1git-fast-import(1) 2================== 3 4NAME 5---- 6git-fast-import - Backend for fast Git data importers 7 8 9SYNOPSIS 10-------- 11[verse] 12frontend | 'git fast-import' [options] 13 14DESCRIPTION 15----------- 16This program is usually not what the end user wants to run directly. 17Most end users want to use one of the existing frontend programs, 18which parses a specific type of foreign source and feeds the contents 19stored there to 'git fast-import'. 20 21fast-import reads a mixed command/data stream from standard input and 22writes one or more packfiles directly into the current repository. 23When EOF is received on standard input, fast import writes out 24updated branch and tag refs, fully updating the current repository 25with the newly imported data. 26 27The fast-import backend itself can import into an empty repository (one that 28has already been initialized by 'git init') or incrementally 29update an existing populated repository. Whether or not incremental 30imports are supported from a particular foreign source depends on 31the frontend program in use. 32 33 34OPTIONS 35------- 36--date-format=<fmt>:: 37 Specify the type of dates the frontend will supply to 38 fast-import within `author`, `committer` and `tagger` commands. 39 See ``Date Formats'' below for details about which formats 40 are supported, and their syntax. 41 42--force:: 43 Force updating modified existing branches, even if doing 44 so would cause commits to be lost (as the new commit does 45 not contain the old commit). 46 47--max-pack-size=<n>:: 48 Maximum size of each output packfile. 49 The default is unlimited. 50 51--big-file-threshold=<n>:: 52 Maximum size of a blob that fast-import will attempt to 53 create a delta for, expressed in bytes. The default is 512m 54 (512 MiB). Some importers may wish to lower this on systems 55 with constrained memory. 56 57--depth=<n>:: 58 Maximum delta depth, for blob and tree deltification. 59 Default is 10. 60 61--active-branches=<n>:: 62 Maximum number of branches to maintain active at once. 63 See ``Memory Utilization'' below for details. Default is 5. 64 65--export-marks=<file>:: 66 Dumps the internal marks table to <file> when complete. 67 Marks are written one per line as `:markid SHA-1`. 68 Frontends can use this file to validate imports after they 69 have been completed, or to save the marks table across 70 incremental runs. As <file> is only opened and truncated 71 at checkpoint (or completion) the same path can also be 72 safely given to \--import-marks. 73 74--import-marks=<file>:: 75 Before processing any input, load the marks specified in 76 <file>. The input file must exist, must be readable, and 77 must use the same format as produced by \--export-marks. 78 Multiple options may be supplied to import more than one 79 set of marks. If a mark is defined to different values, 80 the last file wins. 81 82--import-marks-if-exists=<file>:: 83 Like --import-marks but instead of erroring out, silently 84 skips the file if it does not exist. 85 86--relative-marks:: 87 After specifying --relative-marks the paths specified 88 with --import-marks= and --export-marks= are relative 89 to an internal directory in the current repository. 90 In git-fast-import this means that the paths are relative 91 to the .git/info/fast-import directory. However, other 92 importers may use a different location. 93 94--no-relative-marks:: 95 Negates a previous --relative-marks. Allows for combining 96 relative and non-relative marks by interweaving 97 --(no-)-relative-marks with the --(import|export)-marks= 98 options. 99 100--cat-blob-fd=<fd>:: 101 Specify the file descriptor that will be written to 102 when the `cat-blob` command is encountered in the stream. 103 The default behaviour is to write to `stdout`. 104 105--done:: 106 Require a `done` command at the end of the stream. 107 This option might be useful for detecting errors that 108 cause the frontend to terminate before it has started to 109 write a stream. 110 111--export-pack-edges=<file>:: 112 After creating a packfile, print a line of data to 113 <file> listing the filename of the packfile and the last 114 commit on each branch that was written to that packfile. 115 This information may be useful after importing projects 116 whose total object set exceeds the 4 GiB packfile limit, 117 as these commits can be used as edge points during calls 118 to 'git pack-objects'. 119 120--quiet:: 121 Disable all non-fatal output, making fast-import silent when it 122 is successful. This option disables the output shown by 123 \--stats. 124 125--stats:: 126 Display some basic statistics about the objects fast-import has 127 created, the packfiles they were stored into, and the 128 memory used by fast-import during this run. Showing this output 129 is currently the default, but can be disabled with \--quiet. 130 131 132Performance 133----------- 134The design of fast-import allows it to import large projects in a minimum 135amount of memory usage and processing time. Assuming the frontend 136is able to keep up with fast-import and feed it a constant stream of data, 137import times for projects holding 10+ years of history and containing 138100,000+ individual commits are generally completed in just 1-2 139hours on quite modest (~$2,000 USD) hardware. 140 141Most bottlenecks appear to be in foreign source data access (the 142source just cannot extract revisions fast enough) or disk IO (fast-import 143writes as fast as the disk will take the data). Imports will run 144faster if the source data is stored on a different drive than the 145destination Git repository (due to less IO contention). 146 147 148Development Cost 149---------------- 150A typical frontend for fast-import tends to weigh in at approximately 200 151lines of Perl/Python/Ruby code. Most developers have been able to 152create working importers in just a couple of hours, even though it 153is their first exposure to fast-import, and sometimes even to Git. This is 154an ideal situation, given that most conversion tools are throw-away 155(use once, and never look back). 156 157 158Parallel Operation 159------------------ 160Like 'git push' or 'git fetch', imports handled by fast-import are safe to 161run alongside parallel `git repack -a -d` or `git gc` invocations, 162or any other Git operation (including 'git prune', as loose objects 163are never used by fast-import). 164 165fast-import does not lock the branch or tag refs it is actively importing. 166After the import, during its ref update phase, fast-import tests each 167existing branch ref to verify the update will be a fast-forward 168update (the commit stored in the ref is contained in the new 169history of the commit to be written). If the update is not a 170fast-forward update, fast-import will skip updating that ref and instead 171prints a warning message. fast-import will always attempt to update all 172branch refs, and does not stop on the first failure. 173 174Branch updates can be forced with \--force, but it's recommended that 175this only be used on an otherwise quiet repository. Using \--force 176is not necessary for an initial import into an empty repository. 