diff_match_patch_native 1.0.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- data/.gemtest +0 -0
- data/History.txt +6 -0
- data/Manifest.txt +12 -0
- data/README.rdoc +73 -0
- data/Rakefile +28 -0
- data/ext/diff_match_patch/_dmp.cpp +287 -0
- data/ext/diff_match_patch/_dmp.h +16 -0
- data/ext/diff_match_patch/diff_match_patch.cpp +10 -0
- data/ext/diff_match_patch/extconf.rb +8 -0
- data/lib/diff_match_patch/diff_match_patch.bundle +0 -0
- data/lib/diff_match_patch.rb +18 -0
- data/src/include/diff_match_patch-stl/diff_match_patch.h +2545 -0
- data/test/test_diff_patch_match.rb +278 -0
- metadata +99 -0
@@ -0,0 +1,2545 @@
|
|
1
|
+
/*
|
2
|
+
* Copyright 2008 Google Inc. All Rights Reserved.
|
3
|
+
* Author: fraser@google.com (Neil Fraser)
|
4
|
+
* Author: mikeslemmer@gmail.com (Mike Slemmer)
|
5
|
+
* Author: snhere@gmail.com (Sergey Nozhenko)
|
6
|
+
*
|
7
|
+
* Licensed under the Apache License, Version 2.0 (the "License");
|
8
|
+
* you may not use this file except in compliance with the License.
|
9
|
+
* You may obtain a copy of the License at
|
10
|
+
*
|
11
|
+
* http://www.apache.org/licenses/LICENSE-2.0
|
12
|
+
*
|
13
|
+
* Unless required by applicable law or agreed to in writing, software
|
14
|
+
* distributed under the License is distributed on an "AS IS" BASIS,
|
15
|
+
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
16
|
+
* See the License for the specific language governing permissions and
|
17
|
+
* limitations under the License.
|
18
|
+
*
|
19
|
+
* Diff Match and Patch
|
20
|
+
* http://code.google.com/p/google-diff-match-patch/
|
21
|
+
*/
|
22
|
+
|
23
|
+
#ifndef DIFF_MATCH_PATCH_H
|
24
|
+
#define DIFF_MATCH_PATCH_H
|
25
|
+
|
26
|
+
#include <limits>
|
27
|
+
#include <list>
|
28
|
+
#include <map>
|
29
|
+
#include <vector>
|
30
|
+
#include <time.h>
|
31
|
+
|
32
|
+
|
33
|
+
/*
|
34
|
+
* Functions for diff, match and patch.
|
35
|
+
* Computes the difference between two texts to create a patch.
|
36
|
+
* Applies the patch onto another text, allowing for errors.
|
37
|
+
*
|
38
|
+
* @author fraser@google.com (Neil Fraser)
|
39
|
+
*
|
40
|
+
* Qt/C++ port by mikeslemmer@gmail.com (Mike Slemmer):
|
41
|
+
*
|
42
|
+
* STL-only port by snhere@gmail.com (Sergey Nozhenko)
|
43
|
+
*
|
44
|
+
* Here is a trivial sample program which works properly when linked with this
|
45
|
+
* library:
|
46
|
+
*
|
47
|
+
|
48
|
+
#include "diff_match_patch.h"
|
49
|
+
#include <string>
|
50
|
+
using namespace std;
|
51
|
+
int main(int argc, char **argv) {
|
52
|
+
diff_match_patch<wstring> dmp;
|
53
|
+
wstring str1 = L"First string in diff");
|
54
|
+
wstring str2 = L"Second string in diff");
|
55
|
+
|
56
|
+
wstring strPatch = dmp.patch_toText(dmp.patch_make(str1, str2));
|
57
|
+
pair<wstring, vector<bool> > out
|
58
|
+
= dmp.patch_apply(dmp.patch_fromText(strPatch), str1);
|
59
|
+
wstring strResult = out.first;
|
60
|
+
|
61
|
+
// here, strResult will equal str2 above.
|
62
|
+
return 0;
|
63
|
+
}
|
64
|
+
|
65
|
+
*/
|
66
|
+
|
67
|
+
// Character type dependencies
|
68
|
+
template <class char_t> struct diff_match_patch_traits {};
|
69
|
+
|
70
|
+
/**
|
71
|
+
* Class containing the diff, match and patch methods.
|
72
|
+
* Also contains the behaviour settings.
|
73
|
+
*/
|
74
|
+
template <class stringT, class traits = diff_match_patch_traits<typename stringT::value_type> >
|
75
|
+
class diff_match_patch {
|
76
|
+
public:
|
77
|
+
/**
|
78
|
+
* String and character types
|
79
|
+
*/
|
80
|
+
typedef stringT string_t;
|
81
|
+
typedef typename string_t::value_type char_t;
|
82
|
+
|
83
|
+
/**-
|
84
|
+
* The data structure representing a diff is a Linked list of Diff objects:
|
85
|
+
* {Diff(Operation.DELETE, "Hello"), Diff(Operation.INSERT, "Goodbye"),
|
86
|
+
* Diff(Operation.EQUAL, " world.")}
|
87
|
+
* which means: delete "Hello", add "Goodbye" and keep " world."
|
88
|
+
*/
|
89
|
+
enum Operation {
|
90
|
+
DELETE, INSERT, EQUAL
|
91
|
+
};
|
92
|
+
|
93
|
+
/**
|
94
|
+
* Class representing one diff operation.
|
95
|
+
*/
|
96
|
+
class Diff {
|
97
|
+
public:
|
98
|
+
Operation operation;
|
99
|
+
// One of: INSERT, DELETE or EQUAL.
|
100
|
+
|
101
|
+
string_t text;
|
102
|
+
// The text associated with this diff operation.
|
103
|
+
|
104
|
+
/**
|
105
|
+
* Constructor. Initializes the diff with the provided values.
|
106
|
+
* @param operation One of INSERT, DELETE or EQUAL.
|
107
|
+
* @param text The text being applied.
|
108
|
+
*/
|
109
|
+
Diff(Operation _operation, const string_t &_text) : operation(_operation), text(_text) {}
|
110
|
+
Diff() {}
|
111
|
+
|
112
|
+
/**
|
113
|
+
* Display a human-readable version of this Diff.
|
114
|
+
* @return text version.
|
115
|
+
*/
|
116
|
+
string_t toString() const {
|
117
|
+
string_t prettyText = text;
|
118
|
+
// Replace linebreaks with Pilcrow signs.
|
119
|
+
for (typename string_t::iterator i = prettyText.begin(); i != prettyText.end(); ++i)
|
120
|
+
if (traits::to_wchar(*i) == L'\n') *i = traits::from_wchar(L'\u00b6');
|
121
|
+
return traits::cs(L"Diff(") + strOperation(operation) + traits::cs(L",\"") + prettyText + traits::cs(L"\")");
|
122
|
+
}
|
123
|
+
|
124
|
+
/**
|
125
|
+
* Is this Diff equivalent to another Diff?
|
126
|
+
* @param d Another Diff to compare against
|
127
|
+
* @return true or false
|
128
|
+
*/
|
129
|
+
bool operator==(const Diff &d) const {
|
130
|
+
return (d.operation == this->operation) && (d.text == this->text);
|
131
|
+
}
|
132
|
+
bool operator!=(const Diff &d) const { return !(operator == (d)); }
|
133
|
+
|
134
|
+
static string_t strOperation(Operation op) {
|
135
|
+
switch (op) {
|
136
|
+
case INSERT:
|
137
|
+
return traits::cs(L"INSERT");
|
138
|
+
case DELETE:
|
139
|
+
return traits::cs(L"DELETE");
|
140
|
+
case EQUAL:
|
141
|
+
return traits::cs(L"EQUAL");
|
142
|
+
}
|
143
|
+
throw string_t(traits::cs(L"Invalid operation."));
|
144
|
+
}
|
145
|
+
};
|
146
|
+
|
147
|
+
typedef std::list<Diff> Diffs;
|
148
|
+
|
149
|
+
|
150
|
+
/**
|
151
|
+
* Class representing one patch operation.
|
152
|
+
*/
|
153
|
+
class Patch {
|
154
|
+
public:
|
155
|
+
Diffs diffs;
|
156
|
+
int start1;
|
157
|
+
int start2;
|
158
|
+
int length1;
|
159
|
+
int length2;
|
160
|
+
|
161
|
+
/**
|
162
|
+
* Constructor. Initializes with an empty list of diffs.
|
163
|
+
*/
|
164
|
+
Patch() : start1(0), start2(0), length1(0), length2(0) {}
|
165
|
+
|
166
|
+
bool isNull() const {
|
167
|
+
return start1 == 0 && start2 == 0 && length1 == 0 && length2 == 0 && diffs.size() == 0;
|
168
|
+
}
|
169
|
+
|
170
|
+
/**
|
171
|
+
* Emmulate GNU diff's format.
|
172
|
+
* Header: @@ -382,8 +481,9 @@
|
173
|
+
* Indicies are printed as 1-based, not 0-based.
|
174
|
+
* @return The GNU diff string
|
175
|
+
*/
|
176
|
+
string_t toString() const {
|
177
|
+
string_t coords1, coords2;
|
178
|
+
if (length1 == 0) {
|
179
|
+
coords1 = to_string(start1) + traits::cs(L",0");
|
180
|
+
} else if (length1 == 1) {
|
181
|
+
coords1 = to_string(start1 + 1);
|
182
|
+
} else {
|
183
|
+
coords1 = to_string(start1 + 1) + traits::from_wchar(L',') + to_string(length1);
|
184
|
+
}
|
185
|
+
if (length2 == 0) {
|
186
|
+
coords2 = to_string(start2) + traits::cs(L",0");
|
187
|
+
} else if (length2 == 1) {
|
188
|
+
coords2 = to_string(start2 + 1);
|
189
|
+
} else {
|
190
|
+
coords2 = to_string(start2 + 1) + traits::from_wchar(L',') + to_string(length2);
|
191
|
+
}
|
192
|
+
string_t text(traits::cs(L"@@ -") + coords1 + traits::cs(L" +") + coords2 + traits::cs(L" @@\n"));
|
193
|
+
// Escape the body of the patch with %xx notation.
|
194
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
195
|
+
switch ((*cur_diff).operation) {
|
196
|
+
case INSERT:
|
197
|
+
text += traits::from_wchar(L'+');
|
198
|
+
break;
|
199
|
+
case DELETE:
|
200
|
+
text += traits::from_wchar(L'-');
|
201
|
+
break;
|
202
|
+
case EQUAL:
|
203
|
+
text += traits::from_wchar(L' ');
|
204
|
+
break;
|
205
|
+
}
|
206
|
+
append_percent_encoded(text, (*cur_diff).text);
|
207
|
+
text += traits::from_wchar(L'\n');
|
208
|
+
}
|
209
|
+
|
210
|
+
return text;
|
211
|
+
}
|
212
|
+
};
|
213
|
+
|
214
|
+
typedef std::list<Patch> Patches;
|
215
|
+
|
216
|
+
friend class diff_match_patch_test;
|
217
|
+
|
218
|
+
public:
|
219
|
+
// Defaults.
|
220
|
+
// Set these on your diff_match_patch instance to override the defaults.
|
221
|
+
|
222
|
+
// Number of seconds to map a diff before giving up (0 for infinity).
|
223
|
+
float Diff_Timeout;
|
224
|
+
// Cost of an empty edit operation in terms of edit characters.
|
225
|
+
short Diff_EditCost;
|
226
|
+
// At what point is no match declared (0.0 = perfection, 1.0 = very loose).
|
227
|
+
float Match_Threshold;
|
228
|
+
// How far to search for a match (0 = exact location, 1000+ = broad match).
|
229
|
+
// A match this many characters away from the expected location will add
|
230
|
+
// 1.0 to the score (0.0 is a perfect match).
|
231
|
+
int Match_Distance;
|
232
|
+
// When deleting a large block of text (over ~64 characters), how close does
|
233
|
+
// the contents have to match the expected contents. (0.0 = perfection,
|
234
|
+
// 1.0 = very loose). Note that Match_Threshold controls how closely the
|
235
|
+
// end points of a delete need to match.
|
236
|
+
float Patch_DeleteThreshold;
|
237
|
+
// Chunk size for context length.
|
238
|
+
short Patch_Margin;
|
239
|
+
|
240
|
+
// The number of bits in an int.
|
241
|
+
short Match_MaxBits;
|
242
|
+
|
243
|
+
|
244
|
+
public:
|
245
|
+
|
246
|
+
diff_match_patch() :
|
247
|
+
Diff_Timeout(1.0f),
|
248
|
+
Diff_EditCost(4),
|
249
|
+
Match_Threshold(0.5f),
|
250
|
+
Match_Distance(1000),
|
251
|
+
Patch_DeleteThreshold(0.5f),
|
252
|
+
Patch_Margin(4),
|
253
|
+
Match_MaxBits(32) {
|
254
|
+
}
|
255
|
+
|
256
|
+
// DIFF FUNCTIONS
|
257
|
+
|
258
|
+
/**
|
259
|
+
* Find the differences between two texts.
|
260
|
+
* @param text1 Old string to be diffed.
|
261
|
+
* @param text2 New string to be diffed.
|
262
|
+
* @param checklines Speedup flag. If false, then don't run a
|
263
|
+
* line-level diff first to identify the changed areas.
|
264
|
+
* If true, then run a faster slightly less optimal diff.
|
265
|
+
* Most of the time checklines is wanted, so default to true.
|
266
|
+
* @return Linked List of Diff objects.
|
267
|
+
*/
|
268
|
+
Diffs diff_main(const string_t &text1, const string_t &text2, bool checklines = true) const {
|
269
|
+
// Set a deadline by which time the diff must be complete.
|
270
|
+
clock_t deadline;
|
271
|
+
if (Diff_Timeout <= 0) {
|
272
|
+
deadline = std::numeric_limits<clock_t>::max();
|
273
|
+
} else {
|
274
|
+
deadline = clock() + (clock_t)(Diff_Timeout * CLOCKS_PER_SEC);
|
275
|
+
}
|
276
|
+
Diffs diffs;
|
277
|
+
diff_main(text1, text2, checklines, deadline, diffs);
|
278
|
+
return diffs;
|
279
|
+
}
|
280
|
+
|
281
|
+
/**
|
282
|
+
* Find the differences between two texts. Simplifies the problem by
|
283
|
+
* stripping any common prefix or suffix off the texts before diffing.
|
284
|
+
* @param text1 Old string to be diffed.
|
285
|
+
* @param text2 New string to be diffed.
|
286
|
+
* @param checklines Speedup flag. If false, then don't run a
|
287
|
+
* line-level diff first to identify the changed areas.
|
288
|
+
* If true, then run a faster slightly less optimal diff.
|
289
|
+
* @param deadline Time when the diff should be complete by. Used
|
290
|
+
* internally for recursive calls. Users should set DiffTimeout instead.
|
291
|
+
* @param Linked List of Diff objects.
|
292
|
+
*/
|
293
|
+
private:
|
294
|
+
static void diff_main(const string_t &text1, const string_t &text2, bool checklines, clock_t deadline, Diffs& diffs) {
|
295
|
+
diffs.clear();
|
296
|
+
|
297
|
+
// Check for equality (speedup).
|
298
|
+
if (text1 == text2) {
|
299
|
+
if (!text1.empty()) {
|
300
|
+
diffs.push_back(Diff(EQUAL, text1));
|
301
|
+
}
|
302
|
+
}
|
303
|
+
else {
|
304
|
+
// Trim off common prefix (speedup).
|
305
|
+
int commonlength = diff_commonPrefix(text1, text2);
|
306
|
+
const string_t &commonprefix = text1.substr(0, commonlength);
|
307
|
+
string_t textChopped1 = text1.substr(commonlength);
|
308
|
+
string_t textChopped2 = text2.substr(commonlength);
|
309
|
+
|
310
|
+
// Trim off common suffix (speedup).
|
311
|
+
commonlength = diff_commonSuffix(textChopped1, textChopped2);
|
312
|
+
const string_t &commonsuffix = right(textChopped1, commonlength);
|
313
|
+
textChopped1 = textChopped1.substr(0, textChopped1.length() - commonlength);
|
314
|
+
textChopped2 = textChopped2.substr(0, textChopped2.length() - commonlength);
|
315
|
+
|
316
|
+
// Compute the diff on the middle block.
|
317
|
+
diff_compute(textChopped1, textChopped2, checklines, deadline, diffs);
|
318
|
+
|
319
|
+
// Restore the prefix and suffix.
|
320
|
+
if (!commonprefix.empty()) {
|
321
|
+
diffs.push_front(Diff(EQUAL, commonprefix));
|
322
|
+
}
|
323
|
+
if (!commonsuffix.empty()) {
|
324
|
+
diffs.push_back(Diff(EQUAL, commonsuffix));
|
325
|
+
}
|
326
|
+
|
327
|
+
diff_cleanupMerge(diffs);
|
328
|
+
}
|
329
|
+
}
|
330
|
+
|
331
|
+
/**
|
332
|
+
* Find the differences between two texts. Assumes that the texts do not
|
333
|
+
* have any common prefix or suffix.
|
334
|
+
* @param text1 Old string to be diffed.
|
335
|
+
* @param text2 New string to be diffed.
|
336
|
+
* @param checklines Speedup flag. If false, then don't run a
|
337
|
+
* line-level diff first to identify the changed areas.
|
338
|
+
* If true, then run a faster slightly less optimal diff.
|
339
|
+
* @param deadline Time when the diff should be complete by.
|
340
|
+
* @param Linked List of Diff objects.
