1git-fast-import(1) 2================== 3 4NAME 5---- 6git-fast-import - Backend for fast Git data importers 7 8 9SYNOPSIS 10-------- 11frontend | 'git fast-import' [options] 12 13DESCRIPTION 14----------- 15This program is usually not what the end user wants to run directly. 16Most end users want to use one of the existing frontend programs, 17which parses a specific type of foreign source and feeds the contents 18stored there to 'git-fast-import'. 19 20fast-import reads a mixed command/data stream from standard input and 21writes one or more packfiles directly into the current repository. 22When EOF is received on standard input, fast import writes out 23updated branch and tag refs, fully updating the current repository 24with the newly imported data. 25 26The fast-import backend itself can import into an empty repository (one that 27has already been initialized by 'git-init') or incrementally 28update an existing populated repository. Whether or not incremental 29imports are supported from a particular foreign source depends on 30the frontend program in use. 31 32 33OPTIONS 34------- 35--date-format=<fmt>:: 36 Specify the type of dates the frontend will supply to 37 fast-import within `author`, `committer` and `tagger` commands. 38 See ``Date Formats'' below for details about which formats 39 are supported, and their syntax. 40 41--force:: 42 Force updating modified existing branches, even if doing 43 so would cause commits to be lost (as the new commit does 44 not contain the old commit). 45 46--max-pack-size=<n>:: 47 Maximum size of each output packfile, expressed in MiB. 48 The default is 4096 (4 GiB) as that is the maximum allowed 49 packfile size (due to file format limitations). Some 50 importers may wish to lower this, such as to ensure the 51 resulting packfiles fit on CDs. 52 53--depth=<n>:: 54 Maximum delta depth, for blob and tree deltification. 55 Default is 10. 56 57--active-branches=<n>:: 58 Maximum number of branches to maintain active at once. 59 See ``Memory Utilization'' below for details. Default is 5. 60 61--export-marks=<file>:: 62 Dumps the internal marks table to <file> when complete. 63 Marks are written one per line as `:markid SHA-1`. 64 Frontends can use this file to validate imports after they 65 have been completed, or to save the marks table across 66 incremental runs. As <file> is only opened and truncated 67 at checkpoint (or completion) the same path can also be 68 safely given to \--import-marks. 69 70--import-marks=<file>:: 71 Before processing any input, load the marks specified in 72 <file>. The input file must exist, must be readable, and 73 must use the same format as produced by \--export-marks. 74 Multiple options may be supplied to import more than one 75 set of marks. If a mark is defined to different values, 76 the last file wins. 77 78--relative-marks:: 79 After specifying --relative-marks= the paths specified 80 with --import-marks= and --export-marks= are relative 81 to an internal directory in the current repository. 82 In git-fast-import this means that the paths are relative 83 to the .git/info/fast-import directory. However, other 84 importers may use a different location. 85 86--no-relative-marks:: 87 Negates a previous --relative-marks. Allows for combining 88 relative and non-relative marks by interweaving 89 --(no-)-relative-marks= with the --(import|export)-marks= 90 options. 91 92--export-pack-edges=<file>:: 93 After creating a packfile, print a line of data to 94 <file> listing the filename of the packfile and the last 95 commit on each branch that was written to that packfile. 96 This information may be useful after importing projects 97 whose total object set exceeds the 4 GiB packfile limit, 98 as these commits can be used as edge points during calls 99 to 'git-pack-objects'. 100 101--quiet:: 102 Disable all non-fatal output, making fast-import silent when it 103 is successful. This option disables the output shown by 104 \--stats. 105 106--stats:: 107 Display some basic statistics about the objects fast-import has 108 created, the packfiles they were stored into, and the 109 memory used by fast-import during this run. Showing this output 110 is currently the default, but can be disabled with \--quiet. 111 112 113Performance 114----------- 115The design of fast-import allows it to import large projects in a minimum 116amount of memory usage and processing time. Assuming the frontend 117is able to keep up with fast-import and feed it a constant stream of data, 118import times for projects holding 10+ years of history and containing 119100,000+ individual commits are generally completed in just 1-2 120hours on quite modest (~$2,000 USD) hardware. 121 122Most bottlenecks appear to be in foreign source data access (the 123source just cannot extract revisions fast enough) or disk IO (fast-import 124writes as fast as the disk will take the data). Imports will run 125faster if the source data is stored on a different drive than the 126destination Git repository (due to less IO contention). 127 128 129Development Cost 130---------------- 131A typical frontend for fast-import tends to weigh in at approximately 200 132lines of Perl/Python/Ruby code. Most developers have been able to 133create working importers in just a couple of hours, even though it 134is their first exposure to fast-import, and sometimes even to Git. This is 135an ideal situation, given that most conversion tools are throw-away 136(use once, and never look back). 137 138 139Parallel Operation 140------------------ 141Like 'git-push' or 'git-fetch', imports handled by fast-import are safe to 142run alongside parallel `git repack -a -d` or `git gc` invocations, 143or any other Git operation (including 'git-prune', as loose objects 144are never used by fast-import). 145 146fast-import does not lock the branch or tag refs it is actively importing. 147After the import, during its ref update phase, fast-import tests each 148existing branch ref to verify the update will be a fast-forward 149update (the commit stored in the ref is contained in the new 150history of the commit to be written). If the update is not a 151fast-forward update, fast-import will skip updating that ref and instead 152prints a warning message. fast-import will always attempt to update all 153branch refs, and does not stop on the first failure. 