177 178 179Technical Discussion 180-------------------- 181fast-import tracks a set of branches in memory. Any branch can be created 182or modified at any point during the import process by sending a 183`commit` command on the input stream. This design allows a frontend 184program to process an unlimited number of branches simultaneously, 185generating commits in the order they are available from the source 186data. It also simplifies the frontend programs considerably. 187 188fast-import does not use or alter the current working directory, or any 189file within it. (It does however update the current Git repository, 190as referenced by `GIT_DIR`.) Therefore an import frontend may use 191the working directory for its own purposes, such as extracting file 192revisions from the foreign source. This ignorance of the working 193directory also allows fast-import to run very quickly, as it does not 194need to perform any costly file update operations when switching 195between branches. 196 197Input Format 198------------ 199With the exception of raw file data (which Git does not interpret) 200the fast-import input format is text (ASCII) based. This text based 201format simplifies development and debugging of frontend programs, 202especially when a higher level language such as Perl, Python or 203Ruby is being used. 204 205fast-import is very strict about its input. Where we say SP below we mean 206*exactly* one space. Likewise LF means one (and only one) linefeed 207and HT one (and only one) horizontal tab. 208Supplying additional whitespace characters will cause unexpected 209results, such as branch names or file names with leading or trailing 210spaces in their name, or early termination of fast-import when it encounters 211unexpected input. 212 213Stream Comments 214~~~~~~~~~~~~~~~ 215To aid in debugging frontends fast-import ignores any line that 216begins with `#` (ASCII pound/hash) up to and including the line 217ending `LF`. A comment line may contain any sequence of bytes 218that does not contain an LF and therefore may be used to include 219any detailed debugging information that might be specific to the 220frontend and useful when inspecting a fast-import data stream. 221 222Date Formats 223~~~~~~~~~~~~ 224The following date formats are supported. A frontend should select 225the format it will use for this import by passing the format name 226in the \--date-format=<fmt> command line option. 227 228`raw`:: 229 This is the Git native format and is `<time> SP <offutc>`. 230 It is also fast-import's default format, if \--date-format was 231 not specified. 232+ 233The time of the event is specified by `<time>` as the number of 234seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is 235written as an ASCII decimal integer. 236+ 237The local offset is specified by `<offutc>` as a positive or negative 238offset from UTC. For example EST (which is 5 hours behind UTC) 239would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''. 240The local offset does not affect `<time>`; it is used only as an 241advisement to help formatting routines display the timestamp. 242+ 243If the local offset is not available in the source material, use 244``+0000'', or the most common local offset. For example many 245organizations have a CVS repository which has only ever been accessed 246by users who are located in the same location and timezone. In this 247case a reasonable offset from UTC could be assumed. 248+ 249Unlike the `rfc2822` format, this format is very strict. Any 250variation in formatting will cause fast-import to reject the value. 251 252`rfc2822`:: 253 This is the standard email format as described by RFC 2822. 254+ 255An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git 256parser is accurate, but a little on the lenient side. It is the 257same parser used by 'git am' when applying patches 258received from email. 259+ 260Some malformed strings may be accepted as valid dates. In some of 261these cases Git will still be able to obtain the correct date from 262the malformed string. There are also some types of malformed 263strings which Git will parse wrong, and yet consider valid. 264Seriously malformed strings will be rejected. 265+ 266Unlike the `raw` format above, the timezone/UTC offset information 267contained in an RFC 2822 date string is used to adjust the date 268value to UTC prior to storage. Therefore it is important that 269this information be as accurate as possible. 270+ 271If the source material uses RFC 2822 style dates, 272the frontend should let fast-import handle the parsing and conversion 273(rather than attempting to do it itself) as the Git parser has 274been well tested in the wild. 275+ 276Frontends should prefer the `raw` format if the source material 277already uses UNIX-epoch format, can be coaxed to give dates in that 278format, or its format is easily convertible to it, as there is no 279ambiguity in parsing. 280 281`now`:: 282 Always use the current time and timezone. The literal 283 `now` must always be supplied for `<when>`. 284+ 285This is a toy format. The current time and timezone of this system 286is always copied into the identity string at the time it is being 287created by fast-import. There is no way to specify a different time or 288timezone. 289+ 290This particular format is supplied as it's short to implement and 291may be useful to a process that wants to create a new commit 292right now, without needing to use a working directory or 293'git update-index'. 294+ 295If separate `author` and `committer` commands are used in a `commit` 296the timestamps may not match, as the system clock will be polled 297twice (once for each command). The only way to ensure that both 298author and committer identity information has the same timestamp 299is to omit `author` (thus copying from `committer`) or to use a 300date format other than `now`. 301 302Commands 303~~~~~~~~ 304fast-import accepts several commands to update the current repository 305and control the current import process. More detailed discussion 306(with examples) of each command follows later. 307 308`commit`:: 309 Creates a new branch or updates an existing branch by 310 creating a new commit and updating the branch to point at 311 the newly created commit. 312 313`tag`:: 314 Creates an annotated tag object from an existing commit or 315 branch. Lightweight tags are not supported by this command, 316 as they are not recommended for recording meaningful points 317 in time. 318 319`reset`:: 320 Reset an existing branch (or a new branch) to a specific 321 revision. This command must be used to change a branch to 322 a specific revision without making a commit on it. 323 324`blob`:: 325 Convert raw file data into a blob, for future use in a 326 `commit` command. This command is optional and is not 327 needed to perform an import. 328 329`checkpoint`:: 330 Forces fast-import to close the current packfile, generate its 331 unique SHA-1 checksum and index, and start a new packfile. 