|
341
|
+
*/
|
342
|
+
private:
|
343
|
+
static void diff_compute(string_t text1, string_t text2, bool checklines, clock_t deadline, Diffs& diffs) {
|
344
|
+
if (text1.empty()) {
|
345
|
+
// Just add some text (speedup).
|
346
|
+
diffs.push_back(Diff(INSERT, text2));
|
347
|
+
return;
|
348
|
+
}
|
349
|
+
|
350
|
+
if (text2.empty()) {
|
351
|
+
// Just delete some text (speedup).
|
352
|
+
diffs.push_back(Diff(DELETE, text1));
|
353
|
+
return;
|
354
|
+
}
|
355
|
+
|
356
|
+
{
|
357
|
+
const string_t& longtext = text1.length() > text2.length() ? text1 : text2;
|
358
|
+
const string_t& shorttext = text1.length() > text2.length() ? text2 : text1;
|
359
|
+
const size_t i = longtext.find(shorttext);
|
360
|
+
if (i != string_t::npos) {
|
361
|
+
// Shorter text is inside the longer text (speedup).
|
362
|
+
const Operation op = (text1.length() > text2.length()) ? DELETE : INSERT;
|
363
|
+
diffs.push_back(Diff(op, longtext.substr(0, i)));
|
364
|
+
diffs.push_back(Diff(EQUAL, shorttext));
|
365
|
+
diffs.push_back(Diff(op, safeMid(longtext, i + shorttext.length())));
|
366
|
+
return;
|
367
|
+
}
|
368
|
+
|
369
|
+
if (shorttext.length() == 1) {
|
370
|
+
// Single character string.
|
371
|
+
// After the previous speedup, the character can't be an equality.
|
372
|
+
diffs.push_back(Diff(DELETE, text1));
|
373
|
+
diffs.push_back(Diff(INSERT, text2));
|
374
|
+
return;
|
375
|
+
}
|
376
|
+
// Garbage collect longtext and shorttext by scoping out.
|
377
|
+
}
|
378
|
+
|
379
|
+
// Don't risk returning a non-optimal diff if we have unlimited time.
|
380
|
+
if (deadline != std::numeric_limits<clock_t>::max()) {
|
381
|
+
// Check to see if the problem can be split in two.
|
382
|
+
HalfMatchResult hm;
|
383
|
+
if (diff_halfMatch(text1, text2, hm)) {
|
384
|
+
// A half-match was found, sort out the return data.
|
385
|
+
// Send both pairs off for separate processing.
|
386
|
+
diff_main(hm.text1_a, hm.text2_a, checklines, deadline, diffs);
|
387
|
+
diffs.push_back(Diff(EQUAL, hm.mid_common));
|
388
|
+
Diffs diffs_b;
|
389
|
+
diff_main(hm.text1_b, hm.text2_b, checklines, deadline, diffs_b);
|
390
|
+
diffs.splice(diffs.end(), diffs_b);
|
391
|
+
return;
|
392
|
+
}
|
393
|
+
}
|
394
|
+
|
395
|
+
// Perform a real diff.
|
396
|
+
if (checklines && text1.length() > 100 && text2.length() > 100) {
|
397
|
+
diff_lineMode(text1, text2, deadline, diffs);
|
398
|
+
return;
|
399
|
+
}
|
400
|
+
|
401
|
+
diff_bisect(text1, text2, deadline, diffs);
|
402
|
+
}
|
403
|
+
|
404
|
+
/**
|
405
|
+
* Do a quick line-level diff on both strings, then rediff the parts for
|
406
|
+
* greater accuracy.
|
407
|
+
* This speedup can produce non-minimal diffs.
|
408
|
+
* @param text1 Old string to be diffed.
|
409
|
+
* @param text2 New string to be diffed.
|
410
|
+
* @param deadline Time when the diff should be complete by.
|
411
|
+
* @param Linked List of Diff objects.
|
412
|
+
*/
|
413
|
+
private:
|
414
|
+
static void diff_lineMode(string_t text1, string_t text2, clock_t deadline, Diffs& diffs) {
|
415
|
+
// Scan the text on a line-by-line basis first.
|
416
|
+
Lines linearray;
|
417
|
+
diff_linesToChars(text1, text2, linearray);
|
418
|
+
|
419
|
+
diff_main(text1, text2, false, deadline, diffs);
|
420
|
+
|
421
|
+
// Convert the diff back to original text.
|
422
|
+
diff_charsToLines(diffs, linearray);
|
423
|
+
// Eliminate freak matches (e.g. blank lines)
|
424
|
+
diff_cleanupSemantic(diffs);
|
425
|
+
|
426
|
+
// Rediff any replacement blocks, this time character-by-character.
|
427
|
+
// Add a dummy entry at the end.
|
428
|
+
diffs.push_back(Diff(EQUAL, string_t()));
|
429
|
+
int count_delete = 0;
|
430
|
+
int count_insert = 0;
|
431
|
+
string_t text_delete;
|
432
|
+
string_t text_insert;
|
433
|
+
|
434
|
+
for (typename Diffs::iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
435
|
+
switch ((*cur_diff).operation) {
|
436
|
+
case INSERT:
|
437
|
+
count_insert++;
|
438
|
+
text_insert += (*cur_diff).text;
|
439
|
+
break;
|
440
|
+
case DELETE:
|
441
|
+
count_delete++;
|
442
|
+
text_delete += (*cur_diff).text;
|
443
|
+
break;
|
444
|
+
case EQUAL:
|
445
|
+
// Upon reaching an equality, check for prior redundancies.
|
446
|
+
if (count_delete >= 1 && count_insert >= 1) {
|
447
|
+
// Delete the offending records and add the merged ones.
|
448
|
+
typename Diffs::iterator last = cur_diff;
|
449
|
+
std::advance(cur_diff, -(count_delete + count_insert));
|
450
|
+
cur_diff = diffs.erase(cur_diff, last);
|
451
|
+
|
452
|
+
Diffs new_diffs;
|
453
|
+
diff_main(text_delete, text_insert, false, deadline, new_diffs);
|
454
|
+
diffs.splice(cur_diff++, new_diffs);
|
455
|
+
--cur_diff;
|
456
|
+
}
|
457
|
+
count_insert = 0;
|
458
|
+
count_delete = 0;
|
459
|
+
text_delete.clear();
|
460
|
+
text_insert.clear();
|
461
|
+
break;
|
462
|
+
}
|
463
|
+
}
|
464
|
+
diffs.pop_back(); // Remove the dummy entry at the end.
|
465
|
+
}
|
466
|
+
|
467
|
+
/**
|
468
|
+
* Find the 'middle snake' of a diff, split the problem in two
|
469
|
+
* and return the recursively constructed diff.
|
470
|
+
* See Myers 1986 paper: An O(ND) Difference Algorithm and Its Variations.
|
471
|
+
* @param text1 Old string to be diffed.
|
472
|
+
* @param text2 New string to be diffed.
|
473
|
+
* @return Linked List of Diff objects.
|
474
|
+
*/
|
475
|
+
protected:
|
476
|
+
static Diffs diff_bisect(const string_t &text1, const string_t &text2, clock_t deadline) {
|
477
|
+
Diffs diffs;
|
478
|
+
diff_bisect(text1, text2, deadline, diffs);
|
479
|
+
return diffs;
|
480
|
+
}
|
481
|
+
private:
|
482
|
+
static void diff_bisect(const string_t &text1, const string_t &text2, clock_t deadline, Diffs& diffs) {
|
483
|
+
// Cache the text lengths to prevent multiple calls.
|
484
|
+
const int text1_length = text1.length();
|
485
|
+
const int text2_length = text2.length();
|
486
|
+
const int max_d = (text1_length + text2_length + 1) / 2;
|
487
|
+
const int v_offset = max_d;
|
488
|
+
const int v_length = 2 * max_d;
|
489
|
+
std::vector<int> v1(v_length, -1),
|
490
|
+
v2(v_length, -1);
|
491
|
+
v1[v_offset + 1] = 0;
|
492
|
+
v2[v_offset + 1] = 0;
|
493
|
+
const int delta = text1_length - text2_length;
|
494
|
+
// If the total number of characters is odd, then the front path will
|
495
|
+
// collide with the reverse path.
|
496
|
+
const bool front = (delta % 2 != 0);
|
497
|
+
// Offsets for start and end of k loop.
|
498
|
+
// Prevents mapping of space beyond the grid.
|
499
|
+
int k1start = 0;
|
500
|
+
int k1end = 0;
|
501
|
+
int k2start = 0;
|
502
|
+
int k2end = 0;
|
503
|
+
for (int d = 0; d < max_d; d++) {
|
504
|
+
// Bail out if deadline is reached.
|
505
|
+
if (clock() > deadline) {
|
506
|
+
break;
|
507
|
+
}
|
508
|
+
|
509
|
+
// Walk the front path one step.
|
510
|
+
for (int k1 = -d + k1start; k1 <= d - k1end; k1 += 2) {
|
511
|
+
const int k1_offset = v_offset + k1;
|
512
|
+
int x1;
|
513
|
+
if (k1 == -d || (k1 != d && v1[k1_offset - 1] < v1[k1_offset + 1])) {
|
514
|
+
x1 = v1[k1_offset + 1];
|
515
|
+
} else {
|
516
|
+
x1 = v1[k1_offset - 1] + 1;
|
517
|
+
}
|
518
|
+
int y1 = x1 - k1;
|
519
|
+
while (x1 < text1_length && y1 < text2_length
|
520
|
+
&& text1[x1] == text2[y1]) {
|
521
|
+
x1++;
|
522
|
+
y1++;
|
523
|
+
}
|
524
|
+
v1[k1_offset] = x1;
|
525
|
+
if (x1 > text1_length) {
|
526
|
+
// Ran off the right of the graph.
|
527
|
+
k1end += 2;
|
528
|
+
} else if (y1 > text2_length) {
|
529
|
+
// Ran off the bottom of the graph.
|
530
|
+
k1start += 2;
|
531
|
+
} else if (front) {
|
532
|
+
int k2_offset = v_offset + delta - k1;
|
533
|
+
if (k2_offset >= 0 && k2_offset < v_length && v2[k2_offset] != -1) {
|
534
|
+
// Mirror x2 onto top-left coordinate system.
|
535
|
+
int x2 = text1_length - v2[k2_offset];
|
536
|
+
if (x1 >= x2) {
|
537
|
+
// Overlap detected.
|
538
|
+
diff_bisectSplit(text1, text2, x1, y1, deadline, diffs);
|
539
|
+
return;
|
540
|
+
}
|
541
|
+
}
|
542
|
+
}
|
543
|
+
}
|
544
|
+
|
545
|
+
// Walk the reverse path one step.
|
546
|
+
for (int k2 = -d + k2start; k2 <= d - k2end; k2 += 2) {
|
547
|
+
const int k2_offset = v_offset + k2;
|
548
|
+
int x2;
|
549
|
+
if (k2 == -d || (k2 != d && v2[k2_offset - 1] < v2[k2_offset + 1])) {
|
550
|
+
x2 = v2[k2_offset + 1];
|
551
|
+
} else {
|
552
|
+
x2 = v2[k2_offset - 1] + 1;
|
553
|
+
}
|
554
|
+
int y2 = x2 - k2;
|
555
|
+
while (x2 < text1_length && y2 < text2_length
|
556
|
+
&& text1[text1_length - x2 - 1] == text2[text2_length - y2 - 1]) {
|
557
|
+
x2++;
|
558
|
+
y2++;
|
559
|
+
}
|
560
|
+
v2[k2_offset] = x2;
|
561
|
+
if (x2 > text1_length) {
|
562
|
+
// Ran off the left of the graph.
|
563
|
+
k2end += 2;
|
564
|
+
} else if (y2 > text2_length) {
|
565
|
+
// Ran off the top of the graph.
|
566
|
+
k2start += 2;
|
567
|
+
} else if (!front) {
|
568
|
+
int k1_offset = v_offset + delta - k2;
|
569
|
+
if (k1_offset >= 0 && k1_offset < v_length && v1[k1_offset] != -1) {
|
570
|
+
int x1 = v1[k1_offset];
|
571
|
+
int y1 = v_offset + x1 - k1_offset;
|
572
|
+
// Mirror x2 onto top-left coordinate system.
|
573
|
+
x2 = text1_length - x2;
|
574
|
+
if (x1 >= x2) {
|
575
|
+
// Overlap detected.
|
576
|
+
diff_bisectSplit(text1, text2, x1, y1, deadline, diffs);
|
577
|
+
return;
|
578
|
+
}
|
579
|
+
}
|
580
|
+
}
|
581
|
+
}
|
582
|
+
}
|
583
|
+
// Diff took too long and hit the deadline or
|
584
|
+
// number of diffs equals number of characters, no commonality at all.
|
585
|
+
diffs.clear();
|
586
|
+
diffs.push_back(Diff(DELETE, text1));
|
587
|
+
diffs.push_back(Diff(INSERT, text2));
|
588
|
+
}
|
589
|
+
|
590
|
+
/**
|
591
|
+
* Given the location of the 'middle snake', split the diff in two parts
|
592
|
+
* and recurse.
|
593
|
+
* @param text1 Old string to be diffed.
|
594
|
+
* @param text2 New string to be diffed.
|
595
|
+
* @param x Index of split point in text1.
|
596
|
+
* @param y Index of split point in text2.
|
597
|
+
* @param deadline Time at which to bail if not yet complete.
|
598
|
+
* @return LinkedList of Diff objects.
|
599
|
+
*/
|
600
|
+
private:
|
601
|
+
static void diff_bisectSplit(const string_t &text1, const string_t &text2, int x, int y, clock_t deadline, Diffs& diffs) {
|
602
|
+
string_t text1a = text1.substr(0, x);
|
603
|
+
string_t text2a = text2.substr(0, y);
|
604
|
+
string_t text1b = safeMid(text1, x);
|
605
|
+
string_t text2b = safeMid(text2, y);
|
606
|
+
|
607
|
+
// Compute both diffs serially.
|
608
|
+
diff_main(text1a, text2a, false, deadline, diffs);
|
609
|
+
Diffs diffs_b;
|
610
|
+
diff_main(text1b, text2b, false, deadline, diffs_b);
|
611
|
+
diffs.splice(diffs.end(), diffs_b);
|
612
|
+
}
|
613
|
+
|
614
|
+
/**
|
615
|
+
* Split two texts into a list of strings. Reduce the texts to a string of
|
616
|
+
* hashes where each Unicode character represents one line.
|
617
|
+
* @param text1 First string.
|
618
|
+
* @param text2 Second string.
|
619
|
+
* @return Lines object, containing the encoded text1, the
|
620
|
+
* encoded text2 and the List of pointers to unique strings. The zeroth element
|
621
|
+
* of the List of unique strings is intentionally blank.
|
622
|
+
*/
|
623
|
+
protected:
|
624
|
+
struct LinePtr : std::pair<typename string_t::const_pointer, size_t> {
|
625
|
+
LinePtr() {}
|
626
|
+
LinePtr(typename string_t::const_pointer p, size_t n) : std::pair<typename string_t::const_pointer, size_t>(p, n) {}
|
627
|
+
bool operator<(const LinePtr& p) const
|
628
|
+
{ return this->second < p.second? true : this->second > p.second? false : string_t::traits_type::compare(this->first, p.first, this->second) < 0; }
|
629
|
+
};
|
630
|
+
struct Lines : std::vector<LinePtr> { string_t text1, text2; };
|
631
|
+
|
632
|
+
static void diff_linesToChars(string_t &text1, string_t &text2, Lines& lineArray) {
|
633
|
+
std::map<LinePtr, size_t> lineHash;
|
634
|
+
lineArray.text1.swap(text1), lineArray.text2.swap(text2);
|
635
|
+
// e.g. linearray[4] == "Hello\n"
|
636
|
+
// e.g. linehash.get("Hello\n") == 4
|
637
|
+
|
638
|
+
// "\x00" is a valid character, but various debuggers don't like it.
|
639
|
+
// So we'll insert a junk entry to avoid generating a null character.
|
640
|
+
|
641
|
+
text1 = diff_linesToCharsMunge(lineArray.text1, lineHash);
|
642
|
+
text2 = diff_linesToCharsMunge(lineArray.text2, lineHash);
|
643
|
+
|
644
|
+
lineArray.resize(lineHash.size() + 1);
|
645
|
+
for (typename std::map<LinePtr, size_t>::const_iterator i = lineHash.begin(); i != lineHash.end(); ++i)
|
646
|
+
lineArray[(*i).second] = (*i).first;
|
647
|
+
}
|
648
|
+
|
649
|
+
/**
|
650
|
+
* Split a text into a list of pointers to strings. Reduce the texts to a string of
|
651
|
+
* hashes where each Unicode character represents one line.
|
652
|
+
* @param text String to encode.
|
653
|
+
* @param lineHash Map of string pointers to indices.
|
654
|
+
* @return Encoded string.
|
655
|
+
*/
|
656
|
+
private:
|
657
|
+
static string_t diff_linesToCharsMunge(const string_t &text, std::map<LinePtr, size_t> &lineHash) {
|
658
|
+
string_t chars;
|
659
|
+
// Walk the text, pulling out a substring for each line.
|
660
|
+
// text.split('\n') would would temporarily double our memory footprint.
|
661
|
+
// Modifying text would create many large strings to garbage collect.