154 155Branch updates can be forced with \--force, but its recommended that 156this only be used on an otherwise quiet repository. Using \--force 157is not necessary for an initial import into an empty repository. 158 159 160Technical Discussion 161-------------------- 162fast-import tracks a set of branches in memory. Any branch can be created 163or modified at any point during the import process by sending a 164`commit` command on the input stream. This design allows a frontend 165program to process an unlimited number of branches simultaneously, 166generating commits in the order they are available from the source 167data. It also simplifies the frontend programs considerably. 168 169fast-import does not use or alter the current working directory, or any 170file within it. (It does however update the current Git repository, 171as referenced by `GIT_DIR`.) Therefore an import frontend may use 172the working directory for its own purposes, such as extracting file 173revisions from the foreign source. This ignorance of the working 174directory also allows fast-import to run very quickly, as it does not 175need to perform any costly file update operations when switching 176between branches. 177 178Input Format 179------------ 180With the exception of raw file data (which Git does not interpret) 181the fast-import input format is text (ASCII) based. This text based 182format simplifies development and debugging of frontend programs, 183especially when a higher level language such as Perl, Python or 184Ruby is being used. 185 186fast-import is very strict about its input. Where we say SP below we mean 187*exactly* one space. Likewise LF means one (and only one) linefeed. 188Supplying additional whitespace characters will cause unexpected 189results, such as branch names or file names with leading or trailing 190spaces in their name, or early termination of fast-import when it encounters 191unexpected input. 192 193Stream Comments 194~~~~~~~~~~~~~~~ 195To aid in debugging frontends fast-import ignores any line that 196begins with `#` (ASCII pound/hash) up to and including the line 197ending `LF`. A comment line may contain any sequence of bytes 198that does not contain an LF and therefore may be used to include 199any detailed debugging information that might be specific to the 200frontend and useful when inspecting a fast-import data stream. 201 202Date Formats 203~~~~~~~~~~~~ 204The following date formats are supported. A frontend should select 205the format it will use for this import by passing the format name 206in the \--date-format=<fmt> command line option. 207 208`raw`:: 209 This is the Git native format and is `<time> SP <offutc>`. 210 It is also fast-import's default format, if \--date-format was 211 not specified. 212+ 213The time of the event is specified by `<time>` as the number of 214seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is 215written as an ASCII decimal integer. 216+ 217The local offset is specified by `<offutc>` as a positive or negative 218offset from UTC. For example EST (which is 5 hours behind UTC) 219would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''. 220The local offset does not affect `<time>`; it is used only as an 221advisement to help formatting routines display the timestamp. 222+ 223If the local offset is not available in the source material, use 224``+0000'', or the most common local offset. For example many 225organizations have a CVS repository which has only ever been accessed 226by users who are located in the same location and timezone. In this 227case a reasonable offset from UTC could be assumed. 228+ 229Unlike the `rfc2822` format, this format is very strict. Any 230variation in formatting will cause fast-import to reject the value. 231 232`rfc2822`:: 233 This is the standard email format as described by RFC 2822. 234+ 235An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git 236parser is accurate, but a little on the lenient side. It is the 237same parser used by 'git-am' when applying patches 238received from email. 239+ 240Some malformed strings may be accepted as valid dates. In some of 241these cases Git will still be able to obtain the correct date from 242the malformed string. There are also some types of malformed 243strings which Git will parse wrong, and yet consider valid. 244Seriously malformed strings will be rejected. 245+ 246Unlike the `raw` format above, the timezone/UTC offset information 247contained in an RFC 2822 date string is used to adjust the date 248value to UTC prior to storage. Therefore it is important that 249this information be as accurate as possible. 250+ 251If the source material uses RFC 2822 style dates, 252the frontend should let fast-import handle the parsing and conversion 253(rather than attempting to do it itself) as the Git parser has 254been well tested in the wild. 255+ 256Frontends should prefer the `raw` format if the source material 257already uses UNIX-epoch format, can be coaxed to give dates in that 258format, or its format is easily convertible to it, as there is no 259ambiguity in parsing. 260 261`now`:: 262 Always use the current time and timezone. The literal 263 `now` must always be supplied for `<when>`. 264+ 265This is a toy format. The current time and timezone of this system 266is always copied into the identity string at the time it is being 267created by fast-import. There is no way to specify a different time or 268timezone. 269+ 270This particular format is supplied as its short to implement and 271may be useful to a process that wants to create a new commit 272right now, without needing to use a working directory or 273'git-update-index'. 274+ 275If separate `author` and `committer` commands are used in a `commit` 276the timestamps may not match, as the system clock will be polled 277twice (once for each command). The only way to ensure that both 278author and committer identity information has the same timestamp 279is to omit `author` (thus copying from `committer`) or to use a 280date format other than `now`. 281 282Commands 283~~~~~~~~ 284fast-import accepts several commands to update the current repository 285and control the current import process. More detailed discussion 286(with examples) of each command follows later. 287 288`commit`:: 289 Creates a new branch or updates an existing branch by 290 creating a new commit and updating the branch to point at 291 the newly created commit. 