332 This command is optional and is not needed to perform 333 an import. 334 335`progress`:: 336 Causes fast-import to echo the entire line to its own 337 standard output. This command is optional and is not needed 338 to perform an import. 339 340`done`:: 341 Marks the end of the stream. This command is optional 342 unless the `done` feature was requested using the 343 `--done` command line option or `feature done` command. 344 345`cat-blob`:: 346 Causes fast-import to print a blob in 'cat-file --batch' 347 format to the file descriptor set with `--cat-blob-fd` or 348 `stdout` if unspecified. 349 350`ls`:: 351 Causes fast-import to print a line describing a directory 352 entry in 'ls-tree' format to the file descriptor set with 353 `--cat-blob-fd` or `stdout` if unspecified. 354 355`feature`:: 356 Require that fast-import supports the specified feature, or 357 abort if it does not. 358 359`option`:: 360 Specify any of the options listed under OPTIONS that do not 361 change stream semantic to suit the frontend's needs. This 362 command is optional and is not needed to perform an import. 363 364`commit` 365~~~~~~~~ 366Create or update a branch with a new commit, recording one logical 367change to the project. 368 369.... 370 'commit' SP <ref> LF 371 mark? 372 ('author' (SP <name>)? SP LT <email> GT SP <when> LF)? 373 'committer' (SP <name>)? SP LT <email> GT SP <when> LF 374 data 375 ('from' SP <committish> LF)? 376 ('merge' SP <committish> LF)? 377 (filemodify | filedelete | filecopy | filerename | filedeleteall | notemodify)* 378 LF? 379.... 380 381where `<ref>` is the name of the branch to make the commit on. 382Typically branch names are prefixed with `refs/heads/` in 383Git, so importing the CVS branch symbol `RELENG-1_0` would use 384`refs/heads/RELENG-1_0` for the value of `<ref>`. The value of 385`<ref>` must be a valid refname in Git. As `LF` is not valid in 386a Git refname, no quoting or escaping syntax is supported here. 387 388A `mark` command may optionally appear, requesting fast-import to save a 389reference to the newly created commit for future use by the frontend 390(see below for format). It is very common for frontends to mark 391every commit they create, thereby allowing future branch creation 392from any imported commit. 393 394The `data` command following `committer` must supply the commit 395message (see below for `data` command syntax). To import an empty 396commit message use a 0 length data. Commit messages are free-form 397and are not interpreted by Git. Currently they must be encoded in 398UTF-8, as fast-import does not permit other encodings to be specified. 399 400Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename`, 401`filedeleteall` and `notemodify` commands 402may be included to update the contents of the branch prior to 403creating the commit. These commands may be supplied in any order. 404However it is recommended that a `filedeleteall` command precede 405all `filemodify`, `filecopy`, `filerename` and `notemodify` commands in 406the same commit, as `filedeleteall` wipes the branch clean (see below). 407 408The `LF` after the command is optional (it used to be required). 409 410`author` 411^^^^^^^^ 412An `author` command may optionally appear, if the author information 413might differ from the committer information. If `author` is omitted 414then fast-import will automatically use the committer's information for 415the author portion of the commit. See below for a description of 416the fields in `author`, as they are identical to `committer`. 417 418`committer` 419^^^^^^^^^^^ 420The `committer` command indicates who made this commit, and when 421they made it. 422 423Here `<name>` is the person's display name (for example 424``Com M Itter'') and `<email>` is the person's email address 425(``cm@example.com''). `LT` and `GT` are the literal less-than (\x3c) 426and greater-than (\x3e) symbols. These are required to delimit 427the email address from the other fields in the line. Note that 428`<name>` and `<email>` are free-form and may contain any sequence 429of bytes, except `LT`, `GT` and `LF`. `<name>` is typically UTF-8 encoded. 430 431The time of the change is specified by `<when>` using the date format 432that was selected by the \--date-format=<fmt> command line option. 433See ``Date Formats'' above for the set of supported formats, and 434their syntax. 435 436`from` 437^^^^^^ 438The `from` command is used to specify the commit to initialize 439this branch from. This revision will be the first ancestor of the 440new commit. 441 442Omitting the `from` command in the first commit of a new branch 443will cause fast-import to create that commit with no ancestor. This 444tends to be desired only for the initial commit of a project. 445If the frontend creates all files from scratch when making a new 446branch, a `merge` command may be used instead of `from` to start 447the commit with an empty tree. 448Omitting the `from` command on existing branches is usually desired, 449as the current commit on that branch is automatically assumed to 450be the first ancestor of the new commit. 451 452As `LF` is not valid in a Git refname or SHA-1 expression, no 453quoting or escaping syntax is supported within `<committish>`. 454 455Here `<committish>` is any of the following: 456 457* The name of an existing branch already in fast-import's internal branch 458 table. If fast-import doesn't know the name, it's treated as a SHA-1 459 expression. 460 461* A mark reference, `:<idnum>`, where `<idnum>` is the mark number. 462+ 463The reason fast-import uses `:` to denote a mark reference is this character 464is not legal in a Git branch name. The leading `:` makes it easy 465to distinguish between the mark 42 (`:42`) and the branch 42 (`42` 466or `refs/heads/42`), or an abbreviated SHA-1 which happened to 467consist only of base-10 digits. 468+ 469Marks must be declared (via `mark`) before they can be used. 470 471* A complete 40 byte or abbreviated commit SHA-1 in hex. 472 473* Any valid Git SHA-1 expression that resolves to a commit. See 474 ``SPECIFYING REVISIONS'' in linkgit:gitrevisions[7] for details. 475 476The special case of restarting an incremental import from the 477current branch value should be written as: 478---- 479 from refs/heads/branch^0 480---- 481The `^0` suffix is necessary as fast-import does not permit a branch to 482start from itself, and the branch is created in memory before the 483`from` command is even read from the input. Adding `^0` will force 484fast-import to resolve the commit through Git's revision parsing library, 485rather than its internal branch table, thereby loading in the 486existing value of the branch. 487 488`merge` 489^^^^^^^ 490Includes one additional ancestor commit. If the `from` command is 491omitted when creating a new branch, the first `merge` commit will be 492the first ancestor of the current commit, and the branch will start 493out with no files. An unlimited number of `merge` commands per 494commit are permitted by fast-import, thereby establishing an n-way merge. 