|
662
|
+
typename string_t::size_type lineLen;
|
663
|
+
for (typename string_t::const_pointer lineStart = text.c_str(), textEnd = lineStart + text.size(); lineStart < textEnd; lineStart += lineLen + 1) {
|
664
|
+
lineLen = next_token(text, traits::from_wchar(L'\n'), lineStart);
|
665
|
+
if (lineStart + lineLen == textEnd) --lineLen;
|
666
|
+
chars += (char_t)(*lineHash.insert(std::make_pair(LinePtr(lineStart, lineLen + 1), lineHash.size() + 1)).first).second;
|
667
|
+
}
|
668
|
+
return chars;
|
669
|
+
}
|
670
|
+
|
671
|
+
/**
|
672
|
+
* Rehydrate the text in a diff from a string of line hashes to real lines of
|
673
|
+
* text.
|
674
|
+
* @param diffs LinkedList of Diff objects.
|
675
|
+
* @param lineArray List of pointers to unique strings.
|
676
|
+
*/
|
677
|
+
private:
|
678
|
+
static void diff_charsToLines(Diffs &diffs, const Lines& lineArray) {
|
679
|
+
for (typename Diffs::iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
680
|
+
string_t text;
|
681
|
+
for (int y = 0; y < (int)(*cur_diff).text.length(); y++) {
|
682
|
+
const LinePtr& lp = lineArray[static_cast<size_t>((*cur_diff).text[y])];
|
683
|
+
text.append(lp.first, lp.second);
|
684
|
+
}
|
685
|
+
(*cur_diff).text.swap(text);
|
686
|
+
}
|
687
|
+
}
|
688
|
+
|
689
|
+
/**
|
690
|
+
* Determine the common prefix of two strings.
|
691
|
+
* @param text1 First string.
|
692
|
+
* @param text2 Second string.
|
693
|
+
* @return The number of characters common to the start of each string.
|
694
|
+
*/
|
695
|
+
public:
|
696
|
+
static int diff_commonPrefix(const string_t &text1, const string_t &text2) {
|
697
|
+
// Performance analysis: http://neil.fraser.name/news/2007/10/09/
|
698
|
+
const int n = std::min(text1.length(), text2.length());
|
699
|
+
for (int i = 0; i < n; i++) {
|
700
|
+
if (text1[i] != text2[i]) {
|
701
|
+
return i;
|
702
|
+
}
|
703
|
+
}
|
704
|
+
return n;
|
705
|
+
}
|
706
|
+
|
707
|
+
/**
|
708
|
+
* Determine the common suffix of two strings.
|
709
|
+
* @param text1 First string.
|
710
|
+
* @param text2 Second string.
|
711
|
+
* @return The number of characters common to the end of each string.
|
712
|
+
*/
|
713
|
+
public:
|
714
|
+
static int diff_commonSuffix(const string_t &text1, const string_t &text2) {
|
715
|
+
// Performance analysis: http://neil.fraser.name/news/2007/10/09/
|
716
|
+
const int text1_length = text1.length();
|
717
|
+
const int text2_length = text2.length();
|
718
|
+
const int n = std::min(text1_length, text2_length);
|
719
|
+
for (int i = 1; i <= n; i++) {
|
720
|
+
if (text1[text1_length - i] != text2[text2_length - i]) {
|
721
|
+
return i - 1;
|
722
|
+
}
|
723
|
+
}
|
724
|
+
return n;
|
725
|
+
}
|
726
|
+
|
727
|
+
/**
|
728
|
+
* Determine if the suffix of one string is the prefix of another.
|
729
|
+
* @param text1 First string.
|
730
|
+
* @param text2 Second string.
|
731
|
+
* @return The number of characters common to the end of the first
|
732
|
+
* string and the start of the second string.
|
733
|
+
*/
|
734
|
+
protected:
|
735
|
+
static int diff_commonOverlap(const string_t &text1, const string_t &text2) {
|
736
|
+
// Cache the text lengths to prevent multiple calls.
|
737
|
+
const int text1_length = text1.length();
|
738
|
+
const int text2_length = text2.length();
|
739
|
+
// Eliminate the null case.
|
740
|
+
if (text1_length == 0 || text2_length == 0) {
|
741
|
+
return 0;
|
742
|
+
}
|
743
|
+
// Truncate the longer string.
|
744
|
+
string_t text1_trunc = text1;
|
745
|
+
string_t text2_trunc = text2;
|
746
|
+
if (text1_length > text2_length) {
|
747
|
+
text1_trunc = right(text1, text2_length);
|
748
|
+
} else if (text1_length < text2_length) {
|
749
|
+
text2_trunc = text2.substr(0, text1_length);
|
750
|
+
}
|
751
|
+
const int text_length = std::min(text1_length, text2_length);
|
752
|
+
// Quick check for the worst case.
|
753
|
+
if (text1_trunc == text2_trunc) {
|
754
|
+
return text_length;
|
755
|
+
}
|
756
|
+
|
757
|
+
// Start by looking for a single character match
|
758
|
+
// and increase length until no match is found.
|
759
|
+
// Performance analysis: http://neil.fraser.name/news/2010/11/04/
|
760
|
+
int best = 0;
|
761
|
+
int length = 1;
|
762
|
+
while (true) {
|
763
|
+
string_t pattern = right(text1_trunc, length);
|
764
|
+
size_t found = text2_trunc.find(pattern);
|
765
|
+
if (found == string_t::npos) {
|
766
|
+
return best;
|
767
|
+
}
|
768
|
+
length += found;
|
769
|
+
if (found == 0 || right(text1_trunc, length) == right(text2_trunc, length)) {
|
770
|
+
best = length;
|
771
|
+
length++;
|
772
|
+
}
|
773
|
+
}
|
774
|
+
}
|
775
|
+
|
776
|
+
/**
|
777
|
+
* Do the two texts share a substring which is at least half the length of
|
778
|
+
* the longer text?
|
779
|
+
* This speedup can produce non-minimal diffs.
|
780
|
+
* @param text1 First string.
|
781
|
+
* @param text2 Second string.
|
782
|
+
* @param HalfMatchResult object, containing the prefix of text1, the
|
783
|
+
* suffix of text1, the prefix of text2, the suffix of text2 and the
|
784
|
+
* common middle.
|
785
|
+
* @return Boolean true if there was a match, false otherwise.
|
786
|
+
*/
|
787
|
+
protected:
|
788
|
+
struct HalfMatchResult {
|
789
|
+
string_t text1_a, text1_b, text2_a, text2_b, mid_common;
|
790
|
+
void swap(HalfMatchResult& hm) {
|
791
|
+
text1_a.swap(hm.text1_a), text1_b.swap(hm.text1_b), text2_a.swap(hm.text2_a), text2_b.swap(hm.text2_b), mid_common.swap(hm.mid_common);
|
792
|
+
}
|
793
|
+
};
|
794
|
+
|
795
|
+
static bool diff_halfMatch(const string_t &text1, const string_t &text2, HalfMatchResult& hm) {
|
796
|
+
const string_t longtext = text1.length() > text2.length() ? text1 : text2;
|
797
|
+
const string_t shorttext = text1.length() > text2.length() ? text2 : text1;
|
798
|
+
if (longtext.length() < 4 || shorttext.length() * 2 < longtext.length()) {
|
799
|
+
return false; // Pointless.
|
800
|
+
}
|
801
|
+
|
802
|
+
HalfMatchResult res1, res2;
|
803
|
+
// First check if the second quarter is the seed for a half-match.
|
804
|
+
bool hm1 = diff_halfMatchI(longtext, shorttext,
|
805
|
+
(longtext.length() + 3) / 4, res1);
|
806
|
+
// Check again based on the third quarter.
|
807
|
+
bool hm2 = diff_halfMatchI(longtext, shorttext,
|
808
|
+
(longtext.length() + 1) / 2, res2);
|
809
|
+
if (!hm1 && !hm2) {
|
810
|
+
return false;
|
811
|
+
} else if (!hm2) {
|
812
|
+
hm.swap(res1);
|
813
|
+
} else if (!hm1) {
|
814
|
+
hm.swap(res2);
|
815
|
+
} else {
|
816
|
+
// Both matched. Select the longest.
|
817
|
+
hm.swap(res1.mid_common.length() > res2.mid_common.length() ? res1 : res2);
|
818
|
+
}
|
819
|
+
|
820
|
+
// A half-match was found, sort out the return data.
|
821
|
+
if (text1.length() <= text2.length()) {
|
822
|
+
hm.text1_a.swap(hm.text2_a);
|
823
|
+
hm.text1_b.swap(hm.text2_b);
|
824
|
+
}
|
825
|
+
return true;
|
826
|
+
}
|
827
|
+
|
828
|
+
/**
|
829
|
+
* Does a substring of shorttext exist within longtext such that the
|
830
|
+
* substring is at least half the length of longtext?
|
831
|
+
* @param longtext Longer string.
|
832
|
+
* @param shorttext Shorter string.
|
833
|
+
* @param i Start index of quarter length substring within longtext.
|
834
|
+
* @param HalfMatchResult object, containing the prefix of longtext, the
|
835
|
+
* suffix of longtext, the prefix of shorttext, the suffix of shorttext
|
836
|
+
* and the common middle.
|
837
|
+
* @return Boolean true if there was a match, false otherwise.
|
838
|
+
*/
|
839
|
+
private:
|
840
|
+
static bool diff_halfMatchI(const string_t &longtext, const string_t &shorttext, int i, HalfMatchResult& best) {
|
841
|
+
// Start with a 1/4 length substring at position i as a seed.
|
842
|
+
const string_t seed = safeMid(longtext, i, longtext.length() / 4);
|
843
|
+
size_t j = string_t::npos;
|
844
|
+
while ((j = shorttext.find(seed, j + 1)) != string_t::npos) {
|
845
|
+
const int prefixLength = diff_commonPrefix(safeMid(longtext, i),
|
846
|
+
safeMid(shorttext, j));
|
847
|
+
const int suffixLength = diff_commonSuffix(longtext.substr(0, i),
|
848
|
+
shorttext.substr(0, j));
|
849
|
+
if ((int)best.mid_common.length() < suffixLength + prefixLength) {
|
850
|
+
best.mid_common = safeMid(shorttext, j - suffixLength, suffixLength)
|
851
|
+
+ safeMid(shorttext, j, prefixLength);
|
852
|
+
best.text1_a = longtext.substr(0, i - suffixLength);
|
853
|
+
best.text1_b = safeMid(longtext, i + prefixLength);
|
854
|
+
best.text2_a = shorttext.substr(0, j - suffixLength);
|
855
|
+
best.text2_b = safeMid(shorttext, j + prefixLength);
|
856
|
+
}
|
857
|
+
}
|
858
|
+
return best.mid_common.length() * 2 >= longtext.length();
|
859
|
+
}
|
860
|
+
|
861
|
+
/**
|
862
|
+
* Reduce the number of edits by eliminating semantically trivial equalities.
|
863
|
+
* @param diffs LinkedList of Diff objects.
|
864
|
+
*/
|
865
|
+
public:
|
866
|
+
static void diff_cleanupSemantic(Diffs &diffs) {
|
867
|
+
if (diffs.empty()) {
|
868
|
+
return;
|
869
|
+
}
|
870
|
+
bool changes = false;
|
871
|
+
std::vector<typename Diffs::iterator> equalities; // Stack of equalities.
|
872
|
+
string_t lastequality; // Always equal to equalities.lastElement().text
|
873
|
+
typename Diffs::iterator cur_diff;
|
874
|
+
// Number of characters that changed prior to the equality.
|
875
|
+
int length_insertions1 = 0;
|
876
|
+
int length_deletions1 = 0;
|
877
|
+
// Number of characters that changed after the equality.
|
878
|
+
int length_insertions2 = 0;
|
879
|
+
int length_deletions2 = 0;
|
880
|
+
for (cur_diff = diffs.begin(); cur_diff != diffs.end();) {
|
881
|
+
if ((*cur_diff).operation == EQUAL) {
|
882
|
+
// Equality found.
|
883
|
+
equalities.push_back(cur_diff);
|
884
|
+
length_insertions1 = length_insertions2;
|
885
|
+
length_deletions1 = length_deletions2;
|
886
|
+
length_insertions2 = 0;
|
887
|
+
length_deletions2 = 0;
|
888
|
+
lastequality = (*cur_diff).text;
|
889
|
+
} else {
|
890
|
+
// An insertion or deletion.
|
891
|
+
if ((*cur_diff).operation == INSERT) {
|
892
|
+
length_insertions2 += (*cur_diff).text.length();
|
893
|
+
} else {
|
894
|
+
length_deletions2 += (*cur_diff).text.length();
|
895
|
+
}
|
896
|
+
if (!lastequality.empty()
|
897
|
+
&& ((int)lastequality.length()
|
898
|
+
<= std::max(length_insertions1, length_deletions1))
|
899
|
+
&& ((int)lastequality.length()
|
900
|
+
<= std::max(length_insertions2, length_deletions2))) {
|
901
|
+
// printf("Splitting: '%s'\n", qPrintable(lastequality));
|
902
|
+
// Walk back to offending equality.
|
903
|
+
// Change second copy to insert.
|
904
|
+
(*(cur_diff = equalities.back())).operation = INSERT;
|
905
|
+
// Duplicate record.
|
906
|
+
diffs.insert(cur_diff, Diff(DELETE, lastequality));
|
907
|
+
equalities.pop_back(); // Throw away the equality we just deleted.
|
908
|
+
if (!equalities.empty()) {
|
909
|
+
// Throw away the previous equality (it needs to be reevaluated).
|
910
|
+
equalities.pop_back();
|
911
|
+
}
|
912
|
+
length_insertions1 = 0; // Reset the counters.
|
913
|
+
length_deletions1 = 0;
|
914
|
+
length_insertions2 = 0;
|
915
|
+
length_deletions2 = 0;
|
916
|
+
lastequality = string_t();
|
917
|
+
changes = true;
|
918
|
+
|
919
|
+
if (!equalities.empty())
|
920
|
+
// There is a safe equality we can fall back to.
|
921
|
+
cur_diff = equalities.back();
|
922
|
+
else
|
923
|
+
{
|
924
|
+
// There are no previous equalities, walk back to the start.
|
925
|
+
cur_diff = diffs.begin();
|
926
|
+
continue;
|
927
|
+
}
|
928
|
+
}
|
929
|
+
}
|
930
|
+
++cur_diff;
|
931
|
+
}
|
932
|
+
|
933
|
+
// Normalize the diff.
|
934
|
+
if (changes) {
|
935
|
+
diff_cleanupMerge(diffs);
|
936
|
+
}
|
937
|
+
diff_cleanupSemanticLossless(diffs);
|
938
|
+
|
939
|
+
// Find any overlaps between deletions and insertions.
|
940
|
+
// e.g: <del>abcxx</del><ins>xxdef</ins>
|
941
|
+
// -> <del>abc</del>xx<ins>def</ins>
|
942
|
+
if ((cur_diff = diffs.begin()) != diffs.end()) {
|
943
|
+
for (typename Diffs::iterator prev_diff = cur_diff; ++cur_diff != diffs.end(); prev_diff = cur_diff) {
|
944
|
+
if ((*prev_diff).operation == DELETE &&
|
945
|
+
(*cur_diff).operation == INSERT) {
|
946
|
+
string_t deletion = (*prev_diff).text;
|
947
|
+
string_t insertion = (*cur_diff).text;
|
948
|
+
int overlap_length = diff_commonOverlap(deletion, insertion);
|
949
|
+
if (overlap_length != 0) {
|
950
|
+
// Overlap found. Insert an equality and trim the surrounding edits.
|
951
|
+
diffs.insert(cur_diff, Diff(EQUAL, insertion.substr(0, overlap_length)));
|
952
|
+
(*prev_diff).text =
|
953
|
+
deletion.substr(0, deletion.length() - overlap_length);
|
954
|
+
(*cur_diff).text = safeMid(insertion, overlap_length);
|
955
|
+
// diffs.insert inserts the element before the cursor, so there is
|
956
|
+
// no need to step past the new element.
|
957
|
+
}
|
958
|
+
if (++cur_diff == diffs.end()) break;
|
959
|
+
}
|
960
|
+
}
|
961
|
+
}
|
962
|
+
}
|
963
|
+
|
964
|
+
/**
|
965
|
+
* Look for single edits surrounded on both sides by equalities
|
966
|
+
* which can be shifted sideways to align the edit to a word boundary.
|
967
|
+
* e.g: The c<ins>at c</ins>ame. -> The <ins>cat </ins>came.
|
968
|
+
* @param diffs LinkedList of Diff objects.
|
969
|
+
*/
|
970
|
+
public:
|
971
|
+
static void diff_cleanupSemanticLossless(Diffs &diffs) {
|
972
|
+
string_t equality1, edit, equality2;
|
973
|
+
string_t commonString;
|
974
|
+
int commonOffset;
|
975
|
+
int score, bestScore;
|
976
|
+
string_t bestEquality1, bestEdit, bestEquality2;
|
977
|
+
// Create a new iterator at the start.
|
978
|
+
typename Diffs::iterator prev_diff = diffs.begin(), cur_diff = prev_diff;
|
979
|
+
if (prev_diff == diffs.end() || ++cur_diff == diffs.end()) return;
|
980
|
+
|
981
|
+
// Intentionally ignore the first and last element (don't need checking).
|
982
|
+
for (typename Diffs::iterator next_diff = cur_diff; ++next_diff != diffs.end(); prev_diff = cur_diff, cur_diff = next_diff) {
|
983
|
+
if ((*prev_diff).operation == EQUAL &&
|
984
|
+
(*next_diff).operation == EQUAL) {
|
985
|
+
// This is a single edit surrounded by equalities.