292 293`tag`:: 294 Creates an annotated tag object from an existing commit or 295 branch. Lightweight tags are not supported by this command, 296 as they are not recommended for recording meaningful points 297 in time. 298 299`reset`:: 300 Reset an existing branch (or a new branch) to a specific 301 revision. This command must be used to change a branch to 302 a specific revision without making a commit on it. 303 304`blob`:: 305 Convert raw file data into a blob, for future use in a 306 `commit` command. This command is optional and is not 307 needed to perform an import. 308 309`checkpoint`:: 310 Forces fast-import to close the current packfile, generate its 311 unique SHA-1 checksum and index, and start a new packfile. 312 This command is optional and is not needed to perform 313 an import. 314 315`progress`:: 316 Causes fast-import to echo the entire line to its own 317 standard output. This command is optional and is not needed 318 to perform an import. 319 320`feature`:: 321 Require that fast-import supports the specified feature, or 322 abort if it does not. 323 324`option`:: 325 Specify any of the options listed under OPTIONS that do not 326 change stream semantic to suit the frontend's needs. This 327 command is optional and is not needed to perform an import. 328 329`commit` 330~~~~~~~~ 331Create or update a branch with a new commit, recording one logical 332change to the project. 333 334.... 335 'commit' SP <ref> LF 336 mark? 337 ('author' (SP <name>)? SP LT <email> GT SP <when> LF)? 338 'committer' (SP <name>)? SP LT <email> GT SP <when> LF 339 data 340 ('from' SP <committish> LF)? 341 ('merge' SP <committish> LF)? 342 (filemodify | filedelete | filecopy | filerename | filedeleteall | notemodify)* 343 LF? 344.... 345 346where `<ref>` is the name of the branch to make the commit on. 347Typically branch names are prefixed with `refs/heads/` in 348Git, so importing the CVS branch symbol `RELENG-1_0` would use 349`refs/heads/RELENG-1_0` for the value of `<ref>`. The value of 350`<ref>` must be a valid refname in Git. As `LF` is not valid in 351a Git refname, no quoting or escaping syntax is supported here. 352 353A `mark` command may optionally appear, requesting fast-import to save a 354reference to the newly created commit for future use by the frontend 355(see below for format). It is very common for frontends to mark 356every commit they create, thereby allowing future branch creation 357from any imported commit. 358 359The `data` command following `committer` must supply the commit 360message (see below for `data` command syntax). To import an empty 361commit message use a 0 length data. Commit messages are free-form 362and are not interpreted by Git. Currently they must be encoded in 363UTF-8, as fast-import does not permit other encodings to be specified. 364 365Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename`, 366`filedeleteall` and `notemodify` commands 367may be included to update the contents of the branch prior to 368creating the commit. These commands may be supplied in any order. 369However it is recommended that a `filedeleteall` command precede 370all `filemodify`, `filecopy`, `filerename` and `notemodify` commands in 371the same commit, as `filedeleteall` wipes the branch clean (see below). 372 373The `LF` after the command is optional (it used to be required). 374 375`author` 376^^^^^^^^ 377An `author` command may optionally appear, if the author information 378might differ from the committer information. If `author` is omitted 379then fast-import will automatically use the committer's information for 380the author portion of the commit. See below for a description of 381the fields in `author`, as they are identical to `committer`. 382 383`committer` 384^^^^^^^^^^^ 385The `committer` command indicates who made this commit, and when 386they made it. 387 388Here `<name>` is the person's display name (for example 389``Com M Itter'') and `<email>` is the person's email address 390(``cm@example.com''). `LT` and `GT` are the literal less-than (\x3c) 391and greater-than (\x3e) symbols. These are required to delimit 392the email address from the other fields in the line. Note that 393`<name>` is free-form and may contain any sequence of bytes, except 394`LT` and `LF`. It is typically UTF-8 encoded. 395 396The time of the change is specified by `<when>` using the date format 397that was selected by the \--date-format=<fmt> command line option. 398See ``Date Formats'' above for the set of supported formats, and 399their syntax. 400 401`from` 402^^^^^^ 403The `from` command is used to specify the commit to initialize 404this branch from. This revision will be the first ancestor of the 405new commit. 406 407Omitting the `from` command in the first commit of a new branch 408will cause fast-import to create that commit with no ancestor. This 409tends to be desired only for the initial commit of a project. 410If the frontend creates all files from scratch when making a new 411branch, a `merge` command may be used instead of `from` to start 412the commit with an empty tree. 413Omitting the `from` command on existing branches is usually desired, 414as the current commit on that branch is automatically assumed to 415be the first ancestor of the new commit. 416 417As `LF` is not valid in a Git refname or SHA-1 expression, no 418quoting or escaping syntax is supported within `<committish>`. 419 420Here `<committish>` is any of the following: 421 422* The name of an existing branch already in fast-import's internal branch 423 table. If fast-import doesn't know the name, its treated as a SHA-1 424 expression. 425 426* A mark reference, `:<idnum>`, where `<idnum>` is the mark number. 427+ 428The reason fast-import uses `:` to denote a mark reference is this character 429is not legal in a Git branch name. The leading `:` makes it easy 430to distinguish between the mark 42 (`:42`) and the branch 42 (`42` 431or `refs/heads/42`), or an abbreviated SHA-1 which happened to 432consist only of base-10 digits. 433+ 434Marks must be declared (via `mark`) before they can be used. 435 436* A complete 40 byte or abbreviated commit SHA-1 in hex. 437 438* Any valid Git SHA-1 expression that resolves to a commit. See 439 ``SPECIFYING REVISIONS'' in linkgit:git-rev-parse[1] for details. 