cdf696ff7f6d9873d4d5c069022eae547911ede8
[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
37 --force::
38 Force updating modified existing branches, even if doing
39 so would cause commits to be lost (as the new commit does
40 not contain the old commit).
41
42 --quiet::
43 Disable all non-fatal output, making fast-import silent when it
44 is successful. This option disables the output shown by
45 --stats.
46
47 --stats::
48 Display some basic statistics about the objects fast-import has
49 created, the packfiles they were stored into, and the
50 memory used by fast-import during this run. Showing this output
51 is currently the default, but can be disabled with --quiet.
52
53 Options for Frontends
54 ~~~~~~~~~~~~~~~~~~~~~
55
56 --cat-blob-fd=<fd>::
57 Write responses to `get-mark`, `cat-blob`, and `ls` queries to the
58 file descriptor <fd> instead of `stdout`. Allows `progress`
59 output intended for the end-user to be separated from other
60 output.
61
62 --date-format=<fmt>::
63 Specify the type of dates the frontend will supply to
64 fast-import within `author`, `committer` and `tagger` commands.
65 See ``Date Formats'' below for details about which formats
66 are supported, and their syntax.
67
68 --done::
69 Terminate with error if there is no `done` command at the end of
70 the stream. This option might be useful for detecting errors
71 that cause the frontend to terminate before it has started to
72 write a stream.
73
74 Locations of Marks Files
75 ~~~~~~~~~~~~~~~~~~~~~~~~
76
77 --export-marks=<file>::
78 Dumps the internal marks table to <file> when complete.
79 Marks are written one per line as `:markid SHA-1`.
80 Frontends can use this file to validate imports after they
81 have been completed, or to save the marks table across
82 incremental runs. As <file> is only opened and truncated
83 at checkpoint (or completion) the same path can also be
84 safely given to --import-marks.
85
86 --import-marks=<file>::
87 Before processing any input, load the marks specified in
88 <file>. The input file must exist, must be readable, and
89 must use the same format as produced by --export-marks.
90 Multiple options may be supplied to import more than one
91 set of marks. If a mark is defined to different values,
92 the last file wins.
93
94 --import-marks-if-exists=<file>::
95 Like --import-marks but instead of erroring out, silently
96 skips the file if it does not exist.
97
98 --[no-]relative-marks::
99 After specifying --relative-marks the paths specified
100 with --import-marks= and --export-marks= are relative
101 to an internal directory in the current repository.
102 In git-fast-import this means that the paths are relative
103 to the .git/info/fast-import directory. However, other
104 importers may use a different location.
105 +
106 Relative and non-relative marks may be combined by interweaving
107 --(no-)-relative-marks with the --(import|export)-marks= options.
108
109 Performance and Compression Tuning
110 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
111
112 --active-branches=<n>::
113 Maximum number of branches to maintain active at once.
114 See ``Memory Utilization'' below for details. Default is 5.
115
116 --big-file-threshold=<n>::
117 Maximum size of a blob that fast-import will attempt to
118 create a delta for, expressed in bytes. The default is 512m
119 (512 MiB). Some importers may wish to lower this on systems
120 with constrained memory.
121
122 --depth=<n>::
123 Maximum delta depth, for blob and tree deltification.
124 Default is 50.
125
126 --export-pack-edges=<file>::
127 After creating a packfile, print a line of data to
128 <file> listing the filename of the packfile and the last
129 commit on each branch that was written to that packfile.
130 This information may be useful after importing projects
131 whose total object set exceeds the 4 GiB packfile limit,
132 as these commits can be used as edge points during calls
133 to 'git pack-objects'.
134
135 --max-pack-size=<n>::
136 Maximum size of each output packfile.
137 The default is unlimited.
138
139 fastimport.unpackLimit::
140 See linkgit:git-config[1]
141
142 PERFORMANCE
143 -----------
144 The design of fast-import allows it to import large projects in a minimum
145 amount of memory usage and processing time. Assuming the frontend
146 is able to keep up with fast-import and feed it a constant stream of data,
147 import times for projects holding 10+ years of history and containing
148 100,000+ individual commits are generally completed in just 1-2
149 hours on quite modest (~$2,000 USD) hardware.
150
151 Most bottlenecks appear to be in foreign source data access (the
152 source just cannot extract revisions fast enough) or disk IO (fast-import
153 writes as fast as the disk will take the data). Imports will run
154 faster if the source data is stored on a different drive than the
155 destination Git repository (due to less IO contention).
156
157
158 DEVELOPMENT COST
159 ----------------
160 A typical frontend for fast-import tends to weigh in at approximately 200
161 lines of Perl/Python/Ruby code. Most developers have been able to
162 create working importers in just a couple of hours, even though it
163 is their first exposure to fast-import, and sometimes even to Git. This is
164 an ideal situation, given that most conversion tools are throw-away
165 (use once, and never look back).
166
167
168 PARALLEL OPERATION
169 ------------------
170 Like 'git push' or 'git fetch', imports handled by fast-import are safe to
171 run alongside parallel `git repack -a -d` or `git gc` invocations,
172 or any other Git operation (including 'git prune', as loose objects
173 are never used by fast-import).
174
175 fast-import does not lock the branch or tag refs it is actively importing.
176 After the import, during its ref update phase, fast-import tests each
177 existing branch ref to verify the update will be a fast-forward
178 update (the commit stored in the ref is contained in the new
179 history of the commit to be written). If the update is not a
180 fast-forward update, fast-import will skip updating that ref and instead
181 prints a warning message. fast-import will always attempt to update all
182 branch refs, and does not stop on the first failure.
183
184 Branch updates can be forced with --force, but it's recommended that
185 this only be used on an otherwise quiet repository. Using --force
186 is not necessary for an initial import into an empty repository.
187
188
189 TECHNICAL DISCUSSION
190 --------------------
191 fast-import tracks a set of branches in memory. Any branch can be created
192 or modified at any point during the import process by sending a
193 `commit` command on the input stream. This design allows a frontend
194 program to process an unlimited number of branches simultaneously,
195 generating commits in the order they are available from the source
196 data. It also simplifies the frontend programs considerably.
197
198 fast-import does not use or alter the current working directory, or any
199 file within it. (It does however update the current Git repository,
200 as referenced by `GIT_DIR`.) Therefore an import frontend may use
201 the working directory for its own purposes, such as extracting file
202 revisions from the foreign source. This ignorance of the working
203 directory also allows fast-import to run very quickly, as it does not
204 need to perform any costly file update operations when switching
205 between branches.
206
207 INPUT FORMAT
208 ------------
209 With the exception of raw file data (which Git does not interpret)
210 the fast-import input format is text (ASCII) based. This text based
211 format simplifies development and debugging of frontend programs,
212 especially when a higher level language such as Perl, Python or
213 Ruby is being used.
214
215 fast-import is very strict about its input. Where we say SP below we mean
216 *exactly* one space. Likewise LF means one (and only one) linefeed
217 and HT one (and only one) horizontal tab.
