Merge branch 'tb/test-t9020-no-which'
[git/git.git] / Documentation / git-fast-import.txt
1 git-fast-import(1)
2 ==================
3
4 NAME
5 ----
6 git-fast-import - Backend for fast Git data importers
7
8
9 SYNOPSIS
10 --------
11 [verse]
12 frontend | 'git fast-import' [options]
13
14 DESCRIPTION
15 -----------
16 This program is usually not what the end user wants to run directly.
17 Most end users want to use one of the existing frontend programs,
18 which parses a specific type of foreign source and feeds the contents
19 stored there to 'git fast-import'.
20
21 fast-import reads a mixed command/data stream from standard input and
22 writes one or more packfiles directly into the current repository.
23 When EOF is received on standard input, fast import writes out
24 updated branch and tag refs, fully updating the current repository
25 with the newly imported data.
26
27 The fast-import backend itself can import into an empty repository (one that
28 has already been initialized by 'git init') or incrementally
29 update an existing populated repository. Whether or not incremental
30 imports are supported from a particular foreign source depends on
31 the frontend program in use.
32
33
34 OPTIONS
35 -------
36 --date-format=<fmt>::
37 Specify the type of dates the frontend will supply to
38 fast-import within `author`, `committer` and `tagger` commands.
39 See ``Date Formats'' below for details about which formats
40 are supported, and their syntax.
41
42 --force::
43 Force updating modified existing branches, even if doing
44 so would cause commits to be lost (as the new commit does
45 not contain the old commit).
46
47 --max-pack-size=<n>::
48 Maximum size of each output packfile.
49 The default is unlimited.
50
51 --big-file-threshold=<n>::
52 Maximum size of a blob that fast-import will attempt to
53 create a delta for, expressed in bytes. The default is 512m
54 (512 MiB). Some importers may wish to lower this on systems
55 with constrained memory.
56
57 --depth=<n>::
58 Maximum delta depth, for blob and tree deltification.
59 Default is 10.
60
61 --active-branches=<n>::
62 Maximum number of branches to maintain active at once.
63 See ``Memory Utilization'' below for details. Default is 5.
64
65 --export-marks=<file>::
66 Dumps the internal marks table to <file> when complete.
67 Marks are written one per line as `:markid SHA-1`.
68 Frontends can use this file to validate imports after they
69 have been completed, or to save the marks table across
70 incremental runs. As <file> is only opened and truncated
71 at checkpoint (or completion) the same path can also be
72 safely given to \--import-marks.
73
74 --import-marks=<file>::
75 Before processing any input, load the marks specified in
76 <file>. The input file must exist, must be readable, and
77 must use the same format as produced by \--export-marks.
78 Multiple options may be supplied to import more than one
79 set of marks. If a mark is defined to different values,
80 the last file wins.
81
82 --import-marks-if-exists=<file>::
83 Like --import-marks but instead of erroring out, silently
84 skips the file if it does not exist.
85
86 --relative-marks::
87 After specifying --relative-marks the paths specified
88 with --import-marks= and --export-marks= are relative
89 to an internal directory in the current repository.
90 In git-fast-import this means that the paths are relative
91 to the .git/info/fast-import directory. However, other
92 importers may use a different location.
93
94 --no-relative-marks::
95 Negates a previous --relative-marks. Allows for combining
96 relative and non-relative marks by interweaving
97 --(no-)-relative-marks with the --(import|export)-marks=
98 options.
99
100 --cat-blob-fd=<fd>::
101 Write responses to `cat-blob` and `ls` queries to the
102 file descriptor <fd> instead of `stdout`. Allows `progress`
103 output intended for the end-user to be separated from other
104 output.
105
106 --done::
107 Terminate with error if there is no `done` command at the
108 end of the stream.
109 This option might be useful for detecting errors that
110 cause the frontend to terminate before it has started to
111 write a stream.
112
113 --export-pack-edges=<file>::
114 After creating a packfile, print a line of data to
115 <file> listing the filename of the packfile and the last
116 commit on each branch that was written to that packfile.
117 This information may be useful after importing projects
118 whose total object set exceeds the 4 GiB packfile limit,
119 as these commits can be used as edge points during calls
120 to 'git pack-objects'.
121
122 --quiet::
123 Disable all non-fatal output, making fast-import silent when it
124 is successful. This option disables the output shown by
125 \--stats.
126
127 --stats::
128 Display some basic statistics about the objects fast-import has
129 created, the packfiles they were stored into, and the
130 memory used by fast-import during this run. Showing this output
131 is currently the default, but can be disabled with \--quiet.
132
133
134 Performance
135 -----------
136 The design of fast-import allows it to import large projects in a minimum
137 amount of memory usage and processing time. Assuming the frontend
138 is able to keep up with fast-import and feed it a constant stream of data,
139 import times for projects holding 10+ years of history and containing
140 100,000+ individual commits are generally completed in just 1-2
141 hours on quite modest (~$2,000 USD) hardware.
142
143 Most bottlenecks appear to be in foreign source data access (the
144 source just cannot extract revisions fast enough) or disk IO (fast-import
145 writes as fast as the disk will take the data). Imports will run
146 faster if the source data is stored on a different drive than the
147 destination Git repository (due to less IO contention).
148
149
150 Development Cost
151 ----------------
152 A typical frontend for fast-import tends to weigh in at approximately 200
153 lines of Perl/Python/Ruby code. Most developers have been able to
154 create working importers in just a couple of hours, even though it
155 is their first exposure to fast-import, and sometimes even to Git. This is
156 an ideal situation, given that most conversion tools are throw-away
157 (use once, and never look back).
158
159
160 Parallel Operation
161 ------------------
162 Like 'git push' or 'git fetch', imports handled by fast-import are safe to
163 run alongside parallel `git repack -a -d` or `git gc` invocations,
164 or any other Git operation (including 'git prune', as loose objects
165 are never used by fast-import).
166
167 fast-import does not lock the branch or tag refs it is actively importing.
168 After the import, during its ref update phase, fast-import tests each
169 existing branch ref to verify the update will be a fast-forward
170 update (the commit stored in the ref is contained in the new
171 history of the commit to be written). If the update is not a
172 fast-forward update, fast-import will skip updating that ref and instead
173 prints a warning message. fast-import will always attempt to update all
174 branch refs, and does not stop on the first failure.
175
176 Branch updates can be forced with \--force, but it's recommended that
177 this only be used on an otherwise quiet repository. Using \--force
178 is not necessary for an initial import into an empty repository.
179
180
181 Technical Discussion
182 --------------------
183 fast-import tracks a set of branches in memory. Any branch can be created
184 or modified at any point during the import process by sending a
185 `commit` command on the input stream. This design allows a frontend
186 program to process an unlimited number of branches simultaneously,
187 generating commits in the order they are available from the source
188 data. It also simplifies the frontend programs considerably.
