• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Diff Match and Patch
3  *
4  * Copyright 2006 Google Inc.
5  * http://code.google.com/p/google-diff-match-patch/
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 
20 package name.fraser.neil.plaintext;
21 
22 import java.io.UnsupportedEncodingException;
23 import java.net.URLEncoder;
24 import java.net.URLDecoder;
25 import java.util.ArrayList;
26 import java.util.Arrays;
27 import java.util.HashMap;
28 import java.util.HashSet;
29 import java.util.LinkedList;
30 import java.util.List;
31 import java.util.ListIterator;
32 import java.util.Map;
33 import java.util.Set;
34 import java.util.Stack;
35 import java.util.regex.Matcher;
36 import java.util.regex.Pattern;
37 
38 
39 /*
40  * Functions for diff, match and patch.
41  * Computes the difference between two texts to create a patch.
42  * Applies the patch onto another text, allowing for errors.
43  *
44  * @author fraser@google.com (Neil Fraser)
45  */
46 
47 /**
48  * Class containing the diff, match and patch methods.
49  * Also contains the behaviour settings.
50  */
51 public class diff_match_patch {
52 
53   // Defaults.
54   // Set these on your diff_match_patch instance to override the defaults.
55 
56   /**
57    * Number of seconds to map a diff before giving up (0 for infinity).
58    */
59   public float Diff_Timeout = 1.0f;
60   /**
61    * Cost of an empty edit operation in terms of edit characters.
62    */
63   public short Diff_EditCost = 4;
64   /**
65    * The size beyond which the double-ended diff activates.
66    * Double-ending is twice as fast, but less accurate.
67    */
68   public short Diff_DualThreshold = 32;
69   /**
70    * At what point is no match declared (0.0 = perfection, 1.0 = very loose).
71    */
72   public float Match_Threshold = 0.5f;
73   /**
74    * How far to search for a match (0 = exact location, 1000+ = broad match).
75    * A match this many characters away from the expected location will add
76    * 1.0 to the score (0.0 is a perfect match).
77    */
78   public int Match_Distance = 1000;
79   /**
80    * When deleting a large block of text (over ~64 characters), how close does
81    * the contents have to match the expected contents. (0.0 = perfection,
82    * 1.0 = very loose).  Note that Match_Threshold controls how closely the
83    * end points of a delete need to match.
84    */
85   public float Patch_DeleteThreshold = 0.5f;
86   /**
87    * Chunk size for context length.
88    */
89   public short Patch_Margin = 4;
90 
91   /**
92    * The number of bits in an int.
93    */
94   private int Match_MaxBits = 32;
95 
96   /**
97    * Internal class for returning results from diff_linesToChars().
98    * Other less paranoid languages just use a three-element array.
99    */
100   protected static class LinesToCharsResult {
101     protected String chars1;
102     protected String chars2;
103     protected List<String> lineArray;
104 
LinesToCharsResult(String chars1, String chars2, List<String> lineArray)105     protected LinesToCharsResult(String chars1, String chars2,
106         List<String> lineArray) {
107       this.chars1 = chars1;
108       this.chars2 = chars2;
109       this.lineArray = lineArray;
110     }
111   }
112 
113 
114   //  DIFF FUNCTIONS
115 
116 
117   /**
118    * The data structure representing a diff is a Linked list of Diff objects:
119    * {Diff(Operation.DELETE, "Hello"), Diff(Operation.INSERT, "Goodbye"),
120    *  Diff(Operation.EQUAL, " world.")}
121    * which means: delete "Hello", add "Goodbye" and keep " world."
122    */
123   public enum Operation {
124     DELETE, INSERT, EQUAL
125   }
126 
127 
128   /**
129    * Find the differences between two texts.
130    * Run a faster slightly less optimal diff
131    * This method allows the 'checklines' of diff_main() to be optional.
132    * Most of the time checklines is wanted, so default to true.
133    * @param text1 Old string to be diffed.
134    * @param text2 New string to be diffed.
135    * @return Linked List of Diff objects.
136    */
diff_main(String text1, String text2)137   public LinkedList<Diff> diff_main(String text1, String text2) {
138     return diff_main(text1, text2, true);
139   }
140 
141   /**
142    * Find the differences between two texts.  Simplifies the problem by
143    * stripping any common prefix or suffix off the texts before diffing.
144    * @param text1 Old string to be diffed.
145    * @param text2 New string to be diffed.
146    * @param checklines Speedup flag.  If false, then don't run a
147    *     line-level diff first to identify the changed areas.
148    *     If true, then run a faster slightly less optimal diff
149    * @return Linked List of Diff objects.
150    */
diff_main(String text1, String text2, boolean checklines)151   public LinkedList<Diff> diff_main(String text1, String text2,
152                                     boolean checklines) {
153     // Check for null inputs.
154     if (text1 == null || text2 == null) {
155       throw new IllegalArgumentException("Null inputs. (diff_main)");
156     }
157 
158     // Check for equality (speedup).
159     LinkedList<Diff> diffs;
160     if (text1.equals(text2)) {
161       diffs = new LinkedList<Diff>();
162       diffs.add(new Diff(Operation.EQUAL, text1));
163       return diffs;
164     }
165 
166     // Trim off common prefix (speedup).
167     int commonlength = diff_commonPrefix(text1, text2);
168     String commonprefix = text1.substring(0, commonlength);
169     text1 = text1.substring(commonlength);
170     text2 = text2.substring(commonlength);
171 
172     // Trim off common suffix (speedup).
173     commonlength = diff_commonSuffix(text1, text2);
174     String commonsuffix = text1.substring(text1.length() - commonlength);
175     text1 = text1.substring(0, text1.length() - commonlength);
176     text2 = text2.substring(0, text2.length() - commonlength);
177 
178     // Compute the diff on the middle block.
179     diffs = diff_compute(text1, text2, checklines);
180 
181     // Restore the prefix and suffix.
182     if (commonprefix.length() != 0) {
183       diffs.addFirst(new Diff(Operation.EQUAL, commonprefix));
184     }
185     if (commonsuffix.length() != 0) {
186       diffs.addLast(new Diff(Operation.EQUAL, commonsuffix));
187     }
188 
189     diff_cleanupMerge(diffs);
190     return diffs;
191   }
192 
193 
194   /**
195    * Find the differences between two texts.  Assumes that the texts do not
196    * have any common prefix or suffix.
197    * @param text1 Old string to be diffed.
198    * @param text2 New string to be diffed.
199    * @param checklines Speedup flag.  If false, then don't run a
200    *     line-level diff first to identify the changed areas.
201    *     If true, then run a faster slightly less optimal diff
202    * @return Linked List of Diff objects.
203    */
diff_compute(String text1, String text2, boolean checklines)204   protected LinkedList<Diff> diff_compute(String text1, String text2,
205                                           boolean checklines) {
206     LinkedList<Diff> diffs = new LinkedList<Diff>();
207 
208     if (text1.length() == 0) {
209       // Just add some text (speedup).
210       diffs.add(new Diff(Operation.INSERT, text2));
211       return diffs;
212     }
213 
214     if (text2.length() == 0) {
215       // Just delete some text (speedup).
216       diffs.add(new Diff(Operation.DELETE, text1));
217       return diffs;
218     }
219 
220     String longtext = text1.length() > text2.length() ? text1 : text2;
221     String shorttext = text1.length() > text2.length() ? text2 : text1;
222     int i = longtext.indexOf(shorttext);
223     if (i != -1) {
224       // Shorter text is inside the longer text (speedup).
225       Operation op = (text1.length() > text2.length()) ?
226                      Operation.DELETE : Operation.INSERT;
227       diffs.add(new Diff(op, longtext.substring(0, i)));
228       diffs.add(new Diff(Operation.EQUAL, shorttext));
229       diffs.add(new Diff(op, longtext.substring(i + shorttext.length())));
230       return diffs;
231     }
232     longtext = shorttext = null;  // Garbage collect.
233 
234     // Check to see if the problem can be split in two.
235     String[] hm = diff_halfMatch(text1, text2);
236     if (hm != null) {
237       // A half-match was found, sort out the return data.
238       String text1_a = hm[0];
239       String text1_b = hm[1];
240       String text2_a = hm[2];
241       String text2_b = hm[3];
242       String mid_common = hm[4];
243       // Send both pairs off for separate processing.
244       LinkedList<Diff> diffs_a = diff_main(text1_a, text2_a, checklines);
245       LinkedList<Diff> diffs_b = diff_main(text1_b, text2_b, checklines);
246       // Merge the results.
247       diffs = diffs_a;
248       diffs.add(new Diff(Operation.EQUAL, mid_common));
249       diffs.addAll(diffs_b);
250       return diffs;
251     }
252 
253     // Perform a real diff.
254     if (checklines && (text1.length() < 100 || text2.length() < 100)) {
255       checklines = false;  // Too trivial for the overhead.
256     }
257     List<String> linearray = null;
258     if (checklines) {
259       // Scan the text on a line-by-line basis first.
260       LinesToCharsResult b = diff_linesToChars(text1, text2);
261       text1 = b.chars1;
262       text2 = b.chars2;
263       linearray = b.lineArray;
264     }
265 
266     diffs = diff_map(text1, text2);
267     if (diffs == null) {
268       // No acceptable result.
269       diffs = new LinkedList<Diff>();
270       diffs.add(new Diff(Operation.DELETE, text1));
271       diffs.add(new Diff(Operation.INSERT, text2));
272     }
273 
274     if (checklines) {
275       // Convert the diff back to original text.
276       diff_charsToLines(diffs, linearray);
277       // Eliminate freak matches (e.g. blank lines)
278       diff_cleanupSemantic(diffs);
279 
280       // Rediff any replacement blocks, this time character-by-character.
281       // Add a dummy entry at the end.
282       diffs.add(new Diff(Operation.EQUAL, ""));
283       int count_delete = 0;
284       int count_insert = 0;
285       String text_delete = "";
286       String text_insert = "";
287       ListIterator<Diff> pointer = diffs.listIterator();
288       Diff thisDiff = pointer.next();
289       while (thisDiff != null) {
290         switch (thisDiff.operation) {
291         case INSERT:
292           count_insert++;
293           text_insert += thisDiff.text;
294           break;
295         case DELETE:
296           count_delete++;
297           text_delete += thisDiff.text;
298           break;
299         case EQUAL:
300           // Upon reaching an equality, check for prior redundancies.
301           if (count_delete >= 1 && count_insert >= 1) {
302             // Delete the offending records and add the merged ones.
303             pointer.previous();
304             for (int j = 0; j < count_delete + count_insert; j++) {
305               pointer.previous();
306               pointer.remove();
307             }
308             for (Diff newDiff : diff_main(text_delete, text_insert, false)) {
309               pointer.add(newDiff);
310             }
311           }
312           count_insert = 0;
313           count_delete = 0;
314           text_delete = "";
315           text_insert = "";
316           break;
317         }
318         thisDiff = pointer.hasNext() ? pointer.next() : null;
319       }
320       diffs.removeLast();  // Remove the dummy entry at the end.
321     }
322     return diffs;
323   }
324 
325 
326   /**
327    * Split two texts into a list of strings.  Reduce the texts to a string of
328    * hashes where each Unicode character represents one line.
329    * @param text1 First string.
330    * @param text2 Second string.
331    * @return An object containing the encoded text1, the encoded text2 and
332    *     the List of unique strings.  The zeroth element of the List of
333    *     unique strings is intentionally blank.
334    */
diff_linesToChars(String text1, String text2)335   protected LinesToCharsResult diff_linesToChars(String text1, String text2) {
336     List<String> lineArray = new ArrayList<String>();
337     Map<String, Integer> lineHash = new HashMap<String, Integer>();
338     // e.g. linearray[4] == "Hello\n"
339     // e.g. linehash.get("Hello\n") == 4
340 
341     // "\x00" is a valid character, but various debuggers don't like it.