218 Supplying additional whitespace characters will cause unexpected
219 results, such as branch names or file names with leading or trailing
220 spaces in their name, or early termination of fast-import when it encounters
221 unexpected input.
222
223 Stream Comments
224 ~~~~~~~~~~~~~~~
225 To aid in debugging frontends fast-import ignores any line that
226 begins with `#` (ASCII pound/hash) up to and including the line
227 ending `LF`. A comment line may contain any sequence of bytes
228 that does not contain an LF and therefore may be used to include
229 any detailed debugging information that might be specific to the
230 frontend and useful when inspecting a fast-import data stream.
231
232 Date Formats
233 ~~~~~~~~~~~~
234 The following date formats are supported. A frontend should select
235 the format it will use for this import by passing the format name
236 in the --date-format=<fmt> command-line option.
237
238 `raw`::
239 This is the Git native format and is `<time> SP <offutc>`.
240 It is also fast-import's default format, if --date-format was
241 not specified.
242 +
243 The time of the event is specified by `<time>` as the number of
244 seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is
245 written as an ASCII decimal integer.
246 +
247 The local offset is specified by `<offutc>` as a positive or negative
248 offset from UTC. For example EST (which is 5 hours behind UTC)
249 would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''.
250 The local offset does not affect `<time>`; it is used only as an
251 advisement to help formatting routines display the timestamp.
252 +
253 If the local offset is not available in the source material, use
254 ``+0000'', or the most common local offset. For example many
255 organizations have a CVS repository which has only ever been accessed
256 by users who are located in the same location and time zone. In this
257 case a reasonable offset from UTC could be assumed.
258 +
259 Unlike the `rfc2822` format, this format is very strict. Any
260 variation in formatting will cause fast-import to reject the value.
261
262 `rfc2822`::
263 This is the standard email format as described by RFC 2822.
264 +
265 An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git
266 parser is accurate, but a little on the lenient side. It is the
267 same parser used by 'git am' when applying patches
268 received from email.
269 +
270 Some malformed strings may be accepted as valid dates. In some of
271 these cases Git will still be able to obtain the correct date from
272 the malformed string. There are also some types of malformed
273 strings which Git will parse wrong, and yet consider valid.
274 Seriously malformed strings will be rejected.
275 +
276 Unlike the `raw` format above, the time zone/UTC offset information
277 contained in an RFC 2822 date string is used to adjust the date
278 value to UTC prior to storage. Therefore it is important that
279 this information be as accurate as possible.
280 +
281 If the source material uses RFC 2822 style dates,
282 the frontend should let fast-import handle the parsing and conversion
283 (rather than attempting to do it itself) as the Git parser has
284 been well tested in the wild.
285 +
286 Frontends should prefer the `raw` format if the source material
287 already uses UNIX-epoch format, can be coaxed to give dates in that
288 format, or its format is easily convertible to it, as there is no
289 ambiguity in parsing.
290
291 `now`::
292 Always use the current time and time zone. The literal
293 `now` must always be supplied for `<when>`.
294 +
295 This is a toy format. The current time and time zone of this system
296 is always copied into the identity string at the time it is being
297 created by fast-import. There is no way to specify a different time or
298 time zone.
299 +
300 This particular format is supplied as it's short to implement and
301 may be useful to a process that wants to create a new commit
302 right now, without needing to use a working directory or
303 'git update-index'.
304 +
305 If separate `author` and `committer` commands are used in a `commit`
306 the timestamps may not match, as the system clock will be polled
307 twice (once for each command). The only way to ensure that both
308 author and committer identity information has the same timestamp
309 is to omit `author` (thus copying from `committer`) or to use a
310 date format other than `now`.
311
312 Commands
313 ~~~~~~~~
314 fast-import accepts several commands to update the current repository
315 and control the current import process. More detailed discussion
316 (with examples) of each command follows later.
317
318 `commit`::
319 Creates a new branch or updates an existing branch by
320 creating a new commit and updating the branch to point at
321 the newly created commit.
322
323 `tag`::
324 Creates an annotated tag object from an existing commit or
325 branch. Lightweight tags are not supported by this command,
326 as they are not recommended for recording meaningful points
327 in time.
328
329 `reset`::
330 Reset an existing branch (or a new branch) to a specific
331 revision. This command must be used to change a branch to
332 a specific revision without making a commit on it.
333
334 `blob`::
335 Convert raw file data into a blob, for future use in a
336 `commit` command. This command is optional and is not
337 needed to perform an import.
338
339 `checkpoint`::
340 Forces fast-import to close the current packfile, generate its
341 unique SHA-1 checksum and index, and start a new packfile.
342 This command is optional and is not needed to perform
343 an import.
344
345 `progress`::
346 Causes fast-import to echo the entire line to its own
347 standard output. This command is optional and is not needed
348 to perform an import.
349
350 `done`::
351 Marks the end of the stream. This command is optional
352 unless the `done` feature was requested using the
353 `--done` command-line option or `feature done` command.
354
355 `get-mark`::
356 Causes fast-import to print the SHA-1 corresponding to a mark
357 to the file descriptor set with `--cat-blob-fd`, or `stdout` if
358 unspecified.
359
360 `cat-blob`::
361 Causes fast-import to print a blob in 'cat-file --batch'
362 format to the file descriptor set with `--cat-blob-fd` or
363 `stdout` if unspecified.
364
365 `ls`::
366 Causes fast-import to print a line describing a directory
367 entry in 'ls-tree' format to the file descriptor set with
368 `--cat-blob-fd` or `stdout` if unspecified.
369
370 `feature`::
371 Enable the specified feature. This requires that fast-import
372 supports the specified feature, and aborts if it does not.
373
374 `option`::
375 Specify any of the options listed under OPTIONS that do not
376 change stream semantic to suit the frontend's needs. This
377 command is optional and is not needed to perform an import.
378
379 `commit`
380 ~~~~~~~~
381 Create or update a branch with a new commit, recording one logical
382 change to the project.
383
384 ....
385 'commit' SP <ref> LF
386 mark?
387 ('author' (SP <name>)? SP LT <email> GT SP <when> LF)?
388 'committer' (SP <name>)? SP LT <email> GT SP <when> LF
389 data
390 ('from' SP <commit-ish> LF)?
391 ('merge' SP <commit-ish> LF)?
392 (filemodify | filedelete | filecopy | filerename | filedeleteall | notemodify)*
393 LF?
394 ....
395
396 where `<ref>` is the name of the branch to make the commit on.
397 Typically branch names are prefixed with `refs/heads/` in
398 Git, so importing the CVS branch symbol `RELENG-1_0` would use
399 `refs/heads/RELENG-1_0` for the value of `<ref>`. The value of
400 `<ref>` must be a valid refname in Git. As `LF` is not valid in
401 a Git refname, no quoting or escaping syntax is supported here.