189
190 fast-import does not use or alter the current working directory, or any
191 file within it. (It does however update the current Git repository,
192 as referenced by `GIT_DIR`.) Therefore an import frontend may use
193 the working directory for its own purposes, such as extracting file
194 revisions from the foreign source. This ignorance of the working
195 directory also allows fast-import to run very quickly, as it does not
196 need to perform any costly file update operations when switching
197 between branches.
198
199 Input Format
200 ------------
201 With the exception of raw file data (which Git does not interpret)
202 the fast-import input format is text (ASCII) based. This text based
203 format simplifies development and debugging of frontend programs,
204 especially when a higher level language such as Perl, Python or
205 Ruby is being used.
206
207 fast-import is very strict about its input. Where we say SP below we mean
208 *exactly* one space. Likewise LF means one (and only one) linefeed
209 and HT one (and only one) horizontal tab.
210 Supplying additional whitespace characters will cause unexpected
211 results, such as branch names or file names with leading or trailing
212 spaces in their name, or early termination of fast-import when it encounters
213 unexpected input.
214
215 Stream Comments
216 ~~~~~~~~~~~~~~~
217 To aid in debugging frontends fast-import ignores any line that
218 begins with `#` (ASCII pound/hash) up to and including the line
219 ending `LF`. A comment line may contain any sequence of bytes
220 that does not contain an LF and therefore may be used to include
221 any detailed debugging information that might be specific to the
222 frontend and useful when inspecting a fast-import data stream.
223
224 Date Formats
225 ~~~~~~~~~~~~
226 The following date formats are supported. A frontend should select
227 the format it will use for this import by passing the format name
228 in the \--date-format=<fmt> command line option.
229
230 `raw`::
231 This is the Git native format and is `<time> SP <offutc>`.
232 It is also fast-import's default format, if \--date-format was
233 not specified.
234 +
235 The time of the event is specified by `<time>` as the number of
236 seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is
237 written as an ASCII decimal integer.
238 +
239 The local offset is specified by `<offutc>` as a positive or negative
240 offset from UTC. For example EST (which is 5 hours behind UTC)
241 would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''.
242 The local offset does not affect `<time>`; it is used only as an
243 advisement to help formatting routines display the timestamp.
244 +
245 If the local offset is not available in the source material, use
246 ``+0000'', or the most common local offset. For example many
247 organizations have a CVS repository which has only ever been accessed
248 by users who are located in the same location and timezone. In this
249 case a reasonable offset from UTC could be assumed.
250 +
251 Unlike the `rfc2822` format, this format is very strict. Any
252 variation in formatting will cause fast-import to reject the value.
253
254 `rfc2822`::
255 This is the standard email format as described by RFC 2822.
256 +
257 An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git
258 parser is accurate, but a little on the lenient side. It is the
259 same parser used by 'git am' when applying patches
260 received from email.
261 +
262 Some malformed strings may be accepted as valid dates. In some of
263 these cases Git will still be able to obtain the correct date from
264 the malformed string. There are also some types of malformed
265 strings which Git will parse wrong, and yet consider valid.
266 Seriously malformed strings will be rejected.
267 +
268 Unlike the `raw` format above, the timezone/UTC offset information
269 contained in an RFC 2822 date string is used to adjust the date
270 value to UTC prior to storage. Therefore it is important that
271 this information be as accurate as possible.
272 +
273 If the source material uses RFC 2822 style dates,
274 the frontend should let fast-import handle the parsing and conversion
275 (rather than attempting to do it itself) as the Git parser has
276 been well tested in the wild.
277 +
278 Frontends should prefer the `raw` format if the source material
279 already uses UNIX-epoch format, can be coaxed to give dates in that
280 format, or its format is easily convertible to it, as there is no
281 ambiguity in parsing.
282
283 `now`::
284 Always use the current time and timezone. The literal
285 `now` must always be supplied for `<when>`.
286 +
287 This is a toy format. The current time and timezone of this system
288 is always copied into the identity string at the time it is being
289 created by fast-import. There is no way to specify a different time or
290 timezone.
291 +
292 This particular format is supplied as it's short to implement and
293 may be useful to a process that wants to create a new commit
294 right now, without needing to use a working directory or
295 'git update-index'.
296 +
297 If separate `author` and `committer` commands are used in a `commit`
298 the timestamps may not match, as the system clock will be polled
299 twice (once for each command). The only way to ensure that both
300 author and committer identity information has the same timestamp
301 is to omit `author` (thus copying from `committer`) or to use a
302 date format other than `now`.
303
304 Commands
305 ~~~~~~~~
306 fast-import accepts several commands to update the current repository
307 and control the current import process. More detailed discussion
308 (with examples) of each command follows later.
309
310 `commit`::
311 Creates a new branch or updates an existing branch by
312 creating a new commit and updating the branch to point at
313 the newly created commit.
314
315 `tag`::
316 Creates an annotated tag object from an existing commit or
317 branch. Lightweight tags are not supported by this command,
318 as they are not recommended for recording meaningful points
319 in time.
320
321 `reset`::
322 Reset an existing branch (or a new branch) to a specific
323 revision. This command must be used to change a branch to
324 a specific revision without making a commit on it.
325
326 `blob`::
327 Convert raw file data into a blob, for future use in a
328 `commit` command. This command is optional and is not
329 needed to perform an import.
330
331 `checkpoint`::
332 Forces fast-import to close the current packfile, generate its
333 unique SHA-1 checksum and index, and start a new packfile.
334 This command is optional and is not needed to perform
335 an import.
336
337 `progress`::
338 Causes fast-import to echo the entire line to its own
339 standard output. This command is optional and is not needed
340 to perform an import.
341
342 `done`::
343 Marks the end of the stream. This command is optional
344 unless the `done` feature was requested using the
345 `--done` command line option or `feature done` command.
346
347 `cat-blob`::
348 Causes fast-import to print a blob in 'cat-file --batch'
349 format to the file descriptor set with `--cat-blob-fd` or
350 `stdout` if unspecified.
351
352 `ls`::
353 Causes fast-import to print a line describing a directory
354 entry in 'ls-tree' format to the file descriptor set with
355 `--cat-blob-fd` or `stdout` if unspecified.
356
357 `feature`::
358 Require that fast-import supports the specified feature, or
359 abort if it does not.
360
361 `option`::
362 Specify any of the options listed under OPTIONS that do not
363 change stream semantic to suit the frontend's needs. This
364 command is optional and is not needed to perform an import.
365
366 `commit`
367 ~~~~~~~~
368 Create or update a branch with a new commit, recording one logical
369 change to the project.
370
371 ....
372 'commit' SP <ref> LF
373 mark?
