Merge branch 'jc/mkstemp-more-careful-error-reporting'
[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 -- done::
43 Terminate with error if there is no 'done' command at the
44 end of the stream.
45
46 --force::
47 Force updating modified existing branches, even if doing
48 so would cause commits to be lost (as the new commit does
49 not contain the old commit).
50
51 --max-pack-size=<n>::
52 Maximum size of each output packfile.
53 The default is unlimited.
54
55 --big-file-threshold=<n>::
56 Maximum size of a blob that fast-import will attempt to
57 create a delta for, expressed in bytes. The default is 512m
58 (512 MiB). Some importers may wish to lower this on systems
59 with constrained memory.
60
61 --depth=<n>::
62 Maximum delta depth, for blob and tree deltification.
63 Default is 10.
64
65 --active-branches=<n>::
66 Maximum number of branches to maintain active at once.
67 See ``Memory Utilization'' below for details. Default is 5.
68
69 --export-marks=<file>::
70 Dumps the internal marks table to <file> when complete.
71 Marks are written one per line as `:markid SHA-1`.
72 Frontends can use this file to validate imports after they
73 have been completed, or to save the marks table across
74 incremental runs. As <file> is only opened and truncated
75 at checkpoint (or completion) the same path can also be
76 safely given to \--import-marks.
77
78 --import-marks=<file>::
79 Before processing any input, load the marks specified in
80 <file>. The input file must exist, must be readable, and
81 must use the same format as produced by \--export-marks.
82 Multiple options may be supplied to import more than one
83 set of marks. If a mark is defined to different values,
84 the last file wins.
85
86 --import-marks-if-exists=<file>::
87 Like --import-marks but instead of erroring out, silently
88 skips the file if it does not exist.
89
90 --relative-marks::
91 After specifying --relative-marks the paths specified
92 with --import-marks= and --export-marks= are relative
93 to an internal directory in the current repository.
94 In git-fast-import this means that the paths are relative
95 to the .git/info/fast-import directory. However, other
96 importers may use a different location.
97
98 --no-relative-marks::
99 Negates a previous --relative-marks. Allows for combining
100 relative and non-relative marks by interweaving
101 --(no-)-relative-marks with the --(import|export)-marks=
102 options.
103
104 --cat-blob-fd=<fd>::
105 Write responses to `cat-blob` and `ls` queries to the
106 file descriptor <fd> instead of `stdout`. Allows `progress`
107 output intended for the end-user to be separated from other
108 output.
109
110 --done::
111 Require a `done` command at the end of the stream.
112 This option might be useful for detecting errors that
113 cause the frontend to terminate before it has started to
114 write a stream.
115
116 --export-pack-edges=<file>::
117 After creating a packfile, print a line of data to
118 <file> listing the filename of the packfile and the last
119 commit on each branch that was written to that packfile.
120 This information may be useful after importing projects
121 whose total object set exceeds the 4 GiB packfile limit,
122 as these commits can be used as edge points during calls
123 to 'git pack-objects'.
124
125 --quiet::
126 Disable all non-fatal output, making fast-import silent when it
127 is successful. This option disables the output shown by
128 \--stats.
129
130 --stats::
131 Display some basic statistics about the objects fast-import has
132 created, the packfiles they were stored into, and the
133 memory used by fast-import during this run. Showing this output
134 is currently the default, but can be disabled with \--quiet.
135
136
137 Performance
138 -----------
139 The design of fast-import allows it to import large projects in a minimum
140 amount of memory usage and processing time. Assuming the frontend
141 is able to keep up with fast-import and feed it a constant stream of data,
142 import times for projects holding 10+ years of history and containing
143 100,000+ individual commits are generally completed in just 1-2
144 hours on quite modest (~$2,000 USD) hardware.
145
146 Most bottlenecks appear to be in foreign source data access (the
147 source just cannot extract revisions fast enough) or disk IO (fast-import
148 writes as fast as the disk will take the data). Imports will run
149 faster if the source data is stored on a different drive than the
150 destination Git repository (due to less IO contention).
151
152
153 Development Cost
154 ----------------
155 A typical frontend for fast-import tends to weigh in at approximately 200
156 lines of Perl/Python/Ruby code. Most developers have been able to
157 create working importers in just a couple of hours, even though it
158 is their first exposure to fast-import, and sometimes even to Git. This is
159 an ideal situation, given that most conversion tools are throw-away
160 (use once, and never look back).
161
162
163 Parallel Operation
164 ------------------
165 Like 'git push' or 'git fetch', imports handled by fast-import are safe to
166 run alongside parallel `git repack -a -d` or `git gc` invocations,
167 or any other Git operation (including 'git prune', as loose objects
168 are never used by fast-import).
169
170 fast-import does not lock the branch or tag refs it is actively importing.
171 After the import, during its ref update phase, fast-import tests each
172 existing branch ref to verify the update will be a fast-forward
173 update (the commit stored in the ref is contained in the new
174 history of the commit to be written). If the update is not a
175 fast-forward update, fast-import will skip updating that ref and instead
176 prints a warning message. fast-import will always attempt to update all
177 branch refs, and does not stop on the first failure.
178
179 Branch updates can be forced with \--force, but it's recommended that
180 this only be used on an otherwise quiet repository. Using \--force
181 is not necessary for an initial import into an empty repository.
182
183
184 Technical Discussion
185 --------------------
186 fast-import tracks a set of branches in memory. Any branch can be created
187 or modified at any point during the import process by sending a
188 `commit` command on the input stream. This design allows a frontend
189 program to process an unlimited number of branches simultaneously,
190 generating commits in the order they are available from the source
191 data. It also simplifies the frontend programs considerably.
192
193 fast-import does not use or alter the current working directory, or any
194 file within it. (It does however update the current Git repository,
195 as referenced by `GIT_DIR`.) Therefore an import frontend may use
196 the working directory for its own purposes, such as extracting file
197 revisions from the foreign source. This ignorance of the working
198 directory also allows fast-import to run very quickly, as it does not
199 need to perform any costly file update operations when switching
200 between branches.
201
202 Input Format
203 ------------
204 With the exception of raw file data (which Git does not interpret)
205 the fast-import input format is text (ASCII) based. This text based
206 format simplifies development and debugging of frontend programs,
207 especially when a higher level language such as Perl, Python or
208 Ruby is being used.
209
210 fast-import is very strict about its input. Where we say SP below we mean
211 *exactly* one space. Likewise LF means one (and only one) linefeed
212 and HT one (and only one) horizontal tab.
213 Supplying additional whitespace characters will cause unexpected
214 results, such as branch names or file names with leading or trailing
215 spaces in their name, or early termination of fast-import when it encounters
216 unexpected input.