402
403 A `mark` command may optionally appear, requesting fast-import to save a
404 reference to the newly created commit for future use by the frontend
405 (see below for format). It is very common for frontends to mark
406 every commit they create, thereby allowing future branch creation
407 from any imported commit.
408
409 The `data` command following `committer` must supply the commit
410 message (see below for `data` command syntax). To import an empty
411 commit message use a 0 length data. Commit messages are free-form
412 and are not interpreted by Git. Currently they must be encoded in
413 UTF-8, as fast-import does not permit other encodings to be specified.
414
415 Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename`,
416 `filedeleteall` and `notemodify` commands
417 may be included to update the contents of the branch prior to
418 creating the commit. These commands may be supplied in any order.
419 However it is recommended that a `filedeleteall` command precede
420 all `filemodify`, `filecopy`, `filerename` and `notemodify` commands in
421 the same commit, as `filedeleteall` wipes the branch clean (see below).
422
423 The `LF` after the command is optional (it used to be required).
424
425 `author`
426 ^^^^^^^^
427 An `author` command may optionally appear, if the author information
428 might differ from the committer information. If `author` is omitted
429 then fast-import will automatically use the committer's information for
430 the author portion of the commit. See below for a description of
431 the fields in `author`, as they are identical to `committer`.
432
433 `committer`
434 ^^^^^^^^^^^
435 The `committer` command indicates who made this commit, and when
436 they made it.
437
438 Here `<name>` is the person's display name (for example
439 ``Com M Itter'') and `<email>` is the person's email address
440 (``\cm@example.com''). `LT` and `GT` are the literal less-than (\x3c)
441 and greater-than (\x3e) symbols. These are required to delimit
442 the email address from the other fields in the line. Note that
443 `<name>` and `<email>` are free-form and may contain any sequence
444 of bytes, except `LT`, `GT` and `LF`. `<name>` is typically UTF-8 encoded.
445
446 The time of the change is specified by `<when>` using the date format
447 that was selected by the --date-format=<fmt> command-line option.
448 See ``Date Formats'' above for the set of supported formats, and
449 their syntax.
450
451 `from`
452 ^^^^^^
453 The `from` command is used to specify the commit to initialize
454 this branch from. This revision will be the first ancestor of the
455 new commit. The state of the tree built at this commit will begin
456 with the state at the `from` commit, and be altered by the content
457 modifications in this commit.
458
459 Omitting the `from` command in the first commit of a new branch
460 will cause fast-import to create that commit with no ancestor. This
461 tends to be desired only for the initial commit of a project.
462 If the frontend creates all files from scratch when making a new
463 branch, a `merge` command may be used instead of `from` to start
464 the commit with an empty tree.
465 Omitting the `from` command on existing branches is usually desired,
466 as the current commit on that branch is automatically assumed to
467 be the first ancestor of the new commit.
468
469 As `LF` is not valid in a Git refname or SHA-1 expression, no
470 quoting or escaping syntax is supported within `<commit-ish>`.
471
472 Here `<commit-ish>` is any of the following:
473
474 * The name of an existing branch already in fast-import's internal branch
475 table. If fast-import doesn't know the name, it's treated as a SHA-1
476 expression.
477
478 * A mark reference, `:<idnum>`, where `<idnum>` is the mark number.
479 +
480 The reason fast-import uses `:` to denote a mark reference is this character
481 is not legal in a Git branch name. The leading `:` makes it easy
482 to distinguish between the mark 42 (`:42`) and the branch 42 (`42`
483 or `refs/heads/42`), or an abbreviated SHA-1 which happened to
484 consist only of base-10 digits.
485 +
486 Marks must be declared (via `mark`) before they can be used.
487
488 * A complete 40 byte or abbreviated commit SHA-1 in hex.
489
490 * Any valid Git SHA-1 expression that resolves to a commit. See
491 ``SPECIFYING REVISIONS'' in linkgit:gitrevisions[7] for details.
492
493 * The special null SHA-1 (40 zeros) specifies that the branch is to be
494 removed.
495
496 The special case of restarting an incremental import from the
497 current branch value should be written as:
498 ----
499 from refs/heads/branch^0
500 ----
501 The `^0` suffix is necessary as fast-import does not permit a branch to
502 start from itself, and the branch is created in memory before the
503 `from` command is even read from the input. Adding `^0` will force
504 fast-import to resolve the commit through Git's revision parsing library,
505 rather than its internal branch table, thereby loading in the
506 existing value of the branch.
507
508 `merge`
509 ^^^^^^^
510 Includes one additional ancestor commit. The additional ancestry
511 link does not change the way the tree state is built at this commit.
512 If the `from` command is
513 omitted when creating a new branch, the first `merge` commit will be
514 the first ancestor of the current commit, and the branch will start
515 out with no files. An unlimited number of `merge` commands per
516 commit are permitted by fast-import, thereby establishing an n-way merge.
517
518 Here `<commit-ish>` is any of the commit specification expressions
519 also accepted by `from` (see above).
520
521 `filemodify`
522 ^^^^^^^^^^^^
523 Included in a `commit` command to add a new file or change the
524 content of an existing file. This command has two different means
525 of specifying the content of the file.
526
527 External data format::
528 The data content for the file was already supplied by a prior
529 `blob` command. The frontend just needs to connect it.
530 +
531 ....
532 'M' SP <mode> SP <dataref> SP <path> LF
533 ....
534 +
535 Here usually `<dataref>` must be either a mark reference (`:<idnum>`)
536 set by a prior `blob` command, or a full 40-byte SHA-1 of an
537 existing Git blob object. If `<mode>` is `040000`` then
538 `<dataref>` must be the full 40-byte SHA-1 of an existing
539 Git tree object or a mark reference set with `--import-marks`.
540
541 Inline data format::
542 The data content for the file has not been supplied yet.
543 The frontend wants to supply it as part of this modify
544 command.
545 +
546 ....
547 'M' SP <mode> SP 'inline' SP <path> LF
548 data
549 ....
550 +
551 See below for a detailed description of the `data` command.
552
553 In both formats `<mode>` is the type of file entry, specified
554 in octal. Git only supports the following modes:
555
556 * `100644` or `644`: A normal (not-executable) file. The majority
557 of files in most projects use this mode. If in doubt, this is
558 what you want.
559 * `100755` or `755`: A normal, but executable, file.
560 * `120000`: A symlink, the content of the file will be the link target.
561 * `160000`: A gitlink, SHA-1 of the object refers to a commit in
562 another repository. Git links can only be specified by SHA or through
563 a commit mark. They are used to implement submodules.