|
986
|
+
equality1 = (*prev_diff).text;
|
987
|
+
edit = (*cur_diff).text;
|
988
|
+
equality2 = (*next_diff).text;
|
989
|
+
|
990
|
+
// First, shift the edit as far left as possible.
|
991
|
+
commonOffset = diff_commonSuffix(equality1, edit);
|
992
|
+
if (commonOffset != 0) {
|
993
|
+
commonString = safeMid(edit, edit.length() - commonOffset);
|
994
|
+
equality1 = equality1.substr(0, equality1.length() - commonOffset);
|
995
|
+
edit = commonString + edit.substr(0, edit.length() - commonOffset);
|
996
|
+
equality2 = commonString + equality2;
|
997
|
+
}
|
998
|
+
|
999
|
+
// Second, step character by character right, looking for the best fit.
|
1000
|
+
bestEquality1 = equality1;
|
1001
|
+
bestEdit = edit;
|
1002
|
+
bestEquality2 = equality2;
|
1003
|
+
bestScore = diff_cleanupSemanticScore(equality1, edit)
|
1004
|
+
+ diff_cleanupSemanticScore(edit, equality2);
|
1005
|
+
while (!edit.empty() && !equality2.empty()
|
1006
|
+
&& edit[0] == equality2[0]) {
|
1007
|
+
equality1 += edit[0];
|
1008
|
+
edit = safeMid(edit, 1) + equality2[0];
|
1009
|
+
equality2 = safeMid(equality2, 1);
|
1010
|
+
score = diff_cleanupSemanticScore(equality1, edit)
|
1011
|
+
+ diff_cleanupSemanticScore(edit, equality2);
|
1012
|
+
// The >= encourages trailing rather than leading whitespace on edits.
|
1013
|
+
if (score >= bestScore) {
|
1014
|
+
bestScore = score;
|
1015
|
+
bestEquality1 = equality1;
|
1016
|
+
bestEdit = edit;
|
1017
|
+
bestEquality2 = equality2;
|
1018
|
+
}
|
1019
|
+
}
|
1020
|
+
|
1021
|
+
if ((*prev_diff).text != bestEquality1) {
|
1022
|
+
// We have an improvement, save it back to the diff.
|
1023
|
+
if (!bestEquality1.empty()) {
|
1024
|
+
(*prev_diff).text = bestEquality1;
|
1025
|
+
} else {
|
1026
|
+
diffs.erase(prev_diff);
|
1027
|
+
}
|
1028
|
+
(*cur_diff).text = bestEdit;
|
1029
|
+
if (!bestEquality2.empty()) {
|
1030
|
+
(*next_diff).text = bestEquality2;
|
1031
|
+
} else {
|
1032
|
+
diffs.erase(next_diff); // Delete nextDiff.
|
1033
|
+
next_diff = cur_diff;
|
1034
|
+
cur_diff = prev_diff;
|
1035
|
+
}
|
1036
|
+
}
|
1037
|
+
}
|
1038
|
+
}
|
1039
|
+
}
|
1040
|
+
|
1041
|
+
/**
|
1042
|
+
* Given two strings, compute a score representing whether the internal
|
1043
|
+
* boundary falls on logical boundaries.
|
1044
|
+
* Scores range from 5 (best) to 0 (worst).
|
1045
|
+
* @param one First string.
|
1046
|
+
* @param two Second string.
|
1047
|
+
* @return The score.
|
1048
|
+
*/
|
1049
|
+
private:
|
1050
|
+
static int diff_cleanupSemanticScore(const string_t &one, const string_t &two) {
|
1051
|
+
if (one.empty() || two.empty()) {
|
1052
|
+
// Edges are the best.
|
1053
|
+
return 10;
|
1054
|
+
}
|
1055
|
+
|
1056
|
+
// Each port of this function behaves slightly differently due to
|
1057
|
+
// subtle differences in each language's definition of things like
|
1058
|
+
// 'whitespace'. Since this function's purpose is largely cosmetic,
|
1059
|
+
// the choice has been made to use each language's native features
|
1060
|
+
// rather than force total conformity.
|
1061
|
+
int score = 0;
|
1062
|
+
// One point for non-alphanumeric.
|
1063
|
+
if (!traits::is_alnum(one[one.length() - 1]) || !traits::is_alnum(two[0])) {
|
1064
|
+
score++;
|
1065
|
+
// Two points for whitespace.
|
1066
|
+
if (traits::is_space(one[one.length() - 1]) || traits::is_space(two[0])) {
|
1067
|
+
score++;
|
1068
|
+
// Three points for line breaks.
|
1069
|
+
|
1070
|
+
typename string_t::const_pointer p1 = one.c_str() + one.length() - 1, p2 = two.c_str();
|
1071
|
+
if (is_control(*p1) || is_control(*p2)) {
|
1072
|
+
score++;
|
1073
|
+
// Four points for blank lines.
|
1074
|
+
if (traits::to_wchar(*p1) == L'\n' && p1 != one.c_str() && (traits::to_wchar(*(p1 - 1)) == L'\n'
|
1075
|
+
|| traits::to_wchar(*(p1 - 1)) == L'\r' && p1 - 1 != one.c_str() && traits::to_wchar(*(p1 - 2)) == L'\n')) {
|
1076
|
+
score++;
|
1077
|
+
}
|
1078
|
+
else {
|
1079
|
+
p1 = p2 + two.length();
|
1080
|
+
if (traits::to_wchar(*p2) == L'\r') ++p2;
|
1081
|
+
if (p2 != p1 && traits::to_wchar(*p2) == L'\n') {
|
1082
|
+
if (++p2 != p1 && traits::to_wchar(*p2) == L'\r') ++p2;
|
1083
|
+
if (p2 != p1 && traits::to_wchar(*p2) == L'\n') score++;
|
1084
|
+
}
|
1085
|
+
}
|
1086
|
+
}
|
1087
|
+
}
|
1088
|
+
}
|
1089
|
+
return score;
|
1090
|
+
}
|
1091
|
+
/**
|
1092
|
+
* Reduce the number of edits by eliminating operationally trivial equalities.
|
1093
|
+
* @param diffs LinkedList of Diff objects.
|
1094
|
+
*/
|
1095
|
+
public:
|
1096
|
+
void diff_cleanupEfficiency(Diffs &diffs) const {
|
1097
|
+
if (diffs.empty()) {
|
1098
|
+
return;
|
1099
|
+
}
|
1100
|
+
bool changes = false;
|
1101
|
+
std::vector<typename Diffs::iterator> equalities; // Stack of equalities.
|
1102
|
+
string_t lastequality; // Always equal to equalities.lastElement().text
|
1103
|
+
// Is there an insertion operation before the last equality.
|
1104
|
+
bool pre_ins = false;
|
1105
|
+
// Is there a deletion operation before the last equality.
|
1106
|
+
bool pre_del = false;
|
1107
|
+
// Is there an insertion operation after the last equality.
|
1108
|
+
bool post_ins = false;
|
1109
|
+
// Is there a deletion operation after the last equality.
|
1110
|
+
bool post_del = false;
|
1111
|
+
|
1112
|
+
for (typename Diffs::iterator cur_diff = diffs.begin(); cur_diff != diffs.end();) {
|
1113
|
+
if ((*cur_diff).operation == EQUAL) {
|
1114
|
+
// Equality found.
|
1115
|
+
if ((int)(*cur_diff).text.length() < Diff_EditCost && (post_ins || post_del)) {
|
1116
|
+
// Candidate found.
|
1117
|
+
equalities.push_back(cur_diff);
|
1118
|
+
pre_ins = post_ins;
|
1119
|
+
pre_del = post_del;
|
1120
|
+
lastequality = (*cur_diff).text;
|
1121
|
+
} else {
|
1122
|
+
// Not a candidate, and can never become one.
|
1123
|
+
equalities.clear();
|
1124
|
+
lastequality.clear();
|
1125
|
+
}
|
1126
|
+
post_ins = post_del = false;
|
1127
|
+
} else {
|
1128
|
+
// An insertion or deletion.
|
1129
|
+
if ((*cur_diff).operation == DELETE) {
|
1130
|
+
post_del = true;
|
1131
|
+
} else {
|
1132
|
+
post_ins = true;
|
1133
|
+
}
|
1134
|
+
/*
|
1135
|
+
* Five types to be split:
|
1136
|
+
* <ins>A</ins><del>B</del>XY<ins>C</ins><del>D</del>
|
1137
|
+
* <ins>A</ins>X<ins>C</ins><del>D</del>
|
1138
|
+
* <ins>A</ins><del>B</del>X<ins>C</ins>
|
1139
|
+
* <ins>A</del>X<ins>C</ins><del>D</del>
|
1140
|
+
* <ins>A</ins><del>B</del>X<del>C</del>
|
1141
|
+
*/
|
1142
|
+
if (!lastequality.empty()
|
1143
|
+
&& ((pre_ins && pre_del && post_ins && post_del)
|
1144
|
+
|| (((int)lastequality.length() < Diff_EditCost / 2)
|
1145
|
+
&& ((pre_ins ? 1 : 0) + (pre_del ? 1 : 0)
|
1146
|
+
+ (post_ins ? 1 : 0) + (post_del ? 1 : 0)) == 3))) {
|
1147
|
+
// printf("Splitting: '%s'\n", qPrintable(lastequality));
|
1148
|
+
// Walk back to offending equality.
|
1149
|
+
// Change second copy to insert.
|
1150
|
+
(*(cur_diff = equalities.back())).operation = INSERT;
|
1151
|
+
// Duplicate record.
|
1152
|
+
diffs.insert(cur_diff, Diff(DELETE, lastequality));
|
1153
|
+
equalities.pop_back(); // Throw away the equality we just deleted.
|
1154
|
+
lastequality.clear();
|
1155
|
+
changes = true;
|
1156
|
+
if (pre_ins && pre_del) {
|
1157
|
+
// No changes made which could affect previous entry, keep going.
|
1158
|
+
post_ins = post_del = true;
|
1159
|
+
equalities.clear();
|
1160
|
+
} else {
|
1161
|
+
if (!equalities.empty()) {
|
1162
|
+
// Throw away the previous equality (it needs to be reevaluated).
|
1163
|
+
equalities.pop_back();
|
1164
|
+
}
|
1165
|
+
post_ins = post_del = false;
|
1166
|
+
if (!equalities.empty())
|
1167
|
+
// There is a safe equality we can fall back to.
|
1168
|
+
cur_diff = equalities.back();
|
1169
|
+
else
|
1170
|
+
{
|
1171
|
+
// There are no previous equalities, walk back to the start.
|
1172
|
+
cur_diff = diffs.begin();
|
1173
|
+
continue;
|
1174
|
+
}
|
1175
|
+
}
|
1176
|
+
}
|
1177
|
+
}
|
1178
|
+
++cur_diff;
|
1179
|
+
}
|
1180
|
+
|
1181
|
+
if (changes) {
|
1182
|
+
diff_cleanupMerge(diffs);
|
1183
|
+
}
|
1184
|
+
}
|
1185
|
+
|
1186
|
+
/**
|
1187
|
+
* Reorder and merge like edit sections. Merge equalities.
|
1188
|
+
* Any edit section can move as long as it doesn't cross an equality.
|
1189
|
+
* @param diffs LinkedList of Diff objects.
|
1190
|
+
*/
|
1191
|
+
public:
|
1192
|
+
static void diff_cleanupMerge(Diffs &diffs) {
|
1193
|
+
diffs.push_back(Diff(EQUAL, string_t())); // Add a dummy entry at the end.
|
1194
|
+
typename Diffs::iterator prev_diff, cur_diff;
|
1195
|
+
int count_delete = 0;
|
1196
|
+
int count_insert = 0;
|
1197
|
+
string_t text_delete;
|
1198
|
+
string_t text_insert;
|
1199
|
+
Diff *prevEqual = NULL;
|
1200
|
+
int commonlength;
|
1201
|
+
for (cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1202
|
+
switch ((*cur_diff).operation) {
|
1203
|
+
case INSERT:
|
1204
|
+
count_insert++;
|
1205
|
+
text_insert += (*cur_diff).text;
|
1206
|
+
prevEqual = NULL;
|
1207
|
+
break;
|
1208
|
+
case DELETE:
|
1209
|
+
count_delete++;
|
1210
|
+
text_delete += (*cur_diff).text;
|
1211
|
+
prevEqual = NULL;
|
1212
|
+
break;
|
1213
|
+
case EQUAL:
|
1214
|
+
if (count_delete + count_insert > 1) {
|
1215
|
+
// Delete the offending records.
|
1216
|
+
prev_diff = cur_diff;
|
1217
|
+
std::advance(prev_diff, -(count_delete + count_insert));
|
1218
|
+
diffs.erase(prev_diff, cur_diff);
|
1219
|
+
if (count_delete != 0 && count_insert != 0) {
|
1220
|
+
// Factor out any common prefixies.
|
1221
|
+
commonlength = diff_commonPrefix(text_insert, text_delete);
|
1222
|
+
if (commonlength != 0) {
|
1223
|
+
if (cur_diff != diffs.begin()) {
|
1224
|
+
prev_diff = cur_diff;
|
1225
|
+
if ((*--prev_diff).operation != EQUAL) {
|
1226
|
+
throw string_t(traits::cs(L"Previous diff should have been an equality."));
|
1227
|
+
}
|
1228
|
+
(*prev_diff).text += text_insert.substr(0, commonlength);
|
1229
|
+
} else {
|
1230
|
+
diffs.insert(cur_diff, Diff(EQUAL, text_insert.substr(0, commonlength)));
|
1231
|
+
}
|
1232
|
+
text_insert = safeMid(text_insert, commonlength);
|
1233
|
+
text_delete = safeMid(text_delete, commonlength);
|
1234
|
+
}
|
1235
|
+
// Factor out any common suffixies.
|
1236
|
+
commonlength = diff_commonSuffix(text_insert, text_delete);
|
1237
|
+
if (commonlength != 0) {
|
1238
|
+
(*cur_diff).text = safeMid(text_insert, text_insert.length()
|
1239
|
+
- commonlength) + (*cur_diff).text;
|
1240
|
+
text_insert = text_insert.substr(0, text_insert.length()
|
1241
|
+
- commonlength);
|
1242
|
+
text_delete = text_delete.substr(0, text_delete.length()
|
1243
|
+
- commonlength);
|
1244
|
+
}
|
1245
|
+
}
|
1246
|
+
// Insert the merged records.
|
1247
|
+
if (!text_delete.empty()) {
|
1248
|
+
diffs.insert(cur_diff, Diff(DELETE, text_delete));
|
1249
|
+
}
|
1250
|
+
if (!text_insert.empty()) {
|
1251
|
+
diffs.insert(cur_diff, Diff(INSERT, text_insert));
|
1252
|
+
}
|
1253
|
+
} else if (prevEqual != NULL) {
|
1254
|
+
// Merge this equality with the previous one.
|
1255
|
+
prevEqual->text += (*cur_diff).text;
|
1256
|
+
diffs.erase(cur_diff--);
|
1257
|
+
}
|
1258
|
+
|
1259
|
+
count_insert = 0;
|
1260
|
+
count_delete = 0;
|
1261
|
+
text_delete.clear();
|
1262
|
+
text_insert.clear();
|
1263
|
+
prevEqual = &*cur_diff;
|
1264
|
+
break;
|
1265
|
+
}
|
1266
|
+
|
1267
|
+
}
|
1268
|
+
if (diffs.back().text.empty()) {
|
1269
|
+
diffs.pop_back(); // Remove the dummy entry at the end.
|
1270
|
+
}
|
1271
|
+
|
1272
|
+
/*
|
1273
|
+
* Second pass: look for single edits surrounded on both sides by equalities
|
1274
|
+
* which can be shifted sideways to eliminate an equality.
|
1275
|
+
* e.g: A<ins>BA</ins>C -> <ins>AB</ins>AC
|
1276
|
+
*/
|
1277
|
+
bool changes = false;
|
1278
|
+
// Create a new iterator at the start.
|
1279
|
+
// (As opposed to walking the current one back.)
|
1280
|
+
prev_diff = cur_diff = diffs.begin();
|
1281
|
+
if (prev_diff != diffs.end() && ++cur_diff != diffs.end()) {
|
1282
|
+
// Intentionally ignore the first and last element (don't need checking).
|
1283
|
+
for (typename Diffs::iterator next_diff = cur_diff; ++next_diff != diffs.end(); prev_diff = cur_diff, cur_diff = next_diff) {
|
1284
|
+
if ((*prev_diff).operation == EQUAL &&
|
1285
|
+
(*next_diff).operation == EQUAL) {
|
1286
|
+
// This is a single edit surrounded by equalities.
|
1287
|
+
if ((*cur_diff).text.size() >= (*prev_diff).text.size() &&
|
1288
|
+
(*cur_diff).text.compare((*cur_diff).text.size() - (*prev_diff).text.size(), (*prev_diff).text.size(), (*prev_diff).text) == 0) {
|
1289
|
+
// Shift the edit over the previous equality.
|
1290
|
+
(*cur_diff).text = (*prev_diff).text
|
1291
|
+
+ (*cur_diff).text.substr(0, (*cur_diff).text.length()
|
1292
|
+
- (*prev_diff).text.length());
|
1293
|
+
(*next_diff).text = (*prev_diff).text + (*next_diff).text;
|
1294
|
+
diffs.erase(prev_diff);
|
1295
|
+
cur_diff = next_diff;
|
1296
|
+
changes = true;
|
1297
|
+
if (++next_diff == diffs.end()) break;
|
1298
|
+
} else if ((*cur_diff).text.size() >= (*next_diff).text.size() && (*cur_diff).text.compare(0, (*next_diff).text.size(), (*next_diff).text) == 0) {
|
1299
|
+
// Shift the edit over the next equality.