342     // So we'll insert a junk entry to avoid generating a null character.
343     lineArray.add("");
344 
345     String chars1 = diff_linesToCharsMunge(text1, lineArray, lineHash);
346     String chars2 = diff_linesToCharsMunge(text2, lineArray, lineHash);
347     return new LinesToCharsResult(chars1, chars2, lineArray);
348   }
349 
350 
351   /**
352    * Split a text into a list of strings.  Reduce the texts to a string of
353    * hashes where each Unicode character represents one line.
354    * @param text String to encode.
355    * @param lineArray List of unique strings.
356    * @param lineHash Map of strings to indices.
357    * @return Encoded string.
358    */
diff_linesToCharsMunge(String text, List<String> lineArray, Map<String, Integer> lineHash)359   private String diff_linesToCharsMunge(String text, List<String> lineArray,
360                                         Map<String, Integer> lineHash) {
361     int lineStart = 0;
362     int lineEnd = -1;
363     String line;
364     StringBuilder chars = new StringBuilder();
365     // Walk the text, pulling out a substring for each line.
366     // text.split('\n') would would temporarily double our memory footprint.
367     // Modifying text would create many large strings to garbage collect.
368     while (lineEnd < text.length() - 1) {
369       lineEnd = text.indexOf('\n', lineStart);
370       if (lineEnd == -1) {
371         lineEnd = text.length() - 1;
372       }
373       line = text.substring(lineStart, lineEnd + 1);
374       lineStart = lineEnd + 1;
375 
376       if (lineHash.containsKey(line)) {
377         chars.append(String.valueOf((char) (int) lineHash.get(line)));
378       } else {
379         lineArray.add(line);
380         lineHash.put(line, lineArray.size() - 1);
381         chars.append(String.valueOf((char) (lineArray.size() - 1)));
382       }
383     }
384     return chars.toString();
385   }
386 
387 
388   /**
389    * Rehydrate the text in a diff from a string of line hashes to real lines of
390    * text.
391    * @param diffs LinkedList of Diff objects.
392    * @param lineArray List of unique strings.
393    */
diff_charsToLines(LinkedList<Diff> diffs, List<String> lineArray)394   protected void diff_charsToLines(LinkedList<Diff> diffs,
395                                   List<String> lineArray) {
396     StringBuilder text;
397     for (Diff diff : diffs) {
398       text = new StringBuilder();
399       for (int y = 0; y < diff.text.length(); y++) {
400         text.append(lineArray.get(diff.text.charAt(y)));
401       }
402       diff.text = text.toString();
403     }
404   }
405 
406 
407   /**
408    * Explore the intersection points between the two texts.
409    * @param text1 Old string to be diffed.
410    * @param text2 New string to be diffed.
411    * @return LinkedList of Diff objects or null if no diff available.
412    */
diff_map(String text1, String text2)413   protected LinkedList<Diff> diff_map(String text1, String text2) {
414     long ms_end = System.currentTimeMillis() + (long) (Diff_Timeout * 1000);
415     // Cache the text lengths to prevent multiple calls.
416     int text1_length = text1.length();
417     int text2_length = text2.length();
418     int max_d = text1_length + text2_length - 1;
419     boolean doubleEnd = Diff_DualThreshold * 2 < max_d;
420     List<Set<Long>> v_map1 = new ArrayList<Set<Long>>();
421     List<Set<Long>> v_map2 = new ArrayList<Set<Long>>();
422     Map<Integer, Integer> v1 = new HashMap<Integer, Integer>();
423     Map<Integer, Integer> v2 = new HashMap<Integer, Integer>();
424     v1.put(1, 0);
425     v2.put(1, 0);
426     int x, y;
427     Long footstep = 0L;  // Used to track overlapping paths.
428     Map<Long, Integer> footsteps = new HashMap<Long, Integer>();
429     boolean done = false;
430     // If the total number of characters is odd, then the front path will
431     // collide with the reverse path.
432     boolean front = ((text1_length + text2_length) % 2 == 1);
433     for (int d = 0; d < max_d; d++) {
434       // Bail out if timeout reached.
435       if (Diff_Timeout > 0 && System.currentTimeMillis() > ms_end) {
436         return null;
437       }
438 
439       // Walk the front path one step.
440       v_map1.add(new HashSet<Long>());  // Adds at index 'd'.
441       for (int k = -d; k <= d; k += 2) {
442         if (k == -d || k != d && v1.get(k - 1) < v1.get(k + 1)) {
443           x = v1.get(k + 1);
444         } else {
445           x = v1.get(k - 1) + 1;
446         }
447         y = x - k;
448         if (doubleEnd) {
449           footstep = diff_footprint(x, y);
450           if (front && (footsteps.containsKey(footstep))) {
451             done = true;
452           }
453           if (!front) {
454             footsteps.put(footstep, d);
455           }
456         }
457         while (!done && x < text1_length && y < text2_length
458                && text1.charAt(x) == text2.charAt(y)) {
459           x++;
460           y++;
461           if (doubleEnd) {
462             footstep = diff_footprint(x, y);
463             if (front && (footsteps.containsKey(footstep))) {
464               done = true;
465             }
466             if (!front) {
467               footsteps.put(footstep, d);
468             }
469           }
470         }
471         v1.put(k, x);
472         v_map1.get(d).add(diff_footprint(x, y));
473         if (x == text1_length && y == text2_length) {
474           // Reached the end in single-path mode.
475           return diff_path1(v_map1, text1, text2);
476         } else if (done) {
477           // Front path ran over reverse path.
478           v_map2 = v_map2.subList(0, footsteps.get(footstep) + 1);
479           LinkedList<Diff> a = diff_path1(v_map1, text1.substring(0, x),
480                                           text2.substring(0, y));
481           a.addAll(diff_path2(v_map2, text1.substring(x), text2.substring(y)));
482           return a;
483         }
484       }
485 
486       if (doubleEnd) {
487         // Walk the reverse path one step.
488         v_map2.add(new HashSet<Long>());  // Adds at index 'd'.
489         for (int k = -d; k <= d; k += 2) {
490           if (k == -d || k != d && v2.get(k - 1) < v2.get(k + 1)) {
491             x = v2.get(k + 1);
492           } else {
493             x = v2.get(k - 1) + 1;
494           }
495           y = x - k;
496           footstep = diff_footprint(text1_length - x, text2_length - y);
497           if (!front && (footsteps.containsKey(footstep))) {
498             done = true;
499           }
500           if (front) {
501             footsteps.put(footstep, d);
502           }
503           while (!done && x < text1_length && y < text2_length
504                  && text1.charAt(text1_length - x - 1)
505                  == text2.charAt(text2_length - y - 1)) {
506             x++;
507             y++;
508             footstep = diff_footprint(text1_length - x, text2_length - y);
509             if (!front && (footsteps.containsKey(footstep))) {
510               done = true;
511             }
512             if (front) {
513               footsteps.put(footstep, d);
514             }
515           }
516           v2.put(k, x);
517           v_map2.get(d).add(diff_footprint(x, y));
518           if (done) {
519             // Reverse path ran over front path.
520             v_map1 = v_map1.subList(0, footsteps.get(footstep) + 1);
521             LinkedList<Diff> a
522                 = diff_path1(v_map1, text1.substring(0, text1_length - x),
523                              text2.substring(0, text2_length - y));
524             a.addAll(diff_path2(v_map2, text1.substring(text1_length - x),
525                                 text2.substring(text2_length - y)));
526             return a;
527           }
528         }
529       }
530     }
531     // Number of diffs equals number of characters, no commonality at all.
532     return null;
533   }
534 
535 
536   /**
537    * Work from the middle back to the start to determine the path.
538    * @param v_map List of path sets.
539    * @param text1 Old string fragment to be diffed.
540    * @param text2 New string fragment to be diffed.
541    * @return LinkedList of Diff objects.
542    */
diff_path1(List<Set<Long>> v_map, String text1, String text2)543   protected LinkedList<Diff> diff_path1(List<Set<Long>> v_map,
544                                         String text1, String text2) {
545     LinkedList<Diff> path = new LinkedList<Diff>();
546     int x = text1.length();
547     int y = text2.length();
548     Operation last_op = null;
549     for (int d = v_map.size() - 2; d >= 0; d--) {
550       while (true) {
551         if (v_map.get(d).contains(diff_footprint(x - 1, y))) {
552           x--;
553           if (last_op == Operation.DELETE) {
554             path.getFirst().text = text1.charAt(x) + path.getFirst().text;
555           } else {
556             path.addFirst(new Diff(Operation.DELETE,
557                                    text1.substring(x, x + 1)));
558           }
559           last_op = Operation.DELETE;
560           break;
561         } else if (v_map.get(d).contains(diff_footprint(x, y - 1))) {
562           y--;
563           if (last_op == Operation.INSERT) {
564             path.getFirst().text = text2.charAt(y) + path.getFirst().text;
565           } else {
566             path.addFirst(new Diff(Operation.INSERT,
567                                    text2.substring(y, y + 1)));
568           }
569           last_op = Operation.INSERT;
570           break;
571         } else {
572           x--;
573           y--;
574           assert (text1.charAt(x) == text2.charAt(y))
575                  : "No diagonal.  Can't happen. (diff_path1)";
576           if (last_op == Operation.EQUAL) {
577             path.getFirst().text = text1.charAt(x) + path.getFirst().text;
578           } else {
579             path.addFirst(new Diff(Operation.EQUAL, text1.substring(x, x + 1)));
580           }
581           last_op = Operation.EQUAL;
582         }
583       }
584     }
585     return path;
586   }
587 
588 
589   /**
590    * Work from the middle back to the end to determine the path.
591    * @param v_map List of path sets.
592    * @param text1 Old string fragment to be diffed.
593    * @param text2 New string fragment to be diffed.
594    * @return LinkedList of Diff objects.
595    */
diff_path2(List<Set<Long>> v_map, String text1, String text2)596   protected LinkedList<Diff> diff_path2(List<Set<Long>> v_map,
597                                         String text1, String text2) {
598     LinkedList<Diff> path = new LinkedList<Diff>();
599     int x = text1.length();
600     int y = text2.length();
601     Operation last_op = null;
602     for (int d = v_map.size() - 2; d >= 0; d--) {
603       while (true) {
604         if (v_map.get(d).contains(diff_footprint(x - 1, y))) {
605           x--;
606           if (last_op == Operation.DELETE) {
607             path.getLast().text += text1.charAt(text1.length() - x - 1);
608           } else {
609             path.addLast(new Diff(Operation.DELETE,
610                 text1.substring(text1.length() - x - 1, text1.length() - x)));
611           }
612           last_op = Operation.DELETE;
613           break;
614         } else if (v_map.get(d).contains(diff_footprint(x, y - 1))) {
615           y--;
616           if (last_op == Operation.INSERT) {
617             path.getLast().text += text2.charAt(text2.length() - y - 1);
618           } else {
619             path.addLast(new Diff(Operation.INSERT,
620                 text2.substring(text2.length() - y - 1, text2.length() - y)));
621           }
622           last_op = Operation.INSERT;
623           break;
624         } else {
625           x--;
626           y--;
627           assert (text1.charAt(text1.length() - x - 1)
628                   == text2.charAt(text2.length() - y - 1))
629                  : "No diagonal.  Can't happen. (diff_path2)";
630           if (last_op == Operation.EQUAL) {
631             path.getLast().text += text1.charAt(text1.length() - x - 1);
632           } else {
633             path.addLast(new Diff(Operation.EQUAL,
634                 text1.substring(text1.length() - x - 1, text1.length() - x)));
635           }
636           last_op = Operation.EQUAL;
637         }
638       }
639     }
640     return path;
641   }
642 
643 
644   /**
645    * Compute a good hash of two integers.