440 441The special case of restarting an incremental import from the 442current branch value should be written as: 443---- 444 from refs/heads/branch^0 445---- 446The `{caret}0` suffix is necessary as fast-import does not permit a branch to 447start from itself, and the branch is created in memory before the 448`from` command is even read from the input. Adding `{caret}0` will force 449fast-import to resolve the commit through Git's revision parsing library, 450rather than its internal branch table, thereby loading in the 451existing value of the branch. 452 453`merge` 454^^^^^^^ 455Includes one additional ancestor commit. If the `from` command is 456omitted when creating a new branch, the first `merge` commit will be 457the first ancestor of the current commit, and the branch will start 458out with no files. An unlimited number of `merge` commands per 459commit are permitted by fast-import, thereby establishing an n-way merge. 460However Git's other tools never create commits with more than 15 461additional ancestors (forming a 16-way merge). For this reason 462it is suggested that frontends do not use more than 15 `merge` 463commands per commit; 16, if starting a new, empty branch. 464 465Here `<committish>` is any of the commit specification expressions 466also accepted by `from` (see above). 467 468`filemodify` 469^^^^^^^^^^^^ 470Included in a `commit` command to add a new file or change the 471content of an existing file. This command has two different means 472of specifying the content of the file. 473 474External data format:: 475 The data content for the file was already supplied by a prior 476 `blob` command. The frontend just needs to connect it. 477+ 478.... 479 'M' SP <mode> SP <dataref> SP <path> LF 480.... 481+ 482Here `<dataref>` can be either a mark reference (`:<idnum>`) 483set by a prior `blob` command, or a full 40-byte SHA-1 of an 484existing Git blob object. 485 486Inline data format:: 487 The data content for the file has not been supplied yet. 488 The frontend wants to supply it as part of this modify 489 command. 490+ 491.... 492 'M' SP <mode> SP 'inline' SP <path> LF 493 data 494.... 495+ 496See below for a detailed description of the `data` command. 497 498In both formats `<mode>` is the type of file entry, specified 499in octal. Git only supports the following modes: 500 501* `100644` or `644`: A normal (not-executable) file. The majority 502 of files in most projects use this mode. If in doubt, this is 503 what you want. 504* `100755` or `755`: A normal, but executable, file. 505* `120000`: A symlink, the content of the file will be the link target. 506* `160000`: A gitlink, SHA-1 of the object refers to a commit in 507 another repository. Git links can only be specified by SHA or through 508 a commit mark. They are used to implement submodules. 509 510In both formats `<path>` is the complete path of the file to be added 511(if not already existing) or modified (if already existing). 512 513A `<path>` string must use UNIX-style directory separators (forward 514slash `/`), may contain any byte other than `LF`, and must not 515start with double quote (`"`). 516 517If an `LF` or double quote must be encoded into `<path>` shell-style 518quoting should be used, e.g. `"path/with\n and \" in it"`. 519 520The value of `<path>` must be in canonical form. That is it must not: 521 522* contain an empty directory component (e.g. `foo//bar` is invalid), 523* end with a directory separator (e.g. `foo/` is invalid), 524* start with a directory separator (e.g. `/foo` is invalid), 525* contain the special component `.` or `..` (e.g. `foo/./bar` and 526 `foo/../bar` are invalid). 527 528It is recommended that `<path>` always be encoded using UTF-8. 529 530`filedelete` 531^^^^^^^^^^^^ 532Included in a `commit` command to remove a file or recursively 533delete an entire directory from the branch. If the file or directory 534removal makes its parent directory empty, the parent directory will 535be automatically removed too. This cascades up the tree until the 536first non-empty directory or the root is reached. 537 538.... 539 'D' SP <path> LF 540.... 541 542here `<path>` is the complete path of the file or subdirectory to 543be removed from the branch. 544See `filemodify` above for a detailed description of `<path>`. 545 546`filecopy` 547^^^^^^^^^^^^ 548Recursively copies an existing file or subdirectory to a different 549location within the branch. The existing file or directory must 550exist. If the destination exists it will be completely replaced 551by the content copied from the source. 552 553.... 554 'C' SP <path> SP <path> LF 555.... 556 557here the first `<path>` is the source location and the second 558`<path>` is the destination. See `filemodify` above for a detailed 559description of what `<path>` may look like. To use a source path 560that contains SP the path must be quoted. 561 562A `filecopy` command takes effect immediately. Once the source 563location has been copied to the destination any future commands 564applied to the source location will not impact the destination of 565the copy. 566 567`filerename` 568^^^^^^^^^^^^ 569Renames an existing file or subdirectory to a different location 570within the branch. The existing file or directory must exist. If 571the destination exists it will be replaced by the source directory. 572 573.... 574 'R' SP <path> SP <path> LF 575.... 576 577here the first `<path>` is the source location and the second 578`<path>` is the destination. See `filemodify` above for a detailed 579description of what `<path>` may look like. To use a source path 580that contains SP the path must be quoted. 581 582A `filerename` command takes effect immediately. Once the source 583location has been renamed to the destination any future commands 584applied to the source location will create new files there and not 585impact the destination of the rename. 586 587Note that a `filerename` is the same as a `filecopy` followed by a 588`filedelete` of the source location. There is a slight performance 589advantage to using `filerename`, but the advantage is so small 590that it is never worth trying to convert a delete/add pair in 591source material into a rename for fast-import. This `filerename` 592command is provided just to simplify frontends that already have 593rename information and don't want bother with decomposing it into a 594`filecopy` followed by a `filedelete`. 