374 ('author' (SP <name>)? SP LT <email> GT SP <when> LF)?
375 'committer' (SP <name>)? SP LT <email> GT SP <when> LF
376 data
377 ('from' SP <committish> LF)?
378 ('merge' SP <committish> LF)?
379 (filemodify | filedelete | filecopy | filerename | filedeleteall | notemodify)*
380 LF?
381 ....
382
383 where `<ref>` is the name of the branch to make the commit on.
384 Typically branch names are prefixed with `refs/heads/` in
385 Git, so importing the CVS branch symbol `RELENG-1_0` would use
386 `refs/heads/RELENG-1_0` for the value of `<ref>`. The value of
387 `<ref>` must be a valid refname in Git. As `LF` is not valid in
388 a Git refname, no quoting or escaping syntax is supported here.
389
390 A `mark` command may optionally appear, requesting fast-import to save a
391 reference to the newly created commit for future use by the frontend
392 (see below for format). It is very common for frontends to mark
393 every commit they create, thereby allowing future branch creation
394 from any imported commit.
395
396 The `data` command following `committer` must supply the commit
397 message (see below for `data` command syntax). To import an empty
398 commit message use a 0 length data. Commit messages are free-form
399 and are not interpreted by Git. Currently they must be encoded in
400 UTF-8, as fast-import does not permit other encodings to be specified.
401
402 Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename`,
403 `filedeleteall` and `notemodify` commands
404 may be included to update the contents of the branch prior to
405 creating the commit. These commands may be supplied in any order.
406 However it is recommended that a `filedeleteall` command precede
407 all `filemodify`, `filecopy`, `filerename` and `notemodify` commands in
408 the same commit, as `filedeleteall` wipes the branch clean (see below).
409
410 The `LF` after the command is optional (it used to be required).
411
412 `author`
413 ^^^^^^^^
414 An `author` command may optionally appear, if the author information
415 might differ from the committer information. If `author` is omitted
416 then fast-import will automatically use the committer's information for
417 the author portion of the commit. See below for a description of
418 the fields in `author`, as they are identical to `committer`.
419
420 `committer`
421 ^^^^^^^^^^^
422 The `committer` command indicates who made this commit, and when
423 they made it.
424
425 Here `<name>` is the person's display name (for example
426 ``Com M Itter'') and `<email>` is the person's email address
427 (``\cm@example.com''). `LT` and `GT` are the literal less-than (\x3c)
428 and greater-than (\x3e) symbols. These are required to delimit
429 the email address from the other fields in the line. Note that
430 `<name>` and `<email>` are free-form and may contain any sequence
431 of bytes, except `LT`, `GT` and `LF`. `<name>` is typically UTF-8 encoded.
432
433 The time of the change is specified by `<when>` using the date format
434 that was selected by the \--date-format=<fmt> command line option.
435 See ``Date Formats'' above for the set of supported formats, and
436 their syntax.
437
438 `from`
439 ^^^^^^
440 The `from` command is used to specify the commit to initialize
441 this branch from. This revision will be the first ancestor of the
442 new commit. The state of the tree built at this commit will begin
443 with the state at the `from` commit, and be altered by the content
444 modifications in this commit.
445
446 Omitting the `from` command in the first commit of a new branch
447 will cause fast-import to create that commit with no ancestor. This
448 tends to be desired only for the initial commit of a project.
449 If the frontend creates all files from scratch when making a new
450 branch, a `merge` command may be used instead of `from` to start
451 the commit with an empty tree.
452 Omitting the `from` command on existing branches is usually desired,
453 as the current commit on that branch is automatically assumed to
454 be the first ancestor of the new commit.
455
456 As `LF` is not valid in a Git refname or SHA-1 expression, no
457 quoting or escaping syntax is supported within `<committish>`.
458
459 Here `<committish>` is any of the following:
460
461 * The name of an existing branch already in fast-import's internal branch
462 table. If fast-import doesn't know the name, it's treated as a SHA-1
463 expression.
464
465 * A mark reference, `:<idnum>`, where `<idnum>` is the mark number.
466 +
467 The reason fast-import uses `:` to denote a mark reference is this character
468 is not legal in a Git branch name. The leading `:` makes it easy
469 to distinguish between the mark 42 (`:42`) and the branch 42 (`42`
470 or `refs/heads/42`), or an abbreviated SHA-1 which happened to
471 consist only of base-10 digits.
472 +
473 Marks must be declared (via `mark`) before they can be used.
474
475 * A complete 40 byte or abbreviated commit SHA-1 in hex.
476
477 * Any valid Git SHA-1 expression that resolves to a commit. See
478 ``SPECIFYING REVISIONS'' in linkgit:gitrevisions[7] for details.
479
480 The special case of restarting an incremental import from the
481 current branch value should be written as:
482 ----
483 from refs/heads/branch^0
484 ----
485 The `^0` suffix is necessary as fast-import does not permit a branch to
486 start from itself, and the branch is created in memory before the
487 `from` command is even read from the input. Adding `^0` will force
488 fast-import to resolve the commit through Git's revision parsing library,
489 rather than its internal branch table, thereby loading in the
490 existing value of the branch.
491
492 `merge`
493 ^^^^^^^
494 Includes one additional ancestor commit. The additional ancestry
495 link does not change the way the tree state is built at this commit.
496 If the `from` command is
497 omitted when creating a new branch, the first `merge` commit will be
498 the first ancestor of the current commit, and the branch will start
499 out with no files. An unlimited number of `merge` commands per
500 commit are permitted by fast-import, thereby establishing an n-way merge.
501 However Git's other tools never create commits with more than 15
502 additional ancestors (forming a 16-way merge). For this reason
503 it is suggested that frontends do not use more than 15 `merge`
504 commands per commit; 16, if starting a new, empty branch.
505
506 Here `<committish>` is any of the commit specification expressions
507 also accepted by `from` (see above).
508
509 `filemodify`
510 ^^^^^^^^^^^^
511 Included in a `commit` command to add a new file or change the
512 content of an existing file. This command has two different means
513 of specifying the content of the file.
514
515 External data format::
516 The data content for the file was already supplied by a prior
517 `blob` command. The frontend just needs to connect it.
518 +
519 ....
520 'M' SP <mode> SP <dataref> SP <path> LF
521 ....
522 +
523 Here usually `<dataref>` must be either a mark reference (`:<idnum>`)
524 set by a prior `blob` command, or a full 40-byte SHA-1 of an
525 existing Git blob object. If `<mode>` is `040000`` then
526 `<dataref>` must be the full 40-byte SHA-1 of an existing
527 Git tree object or a mark reference set with `--import-marks`.
528
529 Inline data format::
530 The data content for the file has not been supplied yet.
531 The frontend wants to supply it as part of this modify
532 command.
533 +
534 ....