646    * @param x First int.
647    * @param y Second int.
648    * @return A long made up of both ints.
649    */
diff_footprint(int x, int y)650   protected long diff_footprint(int x, int y) {
651     // The maximum size for a long is 9,223,372,036,854,775,807
652     // The maximum size for an int is 2,147,483,647
653     // Two ints fit nicely in one long.
654     long result = x;
655     result = result << 32;
656     result += y;
657     return result;
658   }
659 
660 
661   /**
662    * Determine the common prefix of two strings
663    * @param text1 First string.
664    * @param text2 Second string.
665    * @return The number of characters common to the start of each string.
666    */
diff_commonPrefix(String text1, String text2)667   public int diff_commonPrefix(String text1, String text2) {
668     // Performance analysis: http://neil.fraser.name/news/2007/10/09/
669     int n = Math.min(text1.length(), text2.length());
670     for (int i = 0; i < n; i++) {
671       if (text1.charAt(i) != text2.charAt(i)) {
672         return i;
673       }
674     }
675     return n;
676   }
677 
678 
679   /**
680    * Determine the common suffix of two strings
681    * @param text1 First string.
682    * @param text2 Second string.
683    * @return The number of characters common to the end of each string.
684    */
diff_commonSuffix(String text1, String text2)685   public int diff_commonSuffix(String text1, String text2) {
686     // Performance analysis: http://neil.fraser.name/news/2007/10/09/
687     int text1_length = text1.length();
688     int text2_length = text2.length();
689     int n = Math.min(text1_length, text2_length);
690     for (int i = 1; i <= n; i++) {
691       if (text1.charAt(text1_length - i) != text2.charAt(text2_length - i)) {
692         return i - 1;
693       }
694     }
695     return n;
696   }
697 
698 
699   /**
700    * Do the two texts share a substring which is at least half the length of
701    * the longer text?
702    * @param text1 First string.
703    * @param text2 Second string.
704    * @return Five element String array, containing the prefix of text1, the
705    *     suffix of text1, the prefix of text2, the suffix of text2 and the
706    *     common middle.  Or null if there was no match.
707    */
diff_halfMatch(String text1, String text2)708   protected String[] diff_halfMatch(String text1, String text2) {
709     String longtext = text1.length() > text2.length() ? text1 : text2;
710     String shorttext = text1.length() > text2.length() ? text2 : text1;
711     if (longtext.length() < 10 || shorttext.length() < 1) {
712       return null;  // Pointless.
713     }
714 
715     // First check if the second quarter is the seed for a half-match.
716     String[] hm1 = diff_halfMatchI(longtext, shorttext,
717                                    (longtext.length() + 3) / 4);
718     // Check again based on the third quarter.
719     String[] hm2 = diff_halfMatchI(longtext, shorttext,
720                                    (longtext.length() + 1) / 2);
721     String[] hm;
722     if (hm1 == null && hm2 == null) {
723       return null;
724     } else if (hm2 == null) {
725       hm = hm1;
726     } else if (hm1 == null) {
727       hm = hm2;
728     } else {
729       // Both matched.  Select the longest.
730       hm = hm1[4].length() > hm2[4].length() ? hm1 : hm2;
731     }
732 
733     // A half-match was found, sort out the return data.
734     if (text1.length() > text2.length()) {
735       return hm;
736       //return new String[]{hm[0], hm[1], hm[2], hm[3], hm[4]};
737     } else {
738       return new String[]{hm[2], hm[3], hm[0], hm[1], hm[4]};
739     }
740   }
741 
742 
743   /**
744    * Does a substring of shorttext exist within longtext such that the
745    * substring is at least half the length of longtext?
746    * @param longtext Longer string.
747    * @param shorttext Shorter string.
748    * @param i Start index of quarter length substring within longtext.
749    * @return Five element String array, containing the prefix of longtext, the
750    *     suffix of longtext, the prefix of shorttext, the suffix of shorttext
751    *     and the common middle.  Or null if there was no match.
752    */
diff_halfMatchI(String longtext, String shorttext, int i)753   private String[] diff_halfMatchI(String longtext, String shorttext, int i) {
754     // Start with a 1/4 length substring at position i as a seed.
755     String seed = longtext.substring(i, i + longtext.length() / 4);
756     int j = -1;
757     String best_common = "";
758     String best_longtext_a = "", best_longtext_b = "";
759     String best_shorttext_a = "", best_shorttext_b = "";
760     while ((j = shorttext.indexOf(seed, j + 1)) != -1) {
761       int prefixLength = diff_commonPrefix(longtext.substring(i),
762                                            shorttext.substring(j));
763       int suffixLength = diff_commonSuffix(longtext.substring(0, i),
764                                            shorttext.substring(0, j));
765       if (best_common.length() < suffixLength + prefixLength) {
766         best_common = shorttext.substring(j - suffixLength, j)
767             + shorttext.substring(j, j + prefixLength);
768         best_longtext_a = longtext.substring(0, i - suffixLength);
769         best_longtext_b = longtext.substring(i + prefixLength);
770         best_shorttext_a = shorttext.substring(0, j - suffixLength);
771         best_shorttext_b = shorttext.substring(j + prefixLength);
772       }
773     }
774     if (best_common.length() >= longtext.length() / 2) {
775       return new String[]{best_longtext_a, best_longtext_b,
776                           best_shorttext_a, best_shorttext_b, best_common};
777     } else {
778       return null;
779     }
780   }
781 
782 
783   /**
784    * Reduce the number of edits by eliminating semantically trivial equalities.
785    * @param diffs LinkedList of Diff objects.
786    */
diff_cleanupSemantic(LinkedList<Diff> diffs)787   public void diff_cleanupSemantic(LinkedList<Diff> diffs) {
788     if (diffs.isEmpty()) {
789       return;
790     }
791     boolean changes = false;
792     Stack<Diff> equalities = new Stack<Diff>();  // Stack of qualities.
793     String lastequality = null; // Always equal to equalities.lastElement().text
794     ListIterator<Diff> pointer = diffs.listIterator();
795     // Number of characters that changed prior to the equality.
796     int length_changes1 = 0;
797     // Number of characters that changed after the equality.
798     int length_changes2 = 0;
799     Diff thisDiff = pointer.next();
800     while (thisDiff != null) {
801       if (thisDiff.operation == Operation.EQUAL) {
802         // equality found
803         equalities.push(thisDiff);
804         length_changes1 = length_changes2;
805         length_changes2 = 0;
806         lastequality = thisDiff.text;
807       } else {
808         // an insertion or deletion
809         length_changes2 += thisDiff.text.length();
810         if (lastequality != null && (lastequality.length() <= length_changes1)
811             && (lastequality.length() <= length_changes2)) {
812           //System.out.println("Splitting: '" + lastequality + "'");
813           // Walk back to offending equality.
814           while (thisDiff != equalities.lastElement()) {
815             thisDiff = pointer.previous();
816           }
817           pointer.next();
818 
819           // Replace equality with a delete.
820           pointer.set(new Diff(Operation.DELETE, lastequality));
821           // Insert a corresponding an insert.
822           pointer.add(new Diff(Operation.INSERT, lastequality));
823 
824           equalities.pop();  // Throw away the equality we just deleted.
825           if (!equalities.empty()) {
826             // Throw away the previous equality (it needs to be reevaluated).
827             equalities.pop();
828           }
829           if (equalities.empty()) {
830             // There are no previous equalities, walk back to the start.
831             while (pointer.hasPrevious()) {
832               pointer.previous();
833             }
834           } else {
835             // There is a safe equality we can fall back to.
836             thisDiff = equalities.lastElement();
837             while (thisDiff != pointer.previous()) {
838               // Intentionally empty loop.
839             }
840           }
841 
842           length_changes1 = 0;  // Reset the counters.
843           length_changes2 = 0;
844           lastequality = null;
845           changes = true;
846         }
847       }
848       thisDiff = pointer.hasNext() ? pointer.next() : null;
849     }
850 
851     if (changes) {
852       diff_cleanupMerge(diffs);
853     }
854     diff_cleanupSemanticLossless(diffs);
855   }
856 
857 
858   /**
859    * Look for single edits surrounded on both sides by equalities
860    * which can be shifted sideways to align the edit to a word boundary.
861    * e.g: The c<ins>at c</ins>ame. -> The <ins>cat </ins>came.
862    * @param diffs LinkedList of Diff objects.
863    */
diff_cleanupSemanticLossless(LinkedList<Diff> diffs)864   public void diff_cleanupSemanticLossless(LinkedList<Diff> diffs) {
865     String equality1, edit, equality2;
866     String commonString;
867     int commonOffset;
868     int score, bestScore;
869     String bestEquality1, bestEdit, bestEquality2;
870     // Create a new iterator at the start.
871     ListIterator<Diff> pointer = diffs.listIterator();
872     Diff prevDiff = pointer.hasNext() ? pointer.next() : null;
873     Diff thisDiff = pointer.hasNext() ? pointer.next() : null;
874     Diff nextDiff = pointer.hasNext() ? pointer.next() : null;
875     // Intentionally ignore the first and last element (don't need checking).
876     while (nextDiff != null) {
877       if (prevDiff.operation == Operation.EQUAL &&
878           nextDiff.operation == Operation.EQUAL) {
879         // This is a single edit surrounded by equalities.
880         equality1 = prevDiff.text;
881         edit = thisDiff.text;
882         equality2 = nextDiff.text;
883 
884         // First, shift the edit as far left as possible.
885         commonOffset = diff_commonSuffix(equality1, edit);
886         if (commonOffset != 0) {
887           commonString = edit.substring(edit.length() - commonOffset);
888           equality1 = equality1.substring(0, equality1.length() - commonOffset);
889           edit = commonString + edit.substring(0, edit.length() - commonOffset);
890           equality2 = commonString + equality2;
891         }
892 
893         // Second, step character by character right, looking for the best fit.
894         bestEquality1 = equality1;
895         bestEdit = edit;
896         bestEquality2 = equality2;
897         bestScore = diff_cleanupSemanticScore(equality1, edit)
898             + diff_cleanupSemanticScore(edit, equality2);
899         while (edit.length() != 0 && equality2.length() != 0
900             && edit.charAt(0) == equality2.charAt(0)) {
901           equality1 += edit.charAt(0);
902           edit = edit.substring(1) + equality2.charAt(0);
903           equality2 = equality2.substring(1);
904           score = diff_cleanupSemanticScore(equality1, edit)
905               + diff_cleanupSemanticScore(edit, equality2);
906           // The >= encourages trailing rather than leading whitespace on edits.
907           if (score >= bestScore) {
908             bestScore = score;
909             bestEquality1 = equality1;
910             bestEdit = edit;
911             bestEquality2 = equality2;
912           }
913         }
914 
915         if (!prevDiff.text.equals(bestEquality1)) {
916           // We have an improvement, save it back to the diff.
917           if (bestEquality1.length() != 0) {
918             prevDiff.text = bestEquality1;
919           } else {
920             pointer.previous(); // Walk past nextDiff.
921             pointer.previous(); // Walk past thisDiff.
922             pointer.previous(); // Walk past prevDiff.
923             pointer.remove(); // Delete prevDiff.
924             pointer.next(); // Walk past thisDiff.
925             pointer.next(); // Walk past nextDiff.
926           }
927           thisDiff.text = bestEdit;
928           if (bestEquality2.length() != 0) {
929             nextDiff.text = bestEquality2;
930           } else {
931             pointer.remove(); // Delete nextDiff.