564 * `040000`: A subdirectory. Subdirectories can only be specified by
565 SHA or through a tree mark set with `--import-marks`.
566
567 In both formats `<path>` is the complete path of the file to be added
568 (if not already existing) or modified (if already existing).
569
570 A `<path>` string must use UNIX-style directory separators (forward
571 slash `/`), may contain any byte other than `LF`, and must not
572 start with double quote (`"`).
573
574 A path can use C-style string quoting; this is accepted in all cases
575 and mandatory if the filename starts with double quote or contains
576 `LF`. In C-style quoting, the complete name should be surrounded with
577 double quotes, and any `LF`, backslash, or double quote characters
578 must be escaped by preceding them with a backslash (e.g.,
579 `"path/with\n, \\ and \" in it"`).
580
581 The value of `<path>` must be in canonical form. That is it must not:
582
583 * contain an empty directory component (e.g. `foo//bar` is invalid),
584 * end with a directory separator (e.g. `foo/` is invalid),
585 * start with a directory separator (e.g. `/foo` is invalid),
586 * contain the special component `.` or `..` (e.g. `foo/./bar` and
587 `foo/../bar` are invalid).
588
589 The root of the tree can be represented by an empty string as `<path>`.
590
591 It is recommended that `<path>` always be encoded using UTF-8.
592
593 `filedelete`
594 ^^^^^^^^^^^^
595 Included in a `commit` command to remove a file or recursively
596 delete an entire directory from the branch. If the file or directory
597 removal makes its parent directory empty, the parent directory will
598 be automatically removed too. This cascades up the tree until the
599 first non-empty directory or the root is reached.
600
601 ....
602 'D' SP <path> LF
603 ....
604
605 here `<path>` is the complete path of the file or subdirectory to
606 be removed from the branch.
607 See `filemodify` above for a detailed description of `<path>`.
608
609 `filecopy`
610 ^^^^^^^^^^
611 Recursively copies an existing file or subdirectory to a different
612 location within the branch. The existing file or directory must
613 exist. If the destination exists it will be completely replaced
614 by the content copied from the source.
615
616 ....
617 'C' SP <path> SP <path> LF
618 ....
619
620 here the first `<path>` is the source location and the second
621 `<path>` is the destination. See `filemodify` above for a detailed
622 description of what `<path>` may look like. To use a source path
623 that contains SP the path must be quoted.
624
625 A `filecopy` command takes effect immediately. Once the source
626 location has been copied to the destination any future commands
627 applied to the source location will not impact the destination of
628 the copy.
629
630 `filerename`
631 ^^^^^^^^^^^^
632 Renames an existing file or subdirectory to a different location
633 within the branch. The existing file or directory must exist. If
634 the destination exists it will be replaced by the source directory.
635
636 ....
637 'R' SP <path> SP <path> LF
638 ....
639
640 here the first `<path>` is the source location and the second
641 `<path>` is the destination. See `filemodify` above for a detailed
642 description of what `<path>` may look like. To use a source path
643 that contains SP the path must be quoted.
644
645 A `filerename` command takes effect immediately. Once the source
646 location has been renamed to the destination any future commands
647 applied to the source location will create new files there and not
648 impact the destination of the rename.
649
650 Note that a `filerename` is the same as a `filecopy` followed by a
651 `filedelete` of the source location. There is a slight performance
652 advantage to using `filerename`, but the advantage is so small
653 that it is never worth trying to convert a delete/add pair in
654 source material into a rename for fast-import. This `filerename`
655 command is provided just to simplify frontends that already have
656 rename information and don't want bother with decomposing it into a
657 `filecopy` followed by a `filedelete`.
658
659 `filedeleteall`
660 ^^^^^^^^^^^^^^^
661 Included in a `commit` command to remove all files (and also all
662 directories) from the branch. This command resets the internal
663 branch structure to have no files in it, allowing the frontend
664 to subsequently add all interesting files from scratch.
665
666 ....
667 'deleteall' LF
668 ....
669
670 This command is extremely useful if the frontend does not know
671 (or does not care to know) what files are currently on the branch,
672 and therefore cannot generate the proper `filedelete` commands to
673 update the content.
674
675 Issuing a `filedeleteall` followed by the needed `filemodify`
676 commands to set the correct content will produce the same results
677 as sending only the needed `filemodify` and `filedelete` commands.
678 The `filedeleteall` approach may however require fast-import to use slightly
679 more memory per active branch (less than 1 MiB for even most large
680 projects); so frontends that can easily obtain only the affected
681 paths for a commit are encouraged to do so.
682
683 `notemodify`
684 ^^^^^^^^^^^^
685 Included in a `commit` `<notes_ref>` command to add a new note
686 annotating a `<commit-ish>` or change this annotation contents.
687 Internally it is similar to filemodify 100644 on `<commit-ish>`
688 path (maybe split into subdirectories). It's not advised to
689 use any other commands to write to the `<notes_ref>` tree except
690 `filedeleteall` to delete all existing notes in this tree.
691 This command has two different means of specifying the content
692 of the note.
693
694 External data format::
695 The data content for the note was already supplied by a prior
696 `blob` command. The frontend just needs to connect it to the
697 commit that is to be annotated.
698 +
699 ....
700 'N' SP <dataref> SP <commit-ish> LF
701 ....
702 +
703 Here `<dataref>` can be either a mark reference (`:<idnum>`)
704 set by a prior `blob` command, or a full 40-byte SHA-1 of an
705 existing Git blob object.
706
707 Inline data format::
708 The data content for the note has not been supplied yet.
709 The frontend wants to supply it as part of this modify
710 command.
711 +
712 ....
713 'N' SP 'inline' SP <commit-ish> LF
714 data
715 ....
716 +
717 See below for a detailed description of the `data` command.
718
719 In both formats `<commit-ish>` is any of the commit specification
720 expressions also accepted by `from` (see above).
721
722 `mark`
723 ~~~~~~
724 Arranges for fast-import to save a reference to the current object, allowing
725 the frontend to recall this object at a future point in time, without
726 knowing its SHA-1. Here the current object is the object creation
727 command the `mark` command appears within. This can be `commit`,
728 `tag`, and `blob`, but `commit` is the most common usage.
729
730 ....
731 'mark' SP ':' <idnum> LF
732 ....
733
734 where `<idnum>` is the number assigned by the frontend to this mark.
735 The value of `<idnum>` is expressed as an ASCII decimal integer.
736 The value 0 is reserved and cannot be used as
737 a mark. Only values greater than or equal to 1 may be used as marks.
738
739 New marks are created automatically. Existing marks can be moved
740 to another object simply by reusing the same `<idnum>` in another
741 `mark` command.
742
743 `tag`
744 ~~~~~
745 Creates an annotated tag referring to a specific commit. To create
746 lightweight (non-annotated) tags see the `reset` command below.
747
748 ....
749 'tag' SP <name> LF
750 'from' SP <commit-ish> LF
751 'tagger' (SP <name>)? SP LT <email> GT SP <when> LF
752 data
753 ....