535 'M' SP <mode> SP 'inline' SP <path> LF
536 data
537 ....
538 +
539 See below for a detailed description of the `data` command.
540
541 In both formats `<mode>` is the type of file entry, specified
542 in octal. Git only supports the following modes:
543
544 * `100644` or `644`: A normal (not-executable) file. The majority
545 of files in most projects use this mode. If in doubt, this is
546 what you want.
547 * `100755` or `755`: A normal, but executable, file.
548 * `120000`: A symlink, the content of the file will be the link target.
549 * `160000`: A gitlink, SHA-1 of the object refers to a commit in
550 another repository. Git links can only be specified by SHA or through
551 a commit mark. They are used to implement submodules.
552 * `040000`: A subdirectory. Subdirectories can only be specified by
553 SHA or through a tree mark set with `--import-marks`.
554
555 In both formats `<path>` is the complete path of the file to be added
556 (if not already existing) or modified (if already existing).
557
558 A `<path>` string must use UNIX-style directory separators (forward
559 slash `/`), may contain any byte other than `LF`, and must not
560 start with double quote (`"`).
561
562 A path can use C-style string quoting; this is accepted in all cases
563 and mandatory if the filename starts with double quote or contains
564 `LF`. In C-style quoting, the complete name should be surrounded with
565 double quotes, and any `LF`, backslash, or double quote characters
566 must be escaped by preceding them with a backslash (e.g.,
567 `"path/with\n, \\ and \" in it"`).
568
569 The value of `<path>` must be in canonical form. That is it must not:
570
571 * contain an empty directory component (e.g. `foo//bar` is invalid),
572 * end with a directory separator (e.g. `foo/` is invalid),
573 * start with a directory separator (e.g. `/foo` is invalid),
574 * contain the special component `.` or `..` (e.g. `foo/./bar` and
575 `foo/../bar` are invalid).
576
577 The root of the tree can be represented by an empty string as `<path>`.
578
579 It is recommended that `<path>` always be encoded using UTF-8.
580
581 `filedelete`
582 ^^^^^^^^^^^^
583 Included in a `commit` command to remove a file or recursively
584 delete an entire directory from the branch. If the file or directory
585 removal makes its parent directory empty, the parent directory will
586 be automatically removed too. This cascades up the tree until the
587 first non-empty directory or the root is reached.
588
589 ....
590 'D' SP <path> LF
591 ....
592
593 here `<path>` is the complete path of the file or subdirectory to
594 be removed from the branch.
595 See `filemodify` above for a detailed description of `<path>`.
596
597 `filecopy`
598 ^^^^^^^^^^^^
599 Recursively copies an existing file or subdirectory to a different
600 location within the branch. The existing file or directory must
601 exist. If the destination exists it will be completely replaced
602 by the content copied from the source.
603
604 ....
605 'C' SP <path> SP <path> LF
606 ....
607
608 here the first `<path>` is the source location and the second
609 `<path>` is the destination. See `filemodify` above for a detailed
610 description of what `<path>` may look like. To use a source path
611 that contains SP the path must be quoted.
612
613 A `filecopy` command takes effect immediately. Once the source
614 location has been copied to the destination any future commands
615 applied to the source location will not impact the destination of
616 the copy.
617
618 `filerename`
619 ^^^^^^^^^^^^
620 Renames an existing file or subdirectory to a different location
621 within the branch. The existing file or directory must exist. If
622 the destination exists it will be replaced by the source directory.
623
624 ....
625 'R' SP <path> SP <path> LF
626 ....
627
628 here the first `<path>` is the source location and the second
629 `<path>` is the destination. See `filemodify` above for a detailed
630 description of what `<path>` may look like. To use a source path
631 that contains SP the path must be quoted.
632
633 A `filerename` command takes effect immediately. Once the source
634 location has been renamed to the destination any future commands
635 applied to the source location will create new files there and not
636 impact the destination of the rename.
637
638 Note that a `filerename` is the same as a `filecopy` followed by a
639 `filedelete` of the source location. There is a slight performance
640 advantage to using `filerename`, but the advantage is so small
641 that it is never worth trying to convert a delete/add pair in
642 source material into a rename for fast-import. This `filerename`
643 command is provided just to simplify frontends that already have
644 rename information and don't want bother with decomposing it into a
645 `filecopy` followed by a `filedelete`.
646
647 `filedeleteall`
648 ^^^^^^^^^^^^^^^
649 Included in a `commit` command to remove all files (and also all
650 directories) from the branch. This command resets the internal
651 branch structure to have no files in it, allowing the frontend
652 to subsequently add all interesting files from scratch.
653
654 ....
655 'deleteall' LF
656 ....
657
658 This command is extremely useful if the frontend does not know
659 (or does not care to know) what files are currently on the branch,
660 and therefore cannot generate the proper `filedelete` commands to
661 update the content.
662
663 Issuing a `filedeleteall` followed by the needed `filemodify`
664 commands to set the correct content will produce the same results
665 as sending only the needed `filemodify` and `filedelete` commands.
666 The `filedeleteall` approach may however require fast-import to use slightly
667 more memory per active branch (less than 1 MiB for even most large
668 projects); so frontends that can easily obtain only the affected
669 paths for a commit are encouraged to do so.
670
671 `notemodify`
672 ^^^^^^^^^^^^
673 Included in a `commit` `<notes_ref>` command to add a new note
674 annotating a `<committish>` or change this annotation contents.
675 Internally it is similar to filemodify 100644 on `<committish>`
676 path (maybe split into subdirectories). It's not advised to
677 use any other commands to write to the `<notes_ref>` tree except
678 `filedeleteall` to delete all existing notes in this tree.
679 This command has two different means of specifying the content
680 of the note.
681
682 External data format::
683 The data content for the note was already supplied by a prior
684 `blob` command. The frontend just needs to connect it to the
685 commit that is to be annotated.
686 +
687 ....
688 'N' SP <dataref> SP <committish> LF
689 ....
690 +
691 Here `<dataref>` can be either a mark reference (`:<idnum>`)
692 set by a prior `blob` command, or a full 40-byte SHA-1 of an
693 existing Git blob object.
694
695 Inline data format::
696 The data content for the note has not been supplied yet.
697 The frontend wants to supply it as part of this modify
698 command.
699 +
700 ....
701 'N' SP 'inline' SP <committish> LF
702 data
703 ....
704 +
705 See below for a detailed description of the `data` command.
706
707 In both formats `<committish>` is any of the commit specification
708 expressions also accepted by `from` (see above).
709
710 `mark`
711 ~~~~~~
712 Arranges for fast-import to save a reference to the current object, allowing
713 the frontend to recall this object at a future point in time, without
714 knowing its SHA-1. Here the current object is the object creation
715 command the `mark` command appears within. This can be `commit`,
716 `tag`, and `blob`, but `commit` is the most common usage.