|
1300
|
+
(*prev_diff).text += (*next_diff).text;
|
1301
|
+
(*cur_diff).text = safeMid((*cur_diff).text, (*next_diff).text.length())
|
1302
|
+
+ (*next_diff).text;
|
1303
|
+
next_diff = diffs.erase(next_diff); // Delete nextDiff.
|
1304
|
+
changes = true;
|
1305
|
+
if (next_diff == diffs.end()) break;
|
1306
|
+
}
|
1307
|
+
}
|
1308
|
+
}
|
1309
|
+
}
|
1310
|
+
// If shifts were made, the diff needs reordering and another shift sweep.
|
1311
|
+
if (changes) {
|
1312
|
+
diff_cleanupMerge(diffs);
|
1313
|
+
}
|
1314
|
+
}
|
1315
|
+
|
1316
|
+
/**
|
1317
|
+
* loc is a location in text1, compute and return the equivalent location in
|
1318
|
+
* text2.
|
1319
|
+
* e.g. "The cat" vs "The big cat", 1->1, 5->8
|
1320
|
+
* @param diffs LinkedList of Diff objects.
|
1321
|
+
* @param loc Location within text1.
|
1322
|
+
* @return Location within text2.
|
1323
|
+
*/
|
1324
|
+
public:
|
1325
|
+
static int diff_xIndex(const Diffs &diffs, int loc) {
|
1326
|
+
int chars1 = 0;
|
1327
|
+
int chars2 = 0;
|
1328
|
+
int last_chars1 = 0;
|
1329
|
+
int last_chars2 = 0;
|
1330
|
+
typename Diffs::const_iterator last_diff = diffs.end(), cur_diff;
|
1331
|
+
for (cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1332
|
+
if ((*cur_diff).operation != INSERT) {
|
1333
|
+
// Equality or deletion.
|
1334
|
+
chars1 += (*cur_diff).text.length();
|
1335
|
+
}
|
1336
|
+
if ((*cur_diff).operation != DELETE) {
|
1337
|
+
// Equality or insertion.
|
1338
|
+
chars2 += (*cur_diff).text.length();
|
1339
|
+
}
|
1340
|
+
if (chars1 > loc) {
|
1341
|
+
// Overshot the location.
|
1342
|
+
last_diff = cur_diff;
|
1343
|
+
break;
|
1344
|
+
}
|
1345
|
+
last_chars1 = chars1;
|
1346
|
+
last_chars2 = chars2;
|
1347
|
+
}
|
1348
|
+
if (last_diff != diffs.end() && (*last_diff).operation == DELETE) {
|
1349
|
+
// The location was deleted.
|
1350
|
+
return last_chars2;
|
1351
|
+
}
|
1352
|
+
// Add the remaining character length.
|
1353
|
+
return last_chars2 + (loc - last_chars1);
|
1354
|
+
}
|
1355
|
+
|
1356
|
+
/**
|
1357
|
+
* Convert a Diff list into a pretty HTML report.
|
1358
|
+
* @param diffs LinkedList of Diff objects.
|
1359
|
+
* @return HTML representation.
|
1360
|
+
*/
|
1361
|
+
public:
|
1362
|
+
static string_t diff_prettyHtml(const Diffs &diffs) {
|
1363
|
+
string_t html;
|
1364
|
+
string_t text;
|
1365
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1366
|
+
typename string_t::size_type n = (*cur_diff).text.size();
|
1367
|
+
typename string_t::const_pointer p, end;
|
1368
|
+
for (p = (*cur_diff).text.c_str(), end = p + n; p != end; ++p)
|
1369
|
+
switch (traits::to_wchar(*p)) {
|
1370
|
+
case L'&': n += 4; break;
|
1371
|
+
case L'<':
|
1372
|
+
case L'>': n += 3; break;
|
1373
|
+
case L'\n': n += 9; break;
|
1374
|
+
}
|
1375
|
+
if (n == (*cur_diff).text.size())
|
1376
|
+
text = (*cur_diff).text;
|
1377
|
+
else {
|
1378
|
+
text.clear();
|
1379
|
+
text.reserve(n);
|
1380
|
+
for (p = (*cur_diff).text.c_str(); p != end; ++p)
|
1381
|
+
switch (traits::to_wchar(*p)) {
|
1382
|
+
case L'&': text += traits::cs(L"&"); break;
|
1383
|
+
case L'<': text += traits::cs(L"<"); break;
|
1384
|
+
case L'>': text += traits::cs(L">"); break;
|
1385
|
+
case L'\n': text += traits::cs(L"¶<br>"); break;
|
1386
|
+
default: text += *p;
|
1387
|
+
}
|
1388
|
+
}
|
1389
|
+
switch ((*cur_diff).operation) {
|
1390
|
+
case INSERT:
|
1391
|
+
html += traits::cs(L"<ins style=\"background:#e6ffe6;\">") + text + traits::cs(L"</ins>");
|
1392
|
+
break;
|
1393
|
+
case DELETE:
|
1394
|
+
html += traits::cs(L"<del style=\"background:#ffe6e6;\">") + text + traits::cs(L"</del>");
|
1395
|
+
break;
|
1396
|
+
case EQUAL:
|
1397
|
+
html += traits::cs(L"<span>") + text + traits::cs(L"</span>");
|
1398
|
+
break;
|
1399
|
+
}
|
1400
|
+
}
|
1401
|
+
return html;
|
1402
|
+
}
|
1403
|
+
|
1404
|
+
/**
|
1405
|
+
* Compute and return the source text (all equalities and deletions).
|
1406
|
+
* @param diffs LinkedList of Diff objects.
|
1407
|
+
* @return Source text.
|
1408
|
+
*/
|
1409
|
+
public:
|
1410
|
+
static string_t diff_text1(const Diffs &diffs) {
|
1411
|
+
string_t text;
|
1412
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1413
|
+
if ((*cur_diff).operation != INSERT) {
|
1414
|
+
text += (*cur_diff).text;
|
1415
|
+
}
|
1416
|
+
}
|
1417
|
+
return text;
|
1418
|
+
}
|
1419
|
+
|
1420
|
+
/**
|
1421
|
+
* Compute and return the destination text (all equalities and insertions).
|
1422
|
+
* @param diffs LinkedList of Diff objects.
|
1423
|
+
* @return Destination text.
|
1424
|
+
*/
|
1425
|
+
public:
|
1426
|
+
static string_t diff_text2(const Diffs &diffs) {
|
1427
|
+
string_t text;
|
1428
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1429
|
+
if ((*cur_diff).operation != DELETE) {
|
1430
|
+
text += (*cur_diff).text;
|
1431
|
+
}
|
1432
|
+
}
|
1433
|
+
return text;
|
1434
|
+
}
|
1435
|
+
|
1436
|
+
/**
|
1437
|
+
* Compute the Levenshtein distance; the number of inserted, deleted or
|
1438
|
+
* substituted characters.
|
1439
|
+
* @param diffs LinkedList of Diff objects.
|
1440
|
+
* @return Number of changes.
|
1441
|
+
*/
|
1442
|
+
public:
|
1443
|
+
static int diff_levenshtein(const Diffs &diffs) {
|
1444
|
+
int levenshtein = 0;
|
1445
|
+
int insertions = 0;
|
1446
|
+
int deletions = 0;
|
1447
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1448
|
+
switch ((*cur_diff).operation) {
|
1449
|
+
case INSERT:
|
1450
|
+
insertions += (*cur_diff).text.length();
|
1451
|
+
break;
|
1452
|
+
case DELETE:
|
1453
|
+
deletions += (*cur_diff).text.length();
|
1454
|
+
break;
|
1455
|
+
case EQUAL:
|
1456
|
+
// A deletion and an insertion is one substitution.
|
1457
|
+
levenshtein += std::max(insertions, deletions);
|
1458
|
+
insertions = 0;
|
1459
|
+
deletions = 0;
|
1460
|
+
break;
|
1461
|
+
}
|
1462
|
+
}
|
1463
|
+
levenshtein += std::max(insertions, deletions);
|
1464
|
+
return levenshtein;
|
1465
|
+
}
|
1466
|
+
|
1467
|
+
/**
|
1468
|
+
* Crush the diff into an encoded string which describes the operations
|
1469
|
+
* required to transform text1 into text2.
|
1470
|
+
* E.g. =3\t-2\t+ing -> Keep 3 chars, delete 2 chars, insert 'ing'.
|
1471
|
+
* Operations are tab-separated. Inserted text is escaped using %xx notation.
|
1472
|
+
* @param diffs Array of diff tuples.
|
1473
|
+
* @return Delta text.
|
1474
|
+
*/
|
1475
|
+
public:
|
1476
|
+
static string_t diff_toDelta(const Diffs &diffs) {
|
1477
|
+
string_t text;
|
1478
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1479
|
+
switch ((*cur_diff).operation) {
|
1480
|
+
case INSERT: {
|
1481
|
+
text += traits::from_wchar(L'+');
|
1482
|
+
append_percent_encoded(text, (*cur_diff).text);
|
1483
|
+
text += traits::from_wchar(L'\t');
|
1484
|
+
break;
|
1485
|
+
}
|
1486
|
+
case DELETE:
|
1487
|
+
text += traits::from_wchar(L'-') + to_string((*cur_diff).text.length()) + traits::from_wchar(L'\t');
|
1488
|
+
break;
|
1489
|
+
case EQUAL:
|
1490
|
+
text += traits::from_wchar(L'=') + to_string((*cur_diff).text.length()) + traits::from_wchar(L'\t');
|
1491
|
+
break;
|
1492
|
+
}
|
1493
|
+
}
|
1494
|
+
if (!text.empty()) {
|
1495
|
+
// Strip off trailing tab character.
|
1496
|
+
text = text.substr(0, text.length() - 1);
|
1497
|
+
}
|
1498
|
+
return text;
|
1499
|
+
}
|
1500
|
+
|
1501
|
+
/**
|
1502
|
+
* Given the original text1, and an encoded string which describes the
|
1503
|
+
* operations required to transform text1 into text2, compute the full diff.
|
1504
|
+
* @param text1 Source string for the diff.
|
1505
|
+
* @param delta Delta text.
|
1506
|
+
* @return Array of diff tuples or null if invalid.
|
1507
|
+
* @throws string_t If invalid input.
|
1508
|
+
*/
|
1509
|
+
public:
|
1510
|
+
static Diffs diff_fromDelta(const string_t &text1, const string_t &delta) {
|
1511
|
+
Diffs diffs;
|
1512
|
+
int pointer = 0; // Cursor in text1
|
1513
|
+
typename string_t::size_type token_len;
|
1514
|
+
for (typename string_t::const_pointer token = delta.c_str(); token - delta.c_str() < (int)delta.length(); token += token_len + 1) {
|
1515
|
+
token_len = next_token(delta, traits::from_wchar(L'\t'), token);
|
1516
|
+
if (token_len == 0) {
|
1517
|
+
// Blank tokens are ok (from a trailing \t).
|
1518
|
+
continue;
|
1519
|
+
}
|
1520
|
+
// Each token begins with a one character parameter which specifies the
|
1521
|
+
// operation of this token (delete, insert, equality).
|
1522
|
+
string_t param(token + 1, token_len - 1);
|
1523
|
+
switch (traits::to_wchar(*token)) {
|
1524
|
+
case L'+':
|
1525
|
+
percent_decode(param);
|
1526
|
+
diffs.push_back(Diff(INSERT, param));
|
1527
|
+
break;
|
1528
|
+
case L'-':
|
1529
|
+
// Fall through.
|
1530
|
+
case L'=': {
|
1531
|
+
int n;
|
1532
|
+
n = to_int(param);
|
1533
|
+
if (n < 0) {
|
1534
|
+
throw string_t(traits::cs(L"Negative number in diff_fromDelta: ") + param);
|
1535
|
+
}
|
1536
|
+
string_t text;
|
1537
|
+
text = safeMid(text1, pointer, n);
|
1538
|
+
pointer += n;
|
1539
|
+
if (traits::to_wchar(*token) == L'=') {
|
1540
|
+
diffs.push_back(Diff(EQUAL, text));
|
1541
|
+
} else {
|
1542
|
+
diffs.push_back(Diff(DELETE, text));
|
1543
|
+
}
|
1544
|
+
break;
|
1545
|
+
}
|
1546
|
+
default:
|
1547
|
+
throw string_t(string_t(traits::cs(L"Invalid diff operation in diff_fromDelta: ")) + *token);
|
1548
|
+
}
|
1549
|
+
}
|
1550
|
+
if (pointer != text1.length()) {
|
1551
|
+
throw string_t(traits::cs(L"Delta length (") + to_string(pointer)
|
1552
|
+
+ traits::cs(L") smaller than source text length (")
|
1553
|
+
+ to_string(text1.length()) + traits::from_wchar(L')'));
|
1554
|
+
}
|
1555
|
+
return diffs;
|
1556
|
+
}
|
1557
|
+
|
1558
|
+
|
1559
|
+
// MATCH FUNCTIONS
|
1560
|
+
|
1561
|
+
|
1562
|
+
/**
|
1563
|
+
* Locate the best instance of 'pattern' in 'text' near 'loc'.
|
1564
|
+
* Returns -1 if no match found.
|
1565
|
+
* @param text The text to search.
|
1566
|
+
* @param pattern The pattern to search for.
|
1567
|
+
* @param loc The location to search around.
|
1568
|
+
* @return Best match index or -1.
|
1569
|
+
*/
|
1570
|
+
public:
|
1571
|
+
int match_main(const string_t &text, const string_t &pattern, int loc) const {
|
1572
|
+
loc = std::max(0, std::min(loc, (int)text.length()));
|
1573
|
+
if (text == pattern) {
|
1574
|
+
// Shortcut (potentially not guaranteed by the algorithm)
|
1575
|
+
return 0;
|
1576
|
+
} else if (text.empty()) {
|
1577
|
+
// Nothing to match.
|
1578
|
+
return -1;
|
1579
|
+
} else if (loc + pattern.length() <= text.length()
|
1580
|
+
&& safeMid(text, loc, pattern.length()) == pattern) {
|
1581
|
+
// Perfect match at the perfect spot! (Includes case of null pattern)
|
1582
|
+
return loc;
|
1583
|
+
} else {
|
1584
|
+
// Do a fuzzy compare.
|
1585
|
+
return match_bitap(text, pattern, loc);
|
1586
|
+
}
|
1587
|
+
}
|
1588
|
+
|
1589
|
+
/**
|
1590
|
+
* Locate the best instance of 'pattern' in 'text' near 'loc' using the
|
1591
|
+
* Bitap algorithm. Returns -1 if no match found.
|
1592
|
+
* @param text The text to search.
|
1593
|
+
* @param pattern The pattern to search for.
|
1594
|
+
* @param loc The location to search around.
|
1595
|
+
* @return Best match index or -1.
|
1596
|
+
*/
|
1597
|
+
protected:
|
1598
|
+
int match_bitap(const string_t &text, const string_t &pattern, int loc) const {
|
1599
|
+
if (!(Match_MaxBits == 0 || (int)pattern.length() <= Match_MaxBits)) {
|
1600
|
+
throw string_t(traits::cs(L"Pattern too long for this application."));
|
1601
|
+
}
|
1602
|
+
|
1603
|
+
// Initialise the alphabet.
|
1604
|
+
std::map<char_t, int> s;
|
1605
|
+
match_alphabet(pattern, s);
|
1606
|
+
|
1607
|
+
// Highest score beyond which we give up.
|
1608
|
+
double score_threshold = Match_Threshold;
|
1609
|
+
// Is there a nearby exact match? (speedup)
|
1610
|
+
size_t best_loc = text.find(pattern, loc);
|
1611
|
+
if (best_loc != string_t::npos) {
|
1612
|
+
score_threshold = std::min(match_bitapScore(0, best_loc, loc, pattern),
|
1613
|
+
score_threshold);
|
1614
|
+
// What about in the other direction? (speedup)
|
1615
|
+
best_loc = text.rfind(pattern, loc + pattern.length());
|
1616
|
+
if (best_loc != string_t::npos) {
|
1617
|
+
score_threshold = std::min(match_bitapScore(0, best_loc, loc, pattern),
|
1618
|
+
score_threshold);
|
1619
|
+
}
|
1620
|
+
}
|
1621
|
+
|
1622
|
+
// Initialise the bit arrays.
|
1623
|
+
int matchmask = 1 << (pattern.length() - 1);
|
1624
|
+
best_loc = -1;
|
1625
|
+
|
1626
|
+
int bin_min, bin_mid;
|
1627
|
+
int bin_max = pattern.length() + text.length();
|
1628
|
+
int *rd;
|
1629
|
+
int *last_rd = NULL;
|
1630
|
+
for (int d = 0; d < (int)pattern.length(); d++) {
|
1631
|
+
// Scan for the best match; each iteration allows for one more error.
|
1632
|
+
// Run a binary search to determine how far from 'loc' we can stray at
|
1633
|
+
// this error level.
|
1634
|
+
bin_min = 0;
|
1635
|
+
bin_mid = bin_max;
|
1636
|
+
while (bin_min < bin_mid) {
|
1637
|
+
if (match_bitapScore(d, loc + bin_mid, loc, pattern)
|
1638
|
+
<= score_threshold) {
|
1639
|
+
bin_min = bin_mid;
|
1640
|
+
} else {
|
1641
|
+
bin_max = bin_mid;
|
1642
|
+
}
|
1643
|
+
bin_mid = (bin_max - bin_min) / 2 + bin_min;
|
1644
|
+
}
|
1645
|
+
// Use the result from this iteration as the maximum for the next.