932             nextDiff = thisDiff;
933             thisDiff = prevDiff;
934           }
935         }
936       }
937       prevDiff = thisDiff;
938       thisDiff = nextDiff;
939       nextDiff = pointer.hasNext() ? pointer.next() : null;
940     }
941   }
942 
943 
944   /**
945    * Given two strings, compute a score representing whether the internal
946    * boundary falls on logical boundaries.
947    * Scores range from 5 (best) to 0 (worst).
948    * @param one First string.
949    * @param two Second string.
950    * @return The score.
951    */
diff_cleanupSemanticScore(String one, String two)952   private int diff_cleanupSemanticScore(String one, String two) {
953     if (one.length() == 0 || two.length() == 0) {
954       // Edges are the best.
955       return 5;
956     }
957 
958     // Each port of this function behaves slightly differently due to
959     // subtle differences in each language's definition of things like
960     // 'whitespace'.  Since this function's purpose is largely cosmetic,
961     // the choice has been made to use each language's native features
962     // rather than force total conformity.
963     int score = 0;
964     // One point for non-alphanumeric.
965     if (!Character.isLetterOrDigit(one.charAt(one.length() - 1))
966         || !Character.isLetterOrDigit(two.charAt(0))) {
967       score++;
968       // Two points for whitespace.
969       if (Character.isWhitespace(one.charAt(one.length() - 1))
970           || Character.isWhitespace(two.charAt(0))) {
971         score++;
972         // Three points for line breaks.
973         if (Character.getType(one.charAt(one.length() - 1)) == Character.CONTROL
974             || Character.getType(two.charAt(0)) == Character.CONTROL) {
975           score++;
976           // Four points for blank lines.
977           if (BLANKLINEEND.matcher(one).find()
978               || BLANKLINESTART.matcher(two).find()) {
979             score++;
980           }
981         }
982       }
983     }
984     return score;
985   }
986 
987 
988   private Pattern BLANKLINEEND
989       = Pattern.compile("\\n\\r?\\n\\Z", Pattern.DOTALL);
990   private Pattern BLANKLINESTART
991       = Pattern.compile("\\A\\r?\\n\\r?\\n", Pattern.DOTALL);
992 
993 
994   /**
995    * Reduce the number of edits by eliminating operationally trivial equalities.
996    * @param diffs LinkedList of Diff objects.
997    */
diff_cleanupEfficiency(LinkedList<Diff> diffs)998   public void diff_cleanupEfficiency(LinkedList<Diff> diffs) {
999     if (diffs.isEmpty()) {
1000       return;
1001     }
1002     boolean changes = false;
1003     Stack<Diff> equalities = new Stack<Diff>();  // Stack of equalities.
1004     String lastequality = null; // Always equal to equalities.lastElement().text
1005     ListIterator<Diff> pointer = diffs.listIterator();
1006     // Is there an insertion operation before the last equality.
1007     boolean pre_ins = false;
1008     // Is there a deletion operation before the last equality.
1009     boolean pre_del = false;
1010     // Is there an insertion operation after the last equality.
1011     boolean post_ins = false;
1012     // Is there a deletion operation after the last equality.
1013     boolean post_del = false;
1014     Diff thisDiff = pointer.next();
1015     Diff safeDiff = thisDiff;  // The last Diff that is known to be unsplitable.
1016     while (thisDiff != null) {
1017       if (thisDiff.operation == Operation.EQUAL) {
1018         // equality found
1019         if (thisDiff.text.length() < Diff_EditCost && (post_ins || post_del)) {
1020           // Candidate found.
1021           equalities.push(thisDiff);
1022           pre_ins = post_ins;
1023           pre_del = post_del;
1024           lastequality = thisDiff.text;
1025         } else {
1026           // Not a candidate, and can never become one.
1027           equalities.clear();
1028           lastequality = null;
1029           safeDiff = thisDiff;
1030         }
1031         post_ins = post_del = false;
1032       } else {
1033         // an insertion or deletion
1034         if (thisDiff.operation == Operation.DELETE) {
1035           post_del = true;
1036         } else {
1037           post_ins = true;
1038         }
1039         /*
1040          * Five types to be split:
1041          * <ins>A</ins><del>B</del>XY<ins>C</ins><del>D</del>
1042          * <ins>A</ins>X<ins>C</ins><del>D</del>
1043          * <ins>A</ins><del>B</del>X<ins>C</ins>
1044          * <ins>A</del>X<ins>C</ins><del>D</del>
1045          * <ins>A</ins><del>B</del>X<del>C</del>
1046          */
1047         if (lastequality != null
1048             && ((pre_ins && pre_del && post_ins && post_del)
1049                 || ((lastequality.length() < Diff_EditCost / 2)
1050                     && ((pre_ins ? 1 : 0) + (pre_del ? 1 : 0)
1051                         + (post_ins ? 1 : 0) + (post_del ? 1 : 0)) == 3))) {
1052           //System.out.println("Splitting: '" + lastequality + "'");
1053           // Walk back to offending equality.
1054           while (thisDiff != equalities.lastElement()) {
1055             thisDiff = pointer.previous();
1056           }
1057           pointer.next();
1058 
1059           // Replace equality with a delete.
1060           pointer.set(new Diff(Operation.DELETE, lastequality));
1061           // Insert a corresponding an insert.
1062           pointer.add(thisDiff = new Diff(Operation.INSERT, lastequality));
1063 
1064           equalities.pop();  // Throw away the equality we just deleted.
1065           lastequality = null;
1066           if (pre_ins && pre_del) {
1067             // No changes made which could affect previous entry, keep going.
1068             post_ins = post_del = true;
1069             equalities.clear();
1070             safeDiff = thisDiff;
1071           } else {
1072             if (!equalities.empty()) {
1073               // Throw away the previous equality (it needs to be reevaluated).
1074               equalities.pop();
1075             }
1076             if (equalities.empty()) {
1077               // There are no previous questionable equalities,
1078               // walk back to the last known safe diff.
1079               thisDiff = safeDiff;
1080             } else {
1081               // There is an equality we can fall back to.
1082               thisDiff = equalities.lastElement();
1083             }
1084             while (thisDiff != pointer.previous()) {
1085               // Intentionally empty loop.
1086             }
1087             post_ins = post_del = false;
1088           }
1089 
1090           changes = true;
1091         }
1092       }
1093       thisDiff = pointer.hasNext() ? pointer.next() : null;
1094     }
1095 
1096     if (changes) {
1097       diff_cleanupMerge(diffs);
1098     }
1099   }
1100 
1101 
1102   /**
1103    * Reorder and merge like edit sections.  Merge equalities.
1104    * Any edit section can move as long as it doesn't cross an equality.
1105    * @param diffs LinkedList of Diff objects.
1106    */
diff_cleanupMerge(LinkedList<Diff> diffs)1107   public void diff_cleanupMerge(LinkedList<Diff> diffs) {
1108     diffs.add(new Diff(Operation.EQUAL, ""));  // Add a dummy entry at the end.
1109     ListIterator<Diff> pointer = diffs.listIterator();
1110     int count_delete = 0;
1111     int count_insert = 0;
1112     String text_delete = "";
1113     String text_insert = "";
1114     Diff thisDiff = pointer.next();
1115     Diff prevEqual = null;
1116     int commonlength;
1117     while (thisDiff != null) {
1118       switch (thisDiff.operation) {
1119       case INSERT:
1120         count_insert++;
1121         text_insert += thisDiff.text;
1122         prevEqual = null;
1123         break;
1124       case DELETE:
1125         count_delete++;
1126         text_delete += thisDiff.text;
1127         prevEqual = null;
1128         break;
1129       case EQUAL:
1130         if (count_delete != 0 || count_insert != 0) {
1131           // Delete the offending records.
1132           pointer.previous();  // Reverse direction.
1133           while (count_delete-- > 0) {
1134             pointer.previous();
1135             pointer.remove();
1136           }
1137           while (count_insert-- > 0) {
1138             pointer.previous();
1139             pointer.remove();
1140           }
1141           if (count_delete != 0 && count_insert != 0) {
1142             // Factor out any common prefixies.
1143             commonlength = diff_commonPrefix(text_insert, text_delete);
1144             if (commonlength != 0) {
1145               if (pointer.hasPrevious()) {
1146                 thisDiff = pointer.previous();
1147                 assert thisDiff.operation == Operation.EQUAL
1148                        : "Previous diff should have been an equality.";
1149                 thisDiff.text += text_insert.substring(0, commonlength);
1150                 pointer.next();
1151               } else {
1152                 pointer.add(new Diff(Operation.EQUAL,
1153                     text_insert.substring(0, commonlength)));
1154               }
1155               text_insert = text_insert.substring(commonlength);
1156               text_delete = text_delete.substring(commonlength);
1157             }
1158             // Factor out any common suffixies.
1159             commonlength = diff_commonSuffix(text_insert, text_delete);
1160             if (commonlength != 0) {
1161               thisDiff = pointer.next();
1162               thisDiff.text = text_insert.substring(text_insert.length()
1163                   - commonlength) + thisDiff.text;
1164               text_insert = text_insert.substring(0, text_insert.length()
1165                   - commonlength);
1166               text_delete = text_delete.substring(0, text_delete.length()
1167                   - commonlength);
1168               pointer.previous();
1169             }
1170           }
1171           // Insert the merged records.
1172           if (text_delete.length() != 0) {
1173             pointer.add(new Diff(Operation.DELETE, text_delete));
1174           }
1175           if (text_insert.length() != 0) {
1176             pointer.add(new Diff(Operation.INSERT, text_insert));
1177           }
1178           // Step forward to the equality.
1179           thisDiff = pointer.hasNext() ? pointer.next() : null;
1180         } else if (prevEqual != null) {
1181           // Merge this equality with the previous one.
1182           prevEqual.text += thisDiff.text;
1183           pointer.remove();
1184           thisDiff = pointer.previous();
1185           pointer.next();  // Forward direction
1186         }
1187         count_insert = 0;
1188         count_delete = 0;
1189         text_delete = "";
1190         text_insert = "";
1191         prevEqual = thisDiff;
1192         break;
1193       }
1194       thisDiff = pointer.hasNext() ? pointer.next() : null;
1195     }
1196     // System.out.println(diff);
1197     if (diffs.getLast().text.length() == 0) {
1198       diffs.removeLast();  // Remove the dummy entry at the end.
1199     }
1200 
1201     /*
1202      * Second pass: look for single edits surrounded on both sides by equalities
1203      * which can be shifted sideways to eliminate an equality.
1204      * e.g: A<ins>BA</ins>C -> <ins>AB</ins>AC
1205      */
1206     boolean changes = false;
1207     // Create a new iterator at the start.
1208     // (As opposed to walking the current one back.)
1209     pointer = diffs.listIterator();
1210     Diff prevDiff = pointer.hasNext() ? pointer.next() : null;
1211     thisDiff = pointer.hasNext() ? pointer.next() : null;
1212     Diff nextDiff = pointer.hasNext() ? pointer.next() : null;
1213     // Intentionally ignore the first and last element (don't need checking).
1214     while (nextDiff != null) {
1215       if (prevDiff.operation == Operation.EQUAL &&
1216           nextDiff.operation == Operation.EQUAL) {
1217         // This is a single edit surrounded by equalities.
1218         if (thisDiff.text.endsWith(prevDiff.text)) {
1219           // Shift the edit over the previous equality.
1220           thisDiff.text = prevDiff.text
1221               + thisDiff.text.substring(0, thisDiff.text.length()
1222                                            - prevDiff.text.length());
1223           nextDiff.text = prevDiff.text + nextDiff.text;
1224           pointer.previous(); // Walk past nextDiff.
1225           pointer.previous(); // Walk past thisDiff.
1226           pointer.previous(); // Walk past prevDiff.
1227           pointer.remove(); // Delete prevDiff.
1228           pointer.next(); // Walk past thisDiff.
1229           thisDiff = pointer.next(); // Walk past nextDiff.