717
718 ....
719 'mark' SP ':' <idnum> LF
720 ....
721
722 where `<idnum>` is the number assigned by the frontend to this mark.
723 The value of `<idnum>` is expressed as an ASCII decimal integer.
724 The value 0 is reserved and cannot be used as
725 a mark. Only values greater than or equal to 1 may be used as marks.
726
727 New marks are created automatically. Existing marks can be moved
728 to another object simply by reusing the same `<idnum>` in another
729 `mark` command.
730
731 `tag`
732 ~~~~~
733 Creates an annotated tag referring to a specific commit. To create
734 lightweight (non-annotated) tags see the `reset` command below.
735
736 ....
737 'tag' SP <name> LF
738 'from' SP <committish> LF
739 'tagger' (SP <name>)? SP LT <email> GT SP <when> LF
740 data
741 ....
742
743 where `<name>` is the name of the tag to create.
744
745 Tag names are automatically prefixed with `refs/tags/` when stored
746 in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would
747 use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the
748 corresponding ref as `refs/tags/RELENG-1_0-FINAL`.
749
750 The value of `<name>` must be a valid refname in Git and therefore
751 may contain forward slashes. As `LF` is not valid in a Git refname,
752 no quoting or escaping syntax is supported here.
753
754 The `from` command is the same as in the `commit` command; see
755 above for details.
756
757 The `tagger` command uses the same format as `committer` within
758 `commit`; again see above for details.
759
760 The `data` command following `tagger` must supply the annotated tag
761 message (see below for `data` command syntax). To import an empty
762 tag message use a 0 length data. Tag messages are free-form and are
763 not interpreted by Git. Currently they must be encoded in UTF-8,
764 as fast-import does not permit other encodings to be specified.
765
766 Signing annotated tags during import from within fast-import is not
767 supported. Trying to include your own PGP/GPG signature is not
768 recommended, as the frontend does not (easily) have access to the
769 complete set of bytes which normally goes into such a signature.
770 If signing is required, create lightweight tags from within fast-import with
771 `reset`, then create the annotated versions of those tags offline
772 with the standard 'git tag' process.
773
774 `reset`
775 ~~~~~~~
776 Creates (or recreates) the named branch, optionally starting from
777 a specific revision. The reset command allows a frontend to issue
778 a new `from` command for an existing branch, or to create a new
779 branch from an existing commit without creating a new commit.
780
781 ....
782 'reset' SP <ref> LF
783 ('from' SP <committish> LF)?
784 LF?
785 ....
786
787 For a detailed description of `<ref>` and `<committish>` see above
788 under `commit` and `from`.
789
790 The `LF` after the command is optional (it used to be required).
791
792 The `reset` command can also be used to create lightweight
793 (non-annotated) tags. For example:
794
795 ====
796 reset refs/tags/938
797 from :938
798 ====
799
800 would create the lightweight tag `refs/tags/938` referring to
801 whatever commit mark `:938` references.
802
803 `blob`
804 ~~~~~~
805 Requests writing one file revision to the packfile. The revision
806 is not connected to any commit; this connection must be formed in
807 a subsequent `commit` command by referencing the blob through an
808 assigned mark.
809
810 ....
811 'blob' LF
812 mark?
813 data
814 ....
815
816 The mark command is optional here as some frontends have chosen
817 to generate the Git SHA-1 for the blob on their own, and feed that
818 directly to `commit`. This is typically more work than it's worth
819 however, as marks are inexpensive to store and easy to use.
820
821 `data`
822 ~~~~~~
823 Supplies raw data (for use as blob/file content, commit messages, or
824 annotated tag messages) to fast-import. Data can be supplied using an exact
825 byte count or delimited with a terminating line. Real frontends
826 intended for production-quality conversions should always use the
827 exact byte count format, as it is more robust and performs better.
828 The delimited format is intended primarily for testing fast-import.
829
830 Comment lines appearing within the `<raw>` part of `data` commands
831 are always taken to be part of the body of the data and are therefore
832 never ignored by fast-import. This makes it safe to import any
833 file/message content whose lines might start with `#`.
834
835 Exact byte count format::
836 The frontend must specify the number of bytes of data.
837 +
838 ....
839 'data' SP <count> LF
840 <raw> LF?
841 ....
842 +
843 where `<count>` is the exact number of bytes appearing within
844 `<raw>`. The value of `<count>` is expressed as an ASCII decimal
845 integer. The `LF` on either side of `<raw>` is not
846 included in `<count>` and will not be included in the imported data.
847 +
848 The `LF` after `<raw>` is optional (it used to be required) but
849 recommended. Always including it makes debugging a fast-import
850 stream easier as the next command always starts in column 0
851 of the next line, even if `<raw>` did not end with an `LF`.
852
853 Delimited format::
854 A delimiter string is used to mark the end of the data.
855 fast-import will compute the length by searching for the delimiter.
856 This format is primarily useful for testing and is not
857 recommended for real data.
858 +
859 ....
860 'data' SP '<<' <delim> LF
861 <raw> LF
862 <delim> LF
863 LF?
864 ....
865 +
866 where `<delim>` is the chosen delimiter string. The string `<delim>`
867 must not appear on a line by itself within `<raw>`, as otherwise
868 fast-import will think the data ends earlier than it really does. The `LF`
869 immediately trailing `<raw>` is part of `<raw>`. This is one of
870 the limitations of the delimited format, it is impossible to supply
871 a data chunk which does not have an LF as its last byte.
872 +
873 The `LF` after `<delim> LF` is optional (it used to be required).
874
875 `checkpoint`
876 ~~~~~~~~~~~~
877 Forces fast-import to close the current packfile, start a new one, and to
878 save out all current branch refs, tags and marks.
879
880 ....
881 'checkpoint' LF
882 LF?
883 ....
884
885 Note that fast-import automatically switches packfiles when the current
886 packfile reaches \--max-pack-size, or 4 GiB, whichever limit is
887 smaller. During an automatic packfile switch fast-import does not update
888 the branch refs, tags or marks.
889
890 As a `checkpoint` can require a significant amount of CPU time and
891 disk IO (to compute the overall pack SHA-1 checksum, generate the
892 corresponding index file, and update the refs) it can easily take
893 several minutes for a single `checkpoint` command to complete.
894
895 Frontends may choose to issue checkpoints during extremely large
896 and long running imports, or when they need to allow another Git
897 process access to a branch. However given that a 30 GiB Subversion
898 repository can be loaded into Git through fast-import in about 3 hours,
899 explicit checkpointing may not be necessary.
900
901 The `LF` after the command is optional (it used to be required).
902
903 `progress`
904 ~~~~~~~~~~
905 Causes fast-import to print the entire `progress` line unmodified to
906 its standard output channel (file descriptor 1) when the command is
907 processed from the input stream. The command otherwise has no impact
908 on the current import, or on any of fast-import's internal state.