595 596`filedeleteall` 597^^^^^^^^^^^^^^^ 598Included in a `commit` command to remove all files (and also all 599directories) from the branch. This command resets the internal 600branch structure to have no files in it, allowing the frontend 601to subsequently add all interesting files from scratch. 602 603.... 604 'deleteall' LF 605.... 606 607This command is extremely useful if the frontend does not know 608(or does not care to know) what files are currently on the branch, 609and therefore cannot generate the proper `filedelete` commands to 610update the content. 611 612Issuing a `filedeleteall` followed by the needed `filemodify` 613commands to set the correct content will produce the same results 614as sending only the needed `filemodify` and `filedelete` commands. 615The `filedeleteall` approach may however require fast-import to use slightly 616more memory per active branch (less than 1 MiB for even most large 617projects); so frontends that can easily obtain only the affected 618paths for a commit are encouraged to do so. 619 620`notemodify` 621^^^^^^^^^^^^ 622Included in a `commit` command to add a new note (annotating a given 623commit) or change the content of an existing note. This command has 624two different means of specifying the content of the note. 625 626External data format:: 627 The data content for the note was already supplied by a prior 628 `blob` command. The frontend just needs to connect it to the 629 commit that is to be annotated. 630+ 631.... 632 'N' SP <dataref> SP <committish> LF 633.... 634+ 635Here `<dataref>` can be either a mark reference (`:<idnum>`) 636set by a prior `blob` command, or a full 40-byte SHA-1 of an 637existing Git blob object. 638 639Inline data format:: 640 The data content for the note has not been supplied yet. 641 The frontend wants to supply it as part of this modify 642 command. 643+ 644.... 645 'N' SP 'inline' SP <committish> LF 646 data 647.... 648+ 649See below for a detailed description of the `data` command. 650 651In both formats `<committish>` is any of the commit specification 652expressions also accepted by `from` (see above). 653 654`mark` 655~~~~~~ 656Arranges for fast-import to save a reference to the current object, allowing 657the frontend to recall this object at a future point in time, without 658knowing its SHA-1. Here the current object is the object creation 659command the `mark` command appears within. This can be `commit`, 660`tag`, and `blob`, but `commit` is the most common usage. 661 662.... 663 'mark' SP ':' <idnum> LF 664.... 665 666where `<idnum>` is the number assigned by the frontend to this mark. 667The value of `<idnum>` is expressed as an ASCII decimal integer. 668The value 0 is reserved and cannot be used as 669a mark. Only values greater than or equal to 1 may be used as marks. 670 671New marks are created automatically. Existing marks can be moved 672to another object simply by reusing the same `<idnum>` in another 673`mark` command. 674 675`tag` 676~~~~~ 677Creates an annotated tag referring to a specific commit. To create 678lightweight (non-annotated) tags see the `reset` command below. 679 680.... 681 'tag' SP <name> LF 682 'from' SP <committish> LF 683 'tagger' (SP <name>)? SP LT <email> GT SP <when> LF 684 data 685.... 686 687where `<name>` is the name of the tag to create. 688 689Tag names are automatically prefixed with `refs/tags/` when stored 690in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would 691use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the 692corresponding ref as `refs/tags/RELENG-1_0-FINAL`. 693 694The value of `<name>` must be a valid refname in Git and therefore 695may contain forward slashes. As `LF` is not valid in a Git refname, 696no quoting or escaping syntax is supported here. 697 698The `from` command is the same as in the `commit` command; see 699above for details. 700 701The `tagger` command uses the same format as `committer` within 702`commit`; again see above for details. 703 704The `data` command following `tagger` must supply the annotated tag 705message (see below for `data` command syntax). To import an empty 706tag message use a 0 length data. Tag messages are free-form and are 707not interpreted by Git. Currently they must be encoded in UTF-8, 708as fast-import does not permit other encodings to be specified. 709 710Signing annotated tags during import from within fast-import is not 711supported. Trying to include your own PGP/GPG signature is not 712recommended, as the frontend does not (easily) have access to the 713complete set of bytes which normally goes into such a signature. 714If signing is required, create lightweight tags from within fast-import with 715`reset`, then create the annotated versions of those tags offline 716with the standard 'git-tag' process. 717 718`reset` 719~~~~~~~ 720Creates (or recreates) the named branch, optionally starting from 721a specific revision. The reset command allows a frontend to issue 722a new `from` command for an existing branch, or to create a new 723branch from an existing commit without creating a new commit. 724 725.... 726 'reset' SP <ref> LF 727 ('from' SP <committish> LF)? 728 LF? 729.... 730 731For a detailed description of `<ref>` and `<committish>` see above 732under `commit` and `from`. 733 734The `LF` after the command is optional (it used to be required). 735 736The `reset` command can also be used to create lightweight 737(non-annotated) tags. For example: 738 739==== 740 reset refs/tags/938 741 from :938 742==== 743 744would create the lightweight tag `refs/tags/938` referring to 745whatever commit mark `:938` references. 746 747`blob` 748~~~~~~ 749Requests writing one file revision to the packfile. The revision 750is not connected to any commit; this connection must be formed in 751a subsequent `commit` command by referencing the blob through an 752assigned mark. 753 754.... 755 'blob' LF 756 mark? 757 data 758.... 759 760The mark command is optional here as some frontends have chosen 761to generate the Git SHA-1 for the blob on their own, and feed that 762directly to `commit`. This is typically more work than its worth 763however, as marks are inexpensive to store and easy to use. 764 765`data` 766~~~~~~ 767Supplies raw data (for use as blob/file content, commit messages, or 768annotated tag messages) to fast-import. Data can be supplied using an exact 769byte count or delimited with a terminating line. Real frontends 770intended for production-quality conversions should always use the 771exact byte count format, as it is more robust and performs better. 