495However Git's other tools never create commits with more than 15 496additional ancestors (forming a 16-way merge). For this reason 497it is suggested that frontends do not use more than 15 `merge` 498commands per commit; 16, if starting a new, empty branch. 499 500Here `<committish>` is any of the commit specification expressions 501also accepted by `from` (see above). 502 503`filemodify` 504^^^^^^^^^^^^ 505Included in a `commit` command to add a new file or change the 506content of an existing file. This command has two different means 507of specifying the content of the file. 508 509External data format:: 510 The data content for the file was already supplied by a prior 511 `blob` command. The frontend just needs to connect it. 512+ 513.... 514 'M' SP <mode> SP <dataref> SP <path> LF 515.... 516+ 517Here usually `<dataref>` must be either a mark reference (`:<idnum>`) 518set by a prior `blob` command, or a full 40-byte SHA-1 of an 519existing Git blob object. If `<mode>` is `040000`` then 520`<dataref>` must be the full 40-byte SHA-1 of an existing 521Git tree object or a mark reference set with `--import-marks`. 522 523Inline data format:: 524 The data content for the file has not been supplied yet. 525 The frontend wants to supply it as part of this modify 526 command. 527+ 528.... 529 'M' SP <mode> SP 'inline' SP <path> LF 530 data 531.... 532+ 533See below for a detailed description of the `data` command. 534 535In both formats `<mode>` is the type of file entry, specified 536in octal. Git only supports the following modes: 537 538* `100644` or `644`: A normal (not-executable) file. The majority 539 of files in most projects use this mode. If in doubt, this is 540 what you want. 541* `100755` or `755`: A normal, but executable, file. 542* `120000`: A symlink, the content of the file will be the link target. 543* `160000`: A gitlink, SHA-1 of the object refers to a commit in 544 another repository. Git links can only be specified by SHA or through 545 a commit mark. They are used to implement submodules. 546* `040000`: A subdirectory. Subdirectories can only be specified by 547 SHA or through a tree mark set with `--import-marks`. 548 549In both formats `<path>` is the complete path of the file to be added 550(if not already existing) or modified (if already existing). 551 552A `<path>` string must use UNIX-style directory separators (forward 553slash `/`), may contain any byte other than `LF`, and must not 554start with double quote (`"`). 555 556If an `LF` or double quote must be encoded into `<path>` shell-style 557quoting should be used, e.g. `"path/with\n and \" in it"`. 558 559The value of `<path>` must be in canonical form. That is it must not: 560 561* contain an empty directory component (e.g. `foo//bar` is invalid), 562* end with a directory separator (e.g. `foo/` is invalid), 563* start with a directory separator (e.g. `/foo` is invalid), 564* contain the special component `.` or `..` (e.g. `foo/./bar` and 565 `foo/../bar` are invalid). 566 567The root of the tree can be represented by an empty string as `<path>`. 568 569It is recommended that `<path>` always be encoded using UTF-8. 570 571`filedelete` 572^^^^^^^^^^^^ 573Included in a `commit` command to remove a file or recursively 574delete an entire directory from the branch. If the file or directory 575removal makes its parent directory empty, the parent directory will 576be automatically removed too. This cascades up the tree until the 577first non-empty directory or the root is reached. 578 579.... 580 'D' SP <path> LF 581.... 582 583here `<path>` is the complete path of the file or subdirectory to 584be removed from the branch. 585See `filemodify` above for a detailed description of `<path>`. 586 587`filecopy` 588^^^^^^^^^^^^ 589Recursively copies an existing file or subdirectory to a different 590location within the branch. The existing file or directory must 591exist. If the destination exists it will be completely replaced 592by the content copied from the source. 593 594.... 595 'C' SP <path> SP <path> LF 596.... 597 598here the first `<path>` is the source location and the second 599`<path>` is the destination. See `filemodify` above for a detailed 600description of what `<path>` may look like. To use a source path 601that contains SP the path must be quoted. 602 603A `filecopy` command takes effect immediately. Once the source 604location has been copied to the destination any future commands 605applied to the source location will not impact the destination of 606the copy. 607 608`filerename` 609^^^^^^^^^^^^ 610Renames an existing file or subdirectory to a different location 611within the branch. The existing file or directory must exist. If 612the destination exists it will be replaced by the source directory. 613 614.... 615 'R' SP <path> SP <path> LF 616.... 617 618here the first `<path>` is the source location and the second 619`<path>` is the destination. See `filemodify` above for a detailed 620description of what `<path>` may look like. To use a source path 621that contains SP the path must be quoted. 622 623A `filerename` command takes effect immediately. Once the source 624location has been renamed to the destination any future commands 625applied to the source location will create new files there and not 626impact the destination of the rename. 627 628Note that a `filerename` is the same as a `filecopy` followed by a 629`filedelete` of the source location. There is a slight performance 630advantage to using `filerename`, but the advantage is so small 631that it is never worth trying to convert a delete/add pair in 632source material into a rename for fast-import. This `filerename` 633command is provided just to simplify frontends that already have 634rename information and don't want bother with decomposing it into a 635`filecopy` followed by a `filedelete`. 636 637`filedeleteall` 638^^^^^^^^^^^^^^^ 639Included in a `commit` command to remove all files (and also all 640directories) from the branch. This command resets the internal 641branch structure to have no files in it, allowing the frontend 642to subsequently add all interesting files from scratch. 643 644.... 645 'deleteall' LF 646.... 647 648This command is extremely useful if the frontend does not know 649(or does not care to know) what files are currently on the branch, 650and therefore cannot generate the proper `filedelete` commands to 651update the content. 652 653Issuing a `filedeleteall` followed by the needed `filemodify` 654commands to set the correct content will produce the same results 655as sending only the needed `filemodify` and `filedelete` commands. 656The `filedeleteall` approach may however require fast-import to use slightly 657more memory per active branch (less than 1 MiB for even most large 658projects); so frontends that can easily obtain only the affected 659paths for a commit are encouraged to do so. 660 661`notemodify` 662^^^^^^^^^^^^ 663Included in a `commit` `<notes_ref>` command to add a new note 664annotating a `<committish>` or change this annotation contents. 