|
1646
|
+
bin_max = bin_mid;
|
1647
|
+
int start = std::max(1, loc - bin_mid + 1);
|
1648
|
+
int finish = std::min(loc + bin_mid, (int)text.length()) + pattern.length();
|
1649
|
+
|
1650
|
+
rd = new int[finish + 2];
|
1651
|
+
rd[finish + 1] = (1 << d) - 1;
|
1652
|
+
for (int j = finish; j >= start; j--) {
|
1653
|
+
int charMatch;
|
1654
|
+
if ((int)text.length() <= j - 1) {
|
1655
|
+
// Out of range.
|
1656
|
+
charMatch = 0;
|
1657
|
+
} else {
|
1658
|
+
charMatch = s[text[j - 1]];
|
1659
|
+
}
|
1660
|
+
if (d == 0) {
|
1661
|
+
// First pass: exact match.
|
1662
|
+
rd[j] = ((rd[j + 1] << 1) | 1) & charMatch;
|
1663
|
+
} else {
|
1664
|
+
// Subsequent passes: fuzzy match.
|
1665
|
+
rd[j] = ((rd[j + 1] << 1) | 1) & charMatch
|
1666
|
+
| (((last_rd[j + 1] | last_rd[j]) << 1) | 1)
|
1667
|
+
| last_rd[j + 1];
|
1668
|
+
}
|
1669
|
+
if ((rd[j] & matchmask) != 0) {
|
1670
|
+
double score = match_bitapScore(d, j - 1, loc, pattern);
|
1671
|
+
// This match will almost certainly be better than any existing
|
1672
|
+
// match. But check anyway.
|
1673
|
+
if (score <= score_threshold) {
|
1674
|
+
// Told you so.
|
1675
|
+
score_threshold = score;
|
1676
|
+
best_loc = j - 1;
|
1677
|
+
if ((int)best_loc > loc) {
|
1678
|
+
// When passing loc, don't exceed our current distance from loc.
|
1679
|
+
start = std::max(1, 2 * loc - (int)best_loc);
|
1680
|
+
} else {
|
1681
|
+
// Already passed loc, downhill from here on in.
|
1682
|
+
break;
|
1683
|
+
}
|
1684
|
+
}
|
1685
|
+
}
|
1686
|
+
}
|
1687
|
+
if (match_bitapScore(d + 1, loc, loc, pattern) > score_threshold) {
|
1688
|
+
// No hope for a (better) match at greater error levels.
|
1689
|
+
break;
|
1690
|
+
}
|
1691
|
+
delete [] last_rd;
|
1692
|
+
last_rd = rd;
|
1693
|
+
}
|
1694
|
+
delete [] last_rd;
|
1695
|
+
delete [] rd;
|
1696
|
+
return best_loc;
|
1697
|
+
}
|
1698
|
+
|
1699
|
+
/**
|
1700
|
+
* Compute and return the score for a match with e errors and x location.
|
1701
|
+
* @param e Number of errors in match.
|
1702
|
+
* @param x Location of match.
|
1703
|
+
* @param loc Expected location of match.
|
1704
|
+
* @param pattern Pattern being sought.
|
1705
|
+
* @return Overall score for match (0.0 = good, 1.0 = bad).
|
1706
|
+
*/
|
1707
|
+
private:
|
1708
|
+
double match_bitapScore(int e, int x, int loc, const string_t &pattern) const {
|
1709
|
+
const float accuracy = static_cast<float> (e) / pattern.length();
|
1710
|
+
const int proximity = (loc - x < 0)? (x - loc) : (loc - x);
|
1711
|
+
if (Match_Distance == 0) {
|
1712
|
+
// Dodge divide by zero error.
|
1713
|
+
return proximity == 0 ? accuracy : 1.0;
|
1714
|
+
}
|
1715
|
+
return accuracy + (proximity / static_cast<float> (Match_Distance));
|
1716
|
+
}
|
1717
|
+
|
1718
|
+
/**
|
1719
|
+
* Initialise the alphabet for the Bitap algorithm.
|
1720
|
+
* @param pattern The text to encode.
|
1721
|
+
* @return Hash of character locations.
|
1722
|
+
*/
|
1723
|
+
protected:
|
1724
|
+
static void match_alphabet(const string_t &pattern, std::map<char_t, int>& s) {
|
1725
|
+
// There is no need to initialize map values, since they are zero-initialized by default
|
1726
|
+
for (size_t i = 0; i < pattern.length(); i++)
|
1727
|
+
s[pattern[i]] |= (1 << (pattern.length() - i - 1));
|
1728
|
+
}
|
1729
|
+
|
1730
|
+
|
1731
|
+
// PATCH FUNCTIONS
|
1732
|
+
|
1733
|
+
|
1734
|
+
/**
|
1735
|
+
* Increase the context until it is unique,
|
1736
|
+
* but don't let the pattern expand beyond Match_MaxBits.
|
1737
|
+
* @param patch The patch to grow.
|
1738
|
+
* @param text Source text.
|
1739
|
+
*/
|
1740
|
+
protected:
|
1741
|
+
void patch_addContext(Patch &patch, const string_t &text) const {
|
1742
|
+
if (text.empty()) {
|
1743
|
+
return;
|
1744
|
+
}
|
1745
|
+
string_t pattern = safeMid(text, patch.start2, patch.length1);
|
1746
|
+
int padding = 0;
|
1747
|
+
|
1748
|
+
// Look for the first and last matches of pattern in text. If two different
|
1749
|
+
// matches are found, increase the pattern length.
|
1750
|
+
while (text.find(pattern) != text.rfind(pattern)
|
1751
|
+
&& (int)pattern.length() < Match_MaxBits - Patch_Margin - Patch_Margin) {
|
1752
|
+
padding += Patch_Margin;
|
1753
|
+
pattern = safeMid(text, std::max(0, patch.start2 - padding),
|
1754
|
+
std::min((int)text.length(), patch.start2 + patch.length1 + padding)
|
1755
|
+
- std::max(0, patch.start2 - padding));
|
1756
|
+
}
|
1757
|
+
// Add one chunk for good luck.
|
1758
|
+
padding += Patch_Margin;
|
1759
|
+
|
1760
|
+
// Add the prefix.
|
1761
|
+
string_t prefix = safeMid(text, std::max(0, patch.start2 - padding),
|
1762
|
+
patch.start2 - std::max(0, patch.start2 - padding));
|
1763
|
+
if (!prefix.empty()) {
|
1764
|
+
patch.diffs.push_front(Diff(EQUAL, prefix));
|
1765
|
+
}
|
1766
|
+
// Add the suffix.
|
1767
|
+
string_t suffix = safeMid(text, patch.start2 + patch.length1,
|
1768
|
+
std::min((int)text.length(), patch.start2 + patch.length1 + padding)
|
1769
|
+
- (patch.start2 + patch.length1));
|
1770
|
+
if (!suffix.empty()) {
|
1771
|
+
patch.diffs.push_back(Diff(EQUAL, suffix));
|
1772
|
+
}
|
1773
|
+
|
1774
|
+
// Roll back the start points.
|
1775
|
+
patch.start1 -= prefix.length();
|
1776
|
+
patch.start2 -= prefix.length();
|
1777
|
+
// Extend the lengths.
|
1778
|
+
patch.length1 += prefix.length() + suffix.length();
|
1779
|
+
patch.length2 += prefix.length() + suffix.length();
|
1780
|
+
}
|
1781
|
+
|
1782
|
+
/**
|
1783
|
+
* Compute a list of patches to turn text1 into text2.
|
1784
|
+
* A set of diffs will be computed.
|
1785
|
+
* @param text1 Old text.
|
1786
|
+
* @param text2 New text.
|
1787
|
+
* @return LinkedList of Patch objects.
|
1788
|
+
*/
|
1789
|
+
public:
|
1790
|
+
Patches patch_make(const string_t &text1, const string_t &text2) const {
|
1791
|
+
// No diffs provided, compute our own.
|
1792
|
+
Diffs diffs = diff_main(text1, text2, true);
|
1793
|
+
if (diffs.size() > 2) {
|
1794
|
+
diff_cleanupSemantic(diffs);
|
1795
|
+
diff_cleanupEfficiency(diffs);
|
1796
|
+
}
|
1797
|
+
|
1798
|
+
return patch_make(text1, diffs);
|
1799
|
+
}
|
1800
|
+
|
1801
|
+
/**
|
1802
|
+
* Compute a list of patches to turn text1 into text2.
|
1803
|
+
* text1 will be derived from the provided diffs.
|
1804
|
+
* @param diffs Array of diff tuples for text1 to text2.
|
1805
|
+
* @return LinkedList of Patch objects.
|
1806
|
+
*/
|
1807
|
+
public:
|
1808
|
+
Patches patch_make(const Diffs &diffs) const {
|
1809
|
+
// No origin string provided, compute our own.
|
1810
|
+
return patch_make(diff_text1(diffs), diffs);
|
1811
|
+
}
|
1812
|
+
|
1813
|
+
/**
|
1814
|
+
* Compute a list of patches to turn text1 into text2.
|
1815
|
+
* text2 is ignored, diffs are the delta between text1 and text2.
|
1816
|
+
* @param text1 Old text.
|
1817
|
+
* @param text2 Ignored.
|
1818
|
+
* @param diffs Array of diff tuples for text1 to text2.
|
1819
|
+
* @return LinkedList of Patch objects.
|
1820
|
+
* @deprecated Prefer patch_make(const string_t &text1, const Diffs &diffs).
|
1821
|
+
*/
|
1822
|
+
public:
|
1823
|
+
Patches patch_make(const string_t &text1, const string_t &/*text2*/, const Diffs &diffs) const {
|
1824
|
+
return patch_make(text1, diffs); // text2 is entirely unused.
|
1825
|
+
}
|
1826
|
+
|
1827
|
+
/**
|
1828
|
+
* Compute a list of patches to turn text1 into text2.
|
1829
|
+
* text2 is not provided, diffs are the delta between text1 and text2.
|
1830
|
+
* @param text1 Old text.
|
1831
|
+
* @param diffs Array of diff tuples for text1 to text2.
|
1832
|
+
* @return LinkedList of Patch objects.
|
1833
|
+
*/
|
1834
|
+
public:
|
1835
|
+
Patches patch_make(const string_t &text1, const Diffs &diffs) const {
|
1836
|
+
Patches patches;
|
1837
|
+
if (!diffs.empty()) { // Get rid of the null case.
|
1838
|
+
Patch patch;
|
1839
|
+
int char_count1 = 0; // Number of characters into the text1 string.
|
1840
|
+
int char_count2 = 0; // Number of characters into the text2 string.
|
1841
|
+
// Start with text1 (prepatch_text) and apply the diffs until we arrive at
|
1842
|
+
// text2 (postpatch_text). We recreate the patches one by one to determine
|
1843
|
+
// context info.
|
1844
|
+
string_t prepatch_text = text1;
|
1845
|
+
string_t postpatch_text = text1;
|
1846
|
+
for (typename Diffs::const_iterator cur_diff = diffs.begin(); cur_diff != diffs.end(); ++cur_diff) {
|
1847
|
+
if (patch.diffs.empty() && (*cur_diff).operation != EQUAL) {
|
1848
|
+
// A new patch starts here.
|
1849
|
+
patch.start1 = char_count1;
|
1850
|
+
patch.start2 = char_count2;
|
1851
|
+
}
|
1852
|
+
|
1853
|
+
switch ((*cur_diff).operation) {
|
1854
|
+
case INSERT:
|
1855
|
+
patch.diffs.push_back(*cur_diff);
|
1856
|
+
patch.length2 += (*cur_diff).text.length();
|
1857
|
+
postpatch_text = postpatch_text.substr(0, char_count2)
|
1858
|
+
+ (*cur_diff).text + safeMid(postpatch_text, char_count2);
|
1859
|
+
break;
|
1860
|
+
case DELETE:
|
1861
|
+
patch.length1 += (*cur_diff).text.length();
|
1862
|
+
patch.diffs.push_back(*cur_diff);
|
1863
|
+
postpatch_text = postpatch_text.substr(0, char_count2)
|
1864
|
+
+ safeMid(postpatch_text, char_count2 + (*cur_diff).text.length());
|
1865
|
+
break;
|
1866
|
+
case EQUAL:
|
1867
|
+
if ((int)(*cur_diff).text.length() <= 2 * Patch_Margin
|
1868
|
+
&& !patch.diffs.empty() && !(*cur_diff == diffs.back())) {
|
1869
|
+
// Small equality inside a patch.
|
1870
|
+
patch.diffs.push_back(*cur_diff);
|
1871
|
+
patch.length1 += (*cur_diff).text.length();
|
1872
|
+
patch.length2 += (*cur_diff).text.length();
|
1873
|
+
}
|
1874
|
+
|
1875
|
+
if ((int)(*cur_diff).text.length() >= 2 * Patch_Margin) {
|
1876
|
+
// Time for a new patch.
|
1877
|
+
if (!patch.diffs.empty()) {
|
1878
|
+
patch_addContext(patch, prepatch_text);
|
1879
|
+
patches.push_back(patch);
|
1880
|
+
patch = Patch();
|
1881
|
+
// Unlike Unidiff, our patch lists have a rolling context.
|
1882
|
+
// http://code.google.com/p/google-diff-match-patch/wiki/Unidiff
|
1883
|
+
// Update prepatch text & pos to reflect the application of the
|
1884
|
+
// just completed patch.
|
1885
|
+
prepatch_text = postpatch_text;
|
1886
|
+
char_count1 = char_count2;
|
1887
|
+
}
|
1888
|
+
}
|
1889
|
+
break;
|
1890
|
+
}
|
1891
|
+
|
1892
|
+
// Update the current character count.
|
1893
|
+
if ((*cur_diff).operation != INSERT) {
|
1894
|
+
char_count1 += (*cur_diff).text.length();
|
1895
|
+
}
|
1896
|
+
if ((*cur_diff).operation != DELETE) {
|
1897
|
+
char_count2 += (*cur_diff).text.length();
|
1898
|
+
}
|
1899
|
+
}
|
1900
|
+
// Pick up the leftover patch if not empty.
|
1901
|
+
if (!patch.diffs.empty()) {
|
1902
|
+
patch_addContext(patch, prepatch_text);
|
1903
|
+
patches.push_back(patch);
|
1904
|
+
}
|
1905
|
+
}
|
1906
|
+
return patches;
|
1907
|
+
}
|
1908
|
+
|
1909
|
+
/**
|
1910
|
+
* Given an array of patches, return another array that is identical.
|
1911
|
+
* @param patches Array of patch objects.
|
1912
|
+
* @return Array of patch objects.
|
1913
|
+
*/
|
1914
|
+
public:
|
1915
|
+
Patches patch_deepCopy(const Patches &patches) const { return patches; }
|
1916
|
+
|
1917
|
+
/**
|
1918
|
+
* Merge a set of patches onto the text. Return a patched text, as well
|
1919
|
+
* as an array of true/false values indicating which patches were applied.
|
1920
|
+
* @param patches Array of patch objects.
|
1921
|
+
* @param text Old text.
|
1922
|
+
* @return Two element Object array, containing the new text and an array of
|
1923
|
+
* boolean values.
|
1924
|
+
*/
|
1925
|
+
public:
|
1926
|
+
std::pair<string_t, std::vector<bool> > patch_apply(const Patches &patches, const string_t &text) const
|
1927
|
+
{ std::pair<string_t, std::vector<bool> > res; patch_apply(patches, text, res); return res; }
|
1928
|
+
void patch_apply(const Patches &patches, const string_t &sourceText, std::pair<string_t, std::vector<bool> >& res) const {
|
1929
|
+
if (patches.empty()) {
|
1930
|
+
res.first = sourceText;
|
1931
|
+
res.second.clear();
|
1932
|
+
return;
|
1933
|
+
}
|
1934
|
+
string_t text = sourceText; // Copy to preserve original.
|
1935
|
+
|
1936
|
+
// Deep copy the patches so that no changes are made to originals.
|
1937
|
+
// Patches patchesCopy = patch_deepCopy(patches);
|
1938
|
+
Patches patchesCopy(patches); // Default copy constructor will do it just fine
|
1939
|
+
|
1940
|
+
string_t nullPadding = patch_addPadding(patchesCopy);
|
1941
|
+
text = nullPadding + text + nullPadding;
|
1942
|
+
patch_splitMax(patchesCopy);
|
1943
|
+
|
1944
|
+
int x = 0;
|
1945
|
+
// delta keeps track of the offset between the expected and actual location
|
1946
|
+
// of the previous patch. If there are patches expected at positions 10 and
|
1947
|
+
// 20, but the first patch was found at 12, delta is 2 and the second patch
|
1948
|
+
// has an effective expected position of 22.
|
1949
|
+
int delta = 0;
|
1950
|
+
std::vector<bool>& results = res.second;
|
1951
|
+
results.resize(patchesCopy.size());
|
1952
|
+
string_t text1, text2;
|
1953
|
+
for (typename Patches::const_iterator cur_patch = patchesCopy.begin(); cur_patch != patchesCopy.end(); ++cur_patch) {
|
1954
|
+
int expected_loc = (*cur_patch).start2 + delta;
|
1955
|
+
text1 = diff_text1((*cur_patch).diffs);
|
1956
|
+
int start_loc;
|
1957
|
+
int end_loc = -1;
|
1958
|
+
if ((int)text1.length() > Match_MaxBits) {
|
1959
|
+
// patch_splitMax will only provide an oversized pattern in the case of
|
1960
|
+
// a monster delete.