909
910 ....
911 'progress' SP <any> LF
912 LF?
913 ....
914
915 The `<any>` part of the command may contain any sequence of bytes
916 that does not contain `LF`. The `LF` after the command is optional.
917 Callers may wish to process the output through a tool such as sed to
918 remove the leading part of the line, for example:
919
920 ====
921 frontend | git fast-import | sed 's/^progress //'
922 ====
923
924 Placing a `progress` command immediately after a `checkpoint` will
925 inform the reader when the `checkpoint` has been completed and it
926 can safely access the refs that fast-import updated.
927
928 `cat-blob`
929 ~~~~~~~~~~
930 Causes fast-import to print a blob to a file descriptor previously
931 arranged with the `--cat-blob-fd` argument. The command otherwise
932 has no impact on the current import; its main purpose is to
933 retrieve blobs that may be in fast-import's memory but not
934 accessible from the target repository.
935
936 ....
937 'cat-blob' SP <dataref> LF
938 ....
939
940 The `<dataref>` can be either a mark reference (`:<idnum>`)
941 set previously or a full 40-byte SHA-1 of a Git blob, preexisting or
942 ready to be written.
943
944 Output uses the same format as `git cat-file --batch`:
945
946 ====
947 <sha1> SP 'blob' SP <size> LF
948 <contents> LF
949 ====
950
951 This command can be used anywhere in the stream that comments are
952 accepted. In particular, the `cat-blob` command can be used in the
953 middle of a commit but not in the middle of a `data` command.
954
955 See ``Responses To Commands'' below for details about how to read
956 this output safely.
957
958 `ls`
959 ~~~~
960 Prints information about the object at a path to a file descriptor
961 previously arranged with the `--cat-blob-fd` argument. This allows
962 printing a blob from the active commit (with `cat-blob`) or copying a
963 blob or tree from a previous commit for use in the current one (with
964 `filemodify`).
965
966 The `ls` command can be used anywhere in the stream that comments are
967 accepted, including the middle of a commit.
968
969 Reading from the active commit::
970 This form can only be used in the middle of a `commit`.
971 The path names a directory entry within fast-import's
972 active commit. The path must be quoted in this case.
973 +
974 ....
975 'ls' SP <path> LF
976 ....
977
978 Reading from a named tree::
979 The `<dataref>` can be a mark reference (`:<idnum>`) or the
980 full 40-byte SHA-1 of a Git tag, commit, or tree object,
981 preexisting or waiting to be written.
982 The path is relative to the top level of the tree
983 named by `<dataref>`.
984 +
985 ....
986 'ls' SP <dataref> SP <path> LF
987 ....
988
989 See `filemodify` above for a detailed description of `<path>`.
990
991 Output uses the same format as `git ls-tree <tree> -- <path>`:
992
993 ====
994 <mode> SP ('blob' | 'tree' | 'commit') SP <dataref> HT <path> LF
995 ====
996
997 The <dataref> represents the blob, tree, or commit object at <path>
998 and can be used in later 'cat-blob', 'filemodify', or 'ls' commands.
999
1000 If there is no file or subtree at that path, 'git fast-import' will
1001 instead report
1002
1003 ====
1004 missing SP <path> LF
1005 ====
1006
1007 See ``Responses To Commands'' below for details about how to read
1008 this output safely.
1009
1010 `feature`
1011 ~~~~~~~~~
1012 Require that fast-import supports the specified feature, or abort if
1013 it does not.
1014
1015 ....
1016 'feature' SP <feature> ('=' <argument>)? LF
1017 ....
1018
1019 The <feature> part of the command may be any one of the following:
1020
1021 date-format::
1022 export-marks::
1023 relative-marks::
1024 no-relative-marks::
1025 force::
1026 Act as though the corresponding command-line option with
1027 a leading '--' was passed on the command line
1028 (see OPTIONS, above).
1029
1030 import-marks::
1031 import-marks-if-exists::
1032 Like --import-marks except in two respects: first, only one
1033 "feature import-marks" or "feature import-marks-if-exists"
1034 command is allowed per stream; second, an --import-marks=
1035 or --import-marks-if-exists command-line option overrides
1036 any of these "feature" commands in the stream; third,
1037 "feature import-marks-if-exists" like a corresponding
1038 command-line option silently skips a nonexistent file.
1039
1040 cat-blob::
1041 ls::
1042 Require that the backend support the 'cat-blob' or 'ls' command.
1043 Versions of fast-import not supporting the specified command
1044 will exit with a message indicating so.
1045 This lets the import error out early with a clear message,
1046 rather than wasting time on the early part of an import
1047 before the unsupported command is detected.
1048
1049 notes::
1050 Require that the backend support the 'notemodify' (N)
1051 subcommand to the 'commit' command.
1052 Versions of fast-import not supporting notes will exit
1053 with a message indicating so.
1054
1055 done::
1056 Error out if the stream ends without a 'done' command.
1057 Without this feature, errors causing the frontend to end
1058 abruptly at a convenient point in the stream can go
1059 undetected. This may occur, for example, if an import
1060 front end dies in mid-operation without emitting SIGTERM
1061 or SIGKILL at its subordinate git fast-import instance.
1062
1063 `option`
1064 ~~~~~~~~
1065 Processes the specified option so that git fast-import behaves in a
1066 way that suits the frontend's needs.
1067 Note that options specified by the frontend are overridden by any
1068 options the user may specify to git fast-import itself.
1069
1070 ....
1071 'option' SP <option> LF
1072 ....
1073
1074 The `<option>` part of the command may contain any of the options
1075 listed in the OPTIONS section that do not change import semantics,
1076 without the leading '--' and is treated in the same way.
1077
1078 Option commands must be the first commands on the input (not counting
1079 feature commands), to give an option command after any non-option
1080 command is an error.
1081
1082 The following commandline options change import semantics and may therefore
1083 not be passed as option:
1084
1085 * date-format
1086 * import-marks
1087 * export-marks
1088 * cat-blob-fd
1089 * force
1090
1091 `done`
1092 ~~~~~~
1093 If the `done` feature is not in use, treated as if EOF was read.
1094 This can be used to tell fast-import to finish early.
1095
1096 If the `--done` command line option or `feature done` command is
1097 in use, the `done` command is mandatory and marks the end of the
1098 stream.
1099
1100 Responses To Commands
1101 ---------------------
1102 New objects written by fast-import are not available immediately.
1103 Most fast-import commands have no visible effect until the next
1104 checkpoint (or completion). The frontend can send commands to
1105 fill fast-import's input pipe without worrying about how quickly
1106 they will take effect, which improves performance by simplifying
1107 scheduling.