217
218 Stream Comments
219 ~~~~~~~~~~~~~~~
220 To aid in debugging frontends fast-import ignores any line that
221 begins with `#` (ASCII pound/hash) up to and including the line
222 ending `LF`. A comment line may contain any sequence of bytes
223 that does not contain an LF and therefore may be used to include
224 any detailed debugging information that might be specific to the
225 frontend and useful when inspecting a fast-import data stream.
226
227 Date Formats
228 ~~~~~~~~~~~~
229 The following date formats are supported. A frontend should select
230 the format it will use for this import by passing the format name
231 in the \--date-format=<fmt> command line option.
232
233 `raw`::
234 This is the Git native format and is `<time> SP <offutc>`.
235 It is also fast-import's default format, if \--date-format was
236 not specified.
237 +
238 The time of the event is specified by `<time>` as the number of
239 seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is
240 written as an ASCII decimal integer.
241 +
242 The local offset is specified by `<offutc>` as a positive or negative
243 offset from UTC. For example EST (which is 5 hours behind UTC)
244 would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''.
245 The local offset does not affect `<time>`; it is used only as an
246 advisement to help formatting routines display the timestamp.
247 +
248 If the local offset is not available in the source material, use
249 ``+0000'', or the most common local offset. For example many
250 organizations have a CVS repository which has only ever been accessed
251 by users who are located in the same location and timezone. In this
252 case a reasonable offset from UTC could be assumed.
253 +
254 Unlike the `rfc2822` format, this format is very strict. Any
255 variation in formatting will cause fast-import to reject the value.
256
257 `rfc2822`::
258 This is the standard email format as described by RFC 2822.
259 +
260 An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git
261 parser is accurate, but a little on the lenient side. It is the
262 same parser used by 'git am' when applying patches
263 received from email.
264 +
265 Some malformed strings may be accepted as valid dates. In some of
266 these cases Git will still be able to obtain the correct date from
267 the malformed string. There are also some types of malformed
268 strings which Git will parse wrong, and yet consider valid.
269 Seriously malformed strings will be rejected.
270 +
271 Unlike the `raw` format above, the timezone/UTC offset information
272 contained in an RFC 2822 date string is used to adjust the date
273 value to UTC prior to storage. Therefore it is important that
274 this information be as accurate as possible.
275 +
276 If the source material uses RFC 2822 style dates,
277 the frontend should let fast-import handle the parsing and conversion
278 (rather than attempting to do it itself) as the Git parser has
279 been well tested in the wild.
280 +
281 Frontends should prefer the `raw` format if the source material
282 already uses UNIX-epoch format, can be coaxed to give dates in that
283 format, or its format is easily convertible to it, as there is no
284 ambiguity in parsing.
285
286 `now`::
287 Always use the current time and timezone. The literal
288 `now` must always be supplied for `<when>`.
289 +
290 This is a toy format. The current time and timezone of this system
291 is always copied into the identity string at the time it is being
292 created by fast-import. There is no way to specify a different time or
293 timezone.
294 +
295 This particular format is supplied as it's short to implement and
296 may be useful to a process that wants to create a new commit
297 right now, without needing to use a working directory or
298 'git update-index'.
299 +
300 If separate `author` and `committer` commands are used in a `commit`
301 the timestamps may not match, as the system clock will be polled
302 twice (once for each command). The only way to ensure that both
303 author and committer identity information has the same timestamp
304 is to omit `author` (thus copying from `committer`) or to use a
305 date format other than `now`.
306
307 Commands
308 ~~~~~~~~
309 fast-import accepts several commands to update the current repository
310 and control the current import process. More detailed discussion
311 (with examples) of each command follows later.
312
313 `commit`::
314 Creates a new branch or updates an existing branch by
315 creating a new commit and updating the branch to point at
316 the newly created commit.
317
318 `tag`::
319 Creates an annotated tag object from an existing commit or
320 branch. Lightweight tags are not supported by this command,
321 as they are not recommended for recording meaningful points
322 in time.
323
324 `reset`::
325 Reset an existing branch (or a new branch) to a specific
326 revision. This command must be used to change a branch to
327 a specific revision without making a commit on it.
328
329 `blob`::
330 Convert raw file data into a blob, for future use in a
331 `commit` command. This command is optional and is not
332 needed to perform an import.
333
334 `checkpoint`::
335 Forces fast-import to close the current packfile, generate its
336 unique SHA-1 checksum and index, and start a new packfile.
337 This command is optional and is not needed to perform
338 an import.
339
340 `progress`::
341 Causes fast-import to echo the entire line to its own
342 standard output. This command is optional and is not needed
343 to perform an import.
344
345 `done`::
346 Marks the end of the stream. This command is optional
347 unless the `done` feature was requested using the
348 `--done` command line option or `feature done` command.
349
350 `cat-blob`::
351 Causes fast-import to print a blob in 'cat-file --batch'
352 format to the file descriptor set with `--cat-blob-fd` or
353 `stdout` if unspecified.
354
355 `ls`::
356 Causes fast-import to print a line describing a directory
357 entry in 'ls-tree' format to the file descriptor set with
358 `--cat-blob-fd` or `stdout` if unspecified.
359
360 `feature`::
361 Require that fast-import supports the specified feature, or
362 abort if it does not.
363
364 `option`::
365 Specify any of the options listed under OPTIONS that do not
366 change stream semantic to suit the frontend's needs. This
367 command is optional and is not needed to perform an import.
368
369 `commit`
370 ~~~~~~~~
371 Create or update a branch with a new commit, recording one logical
372 change to the project.
373
374 ....
375 'commit' SP <ref> LF
376 mark?
377 ('author' (SP <name>)? SP LT <email> GT SP <when> LF)?
378 'committer' (SP <name>)? SP LT <email> GT SP <when> LF
379 data
380 ('from' SP <committish> LF)?
381 ('merge' SP <committish> LF)?
382 (filemodify | filedelete | filecopy | filerename | filedeleteall | notemodify)*
383 LF?
384 ....
385
386 where `<ref>` is the name of the branch to make the commit on.
387 Typically branch names are prefixed with `refs/heads/` in
388 Git, so importing the CVS branch symbol `RELENG-1_0` would use
389 `refs/heads/RELENG-1_0` for the value of `<ref>`. The value of
390 `<ref>` must be a valid refname in Git. As `LF` is not valid in
391 a Git refname, no quoting or escaping syntax is supported here.
392
393 A `mark` command may optionally appear, requesting fast-import to save a
394 reference to the newly created commit for future use by the frontend
395 (see below for format). It is very common for frontends to mark
396 every commit they create, thereby allowing future branch creation
397 from any imported commit.