1230           nextDiff = pointer.hasNext() ? pointer.next() : null;
1231           changes = true;
1232         } else if (thisDiff.text.startsWith(nextDiff.text)) {
1233           // Shift the edit over the next equality.
1234           prevDiff.text += nextDiff.text;
1235           thisDiff.text = thisDiff.text.substring(nextDiff.text.length())
1236               + nextDiff.text;
1237           pointer.remove(); // Delete nextDiff.
1238           nextDiff = pointer.hasNext() ? pointer.next() : null;
1239           changes = true;
1240         }
1241       }
1242       prevDiff = thisDiff;
1243       thisDiff = nextDiff;
1244       nextDiff = pointer.hasNext() ? pointer.next() : null;
1245     }
1246     // If shifts were made, the diff needs reordering and another shift sweep.
1247     if (changes) {
1248       diff_cleanupMerge(diffs);
1249     }
1250   }
1251 
1252 
1253   /**
1254    * loc is a location in text1, compute and return the equivalent location in
1255    * text2.
1256    * e.g. "The cat" vs "The big cat", 1->1, 5->8
1257    * @param diffs LinkedList of Diff objects.
1258    * @param loc Location within text1.
1259    * @return Location within text2.
1260    */
diff_xIndex(LinkedList<Diff> diffs, int loc)1261   public int diff_xIndex(LinkedList<Diff> diffs, int loc) {
1262     int chars1 = 0;
1263     int chars2 = 0;
1264     int last_chars1 = 0;
1265     int last_chars2 = 0;
1266     Diff lastDiff = null;
1267     for (Diff aDiff : diffs) {
1268       if (aDiff.operation != Operation.INSERT) {
1269         // Equality or deletion.
1270         chars1 += aDiff.text.length();
1271       }
1272       if (aDiff.operation != Operation.DELETE) {
1273         // Equality or insertion.
1274         chars2 += aDiff.text.length();
1275       }
1276       if (chars1 > loc) {
1277         // Overshot the location.
1278         lastDiff = aDiff;
1279         break;
1280       }
1281       last_chars1 = chars1;
1282       last_chars2 = chars2;
1283     }
1284     if (lastDiff != null && lastDiff.operation == Operation.DELETE) {
1285       // The location was deleted.
1286       return last_chars2;
1287     }
1288     // Add the remaining character length.
1289     return last_chars2 + (loc - last_chars1);
1290   }
1291 
1292 
1293   /**
1294    * Convert a Diff list into a pretty HTML report.
1295    * @param diffs LinkedList of Diff objects.
1296    * @return HTML representation.
1297    */
diff_prettyHtml(LinkedList<Diff> diffs)1298   public String diff_prettyHtml(LinkedList<Diff> diffs) {
1299     StringBuilder html = new StringBuilder();
1300     int i = 0;
1301     for (Diff aDiff : diffs) {
1302       String text = aDiff.text.replace("&", "&amp;").replace("<", "&lt;")
1303           .replace(">", "&gt;").replace("\n", "&para;<BR>");
1304       switch (aDiff.operation) {
1305       case INSERT:
1306         html.append("<INS STYLE=\"background:#E6FFE6;\" TITLE=\"i=").append(i)
1307             .append("\">").append(text).append("</INS>");
1308         break;
1309       case DELETE:
1310         html.append("<DEL STYLE=\"background:#FFE6E6;\" TITLE=\"i=").append(i)
1311             .append("\">").append(text).append("</DEL>");
1312         break;
1313       case EQUAL:
1314         html.append("<SPAN TITLE=\"i=").append(i).append("\">").append(text)
1315             .append("</SPAN>");
1316         break;
1317       }
1318       if (aDiff.operation != Operation.DELETE) {
1319         i += aDiff.text.length();
1320       }
1321     }
1322     return html.toString();
1323   }
1324 
1325 
1326   /**
1327    * Compute and return the source text (all equalities and deletions).
1328    * @param diffs LinkedList of Diff objects.
1329    * @return Source text.
1330    */
diff_text1(LinkedList<Diff> diffs)1331   public String diff_text1(LinkedList<Diff> diffs) {
1332     StringBuilder text = new StringBuilder();
1333     for (Diff aDiff : diffs) {
1334       if (aDiff.operation != Operation.INSERT) {
1335         text.append(aDiff.text);
1336       }
1337     }
1338     return text.toString();
1339   }
1340 
1341 
1342   /**
1343    * Compute and return the destination text (all equalities and insertions).
1344    * @param diffs LinkedList of Diff objects.
1345    * @return Destination text.
1346    */
diff_text2(LinkedList<Diff> diffs)1347   public String diff_text2(LinkedList<Diff> diffs) {
1348     StringBuilder text = new StringBuilder();
1349     for (Diff aDiff : diffs) {
1350       if (aDiff.operation != Operation.DELETE) {
1351         text.append(aDiff.text);
1352       }
1353     }
1354     return text.toString();
1355   }
1356 
1357 
1358   /**
1359    * Compute the Levenshtein distance; the number of inserted, deleted or
1360    * substituted characters.
1361    * @param diffs LinkedList of Diff objects.
1362    * @return Number of changes.
1363    */
diff_levenshtein(LinkedList<Diff> diffs)1364   public int diff_levenshtein(LinkedList<Diff> diffs) {
1365     int levenshtein = 0;
1366     int insertions = 0;
1367     int deletions = 0;
1368     for (Diff aDiff : diffs) {
1369       switch (aDiff.operation) {
1370       case INSERT:
1371         insertions += aDiff.text.length();
1372         break;
1373       case DELETE:
1374         deletions += aDiff.text.length();
1375         break;
1376       case EQUAL:
1377         // A deletion and an insertion is one substitution.
1378         levenshtein += Math.max(insertions, deletions);
1379         insertions = 0;
1380         deletions = 0;
1381         break;
1382       }
1383     }
1384     levenshtein += Math.max(insertions, deletions);
1385     return levenshtein;
1386   }
1387 
1388 
1389   /**
1390    * Crush the diff into an encoded string which describes the operations
1391    * required to transform text1 into text2.
1392    * E.g. =3\t-2\t+ing  -> Keep 3 chars, delete 2 chars, insert 'ing'.
1393    * Operations are tab-separated.  Inserted text is escaped using %xx notation.
1394    * @param diffs Array of diff tuples.
1395    * @return Delta text.
1396    */
diff_toDelta(LinkedList<Diff> diffs)1397   public String diff_toDelta(LinkedList<Diff> diffs) {
1398     StringBuilder text = new StringBuilder();
1399     for (Diff aDiff : diffs) {
1400       switch (aDiff.operation) {
1401       case INSERT:
1402         try {
1403           text.append("+").append(URLEncoder.encode(aDiff.text, "UTF-8")
1404                                             .replace('+', ' ')).append("\t");
1405         } catch (UnsupportedEncodingException e) {
1406           // Not likely on modern system.
1407           throw new Error("This system does not support UTF-8.", e);
1408         }
1409         break;
1410       case DELETE:
1411         text.append("-").append(aDiff.text.length()).append("\t");
1412         break;
1413       case EQUAL:
1414         text.append("=").append(aDiff.text.length()).append("\t");
1415         break;
1416       }
1417     }
1418     String delta = text.toString();
1419     if (delta.length() != 0) {
1420       // Strip off trailing tab character.
1421       delta = delta.substring(0, delta.length() - 1);
1422       delta = unescapeForEncodeUriCompatability(delta);
1423     }
1424     return delta;
1425   }
1426 
1427 
1428   /**
1429    * Given the original text1, and an encoded string which describes the
1430    * operations required to transform text1 into text2, compute the full diff.
1431    * @param text1 Source string for the diff.
1432    * @param delta Delta text.
1433    * @return Array of diff tuples or null if invalid.
1434    * @throws IllegalArgumentException If invalid input.
1435    */
diff_fromDelta(String text1, String delta)1436   public LinkedList<Diff> diff_fromDelta(String text1, String delta)
1437       throws IllegalArgumentException {
1438     LinkedList<Diff> diffs = new LinkedList<Diff>();
1439     int pointer = 0;  // Cursor in text1
1440     String[] tokens = delta.split("\t");
1441     for (String token : tokens) {
1442       if (token.length() == 0) {
1443         // Blank tokens are ok (from a trailing \t).
1444         continue;
1445       }
1446       // Each token begins with a one character parameter which specifies the
1447       // operation of this token (delete, insert, equality).
1448       String param = token.substring(1);
1449       switch (token.charAt(0)) {
1450       case '+':
1451         // decode would change all "+" to " "
1452         param = param.replace("+", "%2B");
1453         try {
1454           param = URLDecoder.decode(param, "UTF-8");
1455         } catch (UnsupportedEncodingException e) {
1456           // Not likely on modern system.
1457           throw new Error("This system does not support UTF-8.", e);
1458         } catch (IllegalArgumentException e) {
1459           // Malformed URI sequence.
1460           throw new IllegalArgumentException(
1461               "Illegal escape in diff_fromDelta: " + param, e);
1462         }
1463         diffs.add(new Diff(Operation.INSERT, param));
1464         break;
1465       case '-':
1466         // Fall through.
1467       case '=':
1468         int n;
1469         try {
1470           n = Integer.parseInt(param);
1471         } catch (NumberFormatException e) {
1472           throw new IllegalArgumentException(
1473               "Invalid number in diff_fromDelta: " + param, e);
1474         }
1475         if (n < 0) {
1476           throw new IllegalArgumentException(
1477               "Negative number in diff_fromDelta: " + param);
1478         }
1479         String text;
1480         try {
1481           text = text1.substring(pointer, pointer += n);
1482         } catch (StringIndexOutOfBoundsException e) {
1483           throw new IllegalArgumentException("Delta length (" + pointer
1484               + ") larger than source text length (" + text1.length()
1485               + ").", e);
1486         }
1487         if (token.charAt(0) == '=') {
1488           diffs.add(new Diff(Operation.EQUAL, text));
1489         } else {
1490           diffs.add(new Diff(Operation.DELETE, text));
1491         }
1492         break;
1493       default:
1494         // Anything else is an error.
1495         throw new IllegalArgumentException(
1496             "Invalid diff operation in diff_fromDelta: " + token.charAt(0));
1497       }
1498     }
1499     if (pointer != text1.length()) {
1500       throw new IllegalArgumentException("Delta length (" + pointer
1501           + ") smaller than source text length (" + text1.length() + ").");
1502     }
1503     return diffs;
1504   }
1505 
1506 
1507   //  MATCH FUNCTIONS
1508 
1509 
1510   /**
1511    * Locate the best instance of 'pattern' in 'text' near 'loc'.
1512    * Returns -1 if no match found.
1513    * @param text The text to search.
1514    * @param pattern The pattern to search for.
1515    * @param loc The location to search around.
1516    * @return Best match index or -1.
1517    */
match_main(String text, String pattern, int loc)1518   public int match_main(String text, String pattern, int loc) {
1519     // Check for null inputs.
1520     if (text == null || pattern == null) {
1521       throw new IllegalArgumentException("Null inputs. (match_main)");
1522     }
1523 
1524     loc = Math.max(0, Math.min(loc, text.length()));
1525     if (text.equals(pattern)) {
1526       // Shortcut (potentially not guaranteed by the algorithm)
1527       return 0;
1528     } else if (text.length() == 0) {
1529       // Nothing to match.
1530       return -1;
1531     } else if (loc + pattern.length() <= text.length()
1532         && text.substring(loc, loc + pattern.length()).equals(pattern)) {
1533       // Perfect match at the perfect spot!  (Includes case of null pattern)
1534       return loc;
1535     } else {
1536       // Do a fuzzy compare.