754
755 where `<name>` is the name of the tag to create.
756
757 Tag names are automatically prefixed with `refs/tags/` when stored
758 in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would
759 use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the
760 corresponding ref as `refs/tags/RELENG-1_0-FINAL`.
761
762 The value of `<name>` must be a valid refname in Git and therefore
763 may contain forward slashes. As `LF` is not valid in a Git refname,
764 no quoting or escaping syntax is supported here.
765
766 The `from` command is the same as in the `commit` command; see
767 above for details.
768
769 The `tagger` command uses the same format as `committer` within
770 `commit`; again see above for details.
771
772 The `data` command following `tagger` must supply the annotated tag
773 message (see below for `data` command syntax). To import an empty
774 tag message use a 0 length data. Tag messages are free-form and are
775 not interpreted by Git. Currently they must be encoded in UTF-8,
776 as fast-import does not permit other encodings to be specified.
777
778 Signing annotated tags during import from within fast-import is not
779 supported. Trying to include your own PGP/GPG signature is not
780 recommended, as the frontend does not (easily) have access to the
781 complete set of bytes which normally goes into such a signature.
782 If signing is required, create lightweight tags from within fast-import with
783 `reset`, then create the annotated versions of those tags offline
784 with the standard 'git tag' process.
785
786 `reset`
787 ~~~~~~~
788 Creates (or recreates) the named branch, optionally starting from
789 a specific revision. The reset command allows a frontend to issue
790 a new `from` command for an existing branch, or to create a new
791 branch from an existing commit without creating a new commit.
792
793 ....
794 'reset' SP <ref> LF
795 ('from' SP <commit-ish> LF)?
796 LF?
797 ....
798
799 For a detailed description of `<ref>` and `<commit-ish>` see above
800 under `commit` and `from`.
801
802 The `LF` after the command is optional (it used to be required).
803
804 The `reset` command can also be used to create lightweight
805 (non-annotated) tags. For example:
806
807 ====
808 reset refs/tags/938
809 from :938
810 ====
811
812 would create the lightweight tag `refs/tags/938` referring to
813 whatever commit mark `:938` references.
814
815 `blob`
816 ~~~~~~
817 Requests writing one file revision to the packfile. The revision
818 is not connected to any commit; this connection must be formed in
819 a subsequent `commit` command by referencing the blob through an
820 assigned mark.
821
822 ....
823 'blob' LF
824 mark?
825 data
826 ....
827
828 The mark command is optional here as some frontends have chosen
829 to generate the Git SHA-1 for the blob on their own, and feed that
830 directly to `commit`. This is typically more work than it's worth
831 however, as marks are inexpensive to store and easy to use.
832
833 `data`
834 ~~~~~~
835 Supplies raw data (for use as blob/file content, commit messages, or
836 annotated tag messages) to fast-import. Data can be supplied using an exact
837 byte count or delimited with a terminating line. Real frontends
838 intended for production-quality conversions should always use the
839 exact byte count format, as it is more robust and performs better.
840 The delimited format is intended primarily for testing fast-import.
841
842 Comment lines appearing within the `<raw>` part of `data` commands
843 are always taken to be part of the body of the data and are therefore
844 never ignored by fast-import. This makes it safe to import any
845 file/message content whose lines might start with `#`.
846
847 Exact byte count format::
848 The frontend must specify the number of bytes of data.
849 +
850 ....
851 'data' SP <count> LF
852 <raw> LF?
853 ....
854 +
855 where `<count>` is the exact number of bytes appearing within
856 `<raw>`. The value of `<count>` is expressed as an ASCII decimal
857 integer. The `LF` on either side of `<raw>` is not
858 included in `<count>` and will not be included in the imported data.
859 +
860 The `LF` after `<raw>` is optional (it used to be required) but
861 recommended. Always including it makes debugging a fast-import
862 stream easier as the next command always starts in column 0
863 of the next line, even if `<raw>` did not end with an `LF`.
864
865 Delimited format::
866 A delimiter string is used to mark the end of the data.
867 fast-import will compute the length by searching for the delimiter.
868 This format is primarily useful for testing and is not
869 recommended for real data.
870 +
871 ....
872 'data' SP '<<' <delim> LF
873 <raw> LF
874 <delim> LF
875 LF?
876 ....
877 +
878 where `<delim>` is the chosen delimiter string. The string `<delim>`
879 must not appear on a line by itself within `<raw>`, as otherwise
880 fast-import will think the data ends earlier than it really does. The `LF`
881 immediately trailing `<raw>` is part of `<raw>`. This is one of
882 the limitations of the delimited format, it is impossible to supply
883 a data chunk which does not have an LF as its last byte.
884 +
885 The `LF` after `<delim> LF` is optional (it used to be required).
886
887 `checkpoint`
888 ~~~~~~~~~~~~
889 Forces fast-import to close the current packfile, start a new one, and to
890 save out all current branch refs, tags and marks.
891
892 ....
893 'checkpoint' LF
894 LF?
895 ....
896
897 Note that fast-import automatically switches packfiles when the current
898 packfile reaches --max-pack-size, or 4 GiB, whichever limit is
899 smaller. During an automatic packfile switch fast-import does not update
900 the branch refs, tags or marks.
901
902 As a `checkpoint` can require a significant amount of CPU time and
903 disk IO (to compute the overall pack SHA-1 checksum, generate the
904 corresponding index file, and update the refs) it can easily take
905 several minutes for a single `checkpoint` command to complete.
906
907 Frontends may choose to issue checkpoints during extremely large
908 and long running imports, or when they need to allow another Git
909 process access to a branch. However given that a 30 GiB Subversion
910 repository can be loaded into Git through fast-import in about 3 hours,
911 explicit checkpointing may not be necessary.
912
913 The `LF` after the command is optional (it used to be required).
914
915 `progress`
916 ~~~~~~~~~~
917 Causes fast-import to print the entire `progress` line unmodified to
918 its standard output channel (file descriptor 1) when the command is
919 processed from the input stream. The command otherwise has no impact
920 on the current import, or on any of fast-import's internal state.
921
922 ....
923 'progress' SP <any> LF
924 LF?
925 ....
926
927 The `<any>` part of the command may contain any sequence of bytes
928 that does not contain `LF`. The `LF` after the command is optional.
929 Callers may wish to process the output through a tool such as sed to
930 remove the leading part of the line, for example:
931
932 ====
933 frontend | git fast-import | sed 's/^progress //'
934 ====
935
936 Placing a `progress` command immediately after a `checkpoint` will
937 inform the reader when the `checkpoint` has been completed and it
938 can safely access the refs that fast-import updated.
