2.Type() == ntArray && len(n1.Children()) == len(n2.Children()) {
// if arrays are the same length, match all children
// b/c these are arrays, no names should be missing, safe to skip a name check
for _, n1ch := range n1.Children() {
n2ch := n2.Child(n1ch.Addr())
n2ch.SetMatch(n1ch)
n1ch.SetMatch(n2ch)
}
}
}
}
}
// calculate inserts, deletes, and maybe changes by folding tree A into
// tree B, adding unmatched nodes from A to B as deletes
func (d *diff) calcDeltas(t1, t2 node) (dts Deltas) {
// fold t1 into t2, adding deletes to t2
walkSorted(t1, nil, func(p []Addr, n node) bool {
if n.Match() == nil {
n.SetChangeType(DTDelete)
if cmp, ok := n.(compound); ok {
cmp.DropChildNodes()
}
addNode(t2, n, p)
// update t1 array values to reflect deletion so later comparisons will be
// accurate. only place where this really applies is parent of delete is
// an array (object paths will remain accurate)
if parent := n.Parent(); parent != nil {
if arr, ok := parent.(*array); ok {
idx, ok := n.Addr().Value().(int)
if !ok {
panic("expected int type for array address")
}
for i, n := range arr.Children() {
addr := IndexAddr(i - 1)
arr.childNames[addr] = i - 1
if i > idx {
n.SetAddr(addr)
}
}
}
}
// at this point we have the most general insert we know of
if cmp, ok := n.(compound); ok {
// drop any childs node references so subsequent iterations of the
// tree to build up a delta don't iterate any deeper
cmp.DropChildNodes()
}
// by returning false here we stop traversing to any existing children
// avoiding redundant inserts already described by the parent
return false
}
return true
})
walkSorted(t2, nil, func(p []Addr, n node) bool {
// at this point deletes from t1 have been moved here, need to skip 'em
// because n.Match will be a circular reference
if n.ChangeType() == DTDelete {
return false
}
match := n.Match()
if match == nil {
n.SetChangeType(DTInsert)
// update t1 array values to reflect insertion so later comparisons will be
// accurate. only place where this really applies is parent of insert is
// an array (object paths will remain accurate)
if parent := n.Parent(); parent != nil && parent.Type() == ntArray {
if match, ok := parent.Match().(*array); ok && match != nil {
idx, ok := n.Addr().Value().(int)
if !ok {
panic("array address index is not of integer type")
}
for i, n := range match.Children() {
a := IndexAddr(i + 1)
match.childNames[a] = i + 1
if i > idx {
n.SetAddr(a)
}
}
}
}
// at this point we have the most general insert we know of
if cmp, ok := n.(compound); ok {
// drop any childs node references so subsequent iterations of the
// tree to build up a delta don't iterate any deeper
cmp.DropChildNodes()
}
// By returning false here we stop traversing to any existing children
// avoiding redundant inserts already described by the parent
return false
}
if _, ok := n.(compound); !ok {
// check if value is scalar, creating a change delta if so
// TODO (b5): this needs to be a check to see if it's a leaf node
// (eg, empty object is a leaf node)
if delta := compareScalar(match, n, p[len(p)-1]); delta != nil {
n.SetChangeType(DTUpdate)
// TODO (b5) - restore support for change calculation, add tests
// if d.changes {
// // addDelta(root, delta, p)
// dts = append(dts, delta)
// } else {
// // addDelta(root, &Delta{Type: DTDelete, Path: p[len(p)-1], Value: delta.SourceValue}, p)
// // addDelta(root, &Delta{Type: DTInsert, Path: p[len(p)-1], Value: delta.Value}, p)
// // dts = append(dts,
// // &Delta{Type: DTDelete, Path: delta.Path, Value: delta.SourceValue},
// // &Delta{Type: DTInsert, Path: delta.Path, Value: delta.Value},
// // )
// }
}
}
return true
})
// special case where root elements aren't matched. If this happends t1 root
// will never be considered
var script Deltas
if t2.Match() == nil {
del := toDelta(t1)
ins := toDelta(t2)
script = Deltas{del, ins}
} else {
script, _ = d.childDeltas(t2.(compound))
}
sortDeltasAndMaybeCalcStats(script, d.stats)
return script
}
func (d *diff) childDeltas(cmp compound) (changes Deltas, hasChanges bool) {