665Internally it is similar to filemodify 100644 on `<committish>` 666path (maybe split into subdirectories). It's not advised to 667use any other commands to write to the `<notes_ref>` tree except 668`filedeleteall` to delete all existing notes in this tree. 669This command has two different means of specifying the content 670of the note. 671 672External data format:: 673 The data content for the note was already supplied by a prior 674 `blob` command. The frontend just needs to connect it to the 675 commit that is to be annotated. 676+ 677.... 678 'N' SP <dataref> SP <committish> LF 679.... 680+ 681Here `<dataref>` can be either a mark reference (`:<idnum>`) 682set by a prior `blob` command, or a full 40-byte SHA-1 of an 683existing Git blob object. 684 685Inline data format:: 686 The data content for the note has not been supplied yet. 687 The frontend wants to supply it as part of this modify 688 command. 689+ 690.... 691 'N' SP 'inline' SP <committish> LF 692 data 693.... 694+ 695See below for a detailed description of the `data` command. 696 697In both formats `<committish>` is any of the commit specification 698expressions also accepted by `from` (see above). 699 700`mark` 701~~~~~~ 702Arranges for fast-import to save a reference to the current object, allowing 703the frontend to recall this object at a future point in time, without 704knowing its SHA-1. Here the current object is the object creation 705command the `mark` command appears within. This can be `commit`, 706`tag`, and `blob`, but `commit` is the most common usage. 707 708.... 709 'mark' SP ':' <idnum> LF 710.... 711 712where `<idnum>` is the number assigned by the frontend to this mark. 713The value of `<idnum>` is expressed as an ASCII decimal integer. 714The value 0 is reserved and cannot be used as 715a mark. Only values greater than or equal to 1 may be used as marks. 716 717New marks are created automatically. Existing marks can be moved 718to another object simply by reusing the same `<idnum>` in another 719`mark` command. 720 721`tag` 722~~~~~ 723Creates an annotated tag referring to a specific commit. To create 724lightweight (non-annotated) tags see the `reset` command below. 725 726.... 727 'tag' SP <name> LF 728 'from' SP <committish> LF 729 'tagger' (SP <name>)? SP LT <email> GT SP <when> LF 730 data 731.... 732 733where `<name>` is the name of the tag to create. 734 735Tag names are automatically prefixed with `refs/tags/` when stored 736in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would 737use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the 738corresponding ref as `refs/tags/RELENG-1_0-FINAL`. 739 740The value of `<name>` must be a valid refname in Git and therefore 741may contain forward slashes. As `LF` is not valid in a Git refname, 742no quoting or escaping syntax is supported here. 743 744The `from` command is the same as in the `commit` command; see 745above for details. 746 747The `tagger` command uses the same format as `committer` within 748`commit`; again see above for details. 749 750The `data` command following `tagger` must supply the annotated tag 751message (see below for `data` command syntax). To import an empty 752tag message use a 0 length data. Tag messages are free-form and are 753not interpreted by Git. Currently they must be encoded in UTF-8, 754as fast-import does not permit other encodings to be specified. 755 756Signing annotated tags during import from within fast-import is not 757supported. Trying to include your own PGP/GPG signature is not 758recommended, as the frontend does not (easily) have access to the 759complete set of bytes which normally goes into such a signature. 760If signing is required, create lightweight tags from within fast-import with 761`reset`, then create the annotated versions of those tags offline 762with the standard 'git tag' process. 763 764`reset` 765~~~~~~~ 766Creates (or recreates) the named branch, optionally starting from 767a specific revision. The reset command allows a frontend to issue 768a new `from` command for an existing branch, or to create a new 769branch from an existing commit without creating a new commit. 770 771.... 772 'reset' SP <ref> LF 773 ('from' SP <committish> LF)? 774 LF? 775.... 776 777For a detailed description of `<ref>` and `<committish>` see above 778under `commit` and `from`. 779 780The `LF` after the command is optional (it used to be required). 781 782The `reset` command can also be used to create lightweight 783(non-annotated) tags. For example: 784 785==== 786 reset refs/tags/938 787 from :938 788==== 789 790would create the lightweight tag `refs/tags/938` referring to 791whatever commit mark `:938` references. 792 793`blob` 794~~~~~~ 795Requests writing one file revision to the packfile. The revision 796is not connected to any commit; this connection must be formed in 797a subsequent `commit` command by referencing the blob through an 798assigned mark. 799 800.... 801 'blob' LF 802 mark? 803 data 804.... 805 806The mark command is optional here as some frontends have chosen 807to generate the Git SHA-1 for the blob on their own, and feed that 808directly to `commit`. This is typically more work than it's worth 809however, as marks are inexpensive to store and easy to use. 810 811`data` 812~~~~~~ 813Supplies raw data (for use as blob/file content, commit messages, or 814annotated tag messages) to fast-import. Data can be supplied using an exact 815byte count or delimited with a terminating line. Real frontends 816intended for production-quality conversions should always use the 817exact byte count format, as it is more robust and performs better. 818The delimited format is intended primarily for testing fast-import. 819 820Comment lines appearing within the `<raw>` part of `data` commands 821are always taken to be part of the body of the data and are therefore 822never ignored by fast-import. This makes it safe to import any 823file/message content whose lines might start with `#`. 824 825Exact byte count format:: 826 The frontend must specify the number of bytes of data. 827+ 828.... 829 'data' SP <count> LF 830 <raw> LF? 831.... 832+ 833where `<count>` is the exact number of bytes appearing within 834`<raw>`. The value of `<count>` is expressed as an ASCII decimal 835integer. The `LF` on either side of `<raw>` is not 836included in `<count>` and will not be included in the imported data. 837+ 838The `LF` after `<raw>` is optional (it used to be required) but 839recommended. Always including it makes debugging a fast-import 840stream easier as the next command always starts in column 0 841of the next line, even if `<raw>` did not end with an `LF`. 842 843Delimited format:: 844 A delimiter string is used to mark the end of the data. 845 fast-import will compute the length by searching for the delimiter. 846 This format is primarily useful for testing and is not 847 recommended for real data. 848+ 849.... 850 'data' SP '<<' <delim> LF 851 <raw> LF 852 <delim> LF 853 LF? 854.... 