1REPOSURGEON(1) Development Tools REPOSURGEON(1)
2
6 reposurgeon - surgical operations on repositories
7
9 reposurgeon [command...]
10
12 The purpose of reposurgeon is to enable risky operations that VCSes
13 (version-control systems) don't want to let you do, such as (a) editing
14 past comments and metadata, (b) excising commits, (c) coalescing and
15 splitting commits, (d) removing files and subtrees from repo history,
16 (e) merging or grafting two or more repos, and (f) cutting a repo in
17 two by cutting a parent-child link, preserving the branch structure of
18 both child repos.
19 A major use of reposurgeon is to assist a human operator to perform
21 higher-quality conversions among version control systems than can be
22 achieved with fully automated converters.
23 The original motivation for reposurgeon was to clean up artifacts
25 created by repository conversions. It was foreseen that the tool would
26 also have applications when code needs to be removed from repositories
27 for legal or policy reasons.
28 To keep reposurgeon simple and flexible, it normally does not do its
30 own repository reading and writing. Instead, it relies on being able to
31 parse and emit the command streams created by git-fast-export and read
32 by git-fast-import. This means that it can be used on any
33 version-control system that has both fast-export and fast-import
34 utilities. The git-import stream format also implicitly defines a
35 common language of primitive operations for reposurgeon to speak.
36 Fully supported systems (those for which reposurgeon can both read and
38 write repositories) include git, hg, bzr, svn, darcs, bk, RCS, and SRC.
39 For a complete list, with dependencies and technical notes, type prefer
40 to the reposurgeon prompt.
41 Writing to the file-oriented systems RCS and SRC is done via rcs-fast-
43 import(1) and has some serious limitations because those systems cannot
44 represent all the metadata in a git-fast-export stream. Consult that
45 tool's documentation for details and partial workarounds.
46 Writing Subversion repositories also has some significant limitations,
48 discussed in the section on Working With Subversion.
49 Fossil repository files can be read in using the --format=fossil option
51 of the read command and written out with the --format=fossil option of
52 the write. Ignore patterns are not translated in either direction.
53 CVS is supported for read only, not write. For CVS, reposurgeon must be
55 run from within a repository directory (one with a CVSROOT
56 subdirectory). Each module becomes a subdirectory in the the
57 reposurgeon representation of the change history.
58 In order to deal with version-control systems that do not have
60 fast-export equivalents, reposurgeon can also host extractor code that
61 reads repositories directly. For each version-control system supported
62 through an extractor, reposurgeon uses a small amount of knowledge
63 about the system's command-line tools to (in effect) replay repository
64 history into an input stream internally. Repositories under systems
65 supported through extractors can be read by reposurgeon, but not
66 modified by it. In particular, reposurgeon can be used to move a
67 repository history from any VCS supported by an extractor to any VCS
68 supported by a normal importer/exporter pair.
69 Mercurial repository reading is implemented with an extractor class;
71 writing is handled with the stock "hg fastimport" command. A test
72 extractor exists for git, but is normally disabled in favor of the
73 regular exporter.
74 For guidance on the pragmatics of repository conversion, see the DVCS
76 Migration HOWTO[1].
77
79 reposurgeon is a sharp enough tool to cut you. It takes care not to
80 ever write a repository in an actually inconsistent state, and will
81 terminate with an error message rather than proceed when its internal
82 data structures are confused. However, there are lots of things you can
83 do with it - like altering stored commit timestamps so they no longer
84 match the commit sequence - that are likely to cause havoc after you're
85 done. Proceed with caution and check your work.
86 Also note that, if your DVCS does the usual thing of making commit IDs
88 a cryptographic hash of content and parent links, editing a
89 publicly-accessible repository with this tool would be a bad idea. All
90 of the surgical operations in reposurgeon will modify the hash chains.
91 Please also see the notes on system-specific issues under the section
93 called “LIMITATIONS AND GUARANTEES”.
94
96 The program can be run in one of two modes, either as an interactive
97 command interpreter or in batch mode to execute commands given as
98 arguments on the reposurgeon invocation line. The only differences
99 between these modes are (1) the interactive one begins by turning on
100 the 'verbose 1' option, (2) in batch mode all errors (including
101 normally recoverable errors in selection-set syntax) are fatal, and (3)
102 each command-line argument beginning with “--” has that stripped off
103 (which, in particular means that --help and --version will work as
104 expected). Also, in interactive mode, Ctrl-P and Ctrl-N will be
105 available to scroll through your command history and tab completion of
106 both command keywords and name arguments (wherever that makes semantic
107 sense) is available.
108 A git-fast-import stream consists of a sequence of commands which must
110 be executed in the specified sequence to build the repo; to avoid
111 confusion with reposurgeon commands we will refer to the stream
112 commands as events in this documentation. These events are implicitly
113 numbered from 1 upwards. Most commands require specifying a selection
114 of event sequence numbers so reposurgeon will know which events to
115 modify or delete.
116 For all the details of event types and semantics, see the git-fast-
118 import(1) manual page; the rest of this paragraph is a quick start for
119 the impatient. Most events in a stream are commits describing revision
120 states of the repository; these group together under a single change
121 comment one or more fileops (file operations), which usually point to
122 blobs that are revision states of individual files. A fileop may also
123 be a delete operation indicating that a specified previously-existing
124 file was deleted as part of the version commit; there are a couple of
125 other special fileop types of lesser importance.
126 Commands to reposurgeon consist of a command keyword, sometimes
128 preceded by a selection set, sometimes followed by whitespace-separated
129 arguments. It is often possible to omit the selection-set argument and
130 have it default to something reasonable.
131 Here are some motivating examples. The commands will be explained in
133 more detail after the description of selection syntax.
134 :15 edit ;; edit the object associated with mark :15
136 edit ;; edit all editable objects
138 29..71 list ;; list summary index of events 29..71
140 236..$ list ;; List events from 236 to the last
142 <#523> inspect ;; Look for commit #523; they are numbered
144 ;; 1-origin from the beginning of the repository.
145 <2317> inspect ;; Look for a tag with the name 2317, a tip commit
147 ;; of a branch named 2317, or a commit with legacy ID
148 ;; 2317. Inspect what is found. A plain number is
149 ;; probably a legacy ID inherited from a Subversion
150 ;; revision number.
151 /regression/ list ;; list all commits and tags with comments or
153 ;; committer headers or author headers containing
154 ;; the string "regression"
155 1..:97 & =T delete ;; delete tags from event 1 to mark 97
157 [Makefile] inspect ;; Inspect all commits with a file op touching Makefile
159 ;; and all blobs referred to in a fileop
160 ;; touching Makefile.
161 :46 tip ;; Display the branch tip that owns commit :46.
163 @dsc(:55) list ;; Display all commits with ancestry tracing to :55
165 @min([.gitignore]) remove .gitignore delete
167 ;; Remove the first .gitignore fileop in the repo.
168 SELECTION SYNTAX
170 The selection-set specification syntax is an expression-oriented
171 minilanguage. The most basic term in this language is a location. The
172 following sorts of primitive locations are supported:
173 event numbers
175 A plain numeric literal is interpreted as a 1-origin event-sequence
176 number.
177 marks
179 A numeric literal preceded by a colon is interpreted as a mark; see
180 the import stream format documentation for explanation of the
181 semantics of marks.
182 tag and branch names
184 The basename of a branch (including branches in the refs/tags
185 namespace) refers to its tip commit. The name of a tag is
186 equivalent to its mark (that of the tag itself, not the commit it
187 refers to). Tag and branch locations are bracketed with < > (angle
188 brackets) to distinguish them from command keywords.
189 legacy IDs
191 If the contents of name brackets (< >) does not match a tag or
192 branch name, the interpreter next searches legacy IDs of commits.
193 This is especially useful when you have imported a Subversion dump;
194 it means that commits made from it can be referred to by their
195 corresponding Subversion revision numbers.
196 commit numbers
198 A numeric literal within name brackets (< >) preceded by # is
199 interpreted as a 1-origin commit-sequence number.
200 $
202 Refers to the last event.
203 These may be grouped into sets in the following ways:
205 ranges
207 A range is two locations separated by "..", and is the set of
208 events beginning at the left-hand location and ending at the
209 right-hand location (inclusive).
210 lists
212 Comma-separated lists of locations and ranges are accepted, with
213 the obvious meaning.
214 There are some other ways to construct event sets:
216 visibility sets
218 A visibility set is an expression specifying a set of event types.
219 It will consist of a leading equal sign, followed by type letters.
220 These are the type letters:
221 ┌──┬─────────────────────┬───────────────────────┐
223 │B │ blobs │ Most default │
224 │ │ │ selection sets │
225 │ │ │ exclude blobs; they │
226 │ │ │ have to be │
227 │ │ │ manipulated through │
228 │ │ │ the commits they │
229 │ │ │ are attached to. │
230 ├──┼─────────────────────┼───────────────────────┤
231 │C │ commits │ │
232 ├──┼─────────────────────┼───────────────────────┤
233 │D │ all-delete commits │ These are artifacts │
234 │ │ │ produced by some │
235 │ │ │ older │
236 │ │ │ repository-conversion │
237 │ │ │ tools. │
238 ├──┼─────────────────────┼───────────────────────┤
239 │H │ head (branch tip) │ │
240 │ │ commits │ │
241 ├──┼─────────────────────┼───────────────────────┤
242 │O │ orphaned │ │
243 │ │ (parentless) │ │
244 │ │ commits │ │
245 ├──┼─────────────────────┼───────────────────────┤
246 │U │ commits with │ │
247 │ │ callouts as parents │ │
248 ├──┼─────────────────────┼───────────────────────┤
249 │Z │ commits with no │ │
250 │ │ fileops │ │
251 ├──┼─────────────────────┼───────────────────────┤
252 │M │ merge │ │
253 │ │ (multi-parent) │ │
254 │ │ commits │ │
255 ├──┼─────────────────────┼───────────────────────┤
256 │F │ fork (multi-child) │ │
257 │ │ commits │ │
258 ├──┼─────────────────────┼───────────────────────┤
259 │L │ commits with │ │
260 │ │ unclean multi-line │ │
261 │ │ comments (without a │ │
262 │ │ separating empty │ │
263 │ │ line after the │ │
264 │ │ first) │ │
265 ├──┼─────────────────────┼───────────────────────┤
266 │I │ commits for which │ │
267 │ │ metadata cannot be │ │
268 │ │ decoded to UTF-8 │ │
269 ├──┼─────────────────────┼───────────────────────┤
270 │T │ tags │ │
271 ├──┼─────────────────────┼───────────────────────┤
272 │R │ resets │ │
273 ├──┼─────────────────────┼───────────────────────┤
274 │P │ Passthrough │ All event types │
275 │ │ │ simply passed │
276 │ │ │ through, including │
277 │ │ │ comments, progress │
278 │ │ │ commands, and │
279 │ │ │ checkpoint commands. │
280 ├──┼─────────────────────┼───────────────────────┤
281 │N │ Legacy IDs │ Any string matching a │
282 │ │ │ cookie (legacy-ID) │
283 │ │ │ format. │
284 └──┴─────────────────────┴───────────────────────┘
285 references
287 A reference name (bracketed by angle brackets) resolves to a single
288 object, either a commit or tag.
289 ┌──────────────┬────────────────────────────┐
291 │ type │ interpretation │
292 ├──────────────┼────────────────────────────┤
293 │ tag name │ annotated tag with that │
294 │ │ name │
295 ├──────────────┼────────────────────────────┤
296 │ branch name │ the branch tip commit │
297 ├──────────────┼────────────────────────────┤
298 │ legacy ID │ commit with that legacy ID │
299 ├──────────────┼────────────────────────────┤
300 │assigned name │ name equated to a │
301 │ │ selection by assign │
302 └──────────────┴────────────────────────────┘
303 Note that if an annotated tag and a branch have the same name foo,
304 <foo> will resolve to the tag rather than the branch tip commit.
305 dates and action stamps
307 A date or action stamp in angle brackets resolves to a selection
308 set of all matching commits.
309 ┌───────────────────────────┬───────────────────────────┐
311 │ type │ interpretation │
312 ├───────────────────────────┼───────────────────────────┤
313 │ RFC3339 timestamp │ commit or tag with that │
314 │ │ time/date │
315 ├───────────────────────────┼───────────────────────────┤
316 │ action stamp │ commits or tags with that │
317 │ (timestamp!email) │ timestamp and author (or │
318 │ │ committer if no author). │
319 ├───────────────────────────┼───────────────────────────┤
320 │yyyy-mm-dd part of RFC3339 │ all commits and tags with │
321 │timestamp │ that date │
322 └───────────────────────────┴───────────────────────────┘
323 To refine the match to a single commit, use a 1-origin index suffix
324 separated by '#'. Thus "<2000-02-06T09:35:10Z>" can match multiple
325 commits, but "<2000-02-06T09:35:10Z#2>" matches only the second in
326 the set.
327 text search
329 A text search expression is a Python regular expression surrounded
330 by forward slashes (to embed a forward slash in it, use a Python
331 string escape such as \x2f).
332 A text search normally matches against the comment fields of
334 commits and annotated tags, or against their author/committer
335 names, or against the names of tags; also the text of passthrough
336 objects.
337 The scope of a text search can be changed with qualifier letters
339 after the trailing slash. These are as follows:
340 ┌───────┬───────────────────────────┐
342 │letter │ interpretation │
343 ├───────┼───────────────────────────┤
344 │ a │ author name in commit │
345 ├───────┼───────────────────────────┤
346 │ b │ branch name in commit; │
347 │ │ also matches blobs │
348 │ │ referenced by commits on │
349 │ │ matching branches, and │
350 │ │ tags which point to │
351 │ │ commmits on patching │
352 │ │ branches. │
353 ├───────┼───────────────────────────┤
354 │ c │ comment text of commit or │
355 │ │ tag │
356 ├───────┼───────────────────────────┤
357 │ r │ committish reference in │
358 │ │ tag or reset │
359 ├───────┼───────────────────────────┤
360 │ p │ text in passthrough │
361 ├───────┼───────────────────────────┤
362 │ t │ tagger in tag │
363 ├───────┼───────────────────────────┤
364 │ n │ name of tag │
365 ├───────┼───────────────────────────┤
366 │ B │ blob content │
367 └───────┴───────────────────────────┘
368 Multiple qualifier letters can add more search scopes.
369 (The “b” qualifier replaces the branchset syntax in earlier
371 versions of reposurgeon.)