772The delimited format is intended primarily for testing fast-import. 773 774Comment lines appearing within the `<raw>` part of `data` commands 775are always taken to be part of the body of the data and are therefore 776never ignored by fast-import. This makes it safe to import any 777file/message content whose lines might start with `#`. 778 779Exact byte count format:: 780 The frontend must specify the number of bytes of data. 781+ 782.... 783 'data' SP <count> LF 784 <raw> LF? 785.... 786+ 787where `<count>` is the exact number of bytes appearing within 788`<raw>`. The value of `<count>` is expressed as an ASCII decimal 789integer. The `LF` on either side of `<raw>` is not 790included in `<count>` and will not be included in the imported data. 791+ 792The `LF` after `<raw>` is optional (it used to be required) but 793recommended. Always including it makes debugging a fast-import 794stream easier as the next command always starts in column 0 795of the next line, even if `<raw>` did not end with an `LF`. 796 797Delimited format:: 798 A delimiter string is used to mark the end of the data. 799 fast-import will compute the length by searching for the delimiter. 800 This format is primarily useful for testing and is not 801 recommended for real data. 802+ 803.... 804 'data' SP '<<' <delim> LF 805 <raw> LF 806 <delim> LF 807 LF? 808.... 809+ 810where `<delim>` is the chosen delimiter string. The string `<delim>` 811must not appear on a line by itself within `<raw>`, as otherwise 812fast-import will think the data ends earlier than it really does. The `LF` 813immediately trailing `<raw>` is part of `<raw>`. This is one of 814the limitations of the delimited format, it is impossible to supply 815a data chunk which does not have an LF as its last byte. 816+ 817The `LF` after `<delim> LF` is optional (it used to be required). 818 819`checkpoint` 820~~~~~~~~~~~~ 821Forces fast-import to close the current packfile, start a new one, and to 822save out all current branch refs, tags and marks. 823 824.... 825 'checkpoint' LF 826 LF? 827.... 828 829Note that fast-import automatically switches packfiles when the current 830packfile reaches \--max-pack-size, or 4 GiB, whichever limit is 831smaller. During an automatic packfile switch fast-import does not update 832the branch refs, tags or marks. 833 834As a `checkpoint` can require a significant amount of CPU time and 835disk IO (to compute the overall pack SHA-1 checksum, generate the 836corresponding index file, and update the refs) it can easily take 837several minutes for a single `checkpoint` command to complete. 838 839Frontends may choose to issue checkpoints during extremely large 840and long running imports, or when they need to allow another Git 841process access to a branch. However given that a 30 GiB Subversion 842repository can be loaded into Git through fast-import in about 3 hours, 843explicit checkpointing may not be necessary. 844 845The `LF` after the command is optional (it used to be required). 846 847`progress` 848~~~~~~~~~~ 849Causes fast-import to print the entire `progress` line unmodified to 850its standard output channel (file descriptor 1) when the command is 851processed from the input stream. The command otherwise has no impact 852on the current import, or on any of fast-import's internal state. 853 854.... 855 'progress' SP <any> LF 856 LF? 857.... 858 859The `<any>` part of the command may contain any sequence of bytes 860that does not contain `LF`. The `LF` after the command is optional. 861Callers may wish to process the output through a tool such as sed to 862remove the leading part of the line, for example: 863 864==== 865 frontend | git fast-import | sed 's/^progress //' 866==== 867 868Placing a `progress` command immediately after a `checkpoint` will 869inform the reader when the `checkpoint` has been completed and it 870can safely access the refs that fast-import updated. 871 872`feature` 873~~~~~~~~~ 874Require that fast-import supports the specified feature, or abort if 875it does not. 876 877.... 878 'feature' SP <feature> LF 879.... 880 881The <feature> part of the command may be any string matching 882^[a-zA-Z][a-zA-Z-]*$ and should be understood by fast-import. 883 884Feature work identical as their option counterparts with the 885exception of the import-marks feature, see below. 886 887The following features are currently supported: 888 889* date-format 890* import-marks 891* export-marks 892* relative-marks 893* no-relative-marks 894* force 895 896The import-marks behaves differently from when it is specified as 897commandline option in that only one "feature import-marks" is allowed 898per stream. Also, any --import-marks= specified on the commandline 899will override those from the stream (if any). 900 901`option` 902~~~~~~~~ 903Processes the specified option so that git fast-import behaves in a 904way that suits the frontend's needs. 905Note that options specified by the frontend are overridden by any 906options the user may specify to git fast-import itself. 907 908.... 909 'option' SP <option> LF 910.... 911 912The `<option>` part of the command may contain any of the options 913listed in the OPTIONS section that do not change import semantics, 914without the leading '--' and is treated in the same way. 915 916Option commands must be the first commands on the input (not counting 917feature commands), to give an option command after any non-option 918command is an error. 919 920The following commandline options change import semantics and may therefore 921not be passed as option: 922 923* date-format 924* import-marks 925* export-marks 926* force 927 928Crash Reports 929------------- 930If fast-import is supplied invalid input it will terminate with a 931non-zero exit status and create a crash report in the top level of 932the Git repository it was importing into. Crash reports contain 933a snapshot of the internal fast-import state as well as the most 934recent commands that lead up to the crash. 935 936All recent commands (including stream comments, file changes and 937progress commands) are shown in the command history within the crash 938report, but raw file data and commit messages are excluded from the 939crash report. This exclusion saves space within the report file 940and reduces the amount of buffering that fast-import must perform 941during execution. 