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 gitlink:git-init[1]) 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--export-pack-edges=<file>:: 79 After creating a packfile, print a line of data to 80 <file> listing the filename of the packfile and the last 81 commit on each branch that was written to that packfile. 82 This information may be useful after importing projects 83 whose total object set exceeds the 4 GiB packfile limit, 84 as these commits can be used as edge points during calls 85 to gitlink:git-pack-objects[1]. 86 87--quiet:: 88 Disable all non-fatal output, making fast-import silent when it 89 is successful. This option disables the output shown by 90 \--stats. 91 92--stats:: 93 Display some basic statistics about the objects fast-import has 94 created, the packfiles they were stored into, and the 95 memory used by fast-import during this run. Showing this output 96 is currently the default, but can be disabled with \--quiet. 97 98 99Performance 100----------- 101The design of fast-import allows it to import large projects in a minimum 102amount of memory usage and processing time. Assuming the frontend 103is able to keep up with fast-import and feed it a constant stream of data, 104import times for projects holding 10+ years of history and containing 105100,000+ individual commits are generally completed in just 1-2 106hours on quite modest (~$2,000 USD) hardware. 107 108Most bottlenecks appear to be in foreign source data access (the 109source just cannot extract revisions fast enough) or disk IO (fast-import 110writes as fast as the disk will take the data). Imports will run 111faster if the source data is stored on a different drive than the 112destination Git repository (due to less IO contention). 113 114 115Development Cost 116---------------- 117A typical frontend for fast-import tends to weigh in at approximately 200 118lines of Perl/Python/Ruby code. Most developers have been able to 119create working importers in just a couple of hours, even though it 120is their first exposure to fast-import, and sometimes even to Git. This is 121an ideal situation, given that most conversion tools are throw-away 122(use once, and never look back). 123 124 125Parallel Operation 126------------------ 127Like `git-push` or `git-fetch`, imports handled by fast-import are safe to 128run alongside parallel `git repack -a -d` or `git gc` invocations, 129or any other Git operation (including `git prune`, as loose objects 130are never used by fast-import). 131 132fast-import does not lock the branch or tag refs it is actively importing. 133After the import, during its ref update phase, fast-import tests each 134existing branch ref to verify the update will be a fast-forward 135update (the commit stored in the ref is contained in the new 136history of the commit to be written). If the update is not a 137fast-forward update, fast-import will skip updating that ref and instead 138prints a warning message. fast-import will always attempt to update all 139branch refs, and does not stop on the first failure. 140 141Branch updates can be forced with \--force, but its recommended that 142this only be used on an otherwise quiet repository. Using \--force 143is not necessary for an initial import into an empty repository. 144 145 146Technical Discussion 147-------------------- 148fast-import tracks a set of branches in memory. Any branch can be created 149or modified at any point during the import process by sending a 150`commit` command on the input stream. This design allows a frontend 151program to process an unlimited number of branches simultaneously, 152generating commits in the order they are available from the source 153data. It also simplifies the frontend programs considerably. 154 155fast-import does not use or alter the current working directory, or any 156file within it. (It does however update the current Git repository, 157as referenced by `GIT_DIR`.) Therefore an import frontend may use 158the working directory for its own purposes, such as extracting file 159revisions from the foreign source. This ignorance of the working 160directory also allows fast-import to run very quickly, as it does not 161need to perform any costly file update operations when switching 162between branches. 163 164Input Format 165------------ 166With the exception of raw file data (which Git does not interpret) 167the fast-import input format is text (ASCII) based. This text based 168format simplifies development and debugging of frontend programs, 169especially when a higher level language such as Perl, Python or 170Ruby is being used. 171 172fast-import is very strict about its input. Where we say SP below we mean 173*exactly* one space. Likewise LF means one (and only one) linefeed. 174Supplying additional whitespace characters will cause unexpected 175results, such as branch names or file names with leading or trailing 176spaces in their name, or early termination of fast-import when it encounters 177unexpected input. 178 179Stream Comments 180~~~~~~~~~~~~~~~ 181To aid in debugging frontends fast-import ignores any line that 182begins with `#` (ASCII pound/hash) up to and including the line 183ending `LF`. A comment line may contain any sequence of bytes 184that does not contain an LF and therefore may be used to include 185any detailed debugging information that might be specific to the 186frontend and useful when inspecting a fast-import data stream. 