372 paths
374 A "path expression" enclosed in square brackets resolves to the set
375 of all commits and blobs related to a path matching the given
376 expression. The path expression itself is either a path literal or
377 a regular expression surrounded by slashes. Immediately after the
378 trailing / of a path regexp you can put any number of the following
379 characters which act as flags: 'a', 'c', 'D', "M', 'R', 'C', 'N'.
380 By default, a path is related to a commit if the latter has a
382 fileop that touches that file path - modifies that change it,
383 deletes that remove it, renames and copies that have it as a source
384 or target. When the 'c' flag is in use the meaning changes: the
385 paths related to a commit become all paths that would be present in
386 a checkout for that commit.
387 A path literal matches a commit if and only if the path literal is
389 exactly one of the paths related to the commit (no prefix or suffix
390 operation is done). In particular a path literal won't match if it
391 corresponds to a directory in the chosen repository.
392 A regular expression matches a commit if it matches any path
394 related to the commit anywhere in the path. You can use '^' or '$'
395 if you want the expression to only match at the beginning or end of
396 paths. When the 'a' flag is in use, the path expression selects
397 commits whose every path matches the regular expression. This is
398 not always a subset of commits selected without the 'a' flag
399 because it also selects commits with no related paths (e.g. empty
400 commits, deletealls and commits with empty trees). If you want to
401 avoid those, you can use e.g. '[/regex/] & [/regex/a]'.
402 The flags 'D', "M', 'R', 'C', 'N' restrict match checking to the
404 corresponding fileop types. Note that this means an 'a' match is
405 easier (not harder) to achieve. These are no-iops when used with
406 'c'.
407 A path or literal matches a blob if it matches any path that
409 appeared in a modification fileop that referred to that blob. To
410 select purely matching blobs or matching commits, compose a path
411 expression with =B or =C.
412 If you need to embed '[^/]' into your regular expression (e.g. to
414 express "all characters but a slash") you can use a Python string
415 escape such as \x2f.
416 function calls
418 The expression language has named special functions. The sequence
419 for a named function is “@” followed by a function name, followed
420 by an argument in parentheses. Presently the following functions
421 are defined:
422 ┌─────┬────────────────────────────┐
424 │name │ interpretation │
425 ├─────┼────────────────────────────┤
426 │min │ minimum member of a │
427 │ │ selection set │
428 ├─────┼────────────────────────────┤
429 │max │ maximum member of a │
430 │ │ selection set │
431 ├─────┼────────────────────────────┤
432 │amp │ nonempty selection set │
433 │ │ becomes all objects, empty │
434 │ │ set is returned empty │
435 ├─────┼────────────────────────────┤
436 │par │ all parents of commits in │
437 │ │ the argument set │
438 ├─────┼────────────────────────────┤
439 │chn │ all children of commits in │
440 │ │ the argument set │
441 ├─────┼────────────────────────────┤
442 │dsc │ all commits descended from │
443 │ │ the argument set (argument │
444 │ │ set included) │
445 ├─────┼────────────────────────────┤
446 │anc │ all commits whom the │
447 │ │ argument set is descended │
448 │ │ from (argument set │
449 │ │ included) │
450 ├─────┼────────────────────────────┤
451 │pre │ events before the argument │
452 │ │ set; empty if the argument │
453 │ │ set includes the first │
454 │ │ event. │
455 ├─────┼────────────────────────────┤
456 │suc │ events after the argument │
457 │ │ set; empty if the argument │
458 │ │ set includes the last │
459 │ │ event. │
460 ├─────┼────────────────────────────┤
461 │srt │ sort the argument set by │
462 │ │ event number. │
463 └─────┴────────────────────────────┘
464 Set expressions may be combined with the operators | and &; these are,
466 respectively, set union and intersection. The | has lower precedence
467 than intersection, but you may use parentheses '(' and ')' to group
468 expressions in case there is ambiguity (this replaces the curly
469 brackets used in older versions of the syntax).
470 Any set operation may be followed by '?' to add the set members'
472 neighbors and referents. This extends the set to include the parents
473 and children of all commits in the set, and the referents of any tags
474 and resets in the set. Each blob reference in the set is replaced by
475 all commit events that refer to it. The '?' can be repeated to extend
476 the neighborhood depth.
477 Do set negation with prefix ~; it has higher precedence than & and |
479 but lower than ?
480 IMPORT AND EXPORT
482 reposurgeon can hold multiple repository states in core. Each has a
483 name. At any given time, one may be selected for editing. Commands in
484 this group import repositories, export them, and manipulate the in-core
485 list and the selection.
486 read [--format=fossil] [directory|-|<infile]
488 With a directory-name argument, this command attempts to read in
489 the contents of a repository in any supported version-control
490 system under that directory; read with no arguments does this in
491 the current directory. If output is redirected to a plain file, it
492 will be read in as a fast-import stream or Subversion dumpfile.
493 With an argument of “-”, this command reads a fast-import stream or
494 Subversion dumpfile from standard input (this will be useful in
495 filters constructed with command-line arguments).
496 If the contents is a fast-import stream, any "cvs-revision"
498 property on a commit is taken to be a newline-separated list of CVS
499 revision cookies pointing to the commit, and used for reference
500 lifting.
501 If the contents is a fast-import stream, any "legacy-id" property
503 on a commit is taken to be a legacy ID token pointing to the
504 commit, and used for reference-lifting.
505 If the read location is a git repository and contains a
507 .git/cvsauthors file (such as is left in place by git cvsimport -A)
508 that file will be read in as if it had been given to the authors
509 read command.
510 If the read location is a directory, and its repository
512 subdirectory has a file named legacy-map, that file will be read as
513 though passed to a legacy read command.
514 If the read location is a file and the --format=fossil is used, the
516 file is interpreted as a Fossil repository.
517 The just-read-in repo is added to the list of loaded repositories
519 and becomes the current one, selected for surgery. If it was read
520 from a plain file and the file name ends with one of the extensions
521 .fi or .svn, that extension is removed from the load list name.
522 Note: this command does not take a selection set.
524 write [--legacy] [--format=fossil] [--noincremental] [--callout]
526 [>outfile|-]
527 Dump selected events as a fast-import stream representing the
528 edited repository; the default selection set is all events. Where
529 to dump to is standard output if there is no argument or the
530 argument is '-', or the target of an output redirect.
531 Alternatively, if there is no redirect and the argument names a
533 directory, the repository is rebuilt into that directory, with any
534 selection set being ignored; if that target directory is nonempty
535 its contents are backed up to a save directory.
536 If the write location is a file and the --format=fossil is used,
538 the file is written in Fossil repository format.
539 With the --legacy option, the Legacy-ID of each commit is appended
541 to its commit comment at write time. This option is mainly useful
542 for debugging conversion edge cases.
543 If you specify a partial selection set such that some commits are
545 included but their parents are not, the output will include
546 incremental dump cookies for each branch with an origin outside the
547 selection set, just before the first reference to that branch in a
548 commit. An incremental dump cookie looks like "refs/heads/foo^0"
549 and is a clue to export-stream loaders that the branch should be
550 glued to the tip of a pre-existing branch of the same name. The
551 --noincremental option suppresses this behavior.
552 When you specify a partial selection set, including a commit object
554 forces the inclusion of every blob to which it refers and every tag
555 that refers to it.
556 Specifying a partial selection may cause a situation in which some
558 parent marks in merges don't correspond to commits present in the
559 dump. When this happens and --callout option was specified, the
560 write code replaces the merge mark with a callout, the action stamp
561 of the parent commit; otherwise the parent mark is omitted.
562 Importers will fail when reading a stream dump with callouts; it is
563 intended to be used by the graft command.
564 Specifying a write selection set with gaps in it is allowed but
566 unlikely to lead to good results if it is loaded by an importer.
567 Property extensions will be be omitted from the output if the
569 importer for the preferred repository type cannot digest them.
570 Note: to examine small groups of commits without the progress
572 meter, use inspect.
573 choose [reponame]
575 Choose a named repo on which to operate. The name of a repo is
576 normally the basename of the directory or file it was loaded from,
577 but repos loaded from standard input are "unnamed". reposurgeon
578 will add a disambiguating suffix if there have been multiple reads
579 from the same source.
580 With no argument, lists the names of the currently stored
582 repositories and their load times. The second column is '*' for the
583 currently selected repository, '-' for others.
584 drop [reponame]
586 Drop a repo named by the argument from reposurgeon's list, freeing
587 the memory used for its metadata and deleting on-disk blobs. With
588 no argument, drops the currently chosen repo.
589 rename reponame
591 Rename the currently chosen repo; requires an argument. Won't do it
592 if there is already one by the new name.
593 REBUILDS IN PLACE
595 reposurgeon can rebuild an altered repository in place. Untracked files
596 are normally saved and restored when the contents of the new repository
597 is checked out (but see the documentation of the “preserve” command for
598 a caveat).
599 rebuild [directory]
601 Rebuild a repository from the state held by reposurgeon. This
602 command does not take a selection set.
603 The single argument, if present, specifies the target directory in
605 which to do the rebuild; if the repository read was from a repo
606 directory (and not a git-import stream), it defaults to that
607 directory. If the target directory is nonempty its contents are
608 backed up to a save directory. Files and directories on the
609 repository's preserve list are copied back from the backup
610 directory after repo rebuild. The default preserve list depends on
611 the repository type, and can be displayed with the stats command.
612 If reposurgeon has a nonempty legacy map, it will be written to a
614 file named legacy-map in the repository subdirectory as though by a
615 legacy write command. (This will normally be the case for
616 Subversion and CVS conversions.)
617 preserve [file...]
619 Add (presumably untracked) files or directories to the repo's list
620 of paths to be restored from the backup directory after a rebuild.
621 Each argument, if any, is interpreted as a pathname. The current
622 preserve list is displayed afterwards.
623 It is only necessary to use this feature if your version-control
625 system lacks a command to list files under version control. Under
626 systems with such a command (which include git and hg), all files
627 that are neither beneath the repository dot directory nor under
628 reposurgeon temporary directories are preserved automatically.
629 unpreserve [file...]
631 Remove (presumably untracked) files or directories to the repo's
632 list of paths to be restored from the backup directory after a
633 rebuild. Each argument, if any, is interpreted as a pathname. The
634 current preserve list is displayed afterwards.
635 INFORMATION AND REPORTS
637 Commands in this group report information about the selected
638 repository.
639 The output of these commands can individually be redirected to a named
641 output file. Where indicated in the syntax, you can prefix the output
642 filename with “>” and give it as a following argument. If you use “>>”
643 the file is opened for append rather than write.
644 list [>outfile]
646 This is the main command for identifying the events you want to
647 modify. It lists commits in the selection set by event sequence
648 number with summary information. The first column is raw event
649 numbers, the second a timestamp in local time. If the repository
650 has legacy IDs, they will be displayed in the third column. The
651 leading portion of the comment follows.
652 stamp [>outfile]
654 Alternative form of listing that displays full action stamps,
655 usable as references in selections. Supports > redirection.
656 tip [>outfile]
658 Display the branch tip names associated with commits in the
659 selection set. These will not necessarily be the same as their
660 branch fields (which will often be tag names if the repo contains
661 either annotated or lightweight tags).
662 If a commit is at a branch tip, its tip is its branch name. If it
664 has only one child, its tip is the child's tip. If it has multiple
665 children, then if there is a child with a matching branch name its
666 tip is the child's tip. Otherwise this function throws a
667 recoverable error.
668 tags [>outfile]
670 Display tags and resets: three fields, an event number and a type
671 and a name. Branch tip commits associated with tags are also
672 displayed with the type field 'commit'. Supports > redirection.
673 stats [repo-name...] [>outfile]
675 Report size statistics and import/export method information about
676 named repositories, or with no argument the currently chosen
677 repository.
678 count [>outfile]
680 Report a count of items in the selection set. Default set is
681 everything in the currently-selected repo. Supports > redirection.
682 inspect [>outfile]
684 Dump a fast-import stream representing selected events to standard
685 output. Just like a write, except (1) the progress meter is
686 disabled, and (2) there is an identifying header before each event
687 dump.
688 graph [>outfile]
690 Emit a visualization of the commit graph in the DOT markup language
691 used by the graphviz tool suite. This can be fed as input to the
692 main graphviz rendering program dot(1), which will yield a viewable
693 image. Supports > redirection.
694 You may find a script like this useful:
696 graph $1 >/tmp/foo$$
698 shell dot </tmp/foo$$ -Tpng | display -; rm /tmp/foo$$
699 You can substitute in your own preferred image viewer, of course.
701 sizes [>outfile]
703 Print a report on data volume per branch; takes a selection set,
704 defaulting to all events. The numbers tally the size of
705 uncompressed blobs, commit and tag comments, and other metadata
706 strings (a blob is counted each time a commit points at it).
707 The numbers are not an exact measure of storage size: they are
709 intended mainly as a way to get information on how to efficiently
710 partition a repository that has become large enough to be unwieldy.
711 Supports > redirection.
713 lint [>outfile]
715 Look for DAG and metadata configurations that may indicate a
716 problem. Presently checks for: (1) Mid-branch deletes, (2)
717 disconnected commits, (3) parentless commits, (4) the existence of
718 multiple roots, (5) committer and author IDs that don't look
719 well-formed as DVCS IDs, (6) multiple child links with identical
720 branch labels descending from the same commit, (7) time and
721 action-stamp collisions.
722 Options to issue only partial reports are supported; "lint
724 --options" or "lint -?" lists them.
725 The options and output format of this command are unstable; they
727 may change without notice as more sanity checks are added.