939
940 `get-mark`
941 ~~~~~~~~~~
942 Causes fast-import to print the SHA-1 corresponding to a mark to
943 stdout or to the file descriptor previously arranged with the
944 `--cat-blob-fd` argument. The command otherwise has no impact on the
945 current import; its purpose is to retrieve SHA-1s that later commits
946 might want to refer to in their commit messages.
947
948 ....
949 'get-mark' SP ':' <idnum> LF
950 ....
951
952 This command can be used anywhere in the stream that comments are
953 accepted. In particular, the `get-mark` command can be used in the
954 middle of a commit but not in the middle of a `data` command.
955
956 See ``Responses To Commands'' below for details about how to read
957 this output safely.
958
959 `cat-blob`
960 ~~~~~~~~~~
961 Causes fast-import to print a blob to a file descriptor previously
962 arranged with the `--cat-blob-fd` argument. The command otherwise
963 has no impact on the current import; its main purpose is to
964 retrieve blobs that may be in fast-import's memory but not
965 accessible from the target repository.
966
967 ....
968 'cat-blob' SP <dataref> LF
969 ....
970
971 The `<dataref>` can be either a mark reference (`:<idnum>`)
972 set previously or a full 40-byte SHA-1 of a Git blob, preexisting or
973 ready to be written.
974
975 Output uses the same format as `git cat-file --batch`:
976
977 ====
978 <sha1> SP 'blob' SP <size> LF
979 <contents> LF
980 ====
981
982 This command can be used anywhere in the stream that comments are
983 accepted. In particular, the `cat-blob` command can be used in the
984 middle of a commit but not in the middle of a `data` command.
985
986 See ``Responses To Commands'' below for details about how to read
987 this output safely.
988
989 `ls`
990 ~~~~
991 Prints information about the object at a path to a file descriptor
992 previously arranged with the `--cat-blob-fd` argument. This allows
993 printing a blob from the active commit (with `cat-blob`) or copying a
994 blob or tree from a previous commit for use in the current one (with
995 `filemodify`).
996
997 The `ls` command can be used anywhere in the stream that comments are
998 accepted, including the middle of a commit.
999
1000 Reading from the active commit::
1001 This form can only be used in the middle of a `commit`.
1002 The path names a directory entry within fast-import's
1003 active commit. The path must be quoted in this case.
1004 +
1005 ....
1006 'ls' SP <path> LF
1007 ....
1008
1009 Reading from a named tree::
1010 The `<dataref>` can be a mark reference (`:<idnum>`) or the
1011 full 40-byte SHA-1 of a Git tag, commit, or tree object,
1012 preexisting or waiting to be written.
1013 The path is relative to the top level of the tree
1014 named by `<dataref>`.
1015 +
1016 ....
1017 'ls' SP <dataref> SP <path> LF
1018 ....
1019
1020 See `filemodify` above for a detailed description of `<path>`.
1021
1022 Output uses the same format as `git ls-tree <tree> -- <path>`:
1023
1024 ====
1025 <mode> SP ('blob' | 'tree' | 'commit') SP <dataref> HT <path> LF
1026 ====
1027
1028 The <dataref> represents the blob, tree, or commit object at <path>
1029 and can be used in later 'get-mark', 'cat-blob', 'filemodify', or
1030 'ls' commands.
1031
1032 If there is no file or subtree at that path, 'git fast-import' will
1033 instead report
1034
1035 ====
1036 missing SP <path> LF
1037 ====
1038
1039 See ``Responses To Commands'' below for details about how to read
1040 this output safely.
1041
1042 `feature`
1043 ~~~~~~~~~
1044 Require that fast-import supports the specified feature, or abort if
1045 it does not.
1046
1047 ....
1048 'feature' SP <feature> ('=' <argument>)? LF
1049 ....
1050
1051 The <feature> part of the command may be any one of the following:
1052
1053 date-format::
1054 export-marks::
1055 relative-marks::
1056 no-relative-marks::
1057 force::
1058 Act as though the corresponding command-line option with
1059 a leading `--` was passed on the command line
1060 (see OPTIONS, above).
1061
1062 import-marks::
1063 import-marks-if-exists::
1064 Like --import-marks except in two respects: first, only one
1065 "feature import-marks" or "feature import-marks-if-exists"
1066 command is allowed per stream; second, an --import-marks=
1067 or --import-marks-if-exists command-line option overrides
1068 any of these "feature" commands in the stream; third,
1069 "feature import-marks-if-exists" like a corresponding
1070 command-line option silently skips a nonexistent file.
1071
1072 get-mark::
1073 cat-blob::
1074 ls::
1075 Require that the backend support the 'get-mark', 'cat-blob',
1076 or 'ls' command respectively.
1077 Versions of fast-import not supporting the specified command
1078 will exit with a message indicating so.
1079 This lets the import error out early with a clear message,
1080 rather than wasting time on the early part of an import
1081 before the unsupported command is detected.
1082
1083 notes::
1084 Require that the backend support the 'notemodify' (N)
1085 subcommand to the 'commit' command.
1086 Versions of fast-import not supporting notes will exit
1087 with a message indicating so.
1088
1089 done::
1090 Error out if the stream ends without a 'done' command.
1091 Without this feature, errors causing the frontend to end
1092 abruptly at a convenient point in the stream can go
1093 undetected. This may occur, for example, if an import
1094 front end dies in mid-operation without emitting SIGTERM
1095 or SIGKILL at its subordinate git fast-import instance.
1096
1097 `option`
1098 ~~~~~~~~
1099 Processes the specified option so that git fast-import behaves in a
1100 way that suits the frontend's needs.
1101 Note that options specified by the frontend are overridden by any
1102 options the user may specify to git fast-import itself.
1103
1104 ....
1105 'option' SP <option> LF
1106 ....
1107
1108 The `<option>` part of the command may contain any of the options
1109 listed in the OPTIONS section that do not change import semantics,
1110 without the leading `--` and is treated in the same way.
1111
1112 Option commands must be the first commands on the input (not counting
1113 feature commands), to give an option command after any non-option
1114 command is an error.
1115
1116 The following command-line options change import semantics and may therefore
1117 not be passed as option:
1118
1119 * date-format
1120 * import-marks
1121 * export-marks
1122 * cat-blob-fd
1123 * force
1124
1125 `done`
1126 ~~~~~~
1127 If the `done` feature is not in use, treated as if EOF was read.
1128 This can be used to tell fast-import to finish early.
1129
1130 If the `--done` command-line option or `feature done` command is
1131 in use, the `done` command is mandatory and marks the end of the
1132 stream.
1133
1134 RESPONSES TO COMMANDS
1135 ---------------------
1136 New objects written by fast-import are not available immediately.
1137 Most fast-import commands have no visible effect until the next
1138 checkpoint (or completion). The frontend can send commands to
1139 fill fast-import's input pipe without worrying about how quickly
1140 they will take effect, which improves performance by simplifying
1141 scheduling.