855+ 856where `<delim>` is the chosen delimiter string. The string `<delim>` 857must not appear on a line by itself within `<raw>`, as otherwise 858fast-import will think the data ends earlier than it really does. The `LF` 859immediately trailing `<raw>` is part of `<raw>`. This is one of 860the limitations of the delimited format, it is impossible to supply 861a data chunk which does not have an LF as its last byte. 862+ 863The `LF` after `<delim> LF` is optional (it used to be required). 864 865`checkpoint` 866~~~~~~~~~~~~ 867Forces fast-import to close the current packfile, start a new one, and to 868save out all current branch refs, tags and marks. 869 870.... 871 'checkpoint' LF 872 LF? 873.... 874 875Note that fast-import automatically switches packfiles when the current 876packfile reaches \--max-pack-size, or 4 GiB, whichever limit is 877smaller. During an automatic packfile switch fast-import does not update 878the branch refs, tags or marks. 879 880As a `checkpoint` can require a significant amount of CPU time and 881disk IO (to compute the overall pack SHA-1 checksum, generate the 882corresponding index file, and update the refs) it can easily take 883several minutes for a single `checkpoint` command to complete. 884 885Frontends may choose to issue checkpoints during extremely large 886and long running imports, or when they need to allow another Git 887process access to a branch. However given that a 30 GiB Subversion 888repository can be loaded into Git through fast-import in about 3 hours, 889explicit checkpointing may not be necessary. 890 891The `LF` after the command is optional (it used to be required). 892 893`progress` 894~~~~~~~~~~ 895Causes fast-import to print the entire `progress` line unmodified to 896its standard output channel (file descriptor 1) when the command is 897processed from the input stream. The command otherwise has no impact 898on the current import, or on any of fast-import's internal state. 899 900.... 901 'progress' SP <any> LF 902 LF? 903.... 904 905The `<any>` part of the command may contain any sequence of bytes 906that does not contain `LF`. The `LF` after the command is optional. 907Callers may wish to process the output through a tool such as sed to 908remove the leading part of the line, for example: 909 910==== 911 frontend | git fast-import | sed 's/^progress //' 912==== 913 914Placing a `progress` command immediately after a `checkpoint` will 915inform the reader when the `checkpoint` has been completed and it 916can safely access the refs that fast-import updated. 917 918`cat-blob` 919~~~~~~~~~~ 920Causes fast-import to print a blob to a file descriptor previously 921arranged with the `--cat-blob-fd` argument. The command otherwise 922has no impact on the current import; its main purpose is to 923retrieve blobs that may be in fast-import's memory but not 924accessible from the target repository. 925 926.... 927 'cat-blob' SP <dataref> LF 928.... 929 930The `<dataref>` can be either a mark reference (`:<idnum>`) 931set previously or a full 40-byte SHA-1 of a Git blob, preexisting or 932ready to be written. 933 934Output uses the same format as `git cat-file --batch`: 935 936==== 937 <sha1> SP 'blob' SP <size> LF 938 <contents> LF 939==== 940 941This command can be used anywhere in the stream that comments are 942accepted. In particular, the `cat-blob` command can be used in the 943middle of a commit but not in the middle of a `data` command. 944 945`ls` 946~~~~ 947Prints information about the object at a path to a file descriptor 948previously arranged with the `--cat-blob-fd` argument. This allows 949printing a blob from the active commit (with `cat-blob`) or copying a 950blob or tree from a previous commit for use in the current one (with 951`filemodify`). 952 953The `ls` command can be used anywhere in the stream that comments are 954accepted, including the middle of a commit. 955 956Reading from the active commit:: 957 This form can only be used in the middle of a `commit`. 958 The path names a directory entry within fast-import's 959 active commit. The path must be quoted in this case. 960+ 961.... 962 'ls' SP <path> LF 963.... 964 965Reading from a named tree:: 966 The `<dataref>` can be a mark reference (`:<idnum>`) or the 967 full 40-byte SHA-1 of a Git tag, commit, or tree object, 968 preexisting or waiting to be written. 969 The path is relative to the top level of the tree 970 named by `<dataref>`. 971+ 972.... 973 'ls' SP <dataref> SP <path> LF 974.... 975 976See `filemodify` above for a detailed description of `<path>`. 977 978Output uses the same format as `git ls-tree <tree> -- <path>`: 979 980==== 981 <mode> SP ('blob' | 'tree' | 'commit') SP <dataref> HT <path> LF 982==== 983 984The <dataref> represents the blob, tree, or commit object at <path> 985and can be used in later 'cat-blob', 'filemodify', or 'ls' commands. 986 987If there is no file or subtree at that path, 'git fast-import' will 988instead report 989 990==== 991 missing SP <path> LF 992==== 993 994`feature` 995~~~~~~~~~ 996Require that fast-import supports the specified feature, or abort if 997it does not. 998 999....1000 'feature' SP <feature> ('=' <argument>)? LF1001....10021003The <feature> part of the command may be any one of the following:10041005date-format::1006export-marks::1007relative-marks::1008no-relative-marks::1009force::1010 Act as though the corresponding command-line option with1011 a leading '--' was passed on the command line1012 (see OPTIONS, above).10131014import-marks::1015import-marks-if-exists::1016 Like --import-marks except in two respects: first, only one1017 "feature import-marks" or "feature import-marks-if-exists"1018 command is allowed per stream; second, an --import-marks=1019 or --import-marks-if-exists command-line option overrides1020 any of these "feature" commands in the stream; third,1021 "feature import-marks-if-exists" like a corresponding1022 command-line option silently skips a nonexistent file.10231024cat-blob::1025ls::1026 Require that the backend support the 'cat-blob' or 'ls' command.1027 Versions of fast-import not supporting the specified command1028 will exit with a message indicating so.1029 This lets the import error out early with a clear message,1030 rather than wasting time on the early part of an import1031 before the unsupported command is detected.10321033notes::1034 Require that the backend support the 'notemodify' (N)1035 subcommand to the 'commit' command.1036 Versions of fast-import not supporting notes will exit1037 with a message indicating so.10381039done::1040 Error out if the stream ends without a 'done' command.1041 Without this feature, errors causing the frontend to end1042 abruptly at a convenient point in the stream can go1043 undetected.10441045`option`1046~~~~~~~~1047Processes the specified option so that git fast-import behaves in a1048way that suits the frontend's needs.1049Note that options specified by the frontend are overridden by any1050options the user may specify to git fast-import itself.10511052....1053 'option' SP <option> LF1054....10551056The `<option>` part of the command may contain any of the options1057listed in the OPTIONS section that do not change import semantics,1058without