398
399 The `data` command following `committer` must supply the commit
400 message (see below for `data` command syntax). To import an empty
401 commit message use a 0 length data. Commit messages are free-form
402 and are not interpreted by Git. Currently they must be encoded in
403 UTF-8, as fast-import does not permit other encodings to be specified.
404
405 Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename`,
406 `filedeleteall` and `notemodify` commands
407 may be included to update the contents of the branch prior to
408 creating the commit. These commands may be supplied in any order.
409 However it is recommended that a `filedeleteall` command precede
410 all `filemodify`, `filecopy`, `filerename` and `notemodify` commands in
411 the same commit, as `filedeleteall` wipes the branch clean (see below).
412
413 The `LF` after the command is optional (it used to be required).
414
415 `author`
416 ^^^^^^^^
417 An `author` command may optionally appear, if the author information
418 might differ from the committer information. If `author` is omitted
419 then fast-import will automatically use the committer's information for
420 the author portion of the commit. See below for a description of
421 the fields in `author`, as they are identical to `committer`.
422
423 `committer`
424 ^^^^^^^^^^^
425 The `committer` command indicates who made this commit, and when
426 they made it.
427
428 Here `<name>` is the person's display name (for example
429 ``Com M Itter'') and `<email>` is the person's email address
430 (``\cm@example.com''). `LT` and `GT` are the literal less-than (\x3c)
431 and greater-than (\x3e) symbols. These are required to delimit
432 the email address from the other fields in the line. Note that
433 `<name>` and `<email>` are free-form and may contain any sequence
434 of bytes, except `LT`, `GT` and `LF`. `<name>` is typically UTF-8 encoded.
435
436 The time of the change is specified by `<when>` using the date format
437 that was selected by the \--date-format=<fmt> command line option.
438 See ``Date Formats'' above for the set of supported formats, and
439 their syntax.
440
441 `from`
442 ^^^^^^
443 The `from` command is used to specify the commit to initialize
444 this branch from. This revision will be the first ancestor of the
445 new commit. The state of the tree built at this commit will begin
446 with the state at the `from` commit, and be altered by the content
447 modifications in this commit.
448
449 Omitting the `from` command in the first commit of a new branch
450 will cause fast-import to create that commit with no ancestor. This
451 tends to be desired only for the initial commit of a project.
452 If the frontend creates all files from scratch when making a new
453 branch, a `merge` command may be used instead of `from` to start
454 the commit with an empty tree.
455 Omitting the `from` command on existing branches is usually desired,
456 as the current commit on that branch is automatically assumed to
457 be the first ancestor of the new commit.
458
459 As `LF` is not valid in a Git refname or SHA-1 expression, no
460 quoting or escaping syntax is supported within `<committish>`.
461
462 Here `<committish>` is any of the following:
463
464 * The name of an existing branch already in fast-import's internal branch
465 table. If fast-import doesn't know the name, it's treated as a SHA-1
466 expression.
467
468 * A mark reference, `:<idnum>`, where `<idnum>` is the mark number.
469 +
470 The reason fast-import uses `:` to denote a mark reference is this character
471 is not legal in a Git branch name. The leading `:` makes it easy
472 to distinguish between the mark 42 (`:42`) and the branch 42 (`42`
473 or `refs/heads/42`), or an abbreviated SHA-1 which happened to
474 consist only of base-10 digits.
475 +
476 Marks must be declared (via `mark`) before they can be used.
477
478 * A complete 40 byte or abbreviated commit SHA-1 in hex.
479
480 * Any valid Git SHA-1 expression that resolves to a commit. See
481 ``SPECIFYING REVISIONS'' in linkgit:gitrevisions[7] for details.
482
483 The special case of restarting an incremental import from the
484 current branch value should be written as:
485 ----
486 from refs/heads/branch^0
487 ----
488 The `^0` suffix is necessary as fast-import does not permit a branch to
489 start from itself, and the branch is created in memory before the
490 `from` command is even read from the input. Adding `^0` will force
491 fast-import to resolve the commit through Git's revision parsing library,
492 rather than its internal branch table, thereby loading in the
493 existing value of the branch.
494
495 `merge`
496 ^^^^^^^
497 Includes one additional ancestor commit. The additional ancestry
498 link does not change the way the tree state is built at this commit.
499 If the `from` command is
500 omitted when creating a new branch, the first `merge` commit will be
501 the first ancestor of the current commit, and the branch will start
502 out with no files. An unlimited number of `merge` commands per
503 commit are permitted by fast-import, thereby establishing an n-way merge.
504 However Git's other tools never create commits with more than 15
505 additional ancestors (forming a 16-way merge). For this reason
506 it is suggested that frontends do not use more than 15 `merge`
507 commands per commit; 16, if starting a new, empty branch.
508
509 Here `<committish>` is any of the commit specification expressions
510 also accepted by `from` (see above).
511
512 `filemodify`
513 ^^^^^^^^^^^^
514 Included in a `commit` command to add a new file or change the
515 content of an existing file. This command has two different means
516 of specifying the content of the file.
517
518 External data format::
519 The data content for the file was already supplied by a prior
520 `blob` command. The frontend just needs to connect it.
521 +
522 ....
523 'M' SP <mode> SP <dataref> SP <path> LF
524 ....
525 +
526 Here usually `<dataref>` must be either a mark reference (`:<idnum>`)
527 set by a prior `blob` command, or a full 40-byte SHA-1 of an
528 existing Git blob object. If `<mode>` is `040000`` then
529 `<dataref>` must be the full 40-byte SHA-1 of an existing
530 Git tree object or a mark reference set with `--import-marks`.
531
532 Inline data format::
533 The data content for the file has not been supplied yet.
534 The frontend wants to supply it as part of this modify
535 command.
536 +
537 ....
538 'M' SP <mode> SP 'inline' SP <path> LF
539 data
540 ....
541 +
542 See below for a detailed description of the `data` command.
543
544 In both formats `<mode>` is the type of file entry, specified
545 in octal. Git only supports the following modes:
546
547 * `100644` or `644`: A normal (not-executable) file. The majority
548 of files in most projects use this mode. If in doubt, this is
549 what you want.
550 * `100755` or `755`: A normal, but executable, file.
551 * `120000`: A symlink, the content of the file will be the link target.
552 * `160000`: A gitlink, SHA-1 of the object refers to a commit in
553 another repository. Git links can only be specified by SHA or through
554 a commit mark. They are used to implement submodules.