1537       return match_bitap(text, pattern, loc);
1538     }
1539   }
1540 
1541 
1542   /**
1543    * Locate the best instance of 'pattern' in 'text' near 'loc' using the
1544    * Bitap algorithm.  Returns -1 if no match found.
1545    * @param text The text to search.
1546    * @param pattern The pattern to search for.
1547    * @param loc The location to search around.
1548    * @return Best match index or -1.
1549    */
match_bitap(String text, String pattern, int loc)1550   protected int match_bitap(String text, String pattern, int loc) {
1551     assert (Match_MaxBits == 0 || pattern.length() <= Match_MaxBits)
1552         : "Pattern too long for this application.";
1553 
1554     // Initialise the alphabet.
1555     Map<Character, Integer> s = match_alphabet(pattern);
1556 
1557     // Highest score beyond which we give up.
1558     double score_threshold = Match_Threshold;
1559     // Is there a nearby exact match? (speedup)
1560     int best_loc = text.indexOf(pattern, loc);
1561     if (best_loc != -1) {
1562       score_threshold = Math.min(match_bitapScore(0, best_loc, loc, pattern),
1563           score_threshold);
1564       // What about in the other direction? (speedup)
1565       best_loc = text.lastIndexOf(pattern, loc + pattern.length());
1566       if (best_loc != -1) {
1567         score_threshold = Math.min(match_bitapScore(0, best_loc, loc, pattern),
1568             score_threshold);
1569       }
1570     }
1571 
1572     // Initialise the bit arrays.
1573     int matchmask = 1 << (pattern.length() - 1);
1574     best_loc = -1;
1575 
1576     int bin_min, bin_mid;
1577     int bin_max = pattern.length() + text.length();
1578     // Empty initialization added to appease Java compiler.
1579     int[] last_rd = new int[0];
1580     for (int d = 0; d < pattern.length(); d++) {
1581       // Scan for the best match; each iteration allows for one more error.
1582       // Run a binary search to determine how far from 'loc' we can stray at
1583       // this error level.
1584       bin_min = 0;
1585       bin_mid = bin_max;
1586       while (bin_min < bin_mid) {
1587         if (match_bitapScore(d, loc + bin_mid, loc, pattern)
1588             <= score_threshold) {
1589           bin_min = bin_mid;
1590         } else {
1591           bin_max = bin_mid;
1592         }
1593         bin_mid = (bin_max - bin_min) / 2 + bin_min;
1594       }
1595       // Use the result from this iteration as the maximum for the next.
1596       bin_max = bin_mid;
1597       int start = Math.max(1, loc - bin_mid + 1);
1598       int finish = Math.min(loc + bin_mid, text.length()) + pattern.length();
1599 
1600       int[] rd = new int[finish + 2];
1601       rd[finish + 1] = (1 << d) - 1;
1602       for (int j = finish; j >= start; j--) {
1603         int charMatch;
1604         if (text.length() <= j - 1 || !s.containsKey(text.charAt(j - 1))) {
1605           // Out of range.
1606           charMatch = 0;
1607         } else {
1608           charMatch = s.get(text.charAt(j - 1));
1609         }
1610         if (d == 0) {
1611           // First pass: exact match.
1612           rd[j] = ((rd[j + 1] << 1) | 1) & charMatch;
1613         } else {
1614           // Subsequent passes: fuzzy match.
1615           rd[j] = ((rd[j + 1] << 1) | 1) & charMatch
1616               | (((last_rd[j + 1] | last_rd[j]) << 1) | 1) | last_rd[j + 1];
1617         }
1618         if ((rd[j] & matchmask) != 0) {
1619           double score = match_bitapScore(d, j - 1, loc, pattern);
1620           // This match will almost certainly be better than any existing
1621           // match.  But check anyway.
1622           if (score <= score_threshold) {
1623             // Told you so.
1624             score_threshold = score;
1625             best_loc = j - 1;
1626             if (best_loc > loc) {
1627               // When passing loc, don't exceed our current distance from loc.
1628               start = Math.max(1, 2 * loc - best_loc);
1629             } else {
1630               // Already passed loc, downhill from here on in.
1631               break;
1632             }
1633           }
1634         }
1635       }
1636       if (match_bitapScore(d + 1, loc, loc, pattern) > score_threshold) {
1637         // No hope for a (better) match at greater error levels.
1638         break;
1639       }
1640       last_rd = rd;
1641     }
1642     return best_loc;
1643   }
1644 
1645 
1646   /**
1647    * Compute and return the score for a match with e errors and x location.
1648    * @param e Number of errors in match.
1649    * @param x Location of match.
1650    * @param loc Expected location of match.
1651    * @param pattern Pattern being sought.
1652    * @return Overall score for match (0.0 = good, 1.0 = bad).
1653    */
match_bitapScore(int e, int x, int loc, String pattern)1654   private double match_bitapScore(int e, int x, int loc, String pattern) {
1655     float accuracy = (float) e / pattern.length();
1656     int proximity = Math.abs(loc - x);
1657     if (Match_Distance == 0) {
1658       // Dodge divide by zero error.
1659       return proximity == 0 ? accuracy : 1.0;
1660     }
1661     return accuracy + (proximity / (float) Match_Distance);
1662   }
1663 
1664 
1665   /**
1666    * Initialise the alphabet for the Bitap algorithm.
1667    * @param pattern The text to encode.
1668    * @return Hash of character locations.
1669    */
match_alphabet(String pattern)1670   protected Map<Character, Integer> match_alphabet(String pattern) {
1671     Map<Character, Integer> s = new HashMap<Character, Integer>();
1672     char[] char_pattern = pattern.toCharArray();
1673     for (char c : char_pattern) {
1674       s.put(c, 0);
1675     }
1676     int i = 0;
1677     for (char c : char_pattern) {
1678       s.put(c, s.get(c) | (1 << (pattern.length() - i - 1)));
1679       i++;
1680     }
1681     return s;
1682   }
1683 
1684 
1685   //  PATCH FUNCTIONS
1686 
1687 
1688   /**
1689    * Increase the context until it is unique,
1690    * but don't let the pattern expand beyond Match_MaxBits.
1691    * @param patch The patch to grow.
1692    * @param text Source text.
1693    */
patch_addContext(Patch patch, String text)1694   protected void patch_addContext(Patch patch, String text) {
1695     if (text.length() == 0) {
1696       return;
1697     }
1698     String pattern = text.substring(patch.start2, patch.start2 + patch.length1);
1699     int padding = 0;
1700 
1701     // Look for the first and last matches of pattern in text.  If two different
1702     // matches are found, increase the pattern length.
1703     while (text.indexOf(pattern) != text.lastIndexOf(pattern)
1704         && pattern.length() < Match_MaxBits - Patch_Margin - Patch_Margin) {
1705       padding += Patch_Margin;
1706       pattern = text.substring(Math.max(0, patch.start2 - padding),
1707           Math.min(text.length(), patch.start2 + patch.length1 + padding));
1708     }
1709     // Add one chunk for good luck.
1710     padding += Patch_Margin;
1711 
1712     // Add the prefix.
1713     String prefix = text.substring(Math.max(0, patch.start2 - padding),
1714         patch.start2);
1715     if (prefix.length() != 0) {
1716       patch.diffs.addFirst(new Diff(Operation.EQUAL, prefix));
1717     }
1718     // Add the suffix.
1719     String suffix = text.substring(patch.start2 + patch.length1,
1720         Math.min(text.length(), patch.start2 + patch.length1 + padding));
1721     if (suffix.length() != 0) {
1722       patch.diffs.addLast(new Diff(Operation.EQUAL, suffix));
1723     }
1724 
1725     // Roll back the start points.
1726     patch.start1 -= prefix.length();
1727     patch.start2 -= prefix.length();
1728     // Extend the lengths.
1729     patch.length1 += prefix.length() + suffix.length();
1730     patch.length2 += prefix.length() + suffix.length();
1731   }
1732 
1733 
1734   /**
1735    * Compute a list of patches to turn text1 into text2.
1736    * A set of diffs will be computed.
1737    * @param text1 Old text.
1738    * @param text2 New text.
1739    * @return LinkedList of Patch objects.
1740    */
patch_make(String text1, String text2)1741   public LinkedList<Patch> patch_make(String text1, String text2) {
1742     if (text1 == null || text2 == null) {
1743       throw new IllegalArgumentException("Null inputs. (patch_make)");
1744     }
1745     // No diffs provided, compute our own.
1746     LinkedList<Diff> diffs = diff_main(text1, text2, true);
1747     if (diffs.size() > 2) {
1748       diff_cleanupSemantic(diffs);
1749       diff_cleanupEfficiency(diffs);
1750     }
1751     return patch_make(text1, diffs);
1752   }
1753 
1754 
1755   /**
1756    * Compute a list of patches to turn text1 into text2.
1757    * text1 will be derived from the provided diffs.
1758    * @param diffs Array of diff tuples for text1 to text2.
1759    * @return LinkedList of Patch objects.
1760    */
patch_make(LinkedList<Diff> diffs)1761   public LinkedList<Patch> patch_make(LinkedList<Diff> diffs) {
1762     if (diffs == null) {
1763       throw new IllegalArgumentException("Null inputs. (patch_make)");
1764     }
1765     // No origin string provided, compute our own.
1766     String text1 = diff_text1(diffs);
1767     return patch_make(text1, diffs);
1768   }
1769 
1770 
1771   /**
1772    * Compute a list of patches to turn text1 into text2.
1773    * text2 is ignored, diffs are the delta between text1 and text2.
1774    * @param text1 Old text
1775    * @param text2 Ignored.
1776    * @param diffs Array of diff tuples for text1 to text2.
1777    * @return LinkedList of Patch objects.
1778    * @deprecated Prefer patch_make(String text1, LinkedList<Diff> diffs).
1779    */
patch_make(String text1, String text2, LinkedList<Diff> diffs)1780   public LinkedList<Patch> patch_make(String text1, String text2,
1781       LinkedList<Diff> diffs) {
1782     return patch_make(text1, diffs);
1783   }
1784 
1785 
1786   /**
1787    * Compute a list of patches to turn text1 into text2.
1788    * text2 is not provided, diffs are the delta between text1 and text2.
1789    * @param text1 Old text.
1790    * @param diffs Array of diff tuples for text1 to text2.
1791    * @return LinkedList of Patch objects.
1792    */
patch_make(String text1, LinkedList<Diff> diffs)1793   public LinkedList<Patch> patch_make(String text1, LinkedList<Diff> diffs) {
1794     if (text1 == null || diffs == null) {
1795       throw new IllegalArgumentException("Null inputs. (patch_make)");
1796     }
1797 
1798     LinkedList<Patch> patches = new LinkedList<Patch>();
1799     if (diffs.isEmpty()) {
1800       return patches;  // Get rid of the null case.
1801     }
1802     Patch patch = new Patch();
1803     int char_count1 = 0;  // Number of characters into the text1 string.
1804     int char_count2 = 0;  // Number of characters into the text2 string.
1805     // Start with text1 (prepatch_text) and apply the diffs until we arrive at
1806     // text2 (postpatch_text). We recreate the patches one by one to determine
1807     // context info.
1808     String prepatch_text = text1;
1809     String postpatch_text = text1;
1810     for (Diff aDiff : diffs) {
1811       if (patch.diffs.isEmpty() && aDiff.operation != Operation.EQUAL) {
1812         // A new patch starts here.