the leading '--' and is treated in the same way.10591060Option commands must be the first commands on the input (not counting1061feature commands), to give an option command after any non-option1062command is an error.10631064The following commandline options change import semantics and may therefore1065not be passed as option:10661067* date-format1068* import-marks1069* export-marks1070* cat-blob-fd1071* force10721073`done`1074~~~~~~1075If the `done` feature is not in use, treated as if EOF was read.1076This can be used to tell fast-import to finish early.10771078If the `--done` command line option or `feature done` command is1079in use, the `done` command is mandatory and marks the end of the1080stream.10811082Crash Reports1083-------------1084If fast-import is supplied invalid input it will terminate with a1085non-zero exit status and create a crash report in the top level of1086the Git repository it was importing into. Crash reports contain1087a snapshot of the internal fast-import state as well as the most1088recent commands that lead up to the crash.10891090All recent commands (including stream comments, file changes and1091progress commands) are shown in the command history within the crash1092report, but raw file data and commit messages are excluded from the1093crash report. This exclusion saves space within the report file1094and reduces the amount of buffering that fast-import must perform1095during execution.10961097After writing a crash report fast-import will close the current1098packfile and export the marks table. This allows the frontend1099developer to inspect the repository state and resume the import from1100the point where it crashed. The modified branches and tags are not1101updated during a crash, as the import did not complete successfully.1102Branch and tag information can be found in the crash report and1103must be applied manually if the update is needed.11041105An example crash:11061107====1108 $ cat >in <<END_OF_INPUT1109 # my very first test commit1110 commit refs/heads/master1111 committer Shawn O. Pearce <spearce> 19283 -04001112 # who is that guy anyway?1113 data <<EOF1114 this is my commit1115 EOF1116 M 644 inline .gitignore1117 data <<EOF1118 .gitignore1119 EOF1120 M 777 inline bob1121 END_OF_INPUT11221123 $ git fast-import <in1124 fatal: Corrupt mode: M 777 inline bob1125 fast-import: dumping crash report to .git/fast_import_crash_843411261127 $ cat .git/fast_import_crash_84341128 fast-import crash report:1129 fast-import process: 84341130 parent process : 13911131 at Sat Sep 1 00:58:12 200711321133 fatal: Corrupt mode: M 777 inline bob11341135 Most Recent Commands Before Crash1136 ---------------------------------1137 # my very first test commit1138 commit refs/heads/master1139 committer Shawn O. Pearce <spearce> 19283 -04001140 # who is that guy anyway?1141 data <<EOF1142 M 644 inline .gitignore1143 data <<EOF1144 * M 777 inline bob11451146 Active Branch LRU1147 -----------------1148 active_branches = 1 cur, 5 max11491150 pos clock name1151 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1152 1) 0 refs/heads/master11531154 Inactive Branches1155 -----------------1156 refs/heads/master:1157 status : active loaded dirty1158 tip commit : 00000000000000000000000000000000000000001159 old tree : 00000000000000000000000000000000000000001160 cur tree : 00000000000000000000000000000000000000001161 commit clock: 01162 last pack :116311641165 -------------------1166 END OF CRASH REPORT1167====11681169Tips and Tricks1170---------------1171The following tips and tricks have been collected from various1172users of fast-import, and are offered here as suggestions.11731174Use One Mark Per Commit1175~~~~~~~~~~~~~~~~~~~~~~~1176When doing a repository conversion, use a unique mark per commit1177(`mark :<n>`) and supply the \--export-marks option on the command1178line. fast-import will dump a file which lists every mark and the Git1179object SHA-1 that corresponds to it. If the frontend can tie1180the marks back to the source repository, it is easy to verify the1181accuracy and completeness of the import by comparing each Git1182commit to the corresponding source revision.11831184Coming from a system such as Perforce or Subversion this should be1185quite simple, as the fast-import mark can also be the Perforce changeset1186number or the Subversion revision number.11871188Freely Skip Around Branches1189~~~~~~~~~~~~~~~~~~~~~~~~~~~1190Don't bother trying to optimize the frontend to stick to one branch1191at a time during an import. Although doing so might be slightly1192faster for fast-import, it tends to increase the complexity of the frontend1193code considerably.11941195The branch LRU builtin to fast-import tends to behave very well, and the1196cost of activating an inactive branch is so low that bouncing around1197between branches has virtually no impact on import performance.11981199Handling Renames1200~~~~~~~~~~~~~~~~1201When importing a renamed file or directory, simply delete the old1202name(s) and modify the new name(s) during the corresponding commit.1203Git performs rename detection after-the-fact, rather than explicitly1204during a commit.12051206Use Tag Fixup Branches1207~~~~~~~~~~~~~~~~~~~~~~1208Some other SCM systems let the user create a tag from multiple1209files which are not from the same commit/changeset. Or to create1210tags which are a subset of the files available in the repository.12111212Importing these tags as-is in Git is impossible without making at1213least one commit which ``fixes up'' the files to match the content1214of the tag. Use fast-import's `reset` command to reset a dummy branch1215outside of your normal branch space to the base commit for the tag,1216then commit one or more file fixup commits, and finally tag the1217dummy branch.12181219For example since all normal branches are stored under `refs/heads/`1220name the tag fixup branch `TAG_FIXUP`. This way it is impossible for1221the fixup branch used by the importer to have namespace conflicts1222with real branches imported from the source (the name `TAG_FIXUP`1223is not `refs/heads/TAG_FIXUP`).12241225When committing fixups, consider using `merge` to connect the1226commit(s) which are supplying file revisions to the fixup branch.1227Doing so will allow tools such as 'git blame' to track1228through the real commit history and properly annotate the source1229files.12301231After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP`1232to remove the dummy branch.12331234Import Now, Repack Later1235~~~~~~~~~~~~~~~~~~~~~~~~1236As soon as fast-import completes the Git repository is completely valid1237and ready for use. Typically this takes only a very short time,1238even for considerably large projects (100,000+ commits).12391240However repacking the repository is necessary to improve data1241locality and access performance. It can also take hours on extremely1242large projects (especially if -f and a large \--window parameter is1243used). Since repacking is safe to run alongside readers and writers,1244run the repack in the background and let