942 943After writing a crash report fast-import will close the current 944packfile and export the marks table. This allows the frontend 945developer to inspect the repository state and resume the import from 946the point where it crashed. The modified branches and tags are not 947updated during a crash, as the import did not complete successfully. 948Branch and tag information can be found in the crash report and 949must be applied manually if the update is needed. 950 951An example crash: 952 953==== 954 $ cat >in <<END_OF_INPUT 955 # my very first test commit 956 commit refs/heads/master 957 committer Shawn O. Pearce <spearce> 19283 -0400 958 # who is that guy anyway? 959 data <<EOF 960 this is my commit 961 EOF 962 M 644 inline .gitignore 963 data <<EOF 964 .gitignore 965 EOF 966 M 777 inline bob 967 END_OF_INPUT 968 969 $ git fast-import <in 970 fatal: Corrupt mode: M 777 inline bob 971 fast-import: dumping crash report to .git/fast_import_crash_8434 972 973 $ cat .git/fast_import_crash_8434 974 fast-import crash report: 975 fast-import process: 8434 976 parent process : 1391 977 at Sat Sep 1 00:58:12 2007 978 979 fatal: Corrupt mode: M 777 inline bob 980 981 Most Recent Commands Before Crash 982 --------------------------------- 983 # my very first test commit 984 commit refs/heads/master 985 committer Shawn O. Pearce <spearce> 19283 -0400 986 # who is that guy anyway? 987 data <<EOF 988 M 644 inline .gitignore 989 data <<EOF 990 * M 777 inline bob 991 992 Active Branch LRU 993 ----------------- 994 active_branches = 1 cur, 5 max 995 996 pos clock name 997 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 998 1) 0 refs/heads/master 9991000 Inactive Branches1001 -----------------1002 refs/heads/master:1003 status : active loaded dirty1004 tip commit : 00000000000000000000000000000000000000001005 old tree : 00000000000000000000000000000000000000001006 cur tree : 00000000000000000000000000000000000000001007 commit clock: 01008 last pack :100910101011 -------------------1012 END OF CRASH REPORT1013====10141015Tips and Tricks1016---------------1017The following tips and tricks have been collected from various1018users of fast-import, and are offered here as suggestions.10191020Use One Mark Per Commit1021~~~~~~~~~~~~~~~~~~~~~~~1022When doing a repository conversion, use a unique mark per commit1023(`mark :<n>`) and supply the \--export-marks option on the command1024line. fast-import will dump a file which lists every mark and the Git1025object SHA-1 that corresponds to it. If the frontend can tie1026the marks back to the source repository, it is easy to verify the1027accuracy and completeness of the import by comparing each Git1028commit to the corresponding source revision.10291030Coming from a system such as Perforce or Subversion this should be1031quite simple, as the fast-import mark can also be the Perforce changeset1032number or the Subversion revision number.10331034Freely Skip Around Branches1035~~~~~~~~~~~~~~~~~~~~~~~~~~~1036Don't bother trying to optimize the frontend to stick to one branch1037at a time during an import. Although doing so might be slightly1038faster for fast-import, it tends to increase the complexity of the frontend1039code considerably.10401041The branch LRU builtin to fast-import tends to behave very well, and the1042cost of activating an inactive branch is so low that bouncing around1043between branches has virtually no impact on import performance.10441045Handling Renames1046~~~~~~~~~~~~~~~~1047When importing a renamed file or directory, simply delete the old1048name(s) and modify the new name(s) during the corresponding commit.1049Git performs rename detection after-the-fact, rather than explicitly1050during a commit.10511052Use Tag Fixup Branches1053~~~~~~~~~~~~~~~~~~~~~~1054Some other SCM systems let the user create a tag from multiple1055files which are not from the same commit/changeset. Or to create1056tags which are a subset of the files available in the repository.10571058Importing these tags as-is in Git is impossible without making at1059least one commit which ``fixes up'' the files to match the content1060of the tag. Use fast-import's `reset` command to reset a dummy branch1061outside of your normal branch space to the base commit for the tag,1062then commit one or more file fixup commits, and finally tag the1063dummy branch.10641065For example since all normal branches are stored under `refs/heads/`1066name the tag fixup branch `TAG_FIXUP`. This way it is impossible for1067the fixup branch used by the importer to have namespace conflicts1068with real branches imported from the source (the name `TAG_FIXUP`1069is not `refs/heads/TAG_FIXUP`).10701071When committing fixups, consider using `merge` to connect the1072commit(s) which are supplying file revisions to the fixup branch.1073Doing so will allow tools such as 'git-blame' to track1074through the real commit history and properly annotate the source1075files.10761077After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP`1078to remove the dummy branch.10791080Import Now, Repack Later1081~~~~~~~~~~~~~~~~~~~~~~~~1082As soon as fast-import completes the Git repository is completely valid1083and ready for use. Typically this takes only a very short time,1084even for considerably large projects (100,000+ commits).10851086However repacking the repository is necessary to improve data1087locality and access performance. It can also take hours on extremely1088large projects (especially if -f and a large \--window parameter is1089used). Since repacking is safe to run alongside readers and writers,1090run the repack in the background and let it finish when it finishes.1091There is no reason to wait to explore your new Git project!10921093If you choose to wait for the repack, don't try to run benchmarks1094or performance tests until repacking is completed. fast-import outputs1095suboptimal packfiles that are simply never seen in real use1096situations.10971098Repacking Historical Data1099~~~~~~~~~~~~~~~~~~~~~~~~~1100If you are repacking very old imported data (e.g. older than the1101last year), consider expending some extra CPU time and supplying1102\--window=50 (or higher) when you run 