728 when >timespec
730 Interconvert between git timestamps (integer Unix time plus TZ) and
731 RFC3339 format. Takes one argument, autodetects the format. Useful
732 when eyeballing export streams. Also accepts any other supported
733 date format and converts to RFC3339.
734 SURGICAL OPERATIONS
736 These are the operations the rest of reposurgeon is designed to
737 support.
738 squash [policy...]
740 Combine or delete commits in a selection set of events. The default
741 selection set for this command is empty. Has no effect on events
742 other than commits unless the --delete policy is selected; see the
743 'delete' command for discussion.
744 Normally, when a commit is squashed, its file operation list (and
746 any associated blob references) gets either prepended to the
747 beginning of the operation list of each of the commit's children or
748 appended to the operation list of each of the commit's parents.
749 Then children of a deleted commit get it removed from their parent
750 set and its parents added to their parent set.
751 The default is to squash forward, modifying children; but see the
753 list of policy modifiers below for how to change this.
754 Warning
756 It is easy to get the bounds of a squash command wrong, with
757 confusing and destructive results. Beware thinking you can
758 squash on a selection set to merge all commits except the last
759 one into the last one; what you will actually do is to merge
760 all of them to the first commit after the selected set.
761 Normally, any tag pointing to a combined commit will also be pushed
762 forward. But see the list of policy modifiers below for how to
763 change this.
764 Following all operation moves, every one of the altered file
766 operation lists is reduced to a shortest normalized form. The
767 normalized form detects various combinations of modification,
768 deletion, and renaming and simplifies the operation sequence as
769 much as it can without losing any information.
770 After canonicalization, a file op list may still end up containing
772 multiple M operations on the same file. Normally the tool utters a
773 warning when this occurs but does not try to resolve it.
774 The following modifiers change these policies:
776 --delete
778 Simply discards all file ops and tags associated with deleted
779 commit(s).
780 --coalesce
782 Discard all M operations (and associated blobs) except the
783 last.
784 --pushback
786 Append fileops to parents, rather than prepending to children.
787 --pushforward
789 Prepend fileops to children. This is the default; it can be
790 specified in a lift script for explicitness about intentions.
791 --tagforward
793 With the "tagforward" modifier, any tag on the deleted commit
794 is pushed forward to the first child rather than being deleted.
795 This is the default; it can be specified for explicitness.
796 --tagback
798 With the "--tagback" modifier, any tag on the deleted commit is
799 pushed backward to the first parent rather than being deleted.
800 --quiet
802 Suppresses warning messages about deletion of commits with
803 non-delete fileops.
804 --complain
806 The opposite of quiet. Can be specified for explicitness.
807 Under any of these policies except “--delete”, deleting a commit
809 that has children does not back out the changes made by that
810 commit, as they will still be present in the blobs attached to
811 versions past the end of the deletion set. All a delete does when
812 the commit has children is lose the metadata information about when
813 and by who those changes were actually made; after the delete any
814 such changes will be attributed to the first undeleted children of
815 the deleted commits. It is expected that this command will be
816 useful mainly for removing commits mechanically generated by
817 repository converters such as cvs2svn.
818 delete [policy...]
820 Delete a selection set of events. The default selection set for
821 this command is empty. On a set of commits, this is equivalent to a
822 squash with the --delete flag. It unconditionally deletes tags,
823 resets, and passthroughs; blobs can be removed only as a side
824 effect of deleting every commit that points at them.
825 divide parent [child]
827 Attempt to partition a repo by cutting the parent-child link
828 between two specified commits (they must be adjacent). Does not
829 take a general selection set. It is only necessary to specify the
830 parent commit, unless it has multiple children in which case the
831 child commit must follow (separate it with a comma).
832 If the repo was named 'foo', you will normally end up with two
834 repos named 'foo-early' and 'foo-late' (option and feature events
835 at the beginning of the early segment will be duplicated onto the
836 beginning of the late one.). But if the commit graph would remain
837 connected through another path after the cut, the behavior changes.
838 In this case, if the parent and child were on the same branch
839 'qux', the branch segments are renamed 'qux-early' and 'qux-late'
840 but the repo is not divided.
841 expunge [--notagify] [path | /regexp/]...
843 Expunge files from the selected portion of the repo history; the
844 default is the entire history. The arguments to this command may be
845 paths or Python regular expressions matching paths (regexps must be
846 marked by being surrounded with //).
847 All filemodify (M) operations and delete (D) operations involving a
849 matched file in the selected set of events are disconnected from
850 the repo and put in a removal set. Renames are followed as the tool
851 walks forward in the selection set; each triggers a warning
852 message. If a selected file is a copy (C) target, the copy will be
853 deleted and a warning message issued. If a selected file is a copy
854 source, the copy target will be added to the list of paths to be
855 deleted and a warning issued.
856 After file expunges have been performed, any commits with no
858 remaining file operations will be removed, and any tags pointing to
859 them. By default each deleted commit is replaced with a tag of the
860 form 'emptycommit-ident' on the preceding commit unless --notagify
861 is specified as an argument. Commits with deleted fileops pointing
862 both in and outside the path set are not deleted, but are cloned
863 into the removal set.
864 The removal set is not discarded. It is assembled into a new
866 repository named after the old one with the suffix "-expunges"
867 added. Thus, this command can be used to carve a repository into
868 sections by file path matches.
869 tagify [--canonicalize] [--tipdeletes] [--tagify-merges]
871 Search for empty commits and turn them into tags. Takes an optional
872 selection set argument defaulting to all commits. For each commit
873 in the selection set, turn it into a tag with the same message and
874 author information if it has no fileops. By default merge commits
875 are not considered, even if they have no fileops (thus no tree
876 differences with their first parent). To change that, use the
877 --tagify-merges option.
878 The name of the generated tag will be 'emptycommit-ident', where
880 ident is generated from the legacy ID of the deleted commit, or
881 from its mark, or from its index in the repository, with a
882 disambiguation suffix if needed.
883 With the --canonicalize, tagify tries harder to detect trivial
885 commits by first ensuring that all fileops of selected commits will
886 have an actual effect when processed by fast-import.
887 With the --tipdeletes, tagify also considers branch tips with only
889 deleteall fileops to be candidates for tagification. The
890 corresponding tags get names of the form 'tipdelete-branchname'
891 rather than the default 'emptycommit-ident'.
892 With the --tagify-merges, tagify also tagifies merge commits that
894 have no fileops. When this is done the merge link is move to the
895 yagified commit's parent.
896 coalesce [--debug|--changelog] [timefuzz]
898 Scan the selection set for runs of commits with identical comments
899 close to each other in time (this is a common form of scar tissues
900 in repository up-conversions from older file-oriented
901 version-control systems). Merge these cliques by deleting all but
902 the last commit, in order; fileops from the deleted commits are
903 pushed forward to that last one
904 The optional second argument, if present, is a maximum time
906 separation in seconds; the default is 90 seconds.
907 The default selection set for this command is =C, all commits.
909 Occasionally you may want to restrict it, for example to avoid
910 coalescing unrelated cliques of "*** empty log message ***" commits
911 from CVS lifts.
912 With the --debug option, show messages about mismatches.
914 With the --changelog option, any commit with a comment containing
916 the string 'empty log message' (such as is generated by CVS) and
917 containing exactly one file operation modifying a path ending in
918 ChangeLog is treated specially. Such ChangeLog commits are
919 considered to match any commit before them by content, and will
920 coalesce with it if the committer matches and the commit separation
921 is small enough. This option handles a convention used by Free
922 Software Foundation projects.
923 split {at|by} item
925 The first argument is required to be a commit location; the second
926 is a preposition which indicates which splitting method to use. If
927 the preposition is 'at', then the third argument must be an integer
928 1-origin index of a file operation within the commit. If it is
929 'by', then the third argument must be a pathname to be
930 prefix-matched, pathname match is done first).
931 The commit is copied and inserted into a new position in the event
933 sequence, immediately following itself; the duplicate becomes the
934 child of the original, and replaces it as parent of the original's
935 children. Commit metadata is duplicated; the new commit then gets a
936 new mark. If the new commit has a legacy ID, the suffix '.split' is
937 appended to it.
938 Finally, some file operations - starting at the one matched or
940 indexed by the split argument - are moved forward from the original
941 commit into the new one. Legal indices are 2-n, where n is the
942 number of file operations in the original commit.
943 add {D path | M perm mark path | R source target | C source target}
945 To a specified commit, add a specified fileop.
946 For a D operation to be valid there must be an M operation for the
948 path in the commit's ancestry. For an M operation to be valid, the
949 'perm' part must be a token ending with 755 or 644 and the 'mark'
950 must refer to a blob that precedes the commit location. For an R or
951 C operation to be valid, there must be an M operation for the
952 source in the commit's ancestry.
953 remove [index | path | deletes] [to commit]
955 From a specified commit, remove a specified fileop. The op must be
956 one of (a) the keyword “deletes”, (b) a file path, (c) a file path
957 preceded by an op type set (some subset of the letters DMRCN), or
958 (d) a 1-origin numeric index. The “deletes” keyword selects all D
959 fileops in the commit; the others select one each.
960 If the “to” clause is present, the removed op is appended to the
962 commit specified by the following singleton selection set. This
963 option cannot be combined with “deletes”.
964 Note that this command does not attempt to scavenge blobs even if
966 the deleted fileop might be the only reference to them. This
967 behavior may change in a future release.
968 blob
970 Create a blob at mark :1 after renumbering other marks starting
971 from :2. Data is taken from stdin, which may be a here-doc. This
972 can be used with the add command to patch synthetic data into a
973 repository.
974 renumber
976 Renumber the marks in a repository, from :1 up to :<n> where <n> is
977 the count of the last mark. Just in case an importer ever cares
978 about mark ordering or gaps in the sequence.
979 A side effect of this comment is to clean up stray "done"
981 passthroughs that may have entered the repository via graft
982 operations. After a renumber, the repository will have at most one
983 "done" and it will be at the end of the events.
984 mailbox_out [>outfile]
986 Emit a mailbox file of messages in RFC822 format representing the
987 contents of repository metadata. Takes a selection set; members of
988 the set other than commits, annotated tags, and passthroughs are
989 ignored (that is, presently, blobs and resets).
990 The output from this command can optionally be redirected to a
992 named output file. Prefix the filename with “>” and give it as a
993 following argument.
994 May have an option --filter, followed by = and a /-enclosed regular
996 expression. If this is given, only headers with names matching it
997 are emitted. In this context the name of the header includes its
998 trailing colon.
999 mailbox_in [<infile] [--changed >outfile]
1001 Accept a mailbox file of messages in RFC822 format representing the
1002 contents of the metadata in selected commits and annotated tags.
1003 Takes no selection set. If there is an argument it will be taken as
1004 the name of a mailbox file to read from; no argument, or one of
1005 '-'; reads from standard input.
1006 Users should be aware that modifying an Event-Number or Event-Mark
1008 field will change which event the update from that message is
1009 applied to. This is unlikely to have good results.
1010 If the Event-Number and Event-Mark fields are absent, the
1012 mailbox_in logic will attempt to match the commit or tag first by
1013 Legacy-ID, then by a unique committer ID and timestamp pair.
1014 If output is redirected and the modifier “--changed” appears, a
1016 minimal set of modifications actually made is written to the output
1017 file in a form that can be fed back in.
1018 setfield attribute value
1020 In the selected objects (defaulting to none) set every instance of
1021 a named field to a string value. The string may be quoted to
1022 include whitespace, and use backslash escapes interpreted by the
1023 Python string-escape codec, such as \n and \t.
1024 Attempts to set nonexistent attributes are ignored. Valid values
1026 for the attribute are internal Python field names; in particular,
1027 for commits, “comment” and “branch” are legal. Consult the source
1028 code for other interesting values.
1029 append [--rstrip] [>text]
1031 Append text to the comments of commits and tags in the specified
1032 selection set. The text is the first token of the command and may
1033 be a quoted string. C-style escape sequences in the string are
1034 interpreted using Python's string_decode codec.
1035 If the option --rstrip is given, the comment is right-stripped
1037 before the new text is appended.
1038 filter [--shell|--regex|--replace|--dedos]
1040 Run blobs, commit comments, or tag comments in the selection set
1041 through the filter specified on the command line.
1042 In any mode other than --dedos, attempting to specify a selection
1044 set including both blobs and non-blobs (that is, commits or tags)
1045 throws an error. Inline content in commits is filtered when the
1046 selection set contains (only) blobs and the commit is within the
1047 range bounded by the earliest and latest blob in the specification.
1048 When filtering blobs, if the command line contains the magic cookie
1050 '%PATHS%' it is replaced with a space-separated list of all paths
1051 that reference the blob.
1052 With --shell, the remainder of the line specifies a filter as a
1054 shell command. Each blob or comment is presented to the filter on
1055 standard input; the content is replaced with whatever the filter
1056 emits to standard output.
1057 With --regex, the remainder of the line is expected to be a Python
1059 regular expression substitution written as /from/to/ with from and
1060 to being passed as arguments to the standard re.sub() function and
1061 it applied to modify the content. Actually, any non-space character
1062 will work as a delimiter in place of the /; this makes it easier to
1063 use / in patterns. Ordinarily only the first such substitution is
1064 performed; putting 'g' after the slash replaces globally, and a
1065 numeric literal gives the maximum number of substitutions to
1066 perform. Other flags available restrict substitution scope - 'c'
1067 for comment text only, 'C' for committer name only, 'a' for author
1068 names only. Note that parsing of a --regex argument will be
1069 confused by ant substring consisting of whitespace followed by #;
1070 use "\s" rather than whitespace to avoid this.
1071 With --replace, the behavior is like --regexp but the expressions
1073 are not interpreted as regular expressions. (This is slightly
1074 faster).
1075 With --dedos, DOS/Windows-style \r\n line terminators are replaced
1077 with \n.
1078 transcode codec
1080 Transcode blobs, commit comments and committer/author names, or tag
1081 comments and tag committer names in the selection set to UTF-8 from
1082 the character encoding specified on the command line.