1142
1143 For some frontends, though, it is useful to be able to read back
1144 data from the current repository as it is being updated (for
1145 example when the source material describes objects in terms of
1146 patches to be applied to previously imported objects). This can
1147 be accomplished by connecting the frontend and fast-import via
1148 bidirectional pipes:
1149
1150 ====
1151 mkfifo fast-import-output
1152 frontend <fast-import-output |
1153 git fast-import >fast-import-output
1154 ====
1155
1156 A frontend set up this way can use `progress`, `get-mark`, `ls`, and
1157 `cat-blob` commands to read information from the import in progress.
1158
1159 To avoid deadlock, such frontends must completely consume any
1160 pending output from `progress`, `ls`, `get-mark`, and `cat-blob` before
1161 performing writes to fast-import that might block.
1162
1163 CRASH REPORTS
1164 -------------
1165 If fast-import is supplied invalid input it will terminate with a
1166 non-zero exit status and create a crash report in the top level of
1167 the Git repository it was importing into. Crash reports contain
1168 a snapshot of the internal fast-import state as well as the most
1169 recent commands that lead up to the crash.
1170
1171 All recent commands (including stream comments, file changes and
1172 progress commands) are shown in the command history within the crash
1173 report, but raw file data and commit messages are excluded from the
1174 crash report. This exclusion saves space within the report file
1175 and reduces the amount of buffering that fast-import must perform
1176 during execution.
1177
1178 After writing a crash report fast-import will close the current
1179 packfile and export the marks table. This allows the frontend
1180 developer to inspect the repository state and resume the import from
1181 the point where it crashed. The modified branches and tags are not
1182 updated during a crash, as the import did not complete successfully.
1183 Branch and tag information can be found in the crash report and
1184 must be applied manually if the update is needed.
1185
1186 An example crash:
1187
1188 ====
1189 $ cat >in <<END_OF_INPUT
1190 # my very first test commit
1191 commit refs/heads/master
1192 committer Shawn O. Pearce <spearce> 19283 -0400
1193 # who is that guy anyway?
1194 data <<EOF
1195 this is my commit
1196 EOF
1197 M 644 inline .gitignore
1198 data <<EOF
1199 .gitignore
1200 EOF
1201 M 777 inline bob
1202 END_OF_INPUT
1203
1204 $ git fast-import <in
1205 fatal: Corrupt mode: M 777 inline bob
1206 fast-import: dumping crash report to .git/fast_import_crash_8434
1207
1208 $ cat .git/fast_import_crash_8434
1209 fast-import crash report:
1210 fast-import process: 8434
1211 parent process : 1391
1212 at Sat Sep 1 00:58:12 2007
1213
1214 fatal: Corrupt mode: M 777 inline bob
1215
1216 Most Recent Commands Before Crash
1217 ---------------------------------
1218 # my very first test commit
1219 commit refs/heads/master
1220 committer Shawn O. Pearce <spearce> 19283 -0400
1221 # who is that guy anyway?
1222 data <<EOF
1223 M 644 inline .gitignore
1224 data <<EOF
1225 * M 777 inline bob
1226
1227 Active Branch LRU
1228 -----------------
1229 active_branches = 1 cur, 5 max
1230
1231 pos clock name
1232 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1233 1) 0 refs/heads/master
1234
1235 Inactive Branches
1236 -----------------
1237 refs/heads/master:
1238 status : active loaded dirty
1239 tip commit : 0000000000000000000000000000000000000000
1240 old tree : 0000000000000000000000000000000000000000
1241 cur tree : 0000000000000000000000000000000000000000
1242 commit clock: 0
1243 last pack :
1244
1245
1246 -------------------
1247 END OF CRASH REPORT
1248 ====
1249
1250 TIPS AND TRICKS
1251 ---------------
1252 The following tips and tricks have been collected from various
1253 users of fast-import, and are offered here as suggestions.
1254
1255 Use One Mark Per Commit
1256 ~~~~~~~~~~~~~~~~~~~~~~~
1257 When doing a repository conversion, use a unique mark per commit
1258 (`mark :<n>`) and supply the --export-marks option on the command
1259 line. fast-import will dump a file which lists every mark and the Git
1260 object SHA-1 that corresponds to it. If the frontend can tie
1261 the marks back to the source repository, it is easy to verify the
1262 accuracy and completeness of the import by comparing each Git
1263 commit to the corresponding source revision.
1264
1265 Coming from a system such as Perforce or Subversion this should be
1266 quite simple, as the fast-import mark can also be the Perforce changeset
1267 number or the Subversion revision number.
1268
1269 Freely Skip Around Branches
1270 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1271 Don't bother trying to optimize the frontend to stick to one branch
1272 at a time during an import. Although doing so might be slightly
1273 faster for fast-import, it tends to increase the complexity of the frontend
1274 code considerably.
1275
1276 The branch LRU builtin to fast-import tends to behave very well, and the
1277 cost of activating an inactive branch is so low that bouncing around
1278 between branches has virtually no impact on import performance.
1279
1280 Handling Renames
1281 ~~~~~~~~~~~~~~~~
1282 When importing a renamed file or directory, simply delete the old
1283 name(s) and modify the new name(s) during the corresponding commit.
1284 Git performs rename detection after-the-fact, rather than explicitly
1285 during a commit.
1286
1287 Use Tag Fixup Branches
1288 ~~~~~~~~~~~~~~~~~~~~~~
1289 Some other SCM systems let the user create a tag from multiple
1290 files which are not from the same commit/changeset. Or to create
1291 tags which are a subset of the files available in the repository.
1292
1293 Importing these tags as-is in Git is impossible without making at
1294 least one commit which ``fixes up'' the files to match the content
1295 of the tag. Use fast-import's `reset` command to reset a dummy branch
1296 outside of your normal branch space to the base commit for the tag,
1297 then commit one or more file fixup commits, and finally tag the
1298 dummy branch.
1299
1300 For example since all normal branches are stored under `refs/heads/`
1301 name the tag fixup branch `TAG_FIXUP`. This way it is impossible for
1302 the fixup branch used by the importer to have namespace conflicts
1303 with real branches imported from the source (the name `TAG_FIXUP`
1304 is not `refs/heads/TAG_FIXUP`).
1305
1306 When committing fixups, consider using `merge` to connect the
1307 commit(s) which are supplying file revisions to the fixup branch.
1308 Doing so will allow tools such as 'git blame' to track
1309 through the real commit history and properly annotate the source
1310 files.
1311
1312 After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP`
1313 to remove the dummy branch.
1314
1315 Import Now, Repack Later
1316 ~~~~~~~~~~~~~~~~~~~~~~~~
1317 As soon as fast-import completes the Git repository is completely valid
1318 and ready for use. Typically this takes only a very short time,
1319 even for considerably large projects (100,000+ commits).