1813         patch.start1 = char_count1;
1814         patch.start2 = char_count2;
1815       }
1816 
1817       switch (aDiff.operation) {
1818       case INSERT:
1819         patch.diffs.add(aDiff);
1820         patch.length2 += aDiff.text.length();
1821         postpatch_text = postpatch_text.substring(0, char_count2)
1822             + aDiff.text + postpatch_text.substring(char_count2);
1823         break;
1824       case DELETE:
1825         patch.length1 += aDiff.text.length();
1826         patch.diffs.add(aDiff);
1827         postpatch_text = postpatch_text.substring(0, char_count2)
1828             + postpatch_text.substring(char_count2 + aDiff.text.length());
1829         break;
1830       case EQUAL:
1831         if (aDiff.text.length() <= 2 * Patch_Margin
1832             && !patch.diffs.isEmpty() && aDiff != diffs.getLast()) {
1833           // Small equality inside a patch.
1834           patch.diffs.add(aDiff);
1835           patch.length1 += aDiff.text.length();
1836           patch.length2 += aDiff.text.length();
1837         }
1838 
1839         if (aDiff.text.length() >= 2 * Patch_Margin) {
1840           // Time for a new patch.
1841           if (!patch.diffs.isEmpty()) {
1842             patch_addContext(patch, prepatch_text);
1843             patches.add(patch);
1844             patch = new Patch();
1845             // Unlike Unidiff, our patch lists have a rolling context.
1846             // http://code.google.com/p/google-diff-match-patch/wiki/Unidiff
1847             // Update prepatch text & pos to reflect the application of the
1848             // just completed patch.
1849             prepatch_text = postpatch_text;
1850             char_count1 = char_count2;
1851           }
1852         }
1853         break;
1854       }
1855 
1856       // Update the current character count.
1857       if (aDiff.operation != Operation.INSERT) {
1858         char_count1 += aDiff.text.length();
1859       }
1860       if (aDiff.operation != Operation.DELETE) {
1861         char_count2 += aDiff.text.length();
1862       }
1863     }
1864     // Pick up the leftover patch if not empty.
1865     if (!patch.diffs.isEmpty()) {
1866       patch_addContext(patch, prepatch_text);
1867       patches.add(patch);
1868     }
1869 
1870     return patches;
1871   }
1872 
1873 
1874   /**
1875    * Given an array of patches, return another array that is identical.
1876    * @param patches Array of patch objects.
1877    * @return Array of patch objects.
1878    */
patch_deepCopy(LinkedList<Patch> patches)1879   public LinkedList<Patch> patch_deepCopy(LinkedList<Patch> patches) {
1880     LinkedList<Patch> patchesCopy = new LinkedList<Patch>();
1881     for (Patch aPatch : patches) {
1882       Patch patchCopy = new Patch();
1883       for (Diff aDiff : aPatch.diffs) {
1884         Diff diffCopy = new Diff(aDiff.operation, aDiff.text);
1885         patchCopy.diffs.add(diffCopy);
1886       }
1887       patchCopy.start1 = aPatch.start1;
1888       patchCopy.start2 = aPatch.start2;
1889       patchCopy.length1 = aPatch.length1;
1890       patchCopy.length2 = aPatch.length2;
1891       patchesCopy.add(patchCopy);
1892     }
1893     return patchesCopy;
1894   }
1895 
1896 
1897   /**
1898    * Merge a set of patches onto the text.  Return a patched text, as well
1899    * as an array of true/false values indicating which patches were applied.
1900    * @param patches Array of patch objects
1901    * @param text Old text.
1902    * @return Two element Object array, containing the new text and an array of
1903    *      boolean values.
1904    */
patch_apply(LinkedList<Patch> patches, String text)1905   public Object[] patch_apply(LinkedList<Patch> patches, String text) {
1906     if (patches.isEmpty()) {
1907       return new Object[]{text, new boolean[0]};
1908     }
1909 
1910     // Deep copy the patches so that no changes are made to originals.
1911     patches = patch_deepCopy(patches);
1912 
1913     String nullPadding = patch_addPadding(patches);
1914     text = nullPadding + text + nullPadding;
1915     patch_splitMax(patches);
1916 
1917     int x = 0;
1918     // delta keeps track of the offset between the expected and actual location
1919     // of the previous patch.  If there are patches expected at positions 10 and
1920     // 20, but the first patch was found at 12, delta is 2 and the second patch
1921     // has an effective expected position of 22.
1922     int delta = 0;
1923     boolean[] results = new boolean[patches.size()];
1924     for (Patch aPatch : patches) {
1925       int expected_loc = aPatch.start2 + delta;
1926       String text1 = diff_text1(aPatch.diffs);
1927       int start_loc;
1928       int end_loc = -1;
1929       if (text1.length() > this.Match_MaxBits) {
1930         // patch_splitMax will only provide an oversized pattern in the case of
1931         // a monster delete.
1932         start_loc = match_main(text,
1933             text1.substring(0, this.Match_MaxBits), expected_loc);
1934         if (start_loc != -1) {
1935           end_loc = match_main(text,
1936               text1.substring(text1.length() - this.Match_MaxBits),
1937               expected_loc + text1.length() - this.Match_MaxBits);
1938           if (end_loc == -1 || start_loc >= end_loc) {
1939             // Can't find valid trailing context.  Drop this patch.
1940             start_loc = -1;
1941           }
1942         }
1943       } else {
1944         start_loc = match_main(text, text1, expected_loc);
1945       }
1946       if (start_loc == -1) {
1947         // No match found.  :(
1948         results[x] = false;
1949         // Subtract the delta for this failed patch from subsequent patches.
1950         delta -= aPatch.length2 - aPatch.length1;
1951       } else {
1952         // Found a match.  :)
1953         results[x] = true;
1954         delta = start_loc - expected_loc;
1955         String text2;
1956         if (end_loc == -1) {
1957           text2 = text.substring(start_loc,
1958               Math.min(start_loc + text1.length(), text.length()));
1959         } else {
1960           text2 = text.substring(start_loc,
1961               Math.min(end_loc + this.Match_MaxBits, text.length()));
1962         }
1963         if (text1.equals(text2)) {
1964           // Perfect match, just shove the replacement text in.
1965           text = text.substring(0, start_loc) + diff_text2(aPatch.diffs)
1966               + text.substring(start_loc + text1.length());
1967         } else {
1968           // Imperfect match.  Run a diff to get a framework of equivalent
1969           // indices.
1970           LinkedList<Diff> diffs = diff_main(text1, text2, false);
1971           if (text1.length() > this.Match_MaxBits
1972               && diff_levenshtein(diffs) / (float) text1.length()
1973               > this.Patch_DeleteThreshold) {
1974             // The end points match, but the content is unacceptably bad.
1975             results[x] = false;
1976           } else {
1977             diff_cleanupSemanticLossless(diffs);
1978             int index1 = 0;
1979             for (Diff aDiff : aPatch.diffs) {
1980               if (aDiff.operation != Operation.EQUAL) {
1981                 int index2 = diff_xIndex(diffs, index1);
1982                 if (aDiff.operation == Operation.INSERT) {
1983                   // Insertion
1984                   text = text.substring(0, start_loc + index2) + aDiff.text
1985                       + text.substring(start_loc + index2);
1986                 } else if (aDiff.operation == Operation.DELETE) {
1987                   // Deletion
1988                   text = text.substring(0, start_loc + index2)
1989                       + text.substring(start_loc + diff_xIndex(diffs,
1990                       index1 + aDiff.text.length()));
1991                 }
1992               }
1993               if (aDiff.operation != Operation.DELETE) {
1994                 index1 += aDiff.text.length();
1995               }
1996             }
1997           }
1998         }
1999       }
2000       x++;
2001     }
2002     // Strip the padding off.
2003     text = text.substring(nullPadding.length(), text.length()
2004         - nullPadding.length());
2005     return new Object[]{text, results};
2006   }
2007 
2008 
2009   /**
2010    * Add some padding on text start and end so that edges can match something.
2011    * Intended to be called only from within patch_apply.
2012    * @param patches Array of patch objects.
2013    * @return The padding string added to each side.
2014    */
patch_addPadding(LinkedList<Patch> patches)2015   public String patch_addPadding(LinkedList<Patch> patches) {
2016     int paddingLength = this.Patch_Margin;
2017     String nullPadding = "";
2018     for (int x = 1; x <= paddingLength; x++) {
2019       nullPadding += String.valueOf((char) x);
2020     }
2021 
2022     // Bump all the patches forward.
2023     for (Patch aPatch : patches) {
2024       aPatch.start1 += paddingLength;
2025       aPatch.start2 += paddingLength;
2026     }
2027 
2028     // Add some padding on start of first diff.
2029     Patch patch = patches.getFirst();
2030     LinkedList<Diff> diffs = patch.diffs;
2031     if (diffs.isEmpty() || diffs.getFirst().operation != Operation.EQUAL) {
2032       // Add nullPadding equality.
2033       diffs.addFirst(new Diff(Operation.EQUAL, nullPadding));
2034       patch.start1 -= paddingLength;  // Should be 0.
2035       patch.start2 -= paddingLength;  // Should be 0.
2036       patch.length1 += paddingLength;
2037       patch.length2 += paddingLength;
2038     } else if (paddingLength > diffs.getFirst().text.length()) {
2039       // Grow first equality.
2040       Diff firstDiff = diffs.getFirst();
2041       int extraLength = paddingLength - firstDiff.text.length();
2042       firstDiff.text = nullPadding.substring(firstDiff.text.length())
2043           + firstDiff.text;
2044       patch.start1 -= extraLength;
2045       patch.start2 -= extraLength;
2046       patch.length1 += extraLength;
2047       patch.length2 += extraLength;
2048     }
2049 
2050     // Add some padding on end of last diff.
2051     patch = patches.getLast();
2052     diffs = patch.diffs;
2053     if (diffs.isEmpty() || diffs.getLast().operation != Operation.EQUAL) {
2054       // Add nullPadding equality.
2055       diffs.addLast(new Diff(Operation.EQUAL, nullPadding));
2056       patch.length1 += paddingLength;
2057       patch.length2 += paddingLength;
2058     } else if (paddingLength > diffs.getLast().text.length()) {
2059       // Grow last equality.
2060       Diff lastDiff = diffs.getLast();
2061       int extraLength = paddingLength - lastDiff.text.length();
2062       lastDiff.text += nullPadding.substring(0, extraLength);
2063       patch.length1 += extraLength;
2064       patch.length2 += extraLength;
2065     }
2066 
2067     return nullPadding;
2068   }
2069 
2070 
2071   /**
2072    * Look through the patches and break up any which are longer than the
2073    * maximum limit of the match algorithm.
2074    * @param patches LinkedList of Patch objects.
2075    */
patch_splitMax(LinkedList<Patch> patches)2076   public void patch_splitMax(LinkedList<Patch> patches) {
2077     int patch_size;
2078     String precontext, postcontext;
2079     Patch patch;
2080     int start1, start2;
2081     boolean empty;
2082     Operation diff_type;
2083     String diff_text;
2084     ListIterator<Patch> pointer = patches.listIterator();
2085     Patch bigpatch = pointer.hasNext() ? pointer.next() : null;
2086     while (bigpatch != null) {
2087       if (bigpatch.length1 <= Match_MaxBits) {
2088         bigpatch = pointer.hasNext() ? pointer.next() : null;
2089         continue;
2090       }
2091       // Remove the big old patch.
2092       pointer.remove();
2093       patch_size = Match_MaxBits;
2094       start1 = bigpatch.start1;
2095       start2 = bigpatch.start2;
2096       precontext = "";
2097       while (!bigpatch.diffs.isEmpty()) {
2098         // Create one of several smaller patches.
2099         patch = new Patch();
2100         empty = true;
2101         patch.start1 = start1 - precontext.length();
2102         patch.start2 = start2 - precontext.length();
2103         if (precontext.length() != 0) {
2104           patch.length1 = patch.length2 = precontext.length();
2105           patch.diffs.add(new Diff(Operation.EQUAL, precontext));
2106         }
2107         while (!bigpatch.diffs.isEmpty()
2108             && patch.length1 < patch_size - Patch_Margin) {
2109           diff_type = bigpatch.diffs.getFirst().operation;
2110           diff_text = bigpatch.diffs.getFirst().text;
2111           if (diff_type == Operation.INSERT) {
2112             // Insertions are harmless.