1083 Attempting to specify a selection set including both blobs and
1085 non-blobs (that is, commits or tags) throws an error. Inline
1086 content in commits is filtered when the selection set contains
1087 (only) blobs and the commit is within the range bounded by the
1088 earliest and latest blob in the specification.
1089 The encoding argument must name one of the codecs known to the
1091 Python standard codecs library. In particular, 'latin-1' is a valid
1092 codec name.
1093 Errors in this command are fatal, because an error may leave
1095 repository objects in a damaged state.
1096 The theory behind the design of this command is that the repository
1098 might contain a mixture of encodings used to enter commit metadata
1099 by different people at different times. After using =I to identify
1100 metadata containing non-Unicode high bytes in text, a human must
1101 use context to identify which particular encodings were used in
1102 particular event spans and compose appropriate transcode commands
1103 to fix them up.
1104 edit
1106 Report the selection set of events to a tempfile as mailbox_out
1107 does, call an editor on it, and update from the result as
1108 mailbox_in does. If you do not specify an editor name as second
1109 argument, it will be taken from the $EDITOR variable in your
1110 environment.
1111 Normally this command ignores blobs because mailbox_out does.
1113 However, if you specify a selection set consisting of a single
1114 blob, your editor will be called directly on the blob file.
1115 timeoffset offset [timezone]
1117 Apply a time offset to all time/date stamps in the selected set. An
1118 offset argument is required; it may be in the form [+-]ss,
1119 [+-]mm:ss or [+-]hh:mm:ss. The leading sign is required to
1120 distinguish it from a selection expression.
1121 Optionally you may also specify another argument in the form
1123 [+-]hhmm, a timezone literal to apply. To apply a timezone without
1124 an offset, use an offset literal of +0 or -0.
1125 unite [--prune] reponame...
1127 Unite repositories. Name any number of loaded repositories; they
1128 will be united into one union repo and removed from the load list.
1129 The union repo will be selected.
1130 The root of each repo (other than the oldest repo) will be grafted
1132 as a child to the last commit in the dump with a preceding commit
1133 date. This will produce a union repository with one branch for each
1134 part. Running last to first, duplicate tag and branch names will be
1135 disambiguated using the source repository name (thus, recent
1136 duplicates will get priority over older ones). After all grafts,
1137 marks will be renumbered.
1138 The name of the new repo will be the names of all parts
1140 concatenated, separated by '+'. It will have no source directory or
1141 preferred system type.
1142 With the option --prune, at each join D operations for every
1144 ancestral file existing will be prepended to the root commit, then
1145 it will be canonicalized using the rules for squashing the effect
1146 will be that only files with properly matching M, R, and C
1147 operations in the root survive.
1148 graft [--prune] reponame
1150 For when unite doesn't give you enough control. This command may
1151 have either of two forms, selected by the size of the selection
1152 set. The first argument is always required to be the name of a
1153 loaded repo.
1154 If the selection set is of size 1, it must identify a single commit
1156 in the currently chosen repo; in this case the name repo's root
1157 will become a child of the specified commit. If the selection set
1158 is empty, the named repo must contain one or more callouts matching
1159 a commits in the currently chosen repo.
1160 Labels and branches in the named repo are prefixed with its name;
1162 then it is grafted to the selected one. Any other callouts in the
1163 named repo are also resolved in the context of the currently chosen
1164 one. Finally, the named repo is removed from the load list.
1165 With the option --prune, prepend a deleteall operation into the
1167 root of the grafted repository.
1168 path [source] rename [--force] [target]
1170 Rename a path in every fileop of every selected commit. The default
1171 selection set is all commits. The first argument is interpreted as
1172 a Python regular expression to match against paths; the second may
1173 contain back-reference syntax.
1174 Ordinarily, if the target path already exists in the fileops, or is
1176 visible in the ancestry of the commit, this command throws an
1177 error. With the --force option, these checks are skipped.
1178 paths [{sub|sup}] [dirname] [>outfile]
1180 Takes a selection set. Without a modifier, list all paths touched
1181 by fileops in the selection set (which defaults to the entire
1182 repo). This reporting variant does >-redirection.
1183 With the 'sub' modifier, take a second argument that is a directory
1185 name and prepend it to every path. With the 'sup' modifier, strip
1186 any directory argument from the start of the path if it appears
1187 there; with no argument, strip the first directory component from
1188 every path.
1189 merge
1191 Create a merge link. Takes a selection set argument, ignoring all
1192 but the lowest (source) and highest (target) members. Creates a
1193 merge link from the highest member (child) to the lowest (parent).
1194 unmerge
1196 Linearize a commit. Takes a selection set argument, which must
1197 resolve to a single commit, and removes all its parents except for
1198 the first.
1199 It is equivalent to reparent first_parent,commit rebase, where
1201 commit is the same selection set as used with unmerge and
1202 first_parent is a set resolving commit's first parent (see the
1203 reparent command below
1204 The main interest of the unmerge is that you don't have to find and
1206 specify the first parent yourself, saving time and avoiding errors
1207 when nearby surgery would make a manual first parent argument
1208 stale.
1209 reparent [options...] [policy]
1211 Changes the parent list of a commit. Takes a selection set, zero or
1212 more option arguments, and an optional policy argument.
1213 Selection set:
1215 The selection set must resolve to one or more commits. The
1216 selected commit with the highest event number (not necessarily
1217 the last one selected) is the commit to modify. The remainder
1218 of the selected commits, if any, become its parents: the
1219 selected commit with the lowest event number (which is not
1220 necessarily the first one selected) becomes the first parent,
1221 the selected commit with second lowest event number becomes the
1222 second parent, and so on. All original parent links are
1223 removed. Examples:
1224 # this makes 17 the parent of 33
1226 17,33 reparent
1227 # this also makes 17 the parent of 33
1229 33,17 reparent
1230 # this makes 33 a root (parentless) commit
1232 33 reparent
1233 # this makes 33 an octopus merge commit. its first parent
1235 # is commit 15, second parent is 17, and third parent is 22
1236 22,33,15,17 reparent
1237 Options:
1239 --use-order
1241 Use the selection order to determine which selected commit
1242 is the commit to modify and which are the parents (and if
1243 there are multiple parents, their order). The last selected
1244 commit (not necessarily the one with the highest event
1245 number) is the commit to modify, the first selected commit
1246 (not necessarily the one with the lowest event number)
1247 becomes the first parent, the second selected commit
1248 becomes the second parent, and so on. Examples:
1249 # this makes 33 the parent of 17
1251 33|17 reparent --use-order
1252 # this makes 17 an octopus merge commit. its first parent
1254 # is commit 22, second parent is 33, and third parent is 15
1255 22,33,15|17 reparent --use-order
1256 Because ancestor commit events must appear before their
1258 descendants, giving a commit with a low event number a
1259 parent with a high event number triggers a re-sort of the
1260 events. A re-sort assigns different event numbers to some
1261 or all of the events. Re-sorting only works if the
1262 reparenting does not introduce any cycles. To swap the
1263 order of two commits that have an ancestor–descendant
1264 relationship without introducing a cycle during the
1265 process, you must reparent the descendant commit first.
1266 Policy:
1268 By default, the manifest of the reparented commit is computed
1269 before modifying it; a deleteall and some fileops are prepended
1270 so that the manifest stays unchanged even when the first parent
1271 has been changed. This behavior can be changed by specifying a
1272 policy:
1273 rebase
1275 Inhibits the default behavior—no deleteall is issued and
1276 the tree contents of all descendents can be modified as a
1277 result.
1278 branch branchname... {rename|delete} [arg]
1280 Rename or delete a branch (and any associated resets). First
1281 argument must be an existing branch name; second argument must one
1282 of the verbs 'rename' or 'delete'.
1283 For a 'rename', the third argument may be any token that is a
1285 syntactically valid branch name (but not the name of an existing
1286 branch). For a 'delete', no third argument is required.
1287 For either name, if it does not contain a '/' the prefix
1289 'refs/heads' is prepended.
1290 tag tagname... {move|rename|delete} [arg].
1292 Create, move, rename, or delete a tag.
1293 Creation is a special case. First argument is a nane, which must
1295 not be an existing tag. Takes a singleton event second argument
1296 which must point to a commit. A tag object pointing to the commit
1297 is created and inserted just after the last tag in the repo (or
1298 just after the last commit if there are no tags). The tagger,
1299 committish, and comment fields are copied from the commit's
1300 committer, mark, and comment fields.
1301 Otherwise, first argument must be an existing tag name; second
1303 argument must be one of the verbs “move”, “rename”, or “delete”.
1304 For a “move”, a third argument must be a singleton selection set.
1306 For a “rename”, the third argument may be any token that is a
1307 syntactically valid tag name (but not the name of an existing tag).
1308 For a “delete”, no third argument is required.
1309 The behavior of this command is complex because features which
1311 present as tags may be any of three things: (1) True tag objects,
1312 (2) lightweight tags, actually sequences of commits with a common
1313 branchname beginning with “refs/tags” - in this case the tag is
1314 considered to point to the last commit in the sequence, (3) Reset
1315 objects. These may occur in combination; in fact, stream exporters
1316 from systems with annotation tags commonly express each of these as
1317 a true tag object (1) pointing at the tip commit of a sequence (2)
1318 in which the basename of the common branch field is identical to
1319 the tag name. An exporter that generates lightweight-tagged commit
1320 sequences (2) may or may not generate resets pointing at their tip
1321 commits.
1322 This command tries to handle all combinations in a natural way by
1324 doing up to three operations on any true tag, commit sequence, and
1325 reset matching the source name. In a rename, all are renamed
1326 together. In a delete, any matching tag or reset is deleted; then
1327 matching branch fields are changed to match the branch of the
1328 unique descendent of the tagged commit, if there is one. When a tag
1329 is moved, no branch fields are changed and a warning is issued.
1330 Attempts to delete a lightweight tag may fail with the message
1332 “couldn't determine a unique successor”. When this happens, the tag
1333 is on a commit with multiple children that have different branch
1334 labels. There is a hole in the specification of git fast-import
1335 streams that leaves it uncertain how branch labels can be safely
1336 reassigned in this case; rather than do something risky,
1337 reposurgeon throws a recoverable error.
1338 reset resetname... {move|rename|delete} [arg].
1340 Move, rename, or delete a reset. First argument must match an
1341 existing reset name; second argument must be one of the verbs
1342 “move”, “rename”, or “delete”.
1343 For a “move”, a third argument must be a singleton selection set.
1345 For a “rename”, the third argument may be any token token that
1346 matches a syntactically valid reset name (but not the name of an
1347 existing reset). For a “delete”, no third argument is required.
1348 For either name, if it does not contain a “/” the prefix “heads/”
1350 is prepended. If it does not begin with “refs/”, “refs/” is
1351 prepended.
1352 An argument matches a reset's name if it is either the entire
1354 reference (refs/heads/FOO or refs/tags/FOO for some some value of
1355 FOO) or the basename (e.g. FOO), or a suffix of the form heads/FOO
1356 or tags/FOO. An unqualified basename is assumed to refer to a head.
1357 When a reset is renamed, commit branch fields matching the tag are
1359 renamed with it to match. When a reset is deleted, matching branch
1360 fields are changed to match the branch of the unique descendent of
1361 the tip commit of the associated branch, if there is one. When a
1362 reset is moved, no branch fields are changed.
1363 debranch source-branch... [target-branch].
1365 Takes one or two arguments which must be the names of source and
1366 target branches; if the second (target) argument is omitted it
1367 defaults to refs/heads/master. Any trailing segment of a branch
1368 name is accepted as a synonym for it; thus master is the same as
1369 refs/heads/master. Does not take a selection set.
1370 The history of the source branch is merged into the history of the
1372 target branch, becoming the history of a subdirectory with the name
1373 of the source branch. Any resets of the source branch are removed.
1374 strip [blobs|reduce].
1376 Reduce the selected repository to make it a more tractable test
1377 case. Use this when reporting bugs.
1378 With the modifier 'blobs', replace each blob in the repository with
1380 a small, self-identifying stub, leaving all metadata and DAG
1381 topology intact. This is useful when you are reporting a bug, for
1382 reducing large repositories to test cases of manageable size.
1383 A selection set is effective only with the 'blobs' option,
1385 defaulting to all blobs. The 'reduce' mode always acts on the
1386 entire repository.
1387 With the modifier 'reduce', perform a topological reduction that
1389 throws out uninteresting commits. If a commit has all file
1390 modifications (no deletions or copies or renames) and has exactly
1391 one ancestor and one descendant, then it may be boring. To be fully
1392 boring, it must also not be referred to by any tag or reset.
1393 Interesting commits are not boring, or have a non-boring parent or
1394 non-boring child.
1395 With no modifiers, this command strips blobs.
1397 ignores [rename]. [translate]. [defaults].
1399 Intelligent handling of ignore-pattern files. This command fails if
1400 no repository has been selected or no preferred write type has been
1401 set for the repository. It does not take a selection set.
1402 If the rename modifier is present, this command attempts to rename
1404 all ignore-pattern files to whatever is appropriate for the
1405 preferred type - e.g. .gitignore for git, .hgignore for hg, etc.
1406 This option does not cause any translation of the ignore files it
1407 renames.
1408 If the translate modifier is present, syntax translation of each
1410 ignore file is attempted. At present, the only transformation the
1411 code knows is to prepend a 'syntax: glob' header if the preferred
1412 type is hg.
1413 If the defaults modifier is present, the command attempts to
1415 prepend these default patterns to all ignore files. If no ignore
1416 file is created by the first commit, it will be modified to create
1417 one containing the defaults. This command will error out on prefer
1418 types that have no default ignore patterns (git and hg, in
1419 particular). It will also error out when it knows the import tool
1420 has already set default patterns.
1421 REFERENCE LIFTING
1423 This group of commands is meant for fixing up references in commits
1424 that are in the format of older version control systems. The general
1425 workflow is this: first, go over the comment history and change all
1426 old-fashioned commit references into machine-parseable cookies. Then,
1427 automatically turn the machine-parseable cookie into action stamps. The
1428 point of dividing the process this way is that the first part is hard
1429 for a machine to get right, while the second part is prone to errors
1430 when a human does it.
