2 .\" wiggle - apply rejected patches
4 .\" Copyright (C) 2003 Neil Brown <neilb@cse.unsw.edu.au>
5 .\" Copyright (C) 2010-2013 Neil Brown <neilb@suse.de>
6 .\" Copyright (C) 2018-2019 Neil Brown <neil@brown.name>
9 .\" This program is free software; you can redistribute it and/or modify
10 .\" it under the terms of the GNU General Public License as published by
11 .\" the Free Software Foundation; either version 2 of the License, or
12 .\" (at your option) any later version.
14 .\" This program is distributed in the hope that it will be useful,
15 .\" but WITHOUT ANY WARRANTY; without even the implied warranty of
16 .\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 .\" GNU General Public License for more details.
19 .\" You should have received a copy of the GNU General Public License
20 .\" along with this program.
22 .\" Author: Neil Brown
23 .\" Email: <neil@brown.name>
27 wiggle \- apply rejected patches and perform word-wise diffs
29 .BI wiggle " [function] [options] file [files]"
33 is to apply a patch to a file in a similar manner to the
37 The distinctive difference of
39 is that it will attempt to apply a patch even if the "before" part of
40 the patch doesn't match the target file perfectly.
41 This is achieved by breaking the file and patch into words and finding
42 the best alignment of words in the file with words in the patch.
43 Once this alignment has been found, any differences (word-wise) in the
44 patch are applied to the file as best as possible.
48 will (in some cases) detect changes that have already been applied,
52 ensures that every change in the patch is applied to the target
53 file somehow. If a particular change cannot be made in the file, the
54 file is annotated to show where the change should be made in a similar
60 Each annotation contains 3 components: a portion of the original file
61 where the change should be applied, a portion of the patch that
62 couldn't be matched precisely in the file, and the text that should
63 replace that portion of the patch. These are separated by lines
64 containing precisely 7 identical characters, either '<', '|', '=',
65 or '>', possibly followed by a descriptive word. So
70 Some portion of the original file
74 text to replace it with
79 indicates that "text to replace" should be replaced by "text to
80 replace it with" somewhere in the portion of the original file.
83 was not able to find a place to make this change.
86 can also produce conflict reports showing only the words that are
87 involved rather than showing whole lines.
88 In this case the output looks like:
91 <<<---original|||old===new--->>>
98 to apply some patch, and to collect a list of rejects by monitoring
99 the error messages from patch. Then for each file for which a
100 reject was found, run
102 wiggle \-\-replace originalfile originalfile.rej
104 Finally each file must be examined to resolve any unresolved
105 conflicts, and to make sure the applied patch is semantically correct.
107 Alternately, the original patch file can be fed to the
111 wiggle \-B < patchfile
113 This will allow the changes and conflicts to be inspected and, to some
114 extent, modified; and then the results can be saved.
116 The following options are understood by
118 Some of these are explained in more detail in the following sections
119 on MERGE, DIFF, EXTRACT, and BROWSE.
121 .BR \-m ", " \-\-merge
122 Select the "merge" function. This is the default function.
124 .BR \-d ", " \-\-diff
125 Select the "diff" function. This displays the differences between
126 files. This can be given after
128 (see below) in which case a patch or diff of two files can be viewed
129 without the originals.
131 .BR \-x ", " \-\-extract
132 Select the "extract" function. This extracts one branch of a patch or
135 .BR \-B ", " \-\-browse
136 Select the "browse" function. This is similar to "merge" (or "diff")
137 only with a different presentation. Instead of the result simply
138 being sent to standard output, it is presented using an ncurses-based
139 GUI so that each hunk of the patch can be examined to understand what
140 conflicts where involved and what needed to be ignored in order of the
141 patch to be wiggled in to place.
143 .BR \-w ", " \-\-words
144 Request that all operations and display be word based. This is the
145 default for the "diff" function.
148 Request that words be defined as sequences of non-white-space. Without
149 this flag words are sequences of alphanumerics or single non-white-space
150 characters. This flag is enabled by automatically enabled if
152 needs to compare two files which both have more than 50,000 words.