2113             patch.length2 += diff_text.length();
2114             start2 += diff_text.length();
2115             patch.diffs.addLast(bigpatch.diffs.removeFirst());
2116             empty = false;
2117           } else if (diff_type == Operation.DELETE && patch.diffs.size() == 1
2118               && patch.diffs.getFirst().operation == Operation.EQUAL
2119               && diff_text.length() > 2 * patch_size) {
2120             // This is a large deletion.  Let it pass in one chunk.
2121             patch.length1 += diff_text.length();
2122             start1 += diff_text.length();
2123             empty = false;
2124             patch.diffs.add(new Diff(diff_type, diff_text));
2125             bigpatch.diffs.removeFirst();
2126           } else {
2127             // Deletion or equality.  Only take as much as we can stomach.
2128             diff_text = diff_text.substring(0, Math.min(diff_text.length(),
2129                 patch_size - patch.length1 - Patch_Margin));
2130             patch.length1 += diff_text.length();
2131             start1 += diff_text.length();
2132             if (diff_type == Operation.EQUAL) {
2133               patch.length2 += diff_text.length();
2134               start2 += diff_text.length();
2135             } else {
2136               empty = false;
2137             }
2138             patch.diffs.add(new Diff(diff_type, diff_text));
2139             if (diff_text.equals(bigpatch.diffs.getFirst().text)) {
2140               bigpatch.diffs.removeFirst();
2141             } else {
2142               bigpatch.diffs.getFirst().text = bigpatch.diffs.getFirst().text
2143                   .substring(diff_text.length());
2144             }
2145           }
2146         }
2147         // Compute the head context for the next patch.
2148         precontext = diff_text2(patch.diffs);
2149         precontext = precontext.substring(Math.max(0, precontext.length()
2150             - Patch_Margin));
2151         // Append the end context for this patch.
2152         if (diff_text1(bigpatch.diffs).length() > Patch_Margin) {
2153           postcontext = diff_text1(bigpatch.diffs).substring(0, Patch_Margin);
2154         } else {
2155           postcontext = diff_text1(bigpatch.diffs);
2156         }
2157         if (postcontext.length() != 0) {
2158           patch.length1 += postcontext.length();
2159           patch.length2 += postcontext.length();
2160           if (!patch.diffs.isEmpty()
2161               && patch.diffs.getLast().operation == Operation.EQUAL) {
2162             patch.diffs.getLast().text += postcontext;
2163           } else {
2164             patch.diffs.add(new Diff(Operation.EQUAL, postcontext));
2165           }
2166         }
2167         if (!empty) {
2168           pointer.add(patch);
2169         }
2170       }
2171       bigpatch = pointer.hasNext() ? pointer.next() : null;
2172     }
2173   }
2174 
2175 
2176   /**
2177    * Take a list of patches and return a textual representation.
2178    * @param patches List of Patch objects.
2179    * @return Text representation of patches.
2180    */
patch_toText(List<Patch> patches)2181   public String patch_toText(List<Patch> patches) {
2182     StringBuilder text = new StringBuilder();
2183     for (Patch aPatch : patches) {
2184       text.append(aPatch);
2185     }
2186     return text.toString();
2187   }
2188 
2189 
2190   /**
2191    * Parse a textual representation of patches and return a List of Patch
2192    * objects.
2193    * @param textline Text representation of patches.
2194    * @return List of Patch objects.
2195    * @throws IllegalArgumentException If invalid input.
2196    */
patch_fromText(String textline)2197   public List<Patch> patch_fromText(String textline)
2198       throws IllegalArgumentException {
2199     List<Patch> patches = new LinkedList<Patch>();
2200     if (textline.length() == 0) {
2201       return patches;
2202     }
2203     List<String> textList = Arrays.asList(textline.split("\n"));
2204     LinkedList<String> text = new LinkedList<String>(textList);
2205     Patch patch;
2206     Pattern patchHeader
2207         = Pattern.compile("^@@ -(\\d+),?(\\d*) \\+(\\d+),?(\\d*) @@$");
2208     Matcher m;
2209     char sign;
2210     String line;
2211     while (!text.isEmpty()) {
2212       m = patchHeader.matcher(text.getFirst());
2213       if (!m.matches()) {
2214         throw new IllegalArgumentException(
2215             "Invalid patch string: " + text.getFirst());
2216       }
2217       patch = new Patch();
2218       patches.add(patch);
2219       patch.start1 = Integer.parseInt(m.group(1));
2220       if (m.group(2).length() == 0) {
2221         patch.start1--;
2222         patch.length1 = 1;
2223       } else if (m.group(2).equals("0")) {
2224         patch.length1 = 0;
2225       } else {
2226         patch.start1--;
2227         patch.length1 = Integer.parseInt(m.group(2));
2228       }
2229 
2230       patch.start2 = Integer.parseInt(m.group(3));
2231       if (m.group(4).length() == 0) {
2232         patch.start2--;
2233         patch.length2 = 1;
2234       } else if (m.group(4).equals("0")) {
2235         patch.length2 = 0;
2236       } else {
2237         patch.start2--;
2238         patch.length2 = Integer.parseInt(m.group(4));
2239       }
2240       text.removeFirst();
2241 
2242       while (!text.isEmpty()) {
2243         try {
2244           sign = text.getFirst().charAt(0);
2245         } catch (IndexOutOfBoundsException e) {
2246           // Blank line?  Whatever.
2247           text.removeFirst();
2248           continue;
2249         }
2250         line = text.getFirst().substring(1);
2251         line = line.replace("+", "%2B");  // decode would change all "+" to " "
2252         try {
2253           line = URLDecoder.decode(line, "UTF-8");
2254         } catch (UnsupportedEncodingException e) {
2255           // Not likely on modern system.
2256           throw new Error("This system does not support UTF-8.", e);
2257         } catch (IllegalArgumentException e) {
2258           // Malformed URI sequence.
2259           throw new IllegalArgumentException(
2260               "Illegal escape in patch_fromText: " + line, e);
2261         }
2262         if (sign == '-') {
2263           // Deletion.
2264           patch.diffs.add(new Diff(Operation.DELETE, line));
2265         } else if (sign == '+') {
2266           // Insertion.
2267           patch.diffs.add(new Diff(Operation.INSERT, line));
2268         } else if (sign == ' ') {
2269           // Minor equality.
2270           patch.diffs.add(new Diff(Operation.EQUAL, line));
2271         } else if (sign == '@') {
2272           // Start of next patch.
2273           break;
2274         } else {
2275           // WTF?
2276           throw new IllegalArgumentException(
2277               "Invalid patch mode '" + sign + "' in: " + line);
2278         }
2279         text.removeFirst();
2280       }
2281     }
2282     return patches;
2283   }
2284 
2285 
2286   /**
2287    * Class representing one diff operation.
2288    */
2289   public static class Diff {
2290     /**
2291      * One of: INSERT, DELETE or EQUAL.
2292      */
2293     public Operation operation;
2294     /**
2295      * The text associated with this diff operation.
2296      */
2297     public String text;
2298 
2299     /**
2300      * Constructor.  Initializes the diff with the provided values.
2301      * @param operation One of INSERT, DELETE or EQUAL.
2302      * @param text The text being applied.
2303      */
Diff(Operation operation, String text)2304     public Diff(Operation operation, String text) {
2305       // Construct a diff with the specified operation and text.
2306       this.operation = operation;
2307       this.text = text;
2308     }
2309 
2310 
2311     /**
2312      * Display a human-readable version of this Diff.
2313      * @return text version.
2314      */
toString()2315     public String toString() {
2316       String prettyText = this.text.replace('\n', '\u00b6');
2317       return "Diff(" + this.operation + ",\"" + prettyText + "\")";
2318     }
2319 
2320 
2321     /**
2322      * Is this Diff equivalent to another Diff?
2323      * @param d Another Diff to compare against.
2324      * @return true or false.
2325      */
equals(Object d)2326     public boolean equals(Object d) {
2327       try {
2328         return (((Diff) d).operation == this.operation)
2329                && (((Diff) d).text.equals(this.text));
2330       } catch (ClassCastException e) {
2331         return false;
2332       }
2333     }
2334   }
2335 
2336 
2337   /**
2338    * Class representing one patch operation.
2339    */
2340   public static class Patch {
2341     public LinkedList<Diff> diffs;
2342     public int start1;
2343     public int start2;
2344     public int length1;
2345     public int length2;
2346 
2347 
2348     /**
2349      * Constructor.  Initializes with an empty list of diffs.
2350      */
Patch()2351     public Patch() {
2352       this.diffs = new LinkedList<Diff>();
2353     }
2354 
2355 
2356     /**
2357      * Emmulate GNU diff's format.
2358      * Header: @@ -382,8 +481,9 @@
2359      * Indicies are printed as 1-based, not 0-based.
2360      * @return The GNU diff string.
2361      */
toString()2362     public String toString() {
2363       String coords1, coords2;
2364       if (this.length1 == 0) {
2365         coords1 = this.start1 + ",0";
2366       } else if (this.length1 == 1) {
2367         coords1 = Integer.toString(this.start1 + 1);
2368       } else {
2369         coords1 = (this.start1 + 1) + "," + this.length1;
2370       }
2371       if (this.length2 == 0) {
2372         coords2 = this.start2 + ",0";
2373       } else if (this.length2 == 1) {
2374         coords2 = Integer.toString(this.start2 + 1);
2375       } else {
2376         coords2 = (this.start2 + 1) + "," + this.length2;
2377       }
2378       StringBuilder text = new StringBuilder();
2379       text.append("@@ -").append(coords1).append(" +").append(coords2)
2380           .append(" @@\n");
2381       // Escape the body of the patch with %xx notation.
2382       for (Diff aDiff : this.diffs) {
2383         switch (aDiff.operation) {
2384         case INSERT:
2385           text.append('+');
2386           break;
2387         case DELETE:
2388           text.append('-');
2389           break;
2390         case EQUAL:
2391           text.append(' ');
2392           break;
2393         }
2394         try {
2395           text.append(URLEncoder.encode(aDiff.text, "UTF-8").replace('+', ' '))
2396               .append("\n");
2397         } catch (UnsupportedEncodingException e) {
2398           // Not likely on modern system.
2399           throw new Error("This system does not support UTF-8.", e);
2400         }
2401       }
2402       return unescapeForEncodeUriCompatability(text.toString());
2403     }
2404   }
2405 
2406 
2407   /**
2408    * Unescape selected chars for compatability with JavaScript's encodeURI.
2409    * In speed critical applications this could be dropped since the
2410    * receiving application will certainly decode these fine.
2411    * Note that this function is case-sensitive.  Thus "%3f" would not be
2412    * unescaped.  But this is ok because it is only called with the output of
2413    * URLEncoder.encode which returns uppercase hex.
2414    *
2415    * Example: "%3F" -> "?", "%24" -> "$", etc.
2416    *
2417    * @param str The string to escape.
2418    * @return The escaped string.
2419    */
unescapeForEncodeUriCompatability(String str)2420   private static String unescapeForEncodeUriCompatability(String str) {
2421     return str.replace("%21", "!").replace("%7E", "~")
2422         .replace("%27", "'").replace("%28", "(").replace("%29", ")")
2423         .replace("%3B", ";").replace("%2F", "/").replace("%3F", "?")
2424         .replace("%3A", ":").replace("%40", "@").replace("%26", "&")
2425         .replace("%3D", "=").replace("%2B", "+").replace("%24", "$")
2426         .replace("%2C", ",").replace("%23", "#");
2427   }
2428 }
2429