|
1961
|
+
start_loc = match_main(text, text1.substr(0, Match_MaxBits), expected_loc);
|
1962
|
+
if (start_loc != -1) {
|
1963
|
+
end_loc = match_main(text, right(text1, Match_MaxBits),
|
1964
|
+
expected_loc + text1.length() - Match_MaxBits);
|
1965
|
+
if (end_loc == -1 || start_loc >= end_loc) {
|
1966
|
+
// Can't find valid trailing context. Drop this patch.
|
1967
|
+
start_loc = -1;
|
1968
|
+
}
|
1969
|
+
}
|
1970
|
+
} else {
|
1971
|
+
start_loc = match_main(text, text1, expected_loc);
|
1972
|
+
}
|
1973
|
+
if (start_loc == -1) {
|
1974
|
+
// No match found. :(
|
1975
|
+
results[x] = false;
|
1976
|
+
// Subtract the delta for this failed patch from subsequent patches.
|
1977
|
+
delta -= (*cur_patch).length2 - (*cur_patch).length1;
|
1978
|
+
} else {
|
1979
|
+
// Found a match. :)
|
1980
|
+
results[x] = true;
|
1981
|
+
delta = start_loc - expected_loc;
|
1982
|
+
if (end_loc == -1) {
|
1983
|
+
text2 = safeMid(text, start_loc, text1.length());
|
1984
|
+
} else {
|
1985
|
+
text2 = safeMid(text, start_loc, end_loc + Match_MaxBits - start_loc);
|
1986
|
+
}
|
1987
|
+
if (text1 == text2) {
|
1988
|
+
// Perfect match, just shove the replacement text in.
|
1989
|
+
text = text.substr(0, start_loc) + diff_text2((*cur_patch).diffs) + safeMid(text, start_loc + text1.length());
|
1990
|
+
} else {
|
1991
|
+
// Imperfect match. Run a diff to get a framework of equivalent
|
1992
|
+
// indices.
|
1993
|
+
Diffs diffs = diff_main(text1, text2, false);
|
1994
|
+
if ((int)text1.length() > Match_MaxBits
|
1995
|
+
&& diff_levenshtein(diffs) / static_cast<float> (text1.length())
|
1996
|
+
> Patch_DeleteThreshold) {
|
1997
|
+
// The end points match, but the content is unacceptably bad.
|
1998
|
+
results[x] = false;
|
1999
|
+
} else {
|
2000
|
+
diff_cleanupSemanticLossless(diffs);
|
2001
|
+
int index1 = 0;
|
2002
|
+
for (typename Diffs::const_iterator cur_diff = (*cur_patch).diffs.begin(); cur_diff != (*cur_patch).diffs.end(); ++cur_diff) {
|
2003
|
+
if ((*cur_diff).operation != EQUAL) {
|
2004
|
+
int index2 = diff_xIndex(diffs, index1);
|
2005
|
+
if ((*cur_diff).operation == INSERT) {
|
2006
|
+
// Insertion
|
2007
|
+
text = text.substr(0, start_loc + index2) + (*cur_diff).text
|
2008
|
+
+ safeMid(text, start_loc + index2);
|
2009
|
+
} else if ((*cur_diff).operation == DELETE) {
|
2010
|
+
// Deletion
|
2011
|
+
text = text.substr(0, start_loc + index2)
|
2012
|
+
+ safeMid(text, start_loc + diff_xIndex(diffs,
|
2013
|
+
index1 + (*cur_diff).text.length()));
|
2014
|
+
}
|
2015
|
+
}
|
2016
|
+
if ((*cur_diff).operation != DELETE) {
|
2017
|
+
index1 += (*cur_diff).text.length();
|
2018
|
+
}
|
2019
|
+
}
|
2020
|
+
}
|
2021
|
+
}
|
2022
|
+
}
|
2023
|
+
x++;
|
2024
|
+
}
|
2025
|
+
// Strip the padding off.
|
2026
|
+
res.first = safeMid(text, nullPadding.length(), text.length() - 2 * nullPadding.length());
|
2027
|
+
}
|
2028
|
+
|
2029
|
+
/**
|
2030
|
+
* Add some padding on text start and end so that edges can match something.
|
2031
|
+
* Intended to be called only from within patch_apply.
|
2032
|
+
* @param patches Array of patch objects.
|
2033
|
+
* @return The padding string added to each side.
|
2034
|
+
*/
|
2035
|
+
public:
|
2036
|
+
string_t patch_addPadding(Patches &patches) const {
|
2037
|
+
short paddingLength = Patch_Margin;
|
2038
|
+
string_t nullPadding;
|
2039
|
+
for (short x = 1; x <= paddingLength; x++) {
|
2040
|
+
nullPadding += (char_t)x;
|
2041
|
+
}
|
2042
|
+
|
2043
|
+
// Bump all the patches forward.
|
2044
|
+
for (typename Patches::iterator cur_patch = patches.begin(); cur_patch != patches.end(); ++cur_patch) {
|
2045
|
+
(*cur_patch).start1 += paddingLength;
|
2046
|
+
(*cur_patch).start2 += paddingLength;
|
2047
|
+
}
|
2048
|
+
|
2049
|
+
// Add some padding on start of first diff.
|
2050
|
+
Patch &firstPatch = patches.front();
|
2051
|
+
Diffs &firstPatchDiffs = firstPatch.diffs;
|
2052
|
+
if (firstPatchDiffs.empty() || firstPatchDiffs.front().operation != EQUAL) {
|
2053
|
+
// Add nullPadding equality.
|
2054
|
+
firstPatchDiffs.push_front(Diff(EQUAL, nullPadding));
|
2055
|
+
firstPatch.start1 -= paddingLength; // Should be 0.
|
2056
|
+
firstPatch.start2 -= paddingLength; // Should be 0.
|
2057
|
+
firstPatch.length1 += paddingLength;
|
2058
|
+
firstPatch.length2 += paddingLength;
|
2059
|
+
} else if (paddingLength > (int)firstPatchDiffs.front().text.length()) {
|
2060
|
+
// Grow first equality.
|
2061
|
+
Diff &firstDiff = firstPatchDiffs.front();
|
2062
|
+
int extraLength = paddingLength - firstDiff.text.length();
|
2063
|
+
firstDiff.text = safeMid(nullPadding, firstDiff.text.length(),
|
2064
|
+
paddingLength - firstDiff.text.length()) + firstDiff.text;
|
2065
|
+
firstPatch.start1 -= extraLength;
|
2066
|
+
firstPatch.start2 -= extraLength;
|
2067
|
+
firstPatch.length1 += extraLength;
|
2068
|
+
firstPatch.length2 += extraLength;
|
2069
|
+
}
|
2070
|
+
|
2071
|
+
// Add some padding on end of last diff.
|
2072
|
+
Patch &lastPatch = patches.front();
|
2073
|
+
Diffs &lastPatchDiffs = lastPatch.diffs;
|
2074
|
+
if (lastPatchDiffs.empty() || lastPatchDiffs.back().operation != EQUAL) {
|
2075
|
+
// Add nullPadding equality.
|
2076
|
+
lastPatchDiffs.push_back(Diff(EQUAL, nullPadding));
|
2077
|
+
lastPatch.length1 += paddingLength;
|
2078
|
+
lastPatch.length2 += paddingLength;
|
2079
|
+
} else if (paddingLength > (int)lastPatchDiffs.back().text.length()) {
|
2080
|
+
// Grow last equality.
|
2081
|
+
Diff &lastDiff = lastPatchDiffs.back();
|
2082
|
+
int extraLength = paddingLength - lastDiff.text.length();
|
2083
|
+
lastDiff.text += nullPadding.substr(0, extraLength);
|
2084
|
+
lastPatch.length1 += extraLength;
|
2085
|
+
lastPatch.length2 += extraLength;
|
2086
|
+
}
|
2087
|
+
|
2088
|
+
return nullPadding;
|
2089
|
+
}
|
2090
|
+
|
2091
|
+
/**
|
2092
|
+
* Look through the patches and break up any which are longer than the
|
2093
|
+
* maximum limit of the match algorithm.
|
2094
|
+
* Intended to be called only from within patch_apply.
|
2095
|
+
* @param patches LinkedList of Patch objects.
|
2096
|
+
*/
|
2097
|
+
public:
|
2098
|
+
void patch_splitMax(Patches &patches) const {
|
2099
|
+
short patch_size = Match_MaxBits;
|
2100
|
+
string_t precontext, postcontext;
|
2101
|
+
Patch patch;
|
2102
|
+
int start1, start2;
|
2103
|
+
bool empty;
|
2104
|
+
Operation diff_type;
|
2105
|
+
string_t diff_text;
|
2106
|
+
Patch bigpatch;
|
2107
|
+
|
2108
|
+
for (typename Patches::iterator cur_patch = patches.begin(); cur_patch != patches.end();) {
|
2109
|
+
if ((*cur_patch).length1 <= patch_size) { ++cur_patch; continue; }
|
2110
|
+
bigpatch = *cur_patch;
|
2111
|
+
// Remove the big old patch.
|
2112
|
+
cur_patch = patches.erase(cur_patch);
|
2113
|
+
start1 = bigpatch.start1;
|
2114
|
+
start2 = bigpatch.start2;
|
2115
|
+
precontext.clear();
|
2116
|
+
while (!bigpatch.diffs.empty()) {
|
2117
|
+
// Create one of several smaller patches.
|
2118
|
+
patch = Patch();
|
2119
|
+
empty = true;
|
2120
|
+
patch.start1 = start1 - precontext.length();
|
2121
|
+
patch.start2 = start2 - precontext.length();
|
2122
|
+
if (!precontext.empty()) {
|
2123
|
+
patch.length1 = patch.length2 = precontext.length();
|
2124
|
+
patch.diffs.push_back(Diff(EQUAL, precontext));
|
2125
|
+
}
|
2126
|
+
while (!bigpatch.diffs.empty()
|
2127
|
+
&& patch.length1 < patch_size - Patch_Margin) {
|
2128
|
+
diff_type = bigpatch.diffs.front().operation;
|
2129
|
+
diff_text = bigpatch.diffs.front().text;
|
2130
|
+
if (diff_type == INSERT) {
|
2131
|
+
// Insertions are harmless.
|
2132
|
+
patch.length2 += diff_text.length();
|
2133
|
+
start2 += diff_text.length();
|
2134
|
+
patch.diffs.push_back(bigpatch.diffs.front());
|
2135
|
+
bigpatch.diffs.pop_front();
|
2136
|
+
empty = false;
|
2137
|
+
} else if (diff_type == DELETE && patch.diffs.size() == 1
|
2138
|
+
&& patch.diffs.front().operation == EQUAL
|
2139
|
+
&& (int)diff_text.length() > 2 * patch_size) {
|
2140
|
+
// This is a large deletion. Let it pass in one chunk.
|
2141
|
+
patch.length1 += diff_text.length();
|
2142
|
+
start1 += diff_text.length();
|
2143
|
+
empty = false;
|
2144
|
+
patch.diffs.push_back(Diff(diff_type, diff_text));
|
2145
|
+
bigpatch.diffs.pop_front();
|
2146
|
+
} else {
|
2147
|
+
// Deletion or equality. Only take as much as we can stomach.
|
2148
|
+
diff_text = diff_text.substr(0, std::min((int)diff_text.length(),
|
2149
|
+
patch_size - patch.length1 - Patch_Margin));
|
2150
|
+
patch.length1 += diff_text.length();
|
2151
|
+
start1 += diff_text.length();
|
2152
|
+
if (diff_type == EQUAL) {
|
2153
|
+
patch.length2 += diff_text.length();
|
2154
|
+
start2 += diff_text.length();
|
2155
|
+
} else {
|
2156
|
+
empty = false;
|
2157
|
+
}
|
2158
|
+
patch.diffs.push_back(Diff(diff_type, diff_text));
|
2159
|
+
if (diff_text == bigpatch.diffs.front().text) {
|
2160
|
+
bigpatch.diffs.pop_front();
|
2161
|
+
} else {
|
2162
|
+
bigpatch.diffs.front().text = safeMid(bigpatch.diffs.front().text, diff_text.length());
|
2163
|
+
}
|
2164
|
+
}
|
2165
|
+
}
|
2166
|
+
// Compute the head context for the next patch.
|
2167
|
+
precontext = safeMid(diff_text2(patch.diffs), std::max(0, (int)precontext.length() - Patch_Margin));
|
2168
|
+
// Append the end context for this patch.
|
2169
|
+
postcontext = diff_text1(bigpatch.diffs);
|
2170
|
+
if ((int)postcontext.length() > Patch_Margin) {
|
2171
|
+
postcontext = postcontext.substr(0, Patch_Margin);
|
2172
|
+
}
|
2173
|
+
if (!postcontext.empty()) {
|
2174
|
+
patch.length1 += postcontext.length();
|
2175
|
+
patch.length2 += postcontext.length();
|
2176
|
+
if (!patch.diffs.empty()
|
2177
|
+
&& patch.diffs.back().operation == EQUAL) {
|
2178
|
+
patch.diffs.back().text += postcontext;
|
2179
|
+
} else {
|
2180
|
+
patch.diffs.push_back(Diff(EQUAL, postcontext));
|
2181
|
+
}
|
2182
|
+
}
|
2183
|
+
if (!empty) {
|
2184
|
+
patches.insert(cur_patch, patch);
|
2185
|
+
}
|
2186
|
+
}
|
2187
|
+
}
|
2188
|
+
}
|
2189
|
+
|
2190
|
+
/**
|
2191
|
+
* Take a list of patches and return a textual representation.
|
2192
|
+
* @param patches List of Patch objects.
|
2193
|
+
* @return Text representation of patches.
|
2194
|
+
*/
|
2195
|
+
public:
|
2196
|
+
static string_t patch_toText(const Patches &patches) {
|
2197
|
+
string_t text;
|
2198
|
+
for (typename Patches::const_iterator cur_patch = patches.begin(); cur_patch != patches.end(); ++cur_patch) {
|
2199
|
+
text += (*cur_patch).toString();
|
2200
|
+
}
|
2201
|
+
return text;
|
2202
|
+
}
|
2203
|
+
|
2204
|
+
/**
|
2205
|
+
* Parse a textual representation of patches and return a List of Patch
|
2206
|
+
* objects.
|
2207
|
+
* @param textline Text representation of patches.
|
2208
|
+
* @return List of Patch objects.
|
2209
|
+
* @throws string_t If invalid input.
|
2210
|
+
*/
|
2211
|
+
public:
|
2212
|
+
Patches patch_fromText(const string_t &textline) const {
|
2213
|
+
Patches patches;
|
2214
|
+
if (!textline.empty()) {
|
2215
|
+
char_t sign;
|
2216
|
+
string_t line;
|
2217
|
+
typename string_t::const_pointer text = textline.c_str();
|
2218
|
+
typename string_t::size_type text_len, l;
|
2219
|
+
while (text - textline.c_str() < (int)textline.length()) {
|
2220
|
+
if ((text_len = next_token(textline, traits::from_wchar(L'\n'), text)) == 0) { ++text; continue; }
|
2221
|
+
|
2222
|
+
// A replacement for the regexp "^@@ -(\\d+),?(\\d*) \\+(\\d+),?(\\d*) @@$" exact match
|
2223
|
+
string_t start1, length1, start2, length2;
|
2224
|
+
do {
|
2225
|
+
typename string_t::const_pointer t = text;
|
2226
|
+
l = text_len;
|
2227
|
+
if ((l -= 9) > 0 && traits::to_wchar(*t) == L'@' && traits::to_wchar(*++t) == L'@'
|
2228
|
+
&& traits::to_wchar(*++t) == L' ' && traits::to_wchar(*++t) == L'-' && traits::is_digit(*++t)) {
|
2229
|
+
do { start1 += *t; } while (--l > 0 && traits::is_digit(*++t));
|
2230
|
+
if (l > 0 && traits::to_wchar(*t) == L',') ++t, --l;
|
2231
|
+
while (l > 0 && traits::is_digit(*t)) --l, length1 += *t++;
|
2232
|
+
if (l > 0 && traits::to_wchar(*t++) == L' ' && traits::to_wchar(*t++) == L'+' && traits::is_digit(*t)) {
|
2233
|
+
do { start2 += *t; } while (--l >= 0 && traits::is_digit(*++t));
|
2234
|
+
if (l > 0 && traits::to_wchar(*t) == L',') ++t, --l;
|
2235
|
+
while (l > 0 && traits::is_digit(*t)) --l, length2 += *t++;
|
2236
|
+
if (l == 0 && traits::to_wchar(*t++) == L' ' && traits::to_wchar(*t++) == L'@' && traits::to_wchar(*t) == L'@') break; // Success
|
2237
|
+
}
|
2238
|
+
}
|
2239
|
+
throw string_t(traits::cs(L"Invalid patch string: ") + string_t(text, text_len));
|
2240
|
+
} while (false);
|
2241
|
+
|
2242
|
+
Patch patch;
|
2243
|
+
patch.start1 = to_int(start1);
|
2244
|
+
if (length1.empty()) {
|
2245
|
+
patch.start1--;
|
2246
|
+
patch.length1 = 1;
|
2247
|
+
} else if (length1.size() == 1 && traits::to_wchar(length1[0]) == L'0') {
|
2248
|
+
patch.length1 = 0;
|
2249
|
+
} else {
|
2250
|
+
patch.start1--;
|
2251
|
+
patch.length1 = to_int(length1);
|
2252
|
+
}
|
2253
|
+
|
2254
|
+
patch.start2 = to_int(start2);
|
2255
|
+
if (length2.empty()) {
|
2256
|
+
patch.start2--;
|
2257
|
+
patch.length2 = 1;
|
2258
|
+
} else if (length2.size() == 1 && traits::to_wchar(length2[0]) == L'0') {
|
2259
|
+
patch.length2 = 0;
|
2260
|
+
} else {
|
2261
|
+
patch.start2--;
|
2262
|
+
patch.length2 = to_int(length2);
|
2263
|
+
}
|
2264
|
+
|
2265
|
+
for (text += text_len + 1; text - textline.c_str() < (int)textline.length(); text += text_len + 1) {
|
2266
|
+
if ((text_len = next_token(textline, traits::from_wchar(L'\n'), text)) == 0) continue;
|
2267
|
+
|
2268
|
+
sign = *text;
|
2269
|
+
line.assign(text + 1, text_len - 1);
|
2270
|
+
percent_decode(line);
|
2271
|
+
switch (traits::to_wchar(sign)) {
|
2272
|
+
case L'-':
|
2273
|
+
// Deletion.