555 * `040000`: A subdirectory. Subdirectories can only be specified by
556 SHA or through a tree mark set with `--import-marks`.
557
558 In both formats `<path>` is the complete path of the file to be added
559 (if not already existing) or modified (if already existing).
560
561 A `<path>` string must use UNIX-style directory separators (forward
562 slash `/`), may contain any byte other than `LF`, and must not
563 start with double quote (`"`).
564
565 A path can use C-style string quoting; this is accepted in all cases
566 and mandatory if the filename starts with double quote or contains
567 `LF`. In C-style quoting, the complete name should be surrounded with
568 double quotes, and any `LF`, backslash, or double quote characters
569 must be escaped by preceding them with a backslash (e.g.,
570 `"path/with\n, \\ and \" in it"`).
571
572 The value of `<path>` must be in canonical form. That is it must not:
573
574 * contain an empty directory component (e.g. `foo//bar` is invalid),
575 * end with a directory separator (e.g. `foo/` is invalid),
576 * start with a directory separator (e.g. `/foo` is invalid),
577 * contain the special component `.` or `..` (e.g. `foo/./bar` and
578 `foo/../bar` are invalid).
579
580 The root of the tree can be represented by an empty string as `<path>`.
581
582 It is recommended that `<path>` always be encoded using UTF-8.
583
584 `filedelete`
585 ^^^^^^^^^^^^
586 Included in a `commit` command to remove a file or recursively
587 delete an entire directory from the branch. If the file or directory
588 removal makes its parent directory empty, the parent directory will
589 be automatically removed too. This cascades up the tree until the
590 first non-empty directory or the root is reached.
591
592 ....
593 'D' SP <path> LF
594 ....
595
596 here `<path>` is the complete path of the file or subdirectory to
597 be removed from the branch.
598 See `filemodify` above for a detailed description of `<path>`.
599
600 `filecopy`
601 ^^^^^^^^^^^^
602 Recursively copies an existing file or subdirectory to a different
603 location within the branch. The existing file or directory must
604 exist. If the destination exists it will be completely replaced
605 by the content copied from the source.
606
607 ....
608 'C' SP <path> SP <path> LF
609 ....
610
611 here the first `<path>` is the source location and the second
612 `<path>` is the destination. See `filemodify` above for a detailed
613 description of what `<path>` may look like. To use a source path
614 that contains SP the path must be quoted.
615
616 A `filecopy` command takes effect immediately. Once the source
617 location has been copied to the destination any future commands
618 applied to the source location will not impact the destination of
619 the copy.
620
621 `filerename`
622 ^^^^^^^^^^^^
623 Renames an existing file or subdirectory to a different location
624 within the branch. The existing file or directory must exist. If
625 the destination exists it will be replaced by the source directory.
626
627 ....
628 'R' SP <path> SP <path> LF
629 ....
630
631 here the first `<path>` is the source location and the second
632 `<path>` is the destination. See `filemodify` above for a detailed
633 description of what `<path>` may look like. To use a source path
634 that contains SP the path must be quoted.
635
636 A `filerename` command takes effect immediately. Once the source
637 location has been renamed to the destination any future commands
638 applied to the source location will create new files there and not
639 impact the destination of the rename.
640
641 Note that a `filerename` is the same as a `filecopy` followed by a
642 `filedelete` of the source location. There is a slight performance
643 advantage to using `filerename`, but the advantage is so small
644 that it is never worth trying to convert a delete/add pair in
645 source material into a rename for fast-import. This `filerename`
646 command is provided just to simplify frontends that already have
647 rename information and don't want bother with decomposing it into a
648 `filecopy` followed by a `filedelete`.
649
650 `filedeleteall`
651 ^^^^^^^^^^^^^^^
652 Included in a `commit` command to remove all files (and also all
653 directories) from the branch. This command resets the internal
654 branch structure to have no files in it, allowing the frontend
655 to subsequently add all interesting files from scratch.
656
657 ....
658 'deleteall' LF
659 ....
660
661 This command is extremely useful if the frontend does not know
662 (or does not care to know) what files are currently on the branch,
663 and therefore cannot generate the proper `filedelete` commands to
664 update the content.
665
666 Issuing a `filedeleteall` followed by the needed `filemodify`
667 commands to set the correct content will produce the same results
668 as sending only the needed `filemodify` and `filedelete` commands.
669 The `filedeleteall` approach may however require fast-import to use slightly
670 more memory per active branch (less than 1 MiB for even most large
671 projects); so frontends that can easily obtain only the affected
672 paths for a commit are encouraged to do so.
673
674 `notemodify`
675 ^^^^^^^^^^^^
676 Included in a `commit` `<notes_ref>` command to add a new note
677 annotating a `<committish>` or change this annotation contents.
678 Internally it is similar to filemodify 100644 on `<committish>`
679 path (maybe split into subdirectories). It's not advised to
680 use any other commands to write to the `<notes_ref>` tree except
681 `filedeleteall` to delete all existing notes in this tree.
682 This command has two different means of specifying the content
683 of the note.
684
685 External data format::
686 The data content for the note was already supplied by a prior
687 `blob` command. The frontend just needs to connect it to the
688 commit that is to be annotated.
689 +
690 ....
691 'N' SP <dataref> SP <committish> LF
692 ....
693 +
694 Here `<dataref>` can be either a mark reference (`:<idnum>`)
695 set by a prior `blob` command, or a full 40-byte SHA-1 of an
696 existing Git blob object.
697
698 Inline data format::
699 The data content for the note has not been supplied yet.
700 The frontend wants to supply it as part of this modify
701 command.
702 +
703 ....
704 'N' SP 'inline' SP <committish> LF
705 data
706 ....
707 +
708 See below for a detailed description of the `data` command.
709
710 In both formats `<committish>` is any of the commit specification
711 expressions also accepted by `from` (see above).
712
713 `mark`
714 ~~~~~~
715 Arranges for fast-import to save a reference to the current object, allowing
716 the frontend to recall this object at a future point in time, without
717 knowing its SHA-1. Here the current object is the object creation
718 command the `mark` command appears within. This can be `commit`,
719 `tag`, and `blob`, but `commit` is the most common usage.
720
721 ....
722 'mark' SP ':' <idnum> LF
723 ....
724
725 where `<idnum>` is the number assigned by the frontend to this mark.
726 The value of `<idnum>` is expressed as an ASCII decimal integer.
727 The value 0 is reserved and cannot be used as
728 a mark. Only values greater than or equal to 1 may be used as marks.
729
730 New marks are created automatically. Existing marks can be moved
731 to another object simply by reusing the same `<idnum>` in another
732 `mark` command.