ch := cmp.Children()
for _, n := range ch {
dlt := toDelta(n)
if dlt.Type == DTContext {
if childCmp, ok := n.(compound); ok {
if children, childChanges := d.childDeltas(childCmp); childChanges {
hasChanges = true
dlt.Value = nil
dlt.Deltas = children
}
}
} else {
hasChanges = true
}
// If we aren't outputting changes, convert to a delete/insert combo
if dlt.Type == DTUpdate && !d.changes {
changes = append(changes, &Delta{Type: DTDelete, Path: dlt.Path, Value: dlt.SourceValue})
dlt.Type = DTInsert
dlt.SourceValue = nil
}
changes = append(changes, dlt)
}
return changes, hasChanges
}
func sortDeltasAndMaybeCalcStats(deltas Deltas, st *Stats) {
sort.Sort(deltas)
for _, d := range deltas {
if len(d.Deltas) > 0 {
sortDeltasAndMaybeCalcStats(d.Deltas, st)
}
if st != nil {
switch d.Type {
case DTInsert:
st.Inserts++
case DTUpdate:
st.Updates++
case DTDelete:
st.Deletes++
}
}
}
}
// TODO (b5) - restore this. We need this if we want to show moves.
// // calcReorderDeltas creates deltas that describes moves within the same parent
// // it starts by calculates the largest (order preserving) common subsequence between
// // two matched parent compound nodes. Background on LCSS:
// // https://en.wikipedia.org/wiki/Longest_common_subsequence_problem
// //
// // reorder calculation is shingled into sets of maximum 50 values & processed parallel
// // to keep things fast at the expense of missing some common sequences from longer lists
// func calcReorderDeltas(a, b []node) (deltas []*Delta) {
// var wg sync.WaitGroup
// max := len(a)
// if len(b) > max {
// max = len(b)
// }
// aRem := len(a) - 1
// bRem := len(b) - 1
// pageSize := 50
// for i := 0; i <= max/pageSize; i++ {
// var aPage, bPage []node
// start := (i * pageSize)
// if (start + pageSize) > aRem {
// aPage = a[start:]
// } else {
// aPage = a[start : start+pageSize]
// }
// if (start + pageSize) > bRem {
// bPage = b[start:]
// } else {
// bPage = b[start : start+pageSize]
// }
// wg.Add(1)
// go func(a, b []node) {
// if ds := movedBNodes(a, b); ds != nil {
// deltas = append(deltas, ds...)
// }
// wg.Done()
// }(aPage, bPage)
// }
// wg.Wait()
// return
// }
// func movedBNodes(allA, allB []node) []*Delta {
// var a, b []node
// for _, n := range allA {
// if n.Match() != nil {
// a = append(a, n)
// }
// }
// for _, n := range allB {
// if n.Match() != nil {
// b = append(b, n)
// }
// }
// m := len(a) + 1
// n := len(b) + 1
// c := make([][]int, m)
// c[0] = make([]int, n)
// for i := 1; i < m; i++ {
// c[i] = make([]int, n)
// for j := 1; j < n; j++ {
// if a[i-1].Match() != nil && b[j-1].Match() != nil {
// if bytes.Equal(a[i-1].Hash(), b[j-1].Hash()) {
// c[i][j] = c[i-1][j-1] + 1
// } else {
// c[i][j] = c[i][j-1]
// if c[i-1][j] > c[i][j] {
// c[i][j] = c[i-1][j]
// }
// }
// }
// }
// }
// // TODO (b5): a & b *should* be the same length, which would mean a bottom-right
// // common-value that's equal to the length of a should mean list equality
// // which means we need to bail early b/c no moves exist
// if c[m-1][n-1] == len(a) || c[m-1][n-1] == len(b) {
// return nil
// }
// var ass, bss []node
// backtrackB(&ass, c, a, b, m-1, n-1)
// backtrackA(&bss, c, a, b, m-1, n-1)
// amv := intersect(a, ass)
// bmv := intersect(b, bss)
// var deltas []*Delta
// for i := 0; i < len(amv); i++ {
// am := amv[i]
// bm := bmv[i]
// // don't add moves that have the same source & destination paths
// // can be created by matches that move between parents
// if path(am) != path(bm) {
// mv := &Delta{
// Type: DTMove,
// Path: path(bm),
// Value: bm.Value(),
// SourcePath: path(am),
// }
// deltas = append(deltas, mv)
// }
// }
// return deltas
// }
// // intersect produces a set intersection, assuming subset is a subset of set and both nodes are ordered
// func intersect(set, subset []node) (nodes []node) {
// if len(set) == len(subset) {
// return nil
// }
// c := 0
// SET:
// for _, n := range set {
// if c == len(subset) {
// nodes = append(nodes, set[c:]...)