154 .BR \-l ", " \-\-lines
155 Request that all operations and display be line based.
157 .BR \-b ", " \-\-ignore\-blanks
158 De-emphasise white space (space, tab, and newline) is determining
159 differences and changes.
162 Normally white space is treated like a word which can be matched or
163 changed by a patch. When this flag is in force, white space serves
164 only as a separator between other words and is not matched itself.
165 The effect of this is that changes in the amount of white space are
166 not treated as significant.
168 To be precise, any white space is combined with the preceding word
169 or, in the case of leading space on a line, with the following word.
170 However it is not involved in any comparisons of that word. If a patch
171 deletes a word, the attached white space is deleted as well. If a
172 patch adds a word, the attached white space is added as well.
174 An empty line, or one that contains only blanks, will be treated as a
175 single word that will match any other blank line, no matter how many
184 .BR \-p ", " \-\-patch
185 Treat the last named file as a patch instead of a file (with \-\-diff)
186 or a merge (\-\-extract).
193 requires there be exactly one file which is a patch and which can
194 contain patches to multiple files. The patches are merged into each
197 mode, this usage requires the
199 option as writing lots of merged files to standard-out is impractical.
202 When processing a multi-file patch,
204 can be followed by a numeric argument indicating how many file name
205 components should be stripped from files named in the patch file. If no
206 numeric argument is given,
208 will deduce an appropriate number based what files are present in the
212 .BR \-r ", " \-\-replace
213 Normally the merged output is written to standard-output. With
215 the original file is replaced with the merge output.
220 to always save the resulting merge when exiting.
223 Normally when an original file is replaced with the merged result, that
224 file is renamed to have a ".porig" extension, so that it is preserved.
225 If you don't want to keep the original, use this option to suppress
228 .BR \-o ", " \-\-output=
229 Rather than writing the result to stdout or to replace the original
230 file, this requests that the output be written to the given file.
231 This is only meaningful with
235 when given a single merge to browse.
238 This option overrides
242 .BR \-R ", " \-\-reverse
245 function, swap the files before calculating
253 attempts to revert changes rather than apply them.
255 .BR \-i ", " \-\-no\-ignore
256 Normally wiggle will ignore changes in the patch which appear to
257 already have been applied in the original. With this flag those
258 changes are reported as conflicts rather than being ignored.
260 .BR \-W ", " \-\-show\-wiggles
263 conflicts that can be wiggled into place are reported as conflicts
264 with an extra stanza which shows what the result would be if this flag
265 had not been used. The extra stanza is introduce with a line
266 containing 7 ampersand
273 Some portion of the original file
277 text to replace it with
279 Text that would result from a successful wiggle
285 .B \-\-report\-wiggles
286 If a merge is successful in applying all changes, it will normally exit
287 with a success status (0), only reporting failure (1) if a conflict
288 occurred and was annotated. With
289 .B \-\-report\-wiggles
291 will also report failure if any changes had to be wiggled in. This
294 is used for automatic merges as with
296 If any wiggles happen,
298 will report the failure, and the results can be examined to confirm
301 .BR \-h ", " \-\-help
302 Print a simple help message. If given after one of the function
308 help specific to that function is displayed.
310 .BR \-V ", " \-\-version
311 Display the version number of
314 .BR \-v ", " \-\-verbose
315 Enable verbose mode. Currently this makes no difference.
317 .BR \-q ", " \-\-quiet
318 Enable quiet mode. This suppresses the message from the merge
319 function when there are unresolvable conflicts.
322 can divide a text into lines or words when performing it's tasks.
323 A line is simply a string of characters terminated by a newline.
324 A word is either a maximal contiguous string of alphanumerics
325 (including underscore), a maximal contiguous string of space or tab
326 characters, or any other single character.
328 The merge function modifies a given text by finding all changes between
329 two other texts and imposing those changes on the given text.
333 focuses on which words have changed so as to maximise the possibility
334 of finding a good match in the given text for the context of a given
335 change. However it can consider only whole lines instead.