|
2274
|
+
patch.diffs.push_back(Diff(DELETE, line));
|
2275
|
+
continue;
|
2276
|
+
case L'+':
|
2277
|
+
// Insertion.
|
2278
|
+
patch.diffs.push_back(Diff(INSERT, line));
|
2279
|
+
continue;
|
2280
|
+
case L' ':
|
2281
|
+
// Minor equality.
|
2282
|
+
patch.diffs.push_back(Diff(EQUAL, line));
|
2283
|
+
continue;
|
2284
|
+
case L'@':
|
2285
|
+
// Start of next patch.
|
2286
|
+
break;
|
2287
|
+
default:
|
2288
|
+
// WTF?
|
2289
|
+
throw string_t(traits::cs(L"Invalid patch mode '") + (sign + (traits::cs(L"' in: ") + line)));
|
2290
|
+
}
|
2291
|
+
break;
|
2292
|
+
}
|
2293
|
+
|
2294
|
+
patches.push_back(patch);
|
2295
|
+
}
|
2296
|
+
}
|
2297
|
+
return patches;
|
2298
|
+
}
|
2299
|
+
|
2300
|
+
/**
|
2301
|
+
* A safer version of string_t.mid(pos). This one returns "" instead of
|
2302
|
+
* null when the postion equals the string length.
|
2303
|
+
* @param str String to take a substring from.
|
2304
|
+
* @param pos Position to start the substring from.
|
2305
|
+
* @return Substring.
|
2306
|
+
*/
|
2307
|
+
private:
|
2308
|
+
static inline string_t safeMid(const string_t &str, int pos) {
|
2309
|
+
return ((size_t)pos == str.length()) ? string_t() : str.substr(pos);
|
2310
|
+
}
|
2311
|
+
|
2312
|
+
/**
|
2313
|
+
* A safer version of string_t.mid(pos, len). This one returns "" instead of
|
2314
|
+
* null when the postion equals the string length.
|
2315
|
+
* @param str String to take a substring from.
|
2316
|
+
* @param pos Position to start the substring from.
|
2317
|
+
* @param len Length of substring.
|
2318
|
+
* @return Substring.
|
2319
|
+
*/
|
2320
|
+
private:
|
2321
|
+
static inline string_t safeMid(const string_t &str, int pos, int len) {
|
2322
|
+
return ((size_t)pos == str.length()) ? string_t() : str.substr(pos, len);
|
2323
|
+
}
|
2324
|
+
|
2325
|
+
/**
|
2326
|
+
* Utility functions
|
2327
|
+
*/
|
2328
|
+
private:
|
2329
|
+
static string_t to_string(int n) {
|
2330
|
+
string_t str;
|
2331
|
+
bool negative = false;
|
2332
|
+
size_t l = 0;
|
2333
|
+
if (n < 0) n = -n, ++l, negative = true;
|
2334
|
+
int n_ = n; do { ++l; } while ((n_ /= 10) > 0);
|
2335
|
+
str.resize(l);
|
2336
|
+
typename string_t::iterator s = str.end();
|
2337
|
+
const wchar_t digits[] = L"0123456789";
|
2338
|
+
do { *--s = traits::from_wchar(digits[n % 10]); } while ((n /= 10) > 0);
|
2339
|
+
if (negative) *--s = traits::from_wchar(L'-');
|
2340
|
+
return str;
|
2341
|
+
}
|
2342
|
+
|
2343
|
+
static int to_int(const string_t& str) { return traits::to_int(str.c_str()); }
|
2344
|
+
|
2345
|
+
static bool is_control(char_t c) { switch (traits::to_wchar(c)) { case L'\n': case L'\r': return true; } return false; }
|
2346
|
+
|
2347
|
+
static typename string_t::size_type next_token(const string_t& str, char_t delim, typename string_t::const_pointer off) {
|
2348
|
+
typename string_t::const_pointer p = off, end = str.c_str() + str.length();
|
2349
|
+
for (; p != end; ++p) if (*p == delim) break;
|
2350
|
+
return p - off;
|
2351
|
+
}
|
2352
|
+
|
2353
|
+
static void append_percent_encoded(string_t& s1, const string_t& s2) {
|
2354
|
+
const wchar_t safe_chars[] = L"0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-_.~ !*'();/?:@&=+$,#";
|
2355
|
+
|
2356
|
+
size_t safe[0x100], i;
|
2357
|
+
for (i = 0; i < 0x100; ++i) safe[i] = 0;
|
2358
|
+
for (i = 0; i < sizeof(safe_chars) / sizeof(wchar_t); ++i) safe[safe_chars[i]] = i + 1;
|
2359
|
+
|
2360
|
+
size_t n = 0;
|
2361
|
+
typename traits::utf32_t u;
|
2362
|
+
typename string_t::const_pointer c = s2.c_str(), end = c + s2.length();
|
2363
|
+
while (c != end) {
|
2364
|
+
c = traits::to_utf32(c, end, u);
|
2365
|
+
n += u >= 0x10000? 12 : u >= 0x800? 9 : u >= 0x80? 6 : safe[static_cast<unsigned char>(u)]? 1 : 3;
|
2366
|
+
}
|
2367
|
+
if (n == s2.length())
|
2368
|
+
s1.append(s2);
|
2369
|
+
else {
|
2370
|
+
s1.reserve(s1.size() + n);
|
2371
|
+
// Encode as UTF-8, then escape unsafe characters
|
2372
|
+
unsigned char utf8[4];
|
2373
|
+
for (c = s2.c_str(); c != end;) {
|
2374
|
+
c = traits::to_utf32(c, end, u);
|
2375
|
+
unsigned char* pt = utf8;
|
2376
|
+
if (u < 0x80)
|
2377
|
+
*pt++ = (unsigned char)u;
|
2378
|
+
else if (u < 0x800) {
|
2379
|
+
*pt++ = (unsigned char)((u >> 6) | 0xC0);
|
2380
|
+
*pt++ = (unsigned char)((u & 0x3F) | 0x80);
|
2381
|
+
}
|
2382
|
+
else if (u < 0x10000) {
|
2383
|
+
*pt++ = (unsigned char)((u >> 12) | 0xE0);
|
2384
|
+
*pt++ = (unsigned char)(((u >> 6) & 0x3F) | 0x80);
|
2385
|
+
*pt++ = (unsigned char)((u & 0x3F) | 0x80);
|
2386
|
+
}
|
2387
|
+
else {
|
2388
|
+
*pt++ = (unsigned char)((u >> 18) | 0xF0);
|
2389
|
+
*pt++ = (unsigned char)(((u >> 12) & 0x3F) | 0x80);
|
2390
|
+
*pt++ = (unsigned char)(((u >> 6) & 0x3F) | 0x80);
|
2391
|
+
*pt++ = (unsigned char)((u & 0x3F) | 0x80);
|
2392
|
+
}
|
2393
|
+
|
2394
|
+
for (const unsigned char* p = utf8; p < pt; ++p)
|
2395
|
+
if (safe[*p])
|
2396
|
+
s1 += traits::from_wchar(safe_chars[safe[*p] - 1]);
|
2397
|
+
else {
|
2398
|
+
s1 += traits::from_wchar(L'%');
|
2399
|
+
s1 += traits::from_wchar(safe_chars[(*p & 0xF0) >> 4]);
|
2400
|
+
s1 += traits::from_wchar(safe_chars[*p & 0xF]);
|
2401
|
+
}
|
2402
|
+
}
|
2403
|
+
}
|
2404
|
+
}
|
2405
|
+
|
2406
|
+
static unsigned hex_digit_value(char_t c) {
|
2407
|
+
switch (traits::to_wchar(c))
|
2408
|
+
{
|
2409
|
+
case L'0': return 0;
|
2410
|
+
case L'1': return 1;
|
2411
|
+
case L'2': return 2;
|
2412
|
+
case L'3': return 3;
|
2413
|
+
case L'4': return 4;
|
2414
|
+
case L'5': return 5;
|
2415
|
+
case L'6': return 6;
|
2416
|
+
case L'7': return 7;
|
2417
|
+
case L'8': return 8;
|
2418
|
+
case L'9': return 9;
|
2419
|
+
case L'A': case L'a': return 0xA;
|
2420
|
+
case L'B': case L'b': return 0xB;
|
2421
|
+
case L'C': case L'c': return 0xC;
|
2422
|
+
case L'D': case L'd': return 0xD;
|
2423
|
+
case L'E': case L'e': return 0xE;
|
2424
|
+
case L'F': case L'f': return 0xF;
|
2425
|
+
}
|
2426
|
+
throw string_t(string_t(traits::cs(L"Invalid character: ")) + c);
|
2427
|
+
}
|
2428
|
+
|
2429
|
+
static void percent_decode(string_t& str) {
|
2430
|
+
typename string_t::iterator s2 = str.begin(), s3 = s2, s4 = s2;
|
2431
|
+
for (typename string_t::const_pointer s1 = str.c_str(), end = s1 + str.size(); s1 != end; ++s1, ++s2)
|
2432
|
+
if (traits::to_wchar(*s1) != L'%')
|
2433
|
+
*s2 = *s1;
|
2434
|
+
else {
|
2435
|
+
char_t d1 = *++s1;
|
2436
|
+
*s2 = char_t((hex_digit_value(d1) << 4) + hex_digit_value(*++s1));
|
2437
|
+
}
|
2438
|
+
// Decode UTF-8 string in-place
|
2439
|
+
while (s3 != s2) {
|
2440
|
+
unsigned u = *s3;
|
2441
|
+
if (u < 0x80)
|
2442
|
+
;
|
2443
|
+
else if ((u >> 5) == 6) {
|
2444
|
+
if (++s3 == s2 || (*s3 & 0xC0) != 0x80) continue;
|
2445
|
+
u = ((u & 0x1F) << 6) + (*s3 & 0x3F);
|
2446
|
+
}
|
2447
|
+
else if ((u >> 4) == 0xE) {
|
2448
|
+
if (++s3 == s2 || (*s3 & 0xC0) != 0x80) continue;
|
2449
|
+
u = ((u & 0xF) << 12) + ((*s3 & 0x3F) << 6);
|
2450
|
+
if (++s3 == s2 || (*s3 & 0xC0) != 0x80) continue;
|
2451
|
+
u += *s3 & 0x3F;
|
2452
|
+
}
|
2453
|
+
else if ((u >> 3) == 0x1E) {
|
2454
|
+
if (++s3 == s2 || (*s3 & 0xC0) != 0x80) continue;
|
2455
|
+
u = ((u & 7) << 18) + ((*s3 & 0x3F) << 12);
|
2456
|
+
if (++s3 == s2 || (*s3 & 0xC0) != 0x80) continue;
|
2457
|
+
u += (*s3 & 0x3F) << 6;
|
2458
|
+
if (++s3 == s2 || (*s3 & 0xC0) != 0x80) continue;
|
2459
|
+
u += *s3 & 0x3F;
|
2460
|
+
}
|
2461
|
+
else {
|
2462
|
+
++s3;
|
2463
|
+
continue;
|
2464
|
+
}
|
2465
|
+
s4 = traits::from_utf32(u, s4);
|
2466
|
+
++s3;
|
2467
|
+
}
|
2468
|
+
if (s4 != str.end()) str.resize(s4 - str.begin());
|
2469
|
+
}
|
2470
|
+
|
2471
|
+
static string_t right(const string_t& str, typename string_t::size_type n) { return str.substr(str.size() - n); }
|
2472
|
+
};
|
2473
|
+
|
2474
|
+
|
2475
|
+
/**
|
2476
|
+
* Functions dependent on character type
|
2477
|
+
*/
|
2478
|
+
|
2479
|
+
// Unicode helpers
|
2480
|
+
template <class char_t, class utf32_type = unsigned>
|
2481
|
+
struct diff_match_patch_utf32_direct {
|
2482
|
+
typedef utf32_type utf32_t;
|
2483
|
+
template <class iterator> static iterator to_utf32(iterator i, iterator end, utf32_t& u)
|
2484
|
+
{
|
2485
|
+
u = *i++;
|
2486
|
+
return i;
|
2487
|
+
}
|
2488
|
+
template <class iterator> static iterator from_utf32(utf32_t u, iterator o)
|
2489
|
+
{
|
2490
|
+
*o++ = static_cast<char_t>(u);
|
2491
|
+
return o;
|
2492
|
+
}
|
2493
|
+
};
|
2494
|
+
|
2495
|
+
template <class char_t, class utf32_type = unsigned>
|
2496
|
+
struct diff_match_patch_utf32_from_utf16 {
|
2497
|
+
typedef utf32_type utf32_t;
|
2498
|
+
static const unsigned UTF16_SURROGATE_MIN = 0xD800u, UTF16_SURROGATE_MAX = 0xDFFFu, UTF16_HIGH_SURROGATE_MAX = 0xDBFFu, UTF16_LOW_SURROGATE_MIN = 0xDC00u;
|
2499
|
+
static const unsigned UTF16_SURROGATE_OFFSET = (UTF16_SURROGATE_MIN << 10) + UTF16_HIGH_SURROGATE_MAX - 0xFFFFu;
|
2500
|
+
template <class iterator> static iterator to_utf32(iterator i, iterator end, utf32_t& u)
|
2501
|
+
{
|
2502
|
+
u = *i++;
|
2503
|
+
if (UTF16_SURROGATE_MIN <= u && u <= UTF16_HIGH_SURROGATE_MAX && i != end)
|
2504
|
+
u = (u << 10) + *i++ - UTF16_SURROGATE_OFFSET; // Assume it is a UTF-16 surrogate pair
|
2505
|
+
return i;
|
2506
|
+
}
|
2507
|
+
template <class iterator> static iterator from_utf32(utf32_t u, iterator o)
|
2508
|
+
{
|
2509
|
+
if (u > 0xFFFF) { // Encode code points that do not fit in char_t as UTF-16 surrogate pairs
|
2510
|
+
*o++ = static_cast<char_t>((u >> 10) + UTF16_SURROGATE_MIN - (0x10000 >> 10));
|
2511
|
+
*o++ = static_cast<char_t>((u & 0x3FF) + UTF16_LOW_SURROGATE_MIN);
|
2512
|
+
}
|
2513
|
+
else
|
2514
|
+
*o++ = static_cast<char_t>(u);
|
2515
|
+
return o;
|
2516
|
+
}
|
2517
|
+
};
|
2518
|
+
|
2519
|
+
// Specialization of the traits for wchar_t
|
2520
|
+
#include <cwctype>
|
2521
|
+
template <> struct diff_match_patch_traits<wchar_t> : diff_match_patch_utf32_from_utf16<wchar_t> {
|
2522
|
+
static bool is_alnum(wchar_t c) { return std::iswalnum(c)? true : false; }
|
2523
|
+
static bool is_digit(wchar_t c) { return std::iswdigit(c)? true : false; }
|
2524
|
+
static bool is_space(wchar_t c) { return std::iswspace(c)? true : false; }
|
2525
|
+
static int to_int(const wchar_t* s) { return static_cast<int>(std::wcstol(s, NULL, 10)); }
|
2526
|
+
static wchar_t from_wchar(wchar_t c) { return c; }
|
2527
|
+
static wchar_t to_wchar(wchar_t c) { return c; }
|
2528
|
+
static const wchar_t* cs(const wchar_t* s) { return s; }
|
2529
|
+
};
|
2530
|
+
|
2531
|
+
|
2532
|
+
// Possible specialization of the traits for char
|
2533
|
+
#include <cctype>
|
2534
|
+
template <> struct diff_match_patch_traits<char> : diff_match_patch_utf32_direct<char> {
|
2535
|
+
static bool is_alnum(char c) { return std::isalnum(c)? true : false; }
|
2536
|
+
static bool is_digit(char c) { return std::isdigit(c)? true : false; }
|
2537
|
+
static bool is_space(char c) { return std::isspace(c)? true : false; }
|
2538
|
+
static int to_int(const char* s) { return std::atoi(s); }
|
2539
|
+
static char from_wchar(wchar_t c) { return static_cast<char>(c); }
|
2540
|
+
static wchar_t to_wchar(char c) { return static_cast<wchar_t>(c); }
|
2541
|
+
static std::string cs(const wchar_t* s) { return std::string(s, s + wcslen(s)); }
|
2542
|
+
};
|
2543
|
+
|
2544
|
+
|
2545
|
+
#endif // DIFF_MATCH_PATCH_H
|