1320
1321 However repacking the repository is necessary to improve data
1322 locality and access performance. It can also take hours on extremely
1323 large projects (especially if -f and a large --window parameter is
1324 used). Since repacking is safe to run alongside readers and writers,
1325 run the repack in the background and let it finish when it finishes.
1326 There is no reason to wait to explore your new Git project!
1327
1328 If you choose to wait for the repack, don't try to run benchmarks
1329 or performance tests until repacking is completed. fast-import outputs
1330 suboptimal packfiles that are simply never seen in real use
1331 situations.
1332
1333 Repacking Historical Data
1334 ~~~~~~~~~~~~~~~~~~~~~~~~~
1335 If you are repacking very old imported data (e.g. older than the
1336 last year), consider expending some extra CPU time and supplying
1337 --window=50 (or higher) when you run 'git repack'.
1338 This will take longer, but will also produce a smaller packfile.
1339 You only need to expend the effort once, and everyone using your
1340 project will benefit from the smaller repository.
1341
1342 Include Some Progress Messages
1343 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1344 Every once in a while have your frontend emit a `progress` message
1345 to fast-import. The contents of the messages are entirely free-form,
1346 so one suggestion would be to output the current month and year
1347 each time the current commit date moves into the next month.
1348 Your users will feel better knowing how much of the data stream
1349 has been processed.
1350
1351
1352 PACKFILE OPTIMIZATION
1353 ---------------------
1354 When packing a blob fast-import always attempts to deltify against the last
1355 blob written. Unless specifically arranged for by the frontend,
1356 this will probably not be a prior version of the same file, so the
1357 generated delta will not be the smallest possible. The resulting
1358 packfile will be compressed, but will not be optimal.
1359
1360 Frontends which have efficient access to all revisions of a
1361 single file (for example reading an RCS/CVS ,v file) can choose
1362 to supply all revisions of that file as a sequence of consecutive
1363 `blob` commands. This allows fast-import to deltify the different file
1364 revisions against each other, saving space in the final packfile.
1365 Marks can be used to later identify individual file revisions during
1366 a sequence of `commit` commands.
1367
1368 The packfile(s) created by fast-import do not encourage good disk access
1369 patterns. This is caused by fast-import writing the data in the order
1370 it is received on standard input, while Git typically organizes
1371 data within packfiles to make the most recent (current tip) data
1372 appear before historical data. Git also clusters commits together,
1373 speeding up revision traversal through better cache locality.
1374
1375 For this reason it is strongly recommended that users repack the
1376 repository with `git repack -a -d` after fast-import completes, allowing
1377 Git to reorganize the packfiles for faster data access. If blob
1378 deltas are suboptimal (see above) then also adding the `-f` option
1379 to force recomputation of all deltas can significantly reduce the
1380 final packfile size (30-50% smaller can be quite typical).
1381
1382
1383 MEMORY UTILIZATION
1384 ------------------
1385 There are a number of factors which affect how much memory fast-import
1386 requires to perform an import. Like critical sections of core
1387 Git, fast-import uses its own memory allocators to amortize any overheads
1388 associated with malloc. In practice fast-import tends to amortize any
1389 malloc overheads to 0, due to its use of large block allocations.
1390
1391 per object
1392 ~~~~~~~~~~
1393 fast-import maintains an in-memory structure for every object written in
1394 this execution. On a 32 bit system the structure is 32 bytes,
1395 on a 64 bit system the structure is 40 bytes (due to the larger
1396 pointer sizes). Objects in the table are not deallocated until
1397 fast-import terminates. Importing 2 million objects on a 32 bit system
1398 will require approximately 64 MiB of memory.
1399
1400 The object table is actually a hashtable keyed on the object name
1401 (the unique SHA-1). This storage configuration allows fast-import to reuse
1402 an existing or already written object and avoid writing duplicates
1403 to the output packfile. Duplicate blobs are surprisingly common
1404 in an import, typically due to branch merges in the source.
1405
1406 per mark
1407 ~~~~~~~~
1408 Marks are stored in a sparse array, using 1 pointer (4 bytes or 8
1409 bytes, depending on pointer size) per mark. Although the array
1410 is sparse, frontends are still strongly encouraged to use marks
1411 between 1 and n, where n is the total number of marks required for
1412 this import.
1413
1414 per branch
1415 ~~~~~~~~~~
1416 Branches are classified as active and inactive. The memory usage
1417 of the two classes is significantly different.
1418
1419 Inactive branches are stored in a structure which uses 96 or 120
1420 bytes (32 bit or 64 bit systems, respectively), plus the length of
1421 the branch name (typically under 200 bytes), per branch. fast-import will
1422 easily handle as many as 10,000 inactive branches in under 2 MiB
1423 of memory.
1424
1425 Active branches have the same overhead as inactive branches, but
1426 also contain copies of every tree that has been recently modified on
1427 that branch. If subtree `include` has not been modified since the
1428 branch became active, its contents will not be loaded into memory,
1429 but if subtree `src` has been modified by a commit since the branch
1430 became active, then its contents will be loaded in memory.
1431
1432 As active branches store metadata about the files contained on that
1433 branch, their in-memory storage size can grow to a considerable size
1434 (see below).
1435
1436 fast-import automatically moves active branches to inactive status based on
1437 a simple least-recently-used algorithm. The LRU chain is updated on
1438 each `commit` command. The maximum number of active branches can be
1439 increased or decreased on the command line with --active-branches=.
1440
1441 per active tree
1442 ~~~~~~~~~~~~~~~
1443 Trees (aka directories) use just 12 bytes of memory on top of the
1444 memory required for their entries (see ``per active file'' below).
1445 The cost of a tree is virtually 0, as its overhead amortizes out
1446 over the individual file entries.
1447
1448 per active file entry
1449 ~~~~~~~~~~~~~~~~~~~~~
1450 Files (and pointers to subtrees) within active trees require 52 or 64
1451 bytes (32/64 bit platforms) per entry. To conserve space, file and
1452 tree names are pooled in a common string table, allowing the filename
1453 ``Makefile'' to use just 16 bytes (after including the string header
1454 overhead) no matter how many times it occurs within the project.
1455
1456 The active branch LRU, when coupled with the filename string pool
1457 and lazy loading of subtrees, allows fast-import to efficiently import
1458 projects with 2,000+ branches and 45,114+ files in a very limited
1459 memory footprint (less than 2.7 MiB per active branch).
1460
1461 SIGNALS
1462 -------
1463 Sending *SIGUSR1* to the 'git fast-import' process ends the current
1464 packfile early, simulating a `checkpoint` command. The impatient
1465 operator can use this facility to peek at the objects and refs from an
1466 import in progress, at the cost of some added running time and worse
1467 compression.
1468
1469 SEE ALSO
1470 --------
1471 linkgit:git-fast-export[1]
1472
1473 GIT
1474 ---
1475 Part of the linkgit:git[1] suite