1431 A Subversion cookie is a comment substring of the form [[SVN:ddddd]]
1433 (example: [[SVN:2355]] with the revision read directly via the
1434 Subversion exporter, deduced from git-svn metadata, or matching a
1435 $Revision$ header embedded in blob data for the filename.
1436 A CVS cookie is a comment substring of the form
1438 [[CVS:filename:revision]] (example: [[CVS:src/README:1.23]] with the
1439 revision matching a CVS $Id$ or $Revision$ header embedded in blob data
1440 for the filename.
1441 A mark cookie is of the form [[:dddd]] and is simply a reference to the
1443 specified mark. You may want to hand-patch this in when one of previous
1444 forms is inconvenient.
1445 An action stamp is an RFC3339 timestamp, followed by a '!', followed by
1447 an author email address (author rather than committer because that
1448 timestamp is not changed when a patch is replayed on to a branch). It
1449 attempts to refer to a commit without being VCS-specific. Thus, instead
1450 of "commit 304a53c2" or "r2355",
1451 "2011-10-25T15:11:09Z!fred@foonly.com".
1452 The following git aliases allow git to work directly with action
1454 stamps. Append it to your ~/.gitconfig; if you already have an [alias]
1455 section, leave off the first line.
1456 [alias]
1459 # git stamp <commit-ish> - print a reposurgeon-style action stamp
1460 stamp = show -s --format='%cI!%ce'
1461 # git scommit <stamp> <rev-list-args> - list most recent commit that matches <stamp>.
1463 # Must also specify a branch to search or --all, after these arguments.
1464 scommit = "!f(){ d=${1%%!*}; a=${1##*!}; arg=\"--until=$d -1\"; if [ $a != $1 ]; then arg=\"$arg --committer=$a\"; fi; shift; git rev-list $arg ${1:+\"$@\"}; }; f"
1465 # git scommits <stamp> <rev-list-args> - as above, but list all matching commits.
1467 scommits = "!f(){ d=${1%%!*}; a=${1##*!}; arg=\"--until=$d --after $d\"; if [ $a != $1 ]; then arg=\"$arg --committer=$a\"; fi; shift; git rev-list $arg ${1:+\"$@\"}; }; f"
1468 # git smaster <stamp> - list most recent commit on master that matches <stamp>.
1470 smaster = "!f(){ git scommit \"$1\" master --first-parent; }; f"
1471 smasters = "!f(){ git scommits \"$1\" master --first-parent; }; f"
1472 # git shs <stamp> - show the commits on master that match <stamp>.
1474 shs = "!f(){ stamp=$(git smasters $1); shift; git show ${stamp:?not found} $*; }; f"
1475 # git slog <stamp> <log-args> - start git log at <stamp> on master
1477 slog = "!f(){ stamp=$(git smaster $1); shift; git log ${stamp:?not found} $*; }; f"
1478 # git sco <stamp> - check out most recent commit on master that matches <stamp>.
1480 sco = "!f(){ stamp=$(git smaster $1); shift; git checkout ${stamp:?not found} $*; }; f"
1481 There is a rare case in which an action stamp will not refer uniquely
1484 to one commit. It is theoretically possible that the same author might
1485 check in revisions on different branches within the one-second
1486 resolution of the timestamps in a fast-import stream. There is nothing
1487 to be done about this; tools using action stamps need to be aware of
1488 the possibility and throw a warning when it occurs.
1489 In order to support reference lifting, reposurgeon internally builds a
1491 legacy-reference map that associates revision identifiers in older
1492 version-control systems with commits. The contents of this map comes
1493 from three places: (1) cvs2svn:rev properties if the repository was
1494 read from a Subversion dump stream, (2) $Id$ and $Revision$ headers in
1495 repository files, and (3) the .git/cvs-revisions created by git
1496 cvsimport.
1497 The detailed sequence for lifting possible references is this: first,
1499 find possible CVS and Subversion references with the references or =N
1500 visibility set; then replace them with equivalent cookies; then run
1501 references lift to turn the cookies into action stamps (using the
1502 information in the legacy-reference map) without having to do the
1503 lookup by hand.
1504 references [list|edit|lift] [>outfile]
1506 With the modifier 'list', list commit and tag comments for strings
1507 that might be CVS- or Subversion-style revision identifiers. This
1508 will be useful when you want to replace them with equivalent
1509 cookies that can automatically be translated into VCS-independent
1510 action stamps. This reporting command supports >-redirection. It is
1511 equivalent to '=N list'.
1512 With the modifier 'edit', edit the set where revision IDs are
1514 found. This is equivalent to '=N edit'.
1515 With the modifier "lift", attempt to resolve Subversion and CVS
1517 cookies in comments into action stamps using the legacy map. An
1518 action stamp is a timestamp/email/sequence-number combination
1519 uniquely identifying the commit associated with that blob, as
1520 described in the section called “TRANSLATION STYLE”.
1521 It is not guaranteed that every such reference will be resolved, or
1523 even that any at all will be. Normally all references in history
1524 from a Subversion repository will resolve, but CVS references are
1525 less likely to be resolvable.
1526 VARIABLES, MACROS AND EXTENSIONS
1528 Occasionally you will need to issue a large number of complex surgical
1529 commands of very similar form, and it's convenient to be able to
1530 package that form so you don't need to do a lot of error-prone typing.
1531 For those occasions, reposurgeon supports simple forms of named
1532 variables and macro expansion.
1533 assign [name]
1535 Compute a leading selection set and assign it to a symbolic name.
1536 It is an error to assign to a name that is already assigned, or to
1537 any existing branch name. Assignments may be cleared by sequence
1538 mutations (though not ordinary deletions); you will see a warning
1539 when this occurs.
1540 With no selection set and no name, list all assignments.>
1542 Use this to optimize out location and selection computations that
1544 would otherwise be performed repeatedly, e.g. in macro calls.
1545 unassign [name]
1547 Unassign a symbolic name. Throws an error if the name is not
1548 assigned.
1549 names [>outfile]
1551 List the names of all known branches and tags. Tells you what
1552 things are legal within angle brackets and parentheses.
1553 define name body
1555 Define a macro. The first whitespace-separated token is the name;
1556 the remainder of the line is the body, unless it is “{”, which
1557 begins a multi-line macro terminated by a line beginning with “}”.
1558 A later “do” call can invoke this macro.
1560 The command “define” by itself without a name or body produces a
1562 macro list.
1563 do name arguments...
1565 Expand and perform a macro. The first whitespace-separated token is
1566 the name of the macro to be called; remaining tokens replace {0},
1567 {1}... in the macro definition (the conventions used are those of
1568 the Python format method). Tokens may contain whitespace if they
1569 are string-quoted; string quotes are stripped. Macros can call
1570 macros.
1571 If the macro expansion does not itself begin with a selection set,
1573 whatever set was specified before the "do" keyword is available to
1574 the command generated by the expansion.
1575 undefine name]
1577 Undefine the named macro.
1578 Here's an example to illustrate how you might use this. In CVS
1580 repositories of projects that use the GNU ChangeLog convention, a very
1581 common pre-conversion artifact is a commit with the comment "***empty
1582 log message***" that modifies only a ChangeLog entry explaining the
1583 commit immediately previous to it. The following
1584 define changelog <{0}> & /empty log message/ squash --pushback
1586 do changelog 2012-08-14T21:51:35Z
1587 do changelog 2012-08-08T22:52:14Z
1588 do changelog 2012-08-07T04:48:26Z
1589 do changelog 2012-08-08T07:19:09Z
1590 do changelog 2012-07-28T18:40:10Z
1591 is equivalent to the more verbose
1593 <2012-08-14T21:51:35Z> & /empty log message/ squash --pushback
1595 <2012-08-08T22:52:14Z> & /empty log message/ squash --pushback
1596 <2012-08-07T04:48:26Z> & /empty log message/ squash --pushback
1597 <2012-08-08T07:19:09Z> & /empty log message/ squash --pushback
1598 <2012-07-28T18:40:10Z> & /empty log message/ squash --pushback
1599 but you are less likely to make difficult-to-notice errors typing the
1601 first version.
1602 (Also note how the text regexp acts as a failsafe against the
1604 possibility of typing a wrong date that doesn't refer to a commit with
1605 an empty comment. This was a real-world example from the CVS-to-git
1606 conversion of groff.)
1607 When even a macro is not enough, you can write and call custom Python
1609 extensions.
1610 exec name
1612 Execute custom code from standard input (normally a file via <
1613 redirection). Use this to set up custom extension functions for
1614 later eval calls. The code has full access to all internal data
1615 structures. Functions defined are accessible to later eval calls.
1616 This can be called in a script with the extension code in a
1618 here-doc.
1619 eval function-name
1621 Evaluate a line of code in the current interpreter context.
1622 Typically this will be a call to a function defined by a previous
1623 exec. The variables _repository and _selection will have the
1624 obvious values. Note that _selection will be a list of integers,
1625 not objects.
1626 script filename [arg...]
1628 Takes a filename and optional following arguments. Reads each line
1629 from the file and executes it as a command.
1630 During execution of the script, the script name replaces the string
1632 $0 and the optional following arguments (if any) replace the
1633 strings $1, $2 ... $n in the script text. This is done before
1634 tokenization, so the $1 in a string like “foo$1bar” will be
1635 expanded. Additionally, $$ is expanded to the current process ID
1636 (which may be useful for scripts that use tempfiles).
1637 Within scripts (and only within scripts) reposurgeon accepts a
1639 slightly extended syntax: First, a backslash ending a line signals
1640 that the command continues on the next line. Any number of
1641 consecutive lines thus escaped are concatenated, without the ending
1642 backslashes, prior to evaluation. Second, a command that takes an
1643 input filename argument can instead take literal following data in
1644 the syntax of a shell here-document. That is: if the filename is
1645 replaced by "<<EOF", all following lines in the script up to a
1646 terminating line consisting only of "EOF" will be read, placed in a
1647 temporary file, and that file fed to the command and afterwards
1648 deleted. EOF may be replaced by any string. Backslashes have no
1649 special meaning while reading a here-document.
1650 Scripts may have comments. Any line beginning with a '#' is
1652 ignored. If a line has a trailing position that begins with one or
1653 more whitespace characters followed by '#', that trailing portion
1654 is ignored.
1655 ARTIFACT REMOVAL
1657 Some commands automate fixing various kinds of artifacts associated
1658 with repository conversions from older systems.
1659 authors [read|write] [<filename] [>filename]
1661 Apply or dump author-map information for the specified selection
1662 set, defaulting to all events.
1663 Lifts from CVS and Subversion may have only usernames local to the
1665 repository host in committer and author IDs. DVCSes want email
1666 addresses (net-wide identifiers) and complete names. To supply the
1667 map from one to the other, an authors file is expected to consist
1668 of lines each beginning with a local user ID, followed by a '='
1669 (possibly surrounded by whitespace) followed by a full name and
1670 email address, optionally followed by a timezone offset field.
1671 Thus:
1672 ferd = Ferd J. Foonly <foonly@foo.com> -0500
1674 An authors file may have comment lines beginning with '#'; these
1676 are ignored.
1677 When an authors file is applied, email addresses in committer and
1679 author metadata for which the local ID matches between < and @ are
1680 replaced according to the mapping (this handles git-svn lifts).
1681 Alternatively, if the local ID is the entire address, this is also
1682 considered a match (this handles what git-cvsimport and cvs2git do)
1683 With the 'read' modifier, or no modifier, apply author mapping data
1685 (from standard input or a <-redirected file). May be useful if you
1686 are editing a repo or dump created by cvs2git or by git-svn invoked
1687 without -A.
1688 With the 'write' modifier, write a mapping file that could be
1690 interpreted by authors read, with entries for each unique
1691 committer, author, and tagger (to standard output or a <-redirected
1692 mapping file). This may be helpful as a start on building an
1693 authors file, though each part to the right of an equals sign will
1694 need editing.
1695 branchify [path-set]
1697 Specify the list of directories to be treated as potential branches
1698 (to become tags if there are no modifications after the creation
1699 copies) when analyzing a Subversion repo. This list is ignored when
1700 the --nobranch read option is used. It defaults to the 'standard
1701 layout' set of directories, plus any unrecognized directories in
1702 the repository root.
1703 With no arguments, displays the current branchification set.
1705 An asterisk at the end of a path in the set means 'all immediate
1707 subdirectories of this path, unless they are part of another
1708 (longer) path in the branchify set'.
1709 Note that the branchify set is a property of the reposurgeon
1711 interpreter, not of any individual repository, and will persist
1712 across Subversion dumpfile reads. This may lead to unexpected
1713 results if you forget to re-set it.
1714 branchify_map [/regex/branch/...]
1716 Specify the list of regular expressions used for mapping the svn
1717 branches that are detected by branchify. If none of the expressions
1718 match the default behaviour applies. This maps a branch to the name
1719 of the last directory, except for trunk and “*” which are mapped to
1720 master and root.
1721 With no arguments the current regex replacement pairs are shown.
1723 Passing 'reset' will clear the mapping.
1724 The branchify command will match each branch name against regex1
1726 and if it matches rewrite its branch name to branch1. If not it
1727 will try regex2 and so forth until it either found a matching regex
1728 or there are no regexs left. The regular expressions should be in
1729 Python's[2]. format. The branch name can use backreferences (see
1730 the re.sub function in the Python documentation).
1731 Note that the regular expressions are appended to 'refs/' without
1733 either the needed 'heads/' or 'tags/'. This allows for choosing the
1734 right kind of branch type.
1735 While the syntax template above uses slashes, any first character
1737 will be used as a delimeter (and you will need to use a different
1738 one in the common case that the paths contain slashes).
1739 Note that the branchify_map set is a property of the reposurgeon
1741 interpreter, not of any individual repository, and will persist
1742 across Subversion dumpfile reads. This may lead to unexpected
1743 results if you forget to re-set it.