1108
1109 For some frontends, though, it is useful to be able to read back
1110 data from the current repository as it is being updated (for
1111 example when the source material describes objects in terms of
1112 patches to be applied to previously imported objects). This can
1113 be accomplished by connecting the frontend and fast-import via
1114 bidirectional pipes:
1115
1116 ====
1117 mkfifo fast-import-output
1118 frontend <fast-import-output |
1119 git fast-import >fast-import-output
1120 ====
1121
1122 A frontend set up this way can use `progress`, `ls`, and `cat-blob`
1123 commands to read information from the import in progress.
1124
1125 To avoid deadlock, such frontends must completely consume any
1126 pending output from `progress`, `ls`, and `cat-blob` before
1127 performing writes to fast-import that might block.
1128
1129 Crash Reports
1130 -------------
1131 If fast-import is supplied invalid input it will terminate with a
1132 non-zero exit status and create a crash report in the top level of
1133 the Git repository it was importing into. Crash reports contain
1134 a snapshot of the internal fast-import state as well as the most
1135 recent commands that lead up to the crash.
1136
1137 All recent commands (including stream comments, file changes and
1138 progress commands) are shown in the command history within the crash
1139 report, but raw file data and commit messages are excluded from the
1140 crash report. This exclusion saves space within the report file
1141 and reduces the amount of buffering that fast-import must perform
1142 during execution.
1143
1144 After writing a crash report fast-import will close the current
1145 packfile and export the marks table. This allows the frontend
1146 developer to inspect the repository state and resume the import from
1147 the point where it crashed. The modified branches and tags are not
1148 updated during a crash, as the import did not complete successfully.
1149 Branch and tag information can be found in the crash report and
1150 must be applied manually if the update is needed.
1151
1152 An example crash:
1153
1154 ====
1155 $ cat >in <<END_OF_INPUT
1156 # my very first test commit
1157 commit refs/heads/master
1158 committer Shawn O. Pearce <spearce> 19283 -0400
1159 # who is that guy anyway?
1160 data <<EOF
1161 this is my commit
1162 EOF
1163 M 644 inline .gitignore
1164 data <<EOF
1165 .gitignore
1166 EOF
1167 M 777 inline bob
1168 END_OF_INPUT
1169
1170 $ git fast-import <in
1171 fatal: Corrupt mode: M 777 inline bob
1172 fast-import: dumping crash report to .git/fast_import_crash_8434
1173
1174 $ cat .git/fast_import_crash_8434
1175 fast-import crash report:
1176 fast-import process: 8434
1177 parent process : 1391
1178 at Sat Sep 1 00:58:12 2007
1179
1180 fatal: Corrupt mode: M 777 inline bob
1181
1182 Most Recent Commands Before Crash
1183 ---------------------------------
1184 # my very first test commit
1185 commit refs/heads/master
1186 committer Shawn O. Pearce <spearce> 19283 -0400
1187 # who is that guy anyway?
1188 data <<EOF
1189 M 644 inline .gitignore
1190 data <<EOF
1191 * M 777 inline bob
1192
1193 Active Branch LRU
1194 -----------------
1195 active_branches = 1 cur, 5 max
1196
1197 pos clock name
1198 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1199 1) 0 refs/heads/master
1200
1201 Inactive Branches
1202 -----------------
1203 refs/heads/master:
1204 status : active loaded dirty
1205 tip commit : 0000000000000000000000000000000000000000
1206 old tree : 0000000000000000000000000000000000000000
1207 cur tree : 0000000000000000000000000000000000000000
1208 commit clock: 0
1209 last pack :
1210
1211
1212 -------------------
1213 END OF CRASH REPORT
1214 ====
1215
1216 Tips and Tricks
1217 ---------------
1218 The following tips and tricks have been collected from various
1219 users of fast-import, and are offered here as suggestions.
1220
1221 Use One Mark Per Commit
1222 ~~~~~~~~~~~~~~~~~~~~~~~
1223 When doing a repository conversion, use a unique mark per commit
1224 (`mark :<n>`) and supply the \--export-marks option on the command
1225 line. fast-import will dump a file which lists every mark and the Git
1226 object SHA-1 that corresponds to it. If the frontend can tie
1227 the marks back to the source repository, it is easy to verify the
1228 accuracy and completeness of the import by comparing each Git
1229 commit to the corresponding source revision.
1230
1231 Coming from a system such as Perforce or Subversion this should be
1232 quite simple, as the fast-import mark can also be the Perforce changeset
1233 number or the Subversion revision number.
1234
1235 Freely Skip Around Branches
1236 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1237 Don't bother trying to optimize the frontend to stick to one branch
1238 at a time during an import. Although doing so might be slightly
1239 faster for fast-import, it tends to increase the complexity of the frontend
1240 code considerably.
1241
1242 The branch LRU builtin to fast-import tends to behave very well, and the
1243 cost of activating an inactive branch is so low that bouncing around
1244 between branches has virtually no impact on import performance.
1245
1246 Handling Renames
1247 ~~~~~~~~~~~~~~~~
1248 When importing a renamed file or directory, simply delete the old
1249 name(s) and modify the new name(s) during the corresponding commit.
1250 Git performs rename detection after-the-fact, rather than explicitly
1251 during a commit.
1252
1253 Use Tag Fixup Branches
1254 ~~~~~~~~~~~~~~~~~~~~~~
1255 Some other SCM systems let the user create a tag from multiple
1256 files which are not from the same commit/changeset. Or to create
1257 tags which are a subset of the files available in the repository.
1258
1259 Importing these tags as-is in Git is impossible without making at
1260 least one commit which ``fixes up'' the files to match the content
1261 of the tag. Use fast-import's `reset` command to reset a dummy branch
1262 outside of your normal branch space to the base commit for the tag,
1263 then commit one or more file fixup commits, and finally tag the
1264 dummy branch.
1265
1266 For example since all normal branches are stored under `refs/heads/`
1267 name the tag fixup branch `TAG_FIXUP`. This way it is impossible for
1268 the fixup branch used by the importer to have namespace conflicts
1269 with real branches imported from the source (the name `TAG_FIXUP`
1270 is not `refs/heads/TAG_FIXUP`).
1271
1272 When committing fixups, consider using `merge` to connect the
1273 commit(s) which are supplying file revisions to the fixup branch.
1274 Doing so will allow tools such as 'git blame' to track
1275 through the real commit history and properly annotate the source
1276 files.
1277
1278 After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP`
1279 to remove the dummy branch.
1280
1281 Import Now, Repack Later
1282 ~~~~~~~~~~~~~~~~~~~~~~~~
1283 As soon as fast-import completes the Git repository is completely valid
1284 and ready for use. Typically this takes only a very short time,
1285 even for considerably large projects (100,000+ commits).