338 extracts the three texts that it needs from files listed on the
339 command line. Either 1, 2, or 3 files may be listed, and any one of
340 them may be a lone hyphen signifying standard-input.
342 If one file is given and the
344 option is not present, the file is treated as a
346 file, i.e. the output of "merge \-A" or "wiggle". Such a file
347 implicitly contains three streams and these are extracted and
350 If two files are given, then the first simply contains the primary
351 text, and the second is treated as a patch file (the output of "diff\ \-u"
352 or "diff\ \-c", or a ".rej" file from
354 and the two other texts
355 are extracted from that.
357 If one file is given together with the
359 option, the file is treated as a patch file containing the names of
360 the files that it patches. In this case multiple merge operations can
361 happen and each takes one stream from a file named in the patch, and
362 the other two from the patch itself. The
364 option is required and the results are written back to the
367 Finally if three files are listed, they are taken to contain the given
368 text and the two other texts, in order.
370 Normally the result of the merge is written to standard-output.
373 flag is given, the output is written to a file
374 which replaces the original given file. In this case the original file
375 will normally be renamed to have a
377 suffix (for "patched original" which makes sense if you first use
379 to apply a patch, and then use
381 to wiggle the rejects in). This can be suppressed with the
387 option is given with a file name, the output will be written to that
388 file. In this case no backup is created.
390 If no errors occur (such as file access errors)
392 will exit with a status of 0 if all changes were successfully merged,
393 and with an exit status of 1 and a brief message if any changes could
394 not be fully merged and were instead inserted as annotations.
396 .B \-\-report\-wiggles
401 will also exist with status of 1 if any changes had to be wiggled in
402 even though this was successful.
404 The merge function can operate in three different modes with respect
409 option, whole lines are compared and any conflicts
410 are reported as whole lines that need to be replaced.
414 option, individual words are compared and any
415 conflicts are reported just covering the words affected. This uses
416 the \f(CW <<<|||===>>> \fP conflict format.
418 Without either of these options, a hybrid approach is taken.
419 Individual words are compared and merged, but when a conflict is found
420 the whole surrounding line is reported as being in conflict.
423 will ensure that every change between the two other texts is reflected
424 in the result of the merge somehow. There are four different ways
425 that a change can be reflected.
433 is found at a suitable place in the original file, it is
436 This includes the possibility that
443 If a change is found which simply adds
445 and the text immediately preceding and following the insertion are
446 found adjacent in the original file in a suitable place, then
448 is inserted between those adjacent texts.
450 If a change is found which changes
454 and this appears (based on context) to align with
456 in the original, then it is assumed that this change has already been
457 applied, and the change is ignored. When this happens, a message
458 reflecting the number of ignored changes is printed by
460 This optimisation can be suppressed with the
464 If a change is found that does not fit any of the above possibilities,
465 then a conflict is reported as described earlier.
467 The diff function is provided primarily to allow inspection of the
470 calculated between texts and that it uses for performing a merge.
472 The output of the diff function is similar to the unified output of
473 diff. However while diff does not output long stretches of common text,
475 diff mode outputs everything.
477 When calculating a word-based alignment (the default),
479 may need to show these word-based differences. This is done using an
480 extension to the unified-diff format. If a line starts with a
481 vertical bar, then it may contain sections surrounded by special
482 multi-character brackets. The brackets "<<<++" and "++>>>" surround
483 added text while "<<<--" and "-->>>" surround removed text.
486 can be given the two texts to compare in one of three ways.
488 If only one file is given, then it is treated as a patch and the two
489 branches of that patch are compared. This effectively allows a patch
490 to be refined from a line-based patch to a word-based patch.
492 If two files are given, then they are normally assumed to be simple
493 texts to be compared.