1744 EXAMINING TREE STATES
1746 manifest [regular expression] [>outfile]
1747 Takes an optional selection set argument defaulting to all commits,
1748 and an optional Python regular expression. For each commit in the
1749 selection set, print the mapping of all paths in that commit tree
1750 to the corresponding blob marks, mirroring what files would be
1751 created in a checkout of the commit. If a regular expression is
1752 given, only print "path -> mark" lines for paths matching it. This
1753 command supports > redirection.
1754 checkout directory
1756 Takes a selection set which must resolve to a single commit, and a
1757 second argument. The second argument is interpreted as a directory
1758 name. The state of the code tree at that commit is materialized
1759 beneath the directory.
1760 diff [>outfile]
1762 Display the difference between commits. Takes a selection-set
1763 argument which must resolve to exactly two commits. Supports output
1764 redirection.
1765 HOUSEKEEPING
1767 These are backed up by the following housekeeping commands, none of
1768 which take a selection set:
1769 help
1771 Get help on the interpreter commands. Optionally follow with
1772 whitespace and a command name; with no argument, lists all
1773 commands. '?' also invokes this.
1774 shell
1776 Execute the shell command given in the remainder of the line. '!'
1777 also invokes this.
1778 prefer [repotype]
1780 With no arguments, describe capabilities of all supported systems.
1781 With an argument (which must be the name of a supported system)
1782 this has two effects:
1783 First, if there are multiple repositories in a directory you do a
1785 read on, reposurgeon will read the preferred one (otherwise it will
1786 complain that it can't choose among them).
1787 Secondly, this will change reposurgeon's preferred type for output.
1789 This means that you do a write to a directory, it will build a repo
1790 of the preferred type rather than its original type (if it had
1791 one).
1792 If no preferred type has been explicitly selected, reading in a
1794 repository (but not a fast-import stream) will implicitly set the
1795 preferred type to the type of that repository.
1796 In older versions of reposurgeon this command changed the type of
1798 the selected repository, if there is one. That behavior interacted
1799 badly with attempts to interpret legacy IDs and has been removed.
1800 sourcetype [repotype]
1802 Report (with no arguments) or select (with one argument) the
1803 current repository's source type. This type is normally set at
1804 repository-read time, but may remain unset if the source was a
1805 stream file.
1806 The source type affects the interpretation of legacy IDs (for
1808 purposes of the =N visibility set and the 'references' command) by
1809 controlling the regular expressions used to recognize them. If no
1810 preferred output type has been set, it may also change the output
1811 format of stream files made from the repository.
1812 The source type is reliably set whenever a live repository is read,
1814 or when a Subversion stream or Fossil dump is interpreted but not
1815 necessarily by other stream files. Streams generated by cvs-fast-
1816 export(1) using the --reposurgeon are detected as CVS. In some
1817 other cases, the source system is detected from the presence of
1818 magic $-headers in contents blobs.
1819 INSTRUMENTATION
1821 A few commands have been implemented primarily for debugging and
1822 regression-testing purposes, but may be useful in unusual
1823 circumstances.
1824 The output of most of these commands can individually be redirected to
1826 a named output file. Where indicated in the syntax, you can prefix the
1827 output filename with “>” and give it as a following argument.
1828 index [>outfile]
1830 Display four columns of info on objects in the selection set: their
1831 number, their type, the associate mark (or '-' if no mark) and a
1832 summary field varying by type. For a branch or tag it's the
1833 reference; for a commit it's the commit branch; for a blob it's the
1834 repository path of the file in the blob.
1835 The default selection set for this command is =CTRU, all objects
1837 except blobs.
1838 resolve [label-text...]
1840 Does nothing but resolve a selection-set expression and echo the
1841 resulting event-number set to standard output. The remainder of the
1842 line after the command is used as a label for the output.
1843 Implemented mainly for regression testing, but may be useful for
1845 exploring the selection-set language.
1846 verbose [n]
1848 'verbose 1' enables the progress meter and messages, 'verbose 0'
1849 disables them. Higher levels of verbosity are available but
1850 intended for developers only.
1851 quiet [on | off]
1853 Without an argument, this command requests a report of the quiet
1854 boolean; with the argument 'on' or 'off' it is changed. When quiet
1855 is on, time-varying report fields which would otherwise cause
1856 spurious failures in regression testing are suppressed.
1857 print output-text...
1859 Does nothing but ship its argument line to standard output. Useful
1860 in regression tests.
1861 echo [number]
1863 'echo 1' causes each reposurgeon command to be echoed to standard
1864 output just before its output. This can be useful in constructing
1865 regression tests that are easily checked by eyeball.
1866 version [version...]
1868 With no argument, display the program version and the list of VCSes
1869 directly supported. With argument, declare the major version
1870 (single digit) or full version (major.minor) under which the
1871 enclosing script was developed. The program will error out if the
1872 major version has changed (which means the surgical language is not
1873 backwards compatible).
1874 It is good practice to start your lift script with a version
1876 requirement, especially if you are going to archive it for later
1877 reference.
1878 prompt [format...]
1880 Set the command prompt format to the value of the command line;
1881 with an empty command line, display it. The prompt format is
1882 evaluated in Python after each command with the following
1883 dictionary substitutions:
1884 chosen
1886 The name of the selected repository, or None if none is
1887 currently selected.
1888 Thus, one useful format might be 'rs[%(chosen)s]%% '.
1890 More format items may be added in the future. The default prompt
1892 corresponds to the format 'reposurgeon%% '. The format line is
1893 evaluated with shell quotng of tokens, so that spaces can be
1894 included.
1895 history
1897 List the commands you have entered this session.
1898 legacy [read|write] [<filename] [>filename]
1900 Apply or list legacy-reference information. Does not take a
1901 selection set. The 'read' variant reads from standard input or a
1902 <-redirected filename; the 'write' variant writes to standard
1903 output or a >-redirected filename.
1904 A legacy-reference file maps reference cookies to (committer,
1906 commit-date, sequence-number) pairs; these in turn (should)
1907 uniquely identify a commit. The format is two whitespace-separated
1908 fields: the cookie followed by an action stamp identifying the
1909 commit.
1910 It should not normally be necessary to use this command. The legacy
1912 map is automatically preserved through repository reads and
1913 rebuilds, being stored in the file legacy-map under the repository
1914 subdirectory..
1915 set [option]
1917 Turn on an option flag. With no arguments, list all options
1918 Most options are described in conjunction with the specific
1920 operations that the modify. One of general interest is
1921 “compressblobs”; this enables compression on the blob files in the
1922 internal representation reposurgeon uses for editing repositories.
1923 With this option, reading and writing of repositories is slower,
1924 but editing a repository requires less (sometimes much less) disk
1925 space.
1926 clear [option]
1928 Turn off an option flag. With no arguments, list all options
1929 profile
1931 Enable profiling. Profile statistics are dumped to the path given
1932 as argument. Must be one of the initial command-line arguments, and
1933 gathers statistics only on code executed via '-'.
1934 timing
1936 Display statistics on phase timing in repository analysis. Mainly
1937 of interest to developers trying to speed up the program.
1938 exit
1940 Exit, reporting the time. Included here because, while EOT will
1941 also cleanly exit the interpreter, this command reports elapsed
1942 time since start.
1943
1945 reposurgeon uses a built-in extractor class to perform extractions from
1946 Mercurial repositories.
1947 Mercurial branches are exported as branches in the exported repository
1949 and tags are exported as tags. By default, bookmarks are ignored. You
1950 can specify explicit handling for bookmarks by setting
1951 reposurgeon.bookmarks in your .hg/hgrc. Set the value to the prefix
1952 that reposurgeon should use for bookmarks.
1953 For example, if your bookmarks represent branches, put this at the
1955 bottom of your .hg/hgrc:
1956 [reposurgeon]
1958 bookmarks=heads/
1959 If you do that, it's your responsibility to ensure that branch names do
1961 not conflict with bookmark names. You can add a prefix like
1962 bookmarks=heads/feature- to disambiguate as necessary.
1963
1965 reposurgeon can read Subversion dumpfiles or edit a Subversion
1966 repository (and you must point it at a repository, not a checkout
1967 directory). The reposurgeon distribution includes a script named
1968 “repotool” that you can use to make and then incrementally update a
1969 local mirror of a remote repository for editing or conversion purposes.
1970 READING SUBVERSION REPOSITORIES
1972 Certain optional modifiers on the read command change its behavior when
1973 reading Subversion repositories:
1974 --nobranch
1976 Suppress branch analysis.
1977 --ignore-properties
1979 Suppress read-time warnings about discarded property settings.
1980 --user-ignores
1982 Don't generate .gitignore files from svn:ignore properties.
1983 Instead, just pass through .gitignore files found in the history.
1984 --use-uuid
1986 If the --use-uuid read option is set, the repository's UUID will be
1987 used as the hostname when faking up email addresses, a la git-svn.
1988 Otherwise, addresses will be generated the way git cvs-import does
1989 it, simply ciopying the username into the address field.
1990 These modifiers can go anywhere in any order on the read command line
1992 after the read verb. They must be whitespace-separated.
1993 Here are the rules used for mapping subdirectories in a Subversion
1995 repository to branches:
1996 1. At any given time there is a set of eligible paths and path
1998 wildcards which declare potential branches. See the documentation
1999 of the branchify for how to alter this set, which initially
2000 consists of {trunk, tags/*, branches/*, and '*'}.
2001 2. A repository is considered "flat" if it has no directory that
2003 matches a path or path wildcard in the branchify set. All commits
2004 in a flat repository are assigned to branch master, and what would
2005 have been branch structure becomes directory structure. In this
2006 case, we're done; all the other rules apply to non-flat repos.
2007 If you give the option --nobranch when reading a Subversion
2009 repository, branch analysis is skipped and the repository is
2010 treated as though flat (left as a linear sequence of commits on
2011 refs/heads/master). This may be useful if your repository
2012 configuration is highly unusual and you need to do your own branch
2013 surgery. Note that this option will disable partitioning of mixed
2014 commits.
2015 3. If "trunk" is eligible, it always becomes the master branch.
2017 4. If an element of the branchify set ends with *, each immediate
2019 subdirectory of it is considered a potential branch. If '*' is in
2020 the branchify set (which is true by default) all top-level
2021 directories other than /trunk, /tags, and /branches are also
2022 considered potential branches.
2023 5. Each potential branch is checked to see if it has commits on it
2025 after the initial creation or copy. If there are such commits, it
2026 becomes a branch. If not, it becomes a tag in order to preserve the
2027 commit metadata. (In all cases, the name of the tag or branch is
2028 the basename of the directory.)
2029 6. Files in the top-level directory are assigned to a synthetic branch
2031 named 'root'.
2032 Each commit that only creates or deletes directories (in particular,
2034 copy commits for tags and branches, and commits that only change
2035 properties) will be transformed into a tag named after the branch,
2036 containing the date/author/comment metadata from the commit. While this
2037 produces a desirable result for tags, non-tag branches (including
2038 trunk) will also get root tags this way. This apparent misfeature has
2039 been accepted so that reposurgeon will never destroy human-generated
2040 metadata that might have value; it is left up to the user to manually
2041 remove unwanted tags.
2042 Subversion branch deletions are turned into deletealls, clearing the
2044 fileset of the import-stream branch. When a branch finishes with a
2045 deleteall at its tip, the deleteall is transformed into a tag. This
2046 rule cleans up after aborted branch renames.
2047 Occasionally (and usually by mistake) a branchy Subversion repository
2049 will contain revisions that touch multiple branches. These are handled
2050 by partitioning them into multiple import-stream commits, one on each
2051 affected branch. The Legacy-ID of such a split commit will have a
2052 pseudo-decimal part - for example, if Subversion revision 2317 touches
2053 three branches, the three generated commits will have IDs 2317.1,
2054 2317.2, and 2317.3.
2055 The svn:executable and svn:special properties are translated into
2057 permission settings in the input stream; svn:executable becomes 100755
2058 and svn:special becomes 120000 (indicating a symlink; the blob contents
2059 will be the path to which the symlink should resolve).
2060 Any cvs2svn:rev properties generated by cvs2svn are incorporated into
2062 the internal map used for reference-lifting, then discarded.
2063 Normally, per-directory svn:ignore properties become .gitignore files.
2065 Actual .gitignore files in a Subversion directory are presumed to have
2066 been created by git-svn users separately from native Subversion ignore
2067 properties and discarded with a warning. It is up to the user to merge
2068 the content of such files into the target repository by hand. But this
2069 behavior is inverted by the --user-ignores option; if that is on,
2070 .gitignore files are passed through and Subversion svn:ignore
2071 properties are discarded.
2072 (Regardless of the setting of the --user-ignores option, .cvsignore
2074 files found in Subversion repositories always become .gitignores in the
2075 translation. The assumption is that these date from before a CVS-to-SVN
2076 lift and should be preserved to affect behavior when browsing that
2077 section of the repository.)
2078 svn:mergeinfo properties are interpreted. Any svn:mergeinfo property on
2080 a revision A with a merge source range ending in revision B produces a
2081 merge link such that B becomes a parent of A.
2082 All other Subversion properties are discarded. (This may change in a
2084 future release.) The property for which this is most likely to cause
2085 semantic problems is svn:eol-style. However, since property-change-only
2086 commits get turned into annotated tags, the translated tags will retain
2087 information about setting changes.
2088 The sub-second resolution on Subversion commit dates is discarded; Git
2090 wants integer timestamps only.
2091 Because fast-import format cannot represent an empty directory, empty
2093 directories in Subversion repositories will be lost in translation.
2094 Normally, Subversion local usernames are mapped in the style of git
2096 cvs-import; thus user "foo" becomes "foo <foo>", which is sufficient to
2097 pacify git and other systems that require email addresses. With the
2098 option "svn_use_uuid", usernames are mapped in the git-svn style, with
2099 the repository's UUID used as a fake domain in the email address. Both
2100 forms can be remapped to real address using the authors read command.
2101 Reading a Subversion stream enables writing of the legacy map as
2103 'legacy' passthroughs when the repo is written to a stream file.