1286
1287 However repacking the repository is necessary to improve data
1288 locality and access performance. It can also take hours on extremely
1289 large projects (especially if -f and a large \--window parameter is
1290 used). Since repacking is safe to run alongside readers and writers,
1291 run the repack in the background and let it finish when it finishes.
1292 There is no reason to wait to explore your new Git project!
1293
1294 If you choose to wait for the repack, don't try to run benchmarks
1295 or performance tests until repacking is completed. fast-import outputs
1296 suboptimal packfiles that are simply never seen in real use
1297 situations.
1298
1299 Repacking Historical Data
1300 ~~~~~~~~~~~~~~~~~~~~~~~~~
1301 If you are repacking very old imported data (e.g. older than the
1302 last year), consider expending some extra CPU time and supplying
1303 \--window=50 (or higher) when you run 'git repack'.
1304 This will take longer, but will also produce a smaller packfile.
1305 You only need to expend the effort once, and everyone using your
1306 project will benefit from the smaller repository.
1307
1308 Include Some Progress Messages
1309 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1310 Every once in a while have your frontend emit a `progress` message
1311 to fast-import. The contents of the messages are entirely free-form,
1312 so one suggestion would be to output the current month and year
1313 each time the current commit date moves into the next month.
1314 Your users will feel better knowing how much of the data stream
1315 has been processed.
1316
1317
1318 Packfile Optimization
1319 ---------------------
1320 When packing a blob fast-import always attempts to deltify against the last
1321 blob written. Unless specifically arranged for by the frontend,
1322 this will probably not be a prior version of the same file, so the
1323 generated delta will not be the smallest possible. The resulting
1324 packfile will be compressed, but will not be optimal.
1325
1326 Frontends which have efficient access to all revisions of a
1327 single file (for example reading an RCS/CVS ,v file) can choose
1328 to supply all revisions of that file as a sequence of consecutive
1329 `blob` commands. This allows fast-import to deltify the different file
1330 revisions against each other, saving space in the final packfile.
1331 Marks can be used to later identify individual file revisions during
1332 a sequence of `commit` commands.
1333
1334 The packfile(s) created by fast-import do not encourage good disk access
1335 patterns. This is caused by fast-import writing the data in the order
1336 it is received on standard input, while Git typically organizes
1337 data within packfiles to make the most recent (current tip) data
1338 appear before historical data. Git also clusters commits together,
1339 speeding up revision traversal through better cache locality.
1340
1341 For this reason it is strongly recommended that users repack the
1342 repository with `git repack -a -d` after fast-import completes, allowing
1343 Git to reorganize the packfiles for faster data access. If blob
1344 deltas are suboptimal (see above) then also adding the `-f` option
1345 to force recomputation of all deltas can significantly reduce the
1346 final packfile size (30-50% smaller can be quite typical).
1347
1348
1349 Memory Utilization
1350 ------------------
1351 There are a number of factors which affect how much memory fast-import
1352 requires to perform an import. Like critical sections of core
1353 Git, fast-import uses its own memory allocators to amortize any overheads
1354 associated with malloc. In practice fast-import tends to amortize any
1355 malloc overheads to 0, due to its use of large block allocations.
1356
1357 per object
1358 ~~~~~~~~~~
1359 fast-import maintains an in-memory structure for every object written in
1360 this execution. On a 32 bit system the structure is 32 bytes,
1361 on a 64 bit system the structure is 40 bytes (due to the larger
1362 pointer sizes). Objects in the table are not deallocated until
1363 fast-import terminates. Importing 2 million objects on a 32 bit system
1364 will require approximately 64 MiB of memory.
1365
1366 The object table is actually a hashtable keyed on the object name
1367 (the unique SHA-1). This storage configuration allows fast-import to reuse
1368 an existing or already written object and avoid writing duplicates
1369 to the output packfile. Duplicate blobs are surprisingly common
1370 in an import, typically due to branch merges in the source.
1371
1372 per mark
1373 ~~~~~~~~
1374 Marks are stored in a sparse array, using 1 pointer (4 bytes or 8
1375 bytes, depending on pointer size) per mark. Although the array
1376 is sparse, frontends are still strongly encouraged to use marks
1377 between 1 and n, where n is the total number of marks required for
1378 this import.
1379
1380 per branch
1381 ~~~~~~~~~~
1382 Branches are classified as active and inactive. The memory usage
1383 of the two classes is significantly different.
1384
1385 Inactive branches are stored in a structure which uses 96 or 120
1386 bytes (32 bit or 64 bit systems, respectively), plus the length of
1387 the branch name (typically under 200 bytes), per branch. fast-import will
1388 easily handle as many as 10,000 inactive branches in under 2 MiB
1389 of memory.
1390
1391 Active branches have the same overhead as inactive branches, but
1392 also contain copies of every tree that has been recently modified on
1393 that branch. If subtree `include` has not been modified since the
1394 branch became active, its contents will not be loaded into memory,
1395 but if subtree `src` has been modified by a commit since the branch
1396 became active, then its contents will be loaded in memory.
1397
1398 As active branches store metadata about the files contained on that
1399 branch, their in-memory storage size can grow to a considerable size
1400 (see below).
1401
1402 fast-import automatically moves active branches to inactive status based on
1403 a simple least-recently-used algorithm. The LRU chain is updated on
1404 each `commit` command. The maximum number of active branches can be
1405 increased or decreased on the command line with \--active-branches=.
1406
1407 per active tree
1408 ~~~~~~~~~~~~~~~
1409 Trees (aka directories) use just 12 bytes of memory on top of the
1410 memory required for their entries (see ``per active file'' below).
1411 The cost of a tree is virtually 0, as its overhead amortizes out
1412 over the individual file entries.
1413
1414 per active file entry
1415 ~~~~~~~~~~~~~~~~~~~~~
1416 Files (and pointers to subtrees) within active trees require 52 or 64
1417 bytes (32/64 bit platforms) per entry. To conserve space, file and
1418 tree names are pooled in a common string table, allowing the filename
1419 ``Makefile'' to use just 16 bytes (after including the string header
1420 overhead) no matter how many times it occurs within the project.
1421
1422 The active branch LRU, when coupled with the filename string pool
1423 and lazy loading of subtrees, allows fast-import to efficiently import
1424 projects with 2,000+ branches and 45,114+ files in a very limited
1425 memory footprint (less than 2.7 MiB per active branch).
1426
1427 Signals
1428 -------
1429 Sending *SIGUSR1* to the 'git fast-import' process ends the current
1430 packfile early, simulating a `checkpoint` command. The impatient
1431 operator can use this facility to peek at the objects and refs from an
1432 import in progress, at the cost of some added running time and worse
1433 compression.
1434
1435 GIT
1436 ---
1437 Part of the linkgit:git[1] suite