733
734 `tag`
735 ~~~~~
736 Creates an annotated tag referring to a specific commit. To create
737 lightweight (non-annotated) tags see the `reset` command below.
738
739 ....
740 'tag' SP <name> LF
741 'from' SP <committish> LF
742 'tagger' (SP <name>)? SP LT <email> GT SP <when> LF
743 data
744 ....
745
746 where `<name>` is the name of the tag to create.
747
748 Tag names are automatically prefixed with `refs/tags/` when stored
749 in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would
750 use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the
751 corresponding ref as `refs/tags/RELENG-1_0-FINAL`.
752
753 The value of `<name>` must be a valid refname in Git and therefore
754 may contain forward slashes. As `LF` is not valid in a Git refname,
755 no quoting or escaping syntax is supported here.
756
757 The `from` command is the same as in the `commit` command; see
758 above for details.
759
760 The `tagger` command uses the same format as `committer` within
761 `commit`; again see above for details.
762
763 The `data` command following `tagger` must supply the annotated tag
764 message (see below for `data` command syntax). To import an empty
765 tag message use a 0 length data. Tag messages are free-form and are
766 not interpreted by Git. Currently they must be encoded in UTF-8,
767 as fast-import does not permit other encodings to be specified.
768
769 Signing annotated tags during import from within fast-import is not
770 supported. Trying to include your own PGP/GPG signature is not
771 recommended, as the frontend does not (easily) have access to the
772 complete set of bytes which normally goes into such a signature.
773 If signing is required, create lightweight tags from within fast-import with
774 `reset`, then create the annotated versions of those tags offline
775 with the standard 'git tag' process.
776
777 `reset`
778 ~~~~~~~
779 Creates (or recreates) the named branch, optionally starting from
780 a specific revision. The reset command allows a frontend to issue
781 a new `from` command for an existing branch, or to create a new
782 branch from an existing commit without creating a new commit.
783
784 ....
785 'reset' SP <ref> LF
786 ('from' SP <committish> LF)?
787 LF?
788 ....
789
790 For a detailed description of `<ref>` and `<committish>` see above
791 under `commit` and `from`.
792
793 The `LF` after the command is optional (it used to be required).
794
795 The `reset` command can also be used to create lightweight
796 (non-annotated) tags. For example:
797
798 ====
799 reset refs/tags/938
800 from :938
801 ====
802
803 would create the lightweight tag `refs/tags/938` referring to
804 whatever commit mark `:938` references.
805
806 `blob`
807 ~~~~~~
808 Requests writing one file revision to the packfile. The revision
809 is not connected to any commit; this connection must be formed in
810 a subsequent `commit` command by referencing the blob through an
811 assigned mark.
812
813 ....
814 'blob' LF
815 mark?
816 data
817 ....
818
819 The mark command is optional here as some frontends have chosen
820 to generate the Git SHA-1 for the blob on their own, and feed that
821 directly to `commit`. This is typically more work than it's worth
822 however, as marks are inexpensive to store and easy to use.
823
824 `data`
825 ~~~~~~
826 Supplies raw data (for use as blob/file content, commit messages, or
827 annotated tag messages) to fast-import. Data can be supplied using an exact
828 byte count or delimited with a terminating line. Real frontends
829 intended for production-quality conversions should always use the
830 exact byte count format, as it is more robust and performs better.
831 The delimited format is intended primarily for testing fast-import.
832
833 Comment lines appearing within the `<raw>` part of `data` commands
834 are always taken to be part of the body of the data and are therefore
835 never ignored by fast-import. This makes it safe to import any
836 file/message content whose lines might start with `#`.
837
838 Exact byte count format::
839 The frontend must specify the number of bytes of data.
840 +
841 ....
842 'data' SP <count> LF
843 <raw> LF?
844 ....
845 +
846 where `<count>` is the exact number of bytes appearing within
847 `<raw>`. The value of `<count>` is expressed as an ASCII decimal
848 integer. The `LF` on either side of `<raw>` is not
849 included in `<count>` and will not be included in the imported data.
850 +
851 The `LF` after `<raw>` is optional (it used to be required) but
852 recommended. Always including it makes debugging a fast-import
853 stream easier as the next command always starts in column 0
854 of the next line, even if `<raw>` did not end with an `LF`.
855
856 Delimited format::
857 A delimiter string is used to mark the end of the data.
858 fast-import will compute the length by searching for the delimiter.
859 This format is primarily useful for testing and is not
860 recommended for real data.
861 +
862 ....
863 'data' SP '<<' <delim> LF
864 <raw> LF
865 <delim> LF
866 LF?
867 ....
868 +
869 where `<delim>` is the chosen delimiter string. The string `<delim>`
870 must not appear on a line by itself within `<raw>`, as otherwise
871 fast-import will think the data ends earlier than it really does. The `LF`
872 immediately trailing `<raw>` is part of `<raw>`. This is one of
873 the limitations of the delimited format, it is impossible to supply
874 a data chunk which does not have an LF as its last byte.
875 +
876 The `LF` after `<delim> LF` is optional (it used to be required).
877
878 `checkpoint`
879 ~~~~~~~~~~~~
880 Forces fast-import to close the current packfile, start a new one, and to
881 save out all current branch refs, tags and marks.
882
883 ....
884 'checkpoint' LF
885 LF?
886 ....
887
888 Note that fast-import automatically switches packfiles when the current
889 packfile reaches \--max-pack-size, or 4 GiB, whichever limit is
890 smaller. During an automatic packfile switch fast-import does not update
891 the branch refs, tags or marks.
892
893 As a `checkpoint` can require a significant amount of CPU time and
894 disk IO (to compute the overall pack SHA-1 checksum, generate the
895 corresponding index file, and update the refs) it can easily take
896 several minutes for a single `checkpoint` command to complete.
897
898 Frontends may choose to issue checkpoints during extremely large
899 and long running imports, or when they need to allow another Git
900 process access to a branch. However given that a 30 GiB Subversion
901 repository can be loaded into Git through fast-import in about 3 hours,
902 explicit checkpointing may not be necessary.
903
904 The `LF` after the command is optional (it used to be required).
905
906 `progress`
907 ~~~~~~~~~~
908 Causes fast-import to print the entire `progress` line unmodified to
909 its standard output channel (file descriptor 1) when the command is
910 processed from the input stream. The command otherwise has no impact
911 on the current import, or on any of fast-import's internal state.
912
913 ....
914 'progress' SP <any> LF
915 LF?
916 ....
917
918 The `<any>` part of the command may contain any sequence of bytes
919 that does not contain `LF`. The `LF` after the command is optional.