2104 reposurgeon tries hard to silently do the right thing, but there are
2106 Subversion edge cases in which it emits warnings because a human may
2107 need to intervene and perform fixups by hand. Here are the less obvious
2108 messages it may emit:
2109 user-generated .gitignore
2111 This message means means reposurgeon has found a .gitignore file in
2112 the Subversion repository it is analyzing. This probably happened
2113 because somebody was using git-svn as a live gateway, and created
2114 ignores which may or may not be congruent with those in the
2115 generated .gitignore files that the Subversion ignore properties
2116 will be translated into. You'll need to make a policy decision
2117 about which set of ignores to use in the conversion, and possibly
2118 set the --user-ignores option on read to pass through user-created
2119 .gitignore files; in that case this warning will not be emitted.
2120 can't connect nonempty branch XXXX to origin
2122 This is a serious error. reposurgeon has been unable to find a
2123 link from a specified branch to the trunk (master) branch. The
2124 commit graph will not be fully connected and will need manual
2125 repair.
2126 permission information may be lost
2128 A Subversion node change on a file sets or clears properties, but
2129 no ancestor can be found for this file. Executable or symlink
2130 position may be set wrongly on later revisions of this file.
2131 Subversion user-defined properties may also be scrambled or lost.
2132 Usually this error can be ignored.
2133 properties set
2135 reposurgeon has detected a setting of a user-defined property, or
2136 the Subversion properties svn:externals. These properties cannot be
2137 expressed in an import stream; the user is notified in case this is
2138 a showstopper for the conversion or some corrective action is
2139 required, but normally this error can be ignored. This warning is
2140 suppressed by the --ignore-properties option.
2141 branch links detected by file ops only
2143 Branch links are normally deduced by examining Subversion directory
2144 copy operations. A common user error (making a branch with a
2145 non-Subversion directory copy and then doing an svn add on the
2146 contends) can defeat this. While reposurgeon should detect and cope
2147 with most such copies correctly, you should examine the commit
2148 graph to check that the branch is rooted at the correct place.
2149 could not tagify root commit
2151 The earliest commit in your Subversion repository has file
2152 operations, rather than being a pure directory creation. This
2153 probably means your Subversion dump file is malformed, or you may
2154 have attempted to lift from an incremental dump. Proceed with
2155 caution.
2156 deleting parentless tip delete
2158 This message may be triggered by a Subversion branch move followed
2159 by a re-creation under the source name. Check near the indicated
2160 revision to make sure the renamed branch is connected to master.
2161 mid-branch deleteall
2163 A deleteall operation has been found in the middle of a branch
2164 history. This usually indicates that a Subversion tag or branch was
2165 created by mistake, and someone later tried to undo the error by
2166 deleting the tag/branch directory before recreating it with a copy
2167 operation. Examine the topology near the deleteall closely, it may
2168 need hand-hacking. It is fairly likely that both (a) the
2169 reposurgeon translation will be different from what other
2170 translators (such as git-svn) produce, and (b) it will not be
2171 immediately obvious which is right.
2172 lookback for XXX failed, not making branch link
2174 Branch analysis failed, probably due to a set of file copies that
2175 reposurgeon thought it should interpret as a botched branch
2176 creation but couldn't deduce a history for. This is a warning;
2177 check how the directory XXX is converted, it may need post-editing
2178 into a branch.
2179 WRITING SUBVERSION REPOSITORIES
2181 reposurgeon has support for writing Subversion repositories. Due to
2182 mismatches between the ontology of Subversion and that of git import
2183 streams, this support has some significant limitations and bugs.
2184 In summary, Subversion repository histories do not round-trip through
2186 reposurgeon editing. File content changes are preserved but some
2187 metadata is unavoidably lost. Furthermore, writing out a DVCS history
2188 in Subversion also loses significant portions of its metadata. Details
2189 follow.
2190 Writing a Subversion repository or dump stream discards author
2192 information, the committer's name, and the hostname part of the commit
2193 address; only the commit timestamp and the local part of the
2194 committer's email address are preserved, the latter becoming the
2195 Subversion author field. However, reading a Subversion repository and
2196 writing it out again will preserve the author fields.
2197 Import-stream timestamps have 1-second granularity. The sub-second
2199 parts of Subversion commit timestamps will be lost on their way through
2200 reposurgeon.
2201 Empty directories aren't represented in import streams. Consequently,
2203 reading and writing Subversion repositories preserves file content, but
2204 not empty directories. It is also not guaranteed that after editing a
2205 Subversion repository that the sequence of directory creations and
2206 deletions relative to other operations will be identical; the only
2207 guarantee is that enclosing directories will be created before any
2208 files in them are.
2209 When reading a Subversion repository, reposurgeon discards the special
2211 directory-copy nodes associated with branch creations. These can't be
2212 recreated if and when the repository is written back out to Subversion;
2213 rather, each branch copy node from the original translates into a
2214 branch creation plus the first set of file modifications on the branch.
2215 When reading a Subversion repository, reposurgeon also automatically
2217 breaks apart mixed-branch commits. These are not re-united if the
2218 repository is written back out.
2219 When writing to a Subversion repository, all lightweight tags become
2221 Subversion tag copies with empty log comments, named for the tag
2222 basename. The committer name and timestamp are copied from the commit
2223 the tag points to. The distinction between heads and tags is lost.
2224 Because of the preceding two points, it is not guaranteed that even
2226 revision numbers will be stable when a Subversion repository is read in
2227 and then written out!
2228 Subversion repositories are always written with a standard
2230 (trunk/tags/branches) layout. Thus, a repository with a nonstandard
2231 shape that has been analyzed by reposurgeon won't be written out with
2232 the same shape.
2233 When writing a Subversion repository, branch merges are translated into
2235 svn:mergeinfo properties in the simplest possible way - as an
2236 svn:mergeinfo property of the translated merge commit listing the merge
2237 source revisions.
2238 Subversion has a concept of "flows"; that is, named segments of history
2240 corresponding to files or directories that are created when the path is
2241 added, cloned when the path is copied, and deleted when the path is
2242 deleted. This information is not preserved in import streams or the
2243 internal representation that reposurgeon uses. Thus, after editing, the
2244 flow boundaries of a Subversion history may be arbitrarily changed.
2245
2247 reposurgeon recognizes how supported VCSes represent file ignores (CVS
2248 .cvsignore files lurking untranslated in older Subversion repositories,
2249 Subversion ignore properties, .gitignore/.hgignore/.bzrignore file in
2250 other systems) and moves ignore declarations among these containers on
2251 repo input and output. This will be sufficient if the ignore patterns
2252 are exact filenames.
2253 Translation may not, however, be perfect when the ignore patterns are
2255 Unix glob patterns or regular expressions. This compatibility table
2256 describes which patterns will translate; “plain” indicates a plain
2257 filename with no glob or regexp syntax or negation.
2258 RCS has no ignore files or patterns and is therefore not included in
2260 the table.
2261┌─────────────┬───────────────┬──────────────┬───────────────────┬───────────────────┬─────────────────────┬──────────────┬────────────┬────────────┐
2263│ │ from CVS │ from svn │ from git │ from hg │ from bzr │ from │ from SRC │ from bk │
2264│ │ │ │ │ │ │ darcs │ │ │
2265├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2266│ │ │ │ │ │ │ │ │ │
2267│ to │ all │ all │ all │ all │ all │ plain │ all │ all │
2268│ CVS │ │ │ except │ │ except │ │ │ │
2269│ │ │ │ !-prefixed │ │ RE:- │ │ │ │
2270│ │ │ │ but │ │ and │ │ │ │
2271│ │ │ │ nonempty │ │ !-prefixed │ │ │ │
2272├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2273│ │ all except │ │ │ │ │ │ │ │
2274│ to │ !.PP │ all │ all except │ all │ all except │ plain │ all │ all │
2275│ svn │ │ │ !-prefixed │ │ RE:- and │ │ │ │
2276│ │ │ │ │ │ !-prefixed │ │ │ │
2277├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2278│ │ │ │ │ │ │ │ │ │
2279│ to │ all │ all │ all │ all │ all except │ plain │ all │ all │
2280│ git │ │ │ │ except │ RE:-prefixed │ │ │ │
2281│ │ │ │ │ !-prefixed │ │ │ │ │
2282├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2283│ │ │ │ │ │ │ │ │ │
2284│ to │ all │ all │ all except │ all │ all except │ plain │ all │ all │
2285│ hg │ except │ │ !-prefixed │ │ RE:- and │ │ │ │
2286│ │ ! │ │ │ │ !-prefixed │ │ │ │
2287├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2288│ │ │ │ │ │ │ │ │ │
2289│ to │ all │ all │ all │ all │ all │ plain │ all │ all │
2290│ bzr │ │ │ │ │ │ │ │ │
2291├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2292│ │ │ │ │ │ │ │ │ │
2293│ to │ plain │ plain │ plain │ plain │ plain │ all │ all │ all │
2294│ darcs │ │ │ │ │ │ │ │ │
2295├─────────────┼───────────────┼──────────────┼───────────────────┼───────────────────┼─────────────────────┼──────────────┼────────────┼────────────┤
2296│ │ │ │ │ │ │ │ │ │
2297│ to │ all │ all │ all except │ all │ all except │ plain │ all │ all │
2298│ SRC │ except │ │ !-prefixed │ │ RE:- and │ │ │ │
2299│ │ ! │ │ │ │ !-prefixed │ │ │ │
2300└─────────────┴───────────────┴──────────────┴───────────────────┴───────────────────┴─────────────────────┴──────────────┴────────────┴────────────┘
2301 The hg rows and columns of the table describes compatibility to hg's
2303 glob syntax rather than its default regular-expression syntax. When
2304 writing to an hg repository from any other kind, reposurgeon prepends
2305 to the output .hgignore a "syntax: glob" line.
2306
2308 After converting a CVS, SVN, or BitKeeper repository, check for and
2309 remove $-cookies in the head revision(s) of the files. The full
2310 Subversion set is $Date:, $Revision:, $Author:, $HeadURL and $Id:. CVS
2311 uses $Author:, $Date:, $Header:, $Id:, $Log:, $Revision:, also (rarely)
2312 $Locker:, $Name:, $RCSfile:, $Source:, and $State:.
2313 When you need to specify a commit, use the action-stamp format that
2315 references lift generates when it can resolve an SVN or CVS reference
2316 in a comment. It is best that you not vary from this format, even in
2317 trivial ways like omitting the 'Z' or changing the 'T' or '!' or ':'.
2318 Making action stamps uniform and machine-parseable will have good
2319 consequences for future repository-browsing tools.
2320 Sometimes, in converting a repository, you may need to insert an
2322 explanatory comment - for example, if metadata has been garbled or
2323 missing and you need to point to that fact. It's helpful for
2324 repository-browsing tools if there is a uniform syntax for this that is
2325 highly unlikely to show up in repository comments. We recommend
2326 enclosing translation notes in [[ ]]. This has the advantage of being
2327 visually similar to the [ ] traditionally used for editorial comments
2328 in text.
2329 It is good practice to include, in the comment for the root commit of
2331 the repository, a note dating and attributing the conversion work and
2332 explaining these conventions. Example:
2333 [[This repository was converted from Subversion to git on 2011-10-24 by
2335 Eric S. Raymond <esr@thyrsus.com>. Here and elsewhere, conversion notes
2336 are enclosed in double square brackets. Junk commits generated by
2337 cvs2svn have been removed, commit references have been mapped into a
2338 uniform VCS-independent syntax, and some comments edited into
2339 summary-plus-continuation form.]]
2340 It is also good practice to include a generated tag at the point of
2342 conversion. E.g
2343 mailbox_in --create <<EOF
2345 Tag-Name: git-conversion
2346 Marks the spot at which this repository was converted from Subversion to git.
2348 EOF
2349
2351 define lastchange {
2352 @max(=B & [/ChangeLog/] & /{0}/B)? list
2353 }
2354 List the last commit that refers to a ChangeLog file containing a
2356 specified string. (The trick here is that ? extends the singleton set
2357 consisting of the last eligible ChangeLog blob to its set of referring
2358 commits, and listonly notices the commits.)
2359
2361 The event-stream parser in “reposurgeon” supports some extended syntax.
2362 Exporters designed to work with “reposurgeon” may have a --reposurgeon
2363 option that enables emission of extended syntax; notably, this is true
2364 of cvs-fast-export(1). The remainder of this section describes these
2365 syntax extensions. The properties they set are (usually) preserved and
2366 re-output when the stream file is written.
2367 The token “#reposurgeon” at the start of a comment line in a
2369 fast-import stream signals reposurgeon that the remainder is an
2370 extension command to be interpreted by “reposurgeon”.
2371 One such extension command is implemented: #sourcetype, which behaves
2373 identically to the reposurgeon sourcetype command. An exporter for a
2374 version-control system named “frobozz” could, for example, say
2375 #reposurgeon sourcetype frobozz
2377 Within a commit, a magic comment of the form “#legacy-id” declares a
2379 legacy ID from the stream file's source version-control system.
2380 Also accepted is the bzr syntax for setting per-commit properties.
2382 While parsing commit syntax, a line beginning with the token “property”
2383 must contibue with a whitespace-separated property-name token. If it is
2384 then followed by a newline it is taken to set that boolean-valued
2385 property to true. Otherwise it must be followed by a numeric token
2386 specifying a data length, a space, following data (which may contain
2387 newlines) and a terminating newline. For example:
2388 commit refs/heads/master
2390 mark :1
2391 committer Eric S. Raymond <esr@thyrsus.com> 1289147634 -0500
2392 data 16
2393 Example commit.
2394 property legacy-id 2 r1
2396 M 644 inline README
2397 Unlike other extensions, bzr properties are only preserved on stream
2399 output if the preferred type is bzr, because any importer other than
2400 bzr's will choke on them.
2401
2403 In versions before 3.23, “prefer” changed the repository type as well
2404 as the preferred output format.