'git-repack'.1103This will take longer, but will also produce a smaller packfile.1104You only need to expend the effort once, and everyone using your1105project will benefit from the smaller repository.11061107Include Some Progress Messages1108~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1109Every once in a while have your frontend emit a `progress` message1110to fast-import. The contents of the messages are entirely free-form,1111so one suggestion would be to output the current month and year1112each time the current commit date moves into the next month.1113Your users will feel better knowing how much of the data stream1114has been processed.111511161117Packfile Optimization1118---------------------1119When packing a blob fast-import always attempts to deltify against the last1120blob written. Unless specifically arranged for by the frontend,1121this will probably not be a prior version of the same file, so the1122generated delta will not be the smallest possible. The resulting1123packfile will be compressed, but will not be optimal.11241125Frontends which have efficient access to all revisions of a1126single file (for example reading an RCS/CVS ,v file) can choose1127to supply all revisions of that file as a sequence of consecutive1128`blob` commands. This allows fast-import to deltify the different file1129revisions against each other, saving space in the final packfile.1130Marks can be used to later identify individual file revisions during1131a sequence of `commit` commands.11321133The packfile(s) created by fast-import do not encourage good disk access1134patterns. This is caused by fast-import writing the data in the order1135it is received on standard input, while Git typically organizes1136data within packfiles to make the most recent (current tip) data1137appear before historical data. Git also clusters commits together,1138speeding up revision traversal through better cache locality.11391140For this reason it is strongly recommended that users repack the1141repository with `git repack -a -d` after fast-import completes, allowing1142Git to reorganize the packfiles for faster data access. If blob1143deltas are suboptimal (see above) then also adding the `-f` option1144to force recomputation of all deltas can significantly reduce the1145final packfile size (30-50% smaller can be quite typical).114611471148Memory Utilization1149------------------1150There are a number of factors which affect how much memory fast-import1151requires to perform an import. Like critical sections of core1152Git, fast-import uses its own memory allocators to amortize any overheads1153associated with malloc. In practice fast-import tends to amortize any1154malloc overheads to 0, due to its use of large block allocations.11551156per object1157~~~~~~~~~~1158fast-import maintains an in-memory structure for every object written in1159this execution. On a 32 bit system the structure is 32 bytes,1160on a 64 bit system the structure is 40 bytes (due to the larger1161pointer sizes). Objects in the table are not deallocated until1162fast-import terminates. Importing 2 million objects on a 32 bit system1163will require approximately 64 MiB of memory.11641165The object table is actually a hashtable keyed on the object name1166(the unique SHA-1). This storage configuration allows fast-import to reuse1167an existing or already written object and avoid writing duplicates1168to the output packfile. Duplicate blobs are surprisingly common1169in an import, typically due to branch merges in the source.11701171per mark1172~~~~~~~~1173Marks are stored in a sparse array, using 1 pointer (4 bytes or 81174bytes, depending on pointer size) per mark. Although the array1175is sparse, frontends are still strongly encouraged to use marks1176between 1 and n, where n is the total number of marks required for1177this import.11781179per branch1180~~~~~~~~~~1181Branches are classified as active and inactive. The memory usage1182of the two classes is significantly different.11831184Inactive branches are stored in a structure which uses 96 or 1201185bytes (32 bit or 64 bit systems, respectively), plus the length of1186the branch name (typically under 200 bytes), per branch. fast-import will1187easily handle as many as 10,000 inactive branches in under 2 MiB1188of memory.11891190Active branches have the same overhead as inactive branches, but1191also contain copies of every tree that has been recently modified on1192that branch. If subtree `include` has not been modified since the1193branch became active, its contents will not be loaded into memory,1194but if subtree `src` has been modified by a commit since the branch1195became active, then its contents will be loaded in memory.11961197As active branches store metadata about the files contained on that1198branch, their in-memory storage size can grow to a considerable size1199(see below).12001201fast-import automatically moves active branches to inactive status based on1202a simple least-recently-used algorithm. The LRU chain is updated on1203each `commit` command. The maximum number of active branches can be1204increased or decreased on the command line with \--active-branches=.12051206per active tree1207~~~~~~~~~~~~~~~1208Trees (aka directories) use just 12 bytes of memory on top of the1209memory required for their entries (see ``per active file'' below).1210The cost of a tree is virtually 0, as its overhead amortizes out1211over the individual file entries.12121213per active file entry1214~~~~~~~~~~~~~~~~~~~~~1215Files (and pointers to subtrees) within active trees require 52 or 641216bytes (32/64 bit platforms) per entry. To conserve space, file and1217tree names are pooled in a common string table, allowing the filename1218``Makefile'' to use just 16 bytes (after including the string header1219overhead) no matter how many times it occurs within the project.12201221The active branch LRU, when coupled with the filename string pool1222and lazy loading of subtrees, allows fast-import to efficiently import1223projects with 2,000+ branches and 45,114+ files in a very limited1224memory footprint (less than 2.7 MiB per active branch).122512261227Author1228------1229Written by Shawn O. Pearce <spearce@spearce.org>.12301231Documentation1232--------------1233Documentation by Shawn O. Pearce <spearce@spearce.org>.12341235GIT1236---1237Part of the linkgit:git[1] suite