// break
// }
// for _, ssn := range subset[c:] {
// if bytes.Equal(n.Hash(), ssn.Hash()) {
// c++
// continue SET
// }
// }
// nodes = append(nodes, n)
// }
// return
// }
// // backtrack walks the "a" side of a common sequence matrix backward, constructing the
// // secuence of nodes from the "a" (lefthand) node list
// func backtrackA(ss *[]node, c [][]int, a, b []node, i, j int) {
// if i == 0 || j == 0 {
// return
// }
// if bytes.Equal(a[i-1].Hash(), b[j-1].Hash()) {
// // TODO (b5): I think this is where we can backtrack based on which node
// // has the greater weight by taking different paths in the commonalitiy index matrix
// // need to check...
// // if b[j].Weight() > a[i].Weight() {
// // fmt.Printf("append %p, %s\n", b[j-1], path(b[j-1]))
// *ss = append([]node{a[i-1]}, *ss...)
// // } else {
// // ss = append(ss, a[i])
// // }
// backtrackA(ss, c, a, b, i-1, j-1)
// return
// }
// if c[i][j-1] > c[i-1][j] {
// backtrackA(ss, c, a, b, i, j-1)
// return
// }
// backtrackA(ss, c, a, b, i-1, j)
// return
// }
// // backtrack walks the "b" side of a common sequence matrix backward, constructing the
// // secuence of nodes from the "b" (righthand) node list
// func backtrackB(ss *[]node, c [][]int, a, b []node, i, j int) {
// if i == 0 || j == 0 {
// return
// }
// if bytes.Equal(a[i-1].Hash(), b[j-1].Hash()) {
// // TODO (b5): I think this is where we can backtrack based on which node
// // has the greater weight by taking different paths in the commonalitiy index matrix
// // need to check...
// // if b[j].Weight() > a[i].Weight() {
// // fmt.Printf("append %p, %s\n", b[j-1], path(b[j-1]))
// *ss = append([]node{b[j-1]}, *ss...)
// // } else {
// // ss = append(ss, a[i])
// // }
// backtrackB(ss, c, a, b, i-1, j-1)
// return
// }
// if c[i][j-1] > c[i-1][j] {
// backtrackB(ss, c, a, b, i, j-1)
// return
// }
// backtrackB(ss, c, a, b, i-1, j)
// return
// }
// compareScalar compares two scalar values, possibly creating an Update delta
func compareScalar(n1, n2 node, n2Addr Addr) *Delta {
if n1.Type() != n2.Type() {
return &Delta{
Type: DTUpdate,
Path: n2Addr,
Value: n2.Value(),
SourceValue: n1.Value(),
}
}
if !reflect.DeepEqual(n1.Value(), n2.Value()) {
return &Delta{
Type: DTUpdate,
Path: n2Addr,
Value: n2.Value(),
SourceValue: n1.Value(),
}
}
return nil
}
func toDelta(n node) *Delta {
d := &Delta{Type: n.ChangeType(), Path: n.Addr()}
if string(d.Type) == "" {
d.Type = DTContext
}
switch d.Type {
case DTUpdate:
d.Value = n.Value()
d.SourceValue = n.Match().Value()
case DTInsert, DTDelete, DTContext:
d.Value = n.Value()
}
return d
}