495 If two files are given along with the \-\-patch option, then the second
496 file is assumed to be a patch and either the first (with \-1) or the
497 second (with \-2) branch is extracted and compared with text found in
500 This last option causes
502 to apply a "best-fit" algorithm for aligning patch hunks with the
503 file before computing the differences. This algorithm is used when
504 merging a patch with a file, and its value can be seen by comparing
505 the difference produced this way with the difference produced by first
506 extracting one branch of a patch into a file, and then computing the
507 difference of that file with the main file.
509 The extract function of
511 simply exposes the internal functionality for extracting one branch of
512 a patch or a merge file.
514 Precisely one file should be given, and it will be assumed to be a
517 is given, in which case a patch is assumed.
519 The choice of branch in made by providing one of
524 with obvious meanings.
526 The browse function of
528 presents the result of a merge or (with
530 a diff in a text-based GUI that can be
531 navigated using keystrokes similar to
536 The browser allows each of the two or three streams to be viewed individually
537 with colours used to highlight different sorts of text - green for
538 added text, red for deleted text etc. It can also show the patch by
539 itself, the full result of the merge, or the merge and the patch
542 The browser provides a number of context-sensitive help pages which
543 can be accessed by typing '?'
545 The top right of the GUI will report the type of text under the
546 cursor, which is also indicated by the colour of the text. Options
547 are Unchanged, Changed, Unmatched, Extraneous, AlreadyApplied and
548 Conflict. If the meanings of these are clear a little
549 experimentations should help.
551 A limited amount of editing is permitted while in
553 mode. Currently text that is unwanted can be discarded with
555 This will convert a Conflict or Change to Unchanged, and an Unmatched
556 to Changed (which effectively changes it to the empty string).
557 Similarly a text can be marked as wanted with
559 This will convert a Conflict or Extraneous to Changed. Using the same
560 key again will revert the change.
562 Finally, the uppercase
564 will revert all changes on the current line.
566 To make more sweeping changes you can use
568 which runs an editor, preferring
572 if they are set in the environment.
574 If you make any changes, then wiggle will ask you if you want
575 to save the changes, even if
579 Caution should always be exercised when applying a rejected patch with
583 rejects a patch, it does so for a good reason. Even though
585 may be able to find a believable place to apply each textual change,
586 there is no guarantee that the result is correct in any semantic
587 sense. The result should always be inspected to make sure it is
590 .B " wiggle \-\-replace file file.rej"
592 This is the normal usage of
594 and will take any changes in
598 could not apply, and merge them into
601 .B " wiggle \-dp1 file file.rej"
603 This will perform a word-wise comparison between the
607 branch of the diff in
609 and display the differences. This allows you to see where a given
612 .B " wiggle \-\-merge \-\-help"
614 Get help about the merge function of
617 .B " wiggle \-\-browse \-\-patch update.patch"
621 file for patches and present a list of patched files which can be
622 browsed to examine each patch in detail.
625 can be integrated with
627 so that it is used as the default merge tool and diff tool.
628 This can be achieved by adding the following lines to
630 in the user's home directory.
634 name = "Wiggle flexible merging"
635 driver = wiggle \-o %A %A %O %B
640 cmd = wiggle \-B \-o $MERGED $LOCAL $BASE $REMOTE
642 cmd = wiggle \-Bd $LOCAL $REMOTE
656 for merges (which may be dangerous), you can add
663 .BR $HOME/.config/git/attributes .
667 was inspired by the following quote.
670 The problem I find is that I often want to take
671 (file1+patch) -> file2,
672 when I don't have file1. But merge tools want to take
673 (file1|file2) -> file3.
674 I haven't seen a graphical tool which helps you to wiggle a patch
677 \-\- Andrew Morton - 2002
682 cannot read the extended unified-diff output that it produces for
686 cannot read the word-based merge format that it produces for \-\-merge
690 does not understand unicode and so will treat all non-ASCII characters
691 much the same as it treats punctuation - it will treat each one
692 as a separate word. The browser will not display non-ASCII characters
695 Neil Brown at Computer Science and Engineering at
696 The University of New South Wales, Sydney, Australia;
697 and later at SUSE, still in Sydney, Australia.
705 .IR gitattributes (5).