2405 In versions before 3.0, the general command syntax put the command verb
2407 first, then the selection set (if any) then modifiers (VSO). It has
2408 changed to optional selection set first, then command verb, then
2409 modifiers (SVO). The change made parsing simpler, allowed abolishing
2410 some noise keywords, and recapitulates a successful design pattern in
2411 some other Unix tools - notably sed(1).
2412 In versions before 3.0, path expressions only matched commits, not
2414 commits and the associated blobs as well. The names of the “a” and “c”
2415 flags were different.
2416 In reposurgeon versions before 3.0, the delete command had the
2418 semantics of squash; also, the policy flags did not require a “--”
2419 prefix. The “--delete” flag was named “obliterate”.
2420 In reposurgeon versions before 3.0, read and write optionally took file
2422 arguments rather than requiring redirects (and the write command never
2423 wrote into directories). This was changed in order to allow these
2424 commands to have modifiers. These modifiers replaced several global
2425 options that no longer exist.
2426 In reposurgeon versions before 3.0, the earliest factor in a unite
2428 command always kept its tag and branch names unaltered. The new rule
2429 for resolving name conflicts, giving priority to the latest factor,
2430 produces more natural behavior when uniting two repositories end to
2431 end; the master branch of the second (later) one keeps its name.
2432 In reposurgeon versions before 3.0, the tagify command expected
2434 policies as trailing arguments to alter its behaviour. The new syntax
2435 uses similarly named options with leading dashes, that can appear
2436 anywhere after the tagify command
2437 In versions before 2.9. the syntax of "authors", "legacy", "list", and
2439 "mailbox_{in|out}" was different (and "legacy" was "fossils"). They
2440 took plain filename arguments rather that using redirect < and >.
2441
2443 Guarantee: In DVCses that use commit hashes, editing with reposurgeon
2444 never changes the hash of a commit object unless (a) you edit the
2445 commit, or (b) it is a descendant of an edited commit in a VCS that
2446 includes parent hashes in the input of a child object's hash (git and
2447 hg both do this).
2448 Guarantee: reposurgeon only requires main memory proportional to the
2450 size of a repository's metadata history, not its entire content
2451 history. (Exception: the data from inline content is held in memory.)
2452 Guarantee: In the worst case, reposurgeon makes its own copy of every
2454 content blob in the repository's history and thus uses intermediate
2455 disk space approximately equal to the size of a repository's content
2456 history. However, when the repository to be edited is presented as a
2457 stream file, reposurgeon requires no or only very little extra disk
2458 space to represent it; the internal representation of content blobs is
2459 a (seek-offset, length) pair pointing into the stream file.
2460 Guarantee: reposurgeon never modifies the contents of a repository it
2462 reads, nor deletes any repository. The results of surgery are always
2463 expressed in a new repository.
2464 Guarantee: Any line in a fast-import stream that is not a part of a
2466 command reposurgeon parses and understands will be passed through
2467 unaltered. At present the set of potential passthroughs is known to
2468 include the progress, the options, and checkpoint commands as well as
2469 comments led by #.
2470 Guarantee: All reposurgeon operations either preserve all repository
2472 state they are not explicitly told to modify or warn you when they
2473 cannot do so.
2474 Guarantee: reposurgeon handles the bzr commit-properties extension,
2476 correctly passing through property items including those with embedded
2477 newlines. (Such properties are also editable in the mailbox format.)
2478 Limitation: Because reposurgeon relies on other programs to generate
2480 and interpret the fast-import command stream, it is subject to bugs in
2481 those programs.
2482 Limitation: bzr suffers from deep confusion over whether its unit of
2484 work is a repository or a floating branch that might have been cloned
2485 from a repo or created from scratch, and might or might not be destined
2486 to be merged to a repo one day. Its exporter only works on branches,
2487 but its importer creates repos. Thus, a rebuild operation will produce
2488 a subdirectory structure that differs from what you expect. Look for
2489 your content under the subdirectory 'trunk'.
2490 Limitation: under git, signed tags are imported verbatim. However, any
2492 operation that modifies any commit upstream of the target of the tag
2493 will invalidate it.
2494 Limitation: Stock git (at least as of version 1.7.3.2) will choke on
2496 property extension commands. Accordingly, reposurgeon omits them when
2497 rebuilding a repo with git type.
2498 Limitation: Converting an hg repo that uses bookmarks (not branches) to
2500 git can lose information; the branch ref that git assigns to each
2501 commit may not be the same as the hg bookmark that was active when the
2502 commit was originally made under hg. Unfortunately, this is a real
2503 ontological mismatch, not a problem that can be fixed by cleverness in
2504 reposurgeon.
2505 Limitation: Converting an hg repo that uses branches to git can lose
2507 information because git does not store an explicit branch as part of
2508 commit metadata, but colors commits with branch or tag names on the fly
2509 using a specific coloring algorithm, which might not match the explicit
2510 branch assignments to commits in the original hg repo. Reposurgeon
2511 preserves the hg branch information when reading an hg repo, so it is
2512 available from within reposurgeon itself, but there is no way to
2513 preserve it if the repo is written to git.
2514 Limitation: While the Subversion read-side support is in good shape,
2516 the write-side support is more of a sketch or proof-of-concept than a
2517 robust implementation; it only works on very simple cases and does not
2518 round-trip. It may improve in future releases.
2519 Limitation: Not all BitKeeper versions have the fast-import and
2521 fast-export commands that reposurgeon requires. They are present back
2522 to the 7.3 opensource version.
2523 Limitation: reposurgeon may misbehave under a filesystem which smashes
2525 case in filenames, or which nominally preserves case but maps names
2526 differing only by case to the same filesystem node (Mac OS X behaves
2527 like this by default). Problems will arise if any two paths in a repo
2528 differ by case only. To avoid the problem on a Mac, do all your surgery
2529 on an HFS+ file system formatted with case sensitivity specifically
2530 enabled.
2531 Limitation: If whitespace followed by # appears in a string or regexp
2533 command argument, it will be misinterpreted as the beginning of a
2534 line-ending comment and screw up parsing.
2535 Guarantee: As version-control systems add support for the fast-import
2537 format, their repositories will become editable by reposurgeon.
2538
2540 reposurgeon relies on importers and exporters associated with the VCSes
2541 it supports.
2542 git
2544 Core git supports both export and import.
2545 bzr
2547 Requires bzr plus the bzr-fast-import plugin.
2548 hg
2550 Requires core hg, the hg-fastimport plugin, and the third-party
2551 hg-fast-export.py script.
2552 svn
2554 Stock Subversion commands support export and import.
2555 darcs
2557 Stock darcs commands support export and import.
2558 CVS
2560 Requires cvs-fast-export. Note that the quality of CVS lifts may be
2561 poor, with individual lifts requiring serious hand-hacking. This is
2562 due to inherent problems with CVS's file-oriented model.
2563 RCS
2565 Requires cvs-fast-export (yes, that's not a typo; cvs-fast-export
2566 handles RCS collections as well). The caveat for CVS applies.
2567
2569 It is expected that reposurgeon will be extended with more deletion
2570 policies. Policy authors may need to know more about how a commit's
2571 file operation sequence is reduced to normal form after operations from
2572 deleted commits are prepended to it.
2573 Recall that each commit has a list of file operations, each a M
2575 (modify), D (delete), R (rename), C (copy), or 'deleteall' (delete all
2576 files). Only M operations have associated blobs. Normally there is only
2577 one M operation per individual file in a commit's operation list.
2578 To understand how the reduction process works, it's enough to
2580 understand the case where all the operation in the list are working on
2581 the same file. Sublists of operations referring to different files
2582 don't affect each other and reducing them can be thought of as separate
2583 operations. Also, a "deleteall" acts as a D for everything and cancels
2584 all operations before it in the list.
2585 The reduction process walks through the list from the beginning looking
2587 for adjacent pairs of operations it can compose. The following table
2588 describes all possible cases and all but one of the reductions.
2589 ┌──────────────────────────┬────────────────────────────┐
2591 │ M + D → D │ │
2592 │ │ If a file is │
2593 │ │ modified then │
2594 │ │ deleted, the result │
2595 │ │ is as though it had │
2596 │ │ been deleted. If │
2597 │ │ the M was the only │
2598 │ │ modify for the │
2599 │ │ file, it's removed │
2600 │ │ too. │
2601 ├──────────────────────────┼────────────────────────────┤
2602 │M a + R a b → R a b + M b │ │
2603 │ │ The purpose of this │
2604 │ │ transformation is │
2605 │ │ to push renames │
2606 │ │ toward the │
2607 │ │ beginning of the │
2608 │ │ list, where they │
2609 │ │ may become adjacent │
2610 │ │ to another R or C │
2611 │ │ they can be │
2612 │ │ composed with. If │
2613 │ │ the M is the only │
2614 │ │ modify operation │
2615 │ │ for this file, the │
2616 │ │ rename is dropped. │
2617 ├──────────────────────────┼────────────────────────────┤
2618 │ M a + C a b │ │
2619 │ │ No reduction. │
2620 ├──────────────────────────┼────────────────────────────┤
2621 │ M b + R a b → nothing │ │
2622 │ │ Should be │
2623 │ │ impossible, and may │
2624 │ │ indicate repository │
2625 │ │ corruption. │
2626 ├──────────────────────────┼────────────────────────────┤
2627 │ M b + C a b → nothing │ │
2628 │ │ The copy undoes the │
2629 │ │ modification. │
2630 ├──────────────────────────┼────────────────────────────┤
2631 │ D + M → M │ │
2632 │ │ If a file is │
2633 │ │ deleted and │
2634 │ │ modified, the │
2635 │ │ result is as though │
2636 │ │ the deletion had │
2637 │ │ not taken place │
2638 │ │ (because M │
2639 │ │ operations store │
2640 │ │ entire files, not │
2641 │ │ deltas). │
2642 ├──────────────────────────┼────────────────────────────┤
2643 │ D + {D|R|C} │ │
2644 │ │ These cases should │
2645 │ │ be impossible and │
2646 │ │ would suggest the │
2647 │ │ repository has been │
2648 │ │ corrupted. │
2649 ├──────────────────────────┼────────────────────────────┤
2650 │ R a b + D a │ │
2651 │ │ Should never │
2652 │ │ happen, and is │
2653 │ │ another case that │
2654 │ │ would suggest │
2655 │ │ repository │
2656 │ │ corruption. │
2657 ├──────────────────────────┼────────────────────────────┤
2658 │ R a b + D b → D a │ │
2659 │ │ The delete removes │
2660 │ │ the just-renamed │
2661 │ │ file. │
2662 ├──────────────────────────┼────────────────────────────┤
2663 │ {R|C} + M │ │
2664 │ │ No reduction. │
2665 ├──────────────────────────┼────────────────────────────┤
2666 │ R a b + R b c → R a c │ │
2667 │ │ The b terms have to │
2668 │ │ match for these │
2669 │ │ operations to have │
2670 │ │ made sense when │
2671 │ │ they lived in │
2672 │ │ separate commits; │
2673 │ │ if they don't, it │
2674 │ │ indicates │
2675 │ │ repository │
2676 │ │ corruption. │
2677 ├──────────────────────────┼────────────────────────────┤
2678 │ R a b + C b c │ │
2679 │ │ No reduction. │
2680 ├──────────────────────────┼────────────────────────────┤
2681 │ C a b + D a → R a b │ │
2682 │ │ Copy followed by │
2683 │ │ delete of the │
2684 │ │ source is a rename. │
2685 ├──────────────────────────┼────────────────────────────┤
2686 │ C a b + D b → nothing │ │
2687 │ │ This delete undoes │
2688 │ │ the copy. │
2689 ├──────────────────────────┼────────────────────────────┤
2690 │ C a b + R a c │ │
2691 │ │ No reduction. │
2692 ├──────────────────────────┼────────────────────────────┤
2693 │ C a b + R b c → C a c │ │
2694 │ │ Copy followed by a │
2695 │ │ rename of the │
2696 │ │ target reduces to │
2697 │ │ single copy │
2698 ├──────────────────────────┼────────────────────────────┤
2699 │ C + C │ │
2700 │ │ No reduction. │
2701 └──────────────────────────┴────────────────────────────┘
2702
2704 This section will become relevant only if reposurgeon or something
2705 underneath it in the software and hardware stack crashes while in the
2706 middle of writing out a repository, in particular if the target
2707 directory of the rebuild is your current directory.
2708 The tool has two conflicting objectives. On the one hand, we never want
2710 to risk clobbering a pre-existing repo. On the other hand, we want to
2711 be able to run this tool in a directory with a repo and modify it in
2712 place.
2713 We resolve this dilemma by playing a game of three-directory monte.
2715 1. First, we build the repo in a freshly-created staging directory. If
2717 your target directory is named /path/to/foo, the staging directory
2718 will be a peer named /path/to/foo-stageNNNN, where NNNN is a cookie
2719 derived from reposurgeon's process ID.
2720 2. We then make an empty backup directory. This directory will be
2722 named /path/to/foo.~N~, where N is incremented so as not to
2723 conflict with any existing backup directories. reposurgeon never,
2724 under any circumstances, ever deletes a backup directory.
2725 So far, all operations are safe; the worst that can happen up to
2727 this point if the process gets interrupted is that the staging and
2728 backup directories get left behind.
2729 3. The critical region begins. We first move everything in the target
2731 directory to the backup directory.
2732 4. Then we move everything in the staging directory to the target.
2734 5. We finish off by restoring untracked files in the target directory
2736 from the backup directory. That ends the critical region.
2737 During the critical region, all signals that can be ignored are
2739 ignored.
2740
2742 Returns 1 on fatal error, 0 otherwise. In batch mode all errors are
2743 fatal.
2744
2746 bzr(1), cvs(1), darcs(1), git(1), hg(1), rcs(1), svn(1). bk(1).
2747
2749 Eric S. Raymond <esr@thyrsus.com>; project page at
2750 http://www.catb.org/~esr/reposurgeon.
2751
2753 1. DVCS Migration HOWTO
2754 http://www.catb.org/esr/dvcs-migration-guide.html
2755 2. Python's
2757 http://docs.python.org/2/library/re.html
2758reposurgeon 07/15/2018 REPOSURGEON(1)