it finish when it finishes.1245There is no reason to wait to explore your new Git project!12461247If you choose to wait for the repack, don't try to run benchmarks1248or performance tests until repacking is completed. fast-import outputs1249suboptimal packfiles that are simply never seen in real use1250situations.12511252Repacking Historical Data1253~~~~~~~~~~~~~~~~~~~~~~~~~1254If you are repacking very old imported data (e.g. older than the1255last year), consider expending some extra CPU time and supplying1256\--window=50 (or higher) when you run 'git repack'.1257This will take longer, but will also produce a smaller packfile.1258You only need to expend the effort once, and everyone using your1259project will benefit from the smaller repository.12601261Include Some Progress Messages1262~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1263Every once in a while have your frontend emit a `progress` message1264to fast-import. The contents of the messages are entirely free-form,1265so one suggestion would be to output the current month and year1266each time the current commit date moves into the next month.1267Your users will feel better knowing how much of the data stream1268has been processed.126912701271Packfile Optimization1272---------------------1273When packing a blob fast-import always attempts to deltify against the last1274blob written. Unless specifically arranged for by the frontend,1275this will probably not be a prior version of the same file, so the1276generated delta will not be the smallest possible. The resulting1277packfile will be compressed, but will not be optimal.12781279Frontends which have efficient access to all revisions of a1280single file (for example reading an RCS/CVS ,v file) can choose1281to supply all revisions of that file as a sequence of consecutive1282`blob` commands. This allows fast-import to deltify the different file1283revisions against each other, saving space in the final packfile.1284Marks can be used to later identify individual file revisions during1285a sequence of `commit` commands.12861287The packfile(s) created by fast-import do not encourage good disk access1288patterns. This is caused by fast-import writing the data in the order1289it is received on standard input, while Git typically organizes1290data within packfiles to make the most recent (current tip) data1291appear before historical data. Git also clusters commits together,1292speeding up revision traversal through better cache locality.12931294For this reason it is strongly recommended that users repack the1295repository with `git repack -a -d` after fast-import completes, allowing1296Git to reorganize the packfiles for faster data access. If blob1297deltas are suboptimal (see above) then also adding the `-f` option1298to force recomputation of all deltas can significantly reduce the1299final packfile size (30-50% smaller can be quite typical).130013011302Memory Utilization1303------------------1304There are a number of factors which affect how much memory fast-import1305requires to perform an import. Like critical sections of core1306Git, fast-import uses its own memory allocators to amortize any overheads1307associated with malloc. In practice fast-import tends to amortize any1308malloc overheads to 0, due to its use of large block allocations.13091310per object1311~~~~~~~~~~1312fast-import maintains an in-memory structure for every object written in1313this execution. On a 32 bit system the structure is 32 bytes,1314on a 64 bit system the structure is 40 bytes (due to the larger1315pointer sizes). Objects in the table are not deallocated until1316fast-import terminates. Importing 2 million objects on a 32 bit system1317will require approximately 64 MiB of memory.13181319The object table is actually a hashtable keyed on the object name1320(the unique SHA-1). This storage configuration allows fast-import to reuse1321an existing or already written object and avoid writing duplicates1322to the output packfile. Duplicate blobs are surprisingly common1323in an import, typically due to branch merges in the source.13241325per mark1326~~~~~~~~1327Marks are stored in a sparse array, using 1 pointer (4 bytes or 81328bytes, depending on pointer size) per mark. Although the array1329is sparse, frontends are still strongly encouraged to use marks1330between 1 and n, where n is the total number of marks required for1331this import.13321333per branch1334~~~~~~~~~~1335Branches are classified as active and inactive. The memory usage1336of the two classes is significantly different.13371338Inactive branches are stored in a structure which uses 96 or 1201339bytes (32 bit or 64 bit systems, respectively), plus the length of1340the branch name (typically under 200 bytes), per branch. fast-import will1341easily handle as many as 10,000 inactive branches in under 2 MiB1342of memory.13431344Active branches have the same overhead as inactive branches, but1345also contain copies of every tree that has been recently modified on1346that branch. If subtree `include` has not been modified since the1347branch became active, its contents will not be loaded into memory,1348but if subtree `src` has been modified by a commit since the branch1349became active, then its contents will be loaded in memory.13501351As active branches store metadata about the files contained on that1352branch, their in-memory storage size can grow to a considerable size1353(see below).13541355fast-import automatically moves active branches to inactive status based on1356a simple least-recently-used algorithm. The LRU chain is updated on1357each `commit` command. The maximum number of active branches can be1358increased or decreased on the command line with \--active-branches=.13591360per active tree1361~~~~~~~~~~~~~~~1362Trees (aka directories) use just 12 bytes of memory on top of the1363memory required for their entries (see ``per active file'' below).1364The cost of a tree is virtually 0, as its overhead amortizes out1365over the individual file entries.13661367per active file entry1368~~~~~~~~~~~~~~~~~~~~~1369Files (and pointers to subtrees) within active trees require 52 or 641370bytes (32/64 bit platforms) per entry. To conserve space, file and1371tree names are pooled in a common string table, allowing the filename1372``Makefile'' to use just 16 bytes (after including the string header1373overhead) no matter how many times it occurs within the project.13741375The active branch LRU, when coupled with the filename string pool1376and lazy loading of subtrees, allows fast-import to efficiently import1377projects with 2,000+ branches and 45,114+ files in a very limited1378memory footprint (less than 2.7 MiB per active branch).13791380Signals1381-------1382Sending *SIGUSR1* to the 'git fast-import' process ends the current1383packfile early, simulating a `checkpoint` command. The impatient1384operator can use this facility to peek at the objects and refs from an1385import in progress, at the cost of some added running time and worse1386compression.13871388GIT1389---1390Part of the linkgit:git[1] suite