920 Callers may wish to process the output through a tool such as sed to
921 remove the leading part of the line, for example:
922
923 ====
924 frontend | git fast-import | sed 's/^progress //'
925 ====
926
927 Placing a `progress` command immediately after a `checkpoint` will
928 inform the reader when the `checkpoint` has been completed and it
929 can safely access the refs that fast-import updated.
930
931 `cat-blob`
932 ~~~~~~~~~~
933 Causes fast-import to print a blob to a file descriptor previously
934 arranged with the `--cat-blob-fd` argument. The command otherwise
935 has no impact on the current import; its main purpose is to
936 retrieve blobs that may be in fast-import's memory but not
937 accessible from the target repository.
938
939 ....
940 'cat-blob' SP <dataref> LF
941 ....
942
943 The `<dataref>` can be either a mark reference (`:<idnum>`)
944 set previously or a full 40-byte SHA-1 of a Git blob, preexisting or
945 ready to be written.
946
947 Output uses the same format as `git cat-file --batch`:
948
949 ====
950 <sha1> SP 'blob' SP <size> LF
951 <contents> LF
952 ====
953
954 This command can be used anywhere in the stream that comments are
955 accepted. In particular, the `cat-blob` command can be used in the
956 middle of a commit but not in the middle of a `data` command.
957
958 See ``Responses To Commands'' below for details about how to read
959 this output safely.
960
961 `ls`
962 ~~~~
963 Prints information about the object at a path to a file descriptor
964 previously arranged with the `--cat-blob-fd` argument. This allows
965 printing a blob from the active commit (with `cat-blob`) or copying a
966 blob or tree from a previous commit for use in the current one (with
967 `filemodify`).
968
969 The `ls` command can be used anywhere in the stream that comments are
970 accepted, including the middle of a commit.
971
972 Reading from the active commit::
973 This form can only be used in the middle of a `commit`.
974 The path names a directory entry within fast-import's
975 active commit. The path must be quoted in this case.
976 +
977 ....
978 'ls' SP <path> LF
979 ....
980
981 Reading from a named tree::
982 The `<dataref>` can be a mark reference (`:<idnum>`) or the
983 full 40-byte SHA-1 of a Git tag, commit, or tree object,
984 preexisting or waiting to be written.
985 The path is relative to the top level of the tree
986 named by `<dataref>`.
987 +
988 ....
989 'ls' SP <dataref> SP <path> LF
990 ....
991
992 See `filemodify` above for a detailed description of `<path>`.
993
994 Output uses the same format as `git ls-tree <tree> -- <path>`:
995
996 ====
997 <mode> SP ('blob' | 'tree' | 'commit') SP <dataref> HT <path> LF
998 ====
999
1000 The <dataref> represents the blob, tree, or commit object at <path>
1001 and can be used in later 'cat-blob', 'filemodify', or 'ls' commands.
1002
1003 If there is no file or subtree at that path, 'git fast-import' will
1004 instead report
1005
1006 ====
1007 missing SP <path> LF
1008 ====
1009
1010 See ``Responses To Commands'' below for details about how to read
1011 this output safely.
1012
1013 `feature`
1014 ~~~~~~~~~
1015 Require that fast-import supports the specified feature, or abort if
1016 it does not.
1017
1018 ....
1019 'feature' SP <feature> ('=' <argument>)? LF
1020 ....
1021
1022 The <feature> part of the command may be any one of the following:
1023
1024 date-format::
1025 export-marks::
1026 relative-marks::
1027 no-relative-marks::
1028 force::
1029 Act as though the corresponding command-line option with
1030 a leading '--' was passed on the command line
1031 (see OPTIONS, above).
1032
1033 import-marks::
1034 import-marks-if-exists::
1035 Like --import-marks except in two respects: first, only one
1036 "feature import-marks" or "feature import-marks-if-exists"
1037 command is allowed per stream; second, an --import-marks=
1038 or --import-marks-if-exists command-line option overrides
1039 any of these "feature" commands in the stream; third,
1040 "feature import-marks-if-exists" like a corresponding
1041 command-line option silently skips a nonexistent file.
1042
1043 cat-blob::
1044 ls::
1045 Require that the backend support the 'cat-blob' or 'ls' command.
1046 Versions of fast-import not supporting the specified command
1047 will exit with a message indicating so.
1048 This lets the import error out early with a clear message,
1049 rather than wasting time on the early part of an import
1050 before the unsupported command is detected.
1051
1052 notes::
1053 Require that the backend support the 'notemodify' (N)
1054 subcommand to the 'commit' command.
1055 Versions of fast-import not supporting notes will exit
1056 with a message indicating so.
1057
1058 done::
1059 Error out if the stream ends without a 'done' command.
1060 Without this feature, errors causing the frontend to end
1061 abruptly at a convenient point in the stream can go
1062 undetected. This may occur, for example, if an import
1063 front end dies in mid-operation without emitting SIGTERM
1064 or SIGKILL at its subordinate git fast-import instance.
1065
1066 `option`
1067 ~~~~~~~~
1068 Processes the specified option so that git fast-import behaves in a
1069 way that suits the frontend's needs.
1070 Note that options specified by the frontend are overridden by any
1071 options the user may specify to git fast-import itself.
1072
1073 ....
1074 'option' SP <option> LF
1075 ....
1076
1077 The `<option>` part of the command may contain any of the options
1078 listed in the OPTIONS section that do not change import semantics,
1079 without the leading '--' and is treated in the same way.
1080
1081 Option commands must be the first commands on the input (not counting
1082 feature commands), to give an option command after any non-option
1083 command is an error.
1084
1085 The following commandline options change import semantics and may therefore
1086 not be passed as option:
1087
1088 * date-format
1089 * import-marks
1090 * export-marks
1091 * cat-blob-fd
1092 * force
1093
1094 `done`
1095 ~~~~~~
1096 If the `done` feature is not in use, treated as if EOF was read.
1097 This can be used to tell fast-import to finish early.
1098
1099 If the `--done` command line option or `feature done` command is
1100 in use, the `done` command is mandatory and marks the end of the
1101 stream.
1102
1103 Responses To Commands
1104 ---------------------
1105 New objects written by fast-import are not available immediately.
1106 Most fast-import commands have no visible effect until the next
1107 checkpoint (or completion). The frontend can send commands to
1108 fill fast-import's input pipe without worrying about how quickly
1109 they will take effect, which improves performance by simplifying
1110 scheduling.
1111
1112 For some frontends, though, it is useful to be able to read back
1113 data from the current repository as it is being updated (for
1114 example when the source material describes objects in terms of
1115 patches to be applied to previously imported objects). This can
1116 be accomplished by connecting the frontend and fast-import via
1117 bidirectional pipes:
1118
1119 ====
1120 mkfifo fast-import-output
1121 frontend <fast-import-output |
1122 git fast-import >fast-import-output
1123 ====
1124
1125 A frontend set up this way can use `progress`, `ls`, and `cat-blob`
1126 commands to read information from the import in progress.
1127
1128 To avoid deadlock, such frontends must completely consume any
1129 pending output from `progress`, `ls`, and `cat-blob` before
1130 performing writes to fast-import that might block.
1131
1132 Crash Reports
1133 -------------
1134 If fast-import is supplied invalid input it will terminate with a
1135 non-zero exit status and create a crash report in the top level of
1136 the Git repository it was importing into. Crash reports contain
1137 a snapshot of the internal fast-import state as well as the most
1138 recent commands that lead up to the crash.
1139
1140 All recent commands (including stream comments, file changes and
1141 progress commands) are shown in the command history within the crash
1142 report, but raw file data and commit messages are excluded from the
1143 crash report. This exclusion saves space within the report file
1144 and reduces the amount of buffering that fast-import must perform
1145 during execution.
1146
1147 After writing a crash report fast-import will close the current
1148 packfile and export the marks table. This allows the frontend
1149 developer to inspect the repository state and resume the import from
1150 the point where it crashed. The modified branches and tags are not
1151 updated during a crash, as the import did not complete successfully.
1152 Branch and tag information can be found in the crash report and
1153 must be applied manually if the update is needed.
1154
1155 An example crash:
1156
1157 ====
1158 $ cat >in <<END_OF_INPUT
1159 # my very first test commit
1160 commit refs/heads/master
1161 committer Shawn O. Pearce <spearce> 19283 -0400
1162 # who is that guy anyway?
1163 data <<EOF
1164 this is my commit
1165 EOF
1166 M 644 inline .gitignore
1167 data <<EOF
1168 .gitignore
1169 EOF
1170 M 777 inline bob
1171 END_OF_INPUT
1172
1173 $ git fast-import <in
1174 fatal: Corrupt mode: M 777 inline bob
1175 fast-import: dumping crash report to .git/fast_import_crash_8434
1176
1177 $ cat .git/fast_import_crash_8434
1178 fast-import crash report:
1179 fast-import process: 8434
1180 parent process : 1391
1181 at Sat Sep 1 00:58:12 2007
1182
1183 fatal: Corrupt mode: M 777 inline bob
1184
1185 Most Recent Commands Before Crash
1186 ---------------------------------
1187 # my very first test commit
1188 commit refs/heads/master
1189 committer Shawn O. Pearce <spearce> 19283 -0400
1190 # who is that guy anyway?
1191 data <<EOF
1192 M 644 inline .gitignore
1193 data <<EOF
1194 * M 777 inline bob
1195
1196 Active Branch LRU
1197 -----------------
1198 active_branches = 1 cur, 5 max
1199
1200 pos clock name
1201 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1202 1) 0 refs/heads/master
1203
1204 Inactive Branches
1205 -----------------
1206 refs/heads/master:
1207 status : active loaded dirty
1208 tip commit : 0000000000000000000000000000000000000000
1209 old tree : 0000000000000000000000000000000000000000
1210 cur tree : 0000000000000000000000000000000000000000
1211 commit clock: 0
1212 last pack :
1213
1214
1215 -------------------
1216 END OF CRASH REPORT
1217 ====
1218
1219 Tips and Tricks
1220 ---------------
1221 The following tips and tricks have been collected from various
1222 users of fast-import, and are offered here as suggestions.
1223
1224 Use One Mark Per Commit
1225 ~~~~~~~~~~~~~~~~~~~~~~~
1226 When doing a repository conversion, use a unique mark per commit
1227 (`mark :<n>`) and supply the \--export-marks option on the command
1228 line. fast-import will dump a file which lists every mark and the Git
1229 object SHA-1 that corresponds to it. If the frontend can tie
1230 the marks back to the source repository, it is easy to verify the
1231 accuracy and completeness of the import by comparing each Git
1232 commit to the corresponding source revision.
1233
1234 Coming from a system such as Perforce or Subversion this should be
1235 quite simple, as the fast-import mark can also be the Perforce changeset
1236 number or the Subversion revision number.
1237
1238 Freely Skip Around Branches
1239 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
1240 Don't bother trying to optimize the frontend to stick to one branch
1241 at a time during an import. Although doing so might be slightly
1242 faster for fast-import, it tends to increase the complexity of the frontend
1243 code considerably.
1244
1245 The branch LRU builtin to fast-import tends to behave very well, and the
1246 cost of activating an inactive branch is so low that bouncing around
1247 between branches has virtually no impact on import performance.
1248
1249 Handling Renames
1250 ~~~~~~~~~~~~~~~~
1251 When importing a renamed file or directory, simply delete the old
1252 name(s) and modify the new name(s) during the corresponding commit.
1253 Git performs rename detection after-the-fact, rather than explicitly
1254 during a commit.
1255
1256 Use Tag Fixup Branches
1257 ~~~~~~~~~~~~~~~~~~~~~~
1258 Some other SCM systems let the user create a tag from multiple
1259 files which are not from the same commit/changeset. Or to create
1260 tags which are a subset of the files available in the repository.
1261
1262 Importing these tags as-is in Git is impossible without making at
1263 least one commit which ``fixes up'' the files to match the content
1264 of the tag. Use fast-import's `reset` command to reset a dummy branch
1265 outside of your normal branch space to the base commit for the tag,
1266 then commit one or more file fixup commits, and finally tag the
1267 dummy branch.
1268
1269 For example since all normal branches are stored under `refs/heads/`
1270 name the tag fixup branch `TAG_FIXUP`. This way it is impossible for
1271 the fixup branch used by the importer to have namespace conflicts
1272 with real branches imported from the source (the name `TAG_FIXUP`
1273 is not `refs/heads/TAG_FIXUP`).
1274
1275 When committing fixups, consider using `merge` to connect the
1276 commit(s) which are supplying file revisions to the fixup branch.
1277 Doing so will allow tools such as 'git blame' to track
1278 through the real commit history and properly annotate the source
1279 files.
1280
1281 After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP`
1282 to remove the dummy branch.
1283
1284 Import Now, Repack Later
1285 ~~~~~~~~~~~~~~~~~~~~~~~~
1286 As soon as fast-import completes the Git repository is completely valid
1287 and ready for use. Typically this takes only a very short time,
1288 even for considerably large projects (100,000+ commits).
1289
1290 However repacking the repository is necessary to improve data
1291 locality and access performance. It can also take hours on extremely
1292 large projects (especially if -f and a large \--window parameter is
1293 used). Since repacking is safe to run alongside readers and writers,
1294 run the repack in the background and let it finish when it finishes.
1295 There is no reason to wait to explore your new Git project!
1296
1297 If you choose to wait for the repack, don't try to run benchmarks
1298 or performance tests until repacking is completed. fast-import outputs
1299 suboptimal packfiles that are simply never seen in real use
1300 situations.
1301
1302 Repacking Historical Data
1303 ~~~~~~~~~~~~~~~~~~~~~~~~~
1304 If you are repacking very old imported data (e.g. older than the
1305 last year), consider expending some extra CPU time and supplying
1306 \--window=50 (or higher) when you run 'git repack'.
1307 This will take longer, but will also produce a smaller packfile.
1308 You only need to expend the effort once, and everyone using your
1309 project will benefit from the smaller repository.
1310
1311 Include Some Progress Messages
1312 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1313 Every once in a while have your frontend emit a `progress` message
1314 to fast-import. The contents of the messages are entirely free-form,
1315 so one suggestion would be to output the current month and year
1316 each time the current commit date moves into the next month.
1317 Your users will feel better knowing how much of the data stream
1318 has been processed.
1319
1320
1321 Packfile Optimization
1322 ---------------------
1323 When packing a blob fast-import always attempts to deltify against the last
1324 blob written. Unless specifically arranged for by the frontend,
1325 this will probably not be a prior version of the same file, so the
1326 generated delta will not be the smallest possible. The resulting
1327 packfile will be compressed, but will not be optimal.
1328
1329 Frontends which have efficient access to all revisions of a
1330 single file (for example reading an RCS/CVS ,v file) can choose
1331 to supply all revisions of that file as a sequence of consecutive
1332 `blob` commands. This allows fast-import to deltify the different file
1333 revisions against each other, saving space in the final packfile.
1334 Marks can be used to later identify individual file revisions during
1335 a sequence of `commit` commands.
1336
1337 The packfile(s) created by fast-import do not encourage good disk access
1338 patterns. This is caused by fast-import writing the data in the order
1339 it is received on standard input, while Git typically organizes
1340 data within packfiles to make the most recent (current tip) data
1341 appear before historical data. Git also clusters commits together,
1342 speeding up revision traversal through better cache locality.
1343
1344 For this reason it is strongly recommended that users repack the
1345 repository with `git repack -a -d` after fast-import completes, allowing
1346 Git to reorganize the packfiles for faster data access. If blob
1347 deltas are suboptimal (see above) then also adding the `-f` option
1348 to force recomputation of all deltas can significantly reduce the
1349 final packfile size (30-50% smaller can be quite typical).
1350
1351
1352 Memory Utilization
1353 ------------------
1354 There are a number of factors which affect how much memory fast-import
1355 requires to perform an import. Like critical sections of core
1356 Git, fast-import uses its own memory allocators to amortize any overheads
1357 associated with malloc. In practice fast-import tends to amortize any
1358 malloc overheads to 0, due to its use of large block allocations.
1359
1360 per object
1361 ~~~~~~~~~~
1362 fast-import maintains an in-memory structure for every object written in
1363 this execution. On a 32 bit system the structure is 32 bytes,
1364 on a 64 bit system the structure is 40 bytes (due to the larger
1365 pointer sizes). Objects in the table are not deallocated until
1366 fast-import terminates. Importing 2 million objects on a 32 bit system
1367 will require approximately 64 MiB of memory.
1368
1369 The object table is actually a hashtable keyed on the object name
1370 (the unique SHA-1). This storage configuration allows fast-import to reuse
1371 an existing or already written object and avoid writing duplicates
1372 to the output packfile. Duplicate blobs are surprisingly common
1373 in an import, typically due to branch merges in the source.
1374
1375 per mark
1376 ~~~~~~~~
1377 Marks are stored in a sparse array, using 1 pointer (4 bytes or 8
1378 bytes, depending on pointer size) per mark. Although the array
1379 is sparse, frontends are still strongly encouraged to use marks
1380 between 1 and n, where n is the total number of marks required for
1381 this import.
1382
1383 per branch
1384 ~~~~~~~~~~
1385 Branches are classified as active and inactive. The memory usage
1386 of the two classes is significantly different.
1387
1388 Inactive branches are stored in a structure which uses 96 or 120
1389 bytes (32 bit or 64 bit systems, respectively), plus the length of
1390 the branch name (typically under 200 bytes), per branch. fast-import will
1391 easily handle as many as 10,000 inactive branches in under 2 MiB
1392 of memory.
1393
1394 Active branches have the same overhead as inactive branches, but
1395 also contain copies of every tree that has been recently modified on
1396 that branch. If subtree `include` has not been modified since the
1397 branch became active, its contents will not be loaded into memory,
1398 but if subtree `src` has been modified by a commit since the branch
1399 became active, then its contents will be loaded in memory.
1400
1401 As active branches store metadata about the files contained on that
1402 branch, their in-memory storage size can grow to a considerable size
1403 (see below).
1404
1405 fast-import automatically moves active branches to inactive status based on
1406 a simple least-recently-used algorithm. The LRU chain is updated on
1407 each `commit` command. The maximum number of active branches can be
1408 increased or decreased on the command line with \--active-branches=.
1409
1410 per active tree
1411 ~~~~~~~~~~~~~~~
1412 Trees (aka directories) use just 12 bytes of memory on top of the
1413 memory required for their entries (see ``per active file'' below).
1414 The cost of a tree is virtually 0, as its overhead amortizes out
1415 over the individual file entries.
1416
1417 per active file entry
1418 ~~~~~~~~~~~~~~~~~~~~~
1419 Files (and pointers to subtrees) within active trees require 52 or 64
1420 bytes (32/64 bit platforms) per entry. To conserve space, file and
1421 tree names are pooled in a common string table, allowing the filename
1422 ``Makefile'' to use just 16 bytes (after including the string header
1423 overhead) no matter how many times it occurs within the project.
1424
1425 The active branch LRU, when coupled with the filename string pool
1426 and lazy loading of subtrees, allows fast-import to efficiently import
1427 projects with 2,000+ branches and 45,114+ files in a very limited
1428 memory footprint (less than 2.7 MiB per active branch).
1429
1430 Signals
1431 -------
1432 Sending *SIGUSR1* to the 'git fast-import' process ends the current
1433 packfile early, simulating a `checkpoint` command. The impatient
1434 operator can use this facility to peek at the objects and refs from an
1435 import in progress, at the cost of some added running time and worse
1436 compression.
1437
1438 GIT
1439 ---
1440 Part of the linkgit:git[1] suite