187 188Date Formats 189~~~~~~~~~~~~ 190The following date formats are supported. A frontend should select 191the format it will use for this import by passing the format name 192in the \--date-format=<fmt> command line option. 193 194`raw`:: 195 This is the Git native format and is `<time> SP <offutc>`. 196 It is also fast-import's default format, if \--date-format was 197 not specified. 198+ 199The time of the event is specified by `<time>` as the number of 200seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is 201written as an ASCII decimal integer. 202+ 203The local offset is specified by `<offutc>` as a positive or negative 204offset from UTC. For example EST (which is 5 hours behind UTC) 205would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''. 206The local offset does not affect `<time>`; it is used only as an 207advisement to help formatting routines display the timestamp. 208+ 209If the local offset is not available in the source material, use 210``+0000'', or the most common local offset. For example many 211organizations have a CVS repository which has only ever been accessed 212by users who are located in the same location and timezone. In this 213case a reasonable offset from UTC could be assumed. 214+ 215Unlike the `rfc2822` format, this format is very strict. Any 216variation in formatting will cause fast-import to reject the value. 217 218`rfc2822`:: 219 This is the standard email format as described by RFC 2822. 220+ 221An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git 222parser is accurate, but a little on the lenient side. It is the 223same parser used by gitlink:git-am[1] when applying patches 224received from email. 225+ 226Some malformed strings may be accepted as valid dates. In some of 227these cases Git will still be able to obtain the correct date from 228the malformed string. There are also some types of malformed 229strings which Git will parse wrong, and yet consider valid. 230Seriously malformed strings will be rejected. 231+ 232Unlike the `raw` format above, the timezone/UTC offset information 233contained in an RFC 2822 date string is used to adjust the date 234value to UTC prior to storage. Therefore it is important that 235this information be as accurate as possible. 236+ 237If the source material uses RFC 2822 style dates, 238the frontend should let fast-import handle the parsing and conversion 239(rather than attempting to do it itself) as the Git parser has 240been well tested in the wild. 241+ 242Frontends should prefer the `raw` format if the source material 243already uses UNIX-epoch format, can be coaxed to give dates in that 244format, or its format is easiliy convertible to it, as there is no 245ambiguity in parsing. 246 247`now`:: 248 Always use the current time and timezone. The literal 249 `now` must always be supplied for `<when>`. 250+ 251This is a toy format. The current time and timezone of this system 252is always copied into the identity string at the time it is being 253created by fast-import. There is no way to specify a different time or 254timezone. 255+ 256This particular format is supplied as its short to implement and 257may be useful to a process that wants to create a new commit 258right now, without needing to use a working directory or 259gitlink:git-update-index[1]. 260+ 261If separate `author` and `committer` commands are used in a `commit` 262the timestamps may not match, as the system clock will be polled 263twice (once for each command). The only way to ensure that both 264author and committer identity information has the same timestamp 265is to omit `author` (thus copying from `committer`) or to use a 266date format other than `now`. 267 268Commands 269~~~~~~~~ 270fast-import accepts several commands to update the current repository 271and control the current import process. More detailed discussion 272(with examples) of each command follows later. 273 274`commit`:: 275 Creates a new branch or updates an existing branch by 276 creating a new commit and updating the branch to point at 277 the newly created commit. 278 279`tag`:: 280 Creates an annotated tag object from an existing commit or 281 branch. Lightweight tags are not supported by this command, 282 as they are not recommended for recording meaningful points 283 in time. 284 285`reset`:: 286 Reset an existing branch (or a new branch) to a specific 287 revision. This command must be used to change a branch to 288 a specific revision without making a commit on it. 289 290`blob`:: 291 Convert raw file data into a blob, for future use in a 292 `commit` command. This command is optional and is not 293 needed to perform an import. 294 295`checkpoint`:: 296 Forces fast-import to close the current packfile, generate its 297 unique SHA-1 checksum and index, and start a new packfile. 298 This command is optional and is not needed to perform 299 an import. 300 301`commit` 302~~~~~~~~ 303Create or update a branch with a new commit, recording one logical 304change to the project. 305 306.... 307 'commit' SP <ref> LF 308 mark? 309 ('author' SP <name> SP LT <email> GT SP <when> LF)? 310 'committer' SP <name> SP LT <email> GT SP <when> LF 311 data 312 ('from' SP <committish> LF)? 313 ('merge' SP <committish> LF)? 314 (filemodify | filedelete | filecopy | filerename | filedeleteall)* 315 LF 316.... 317 318where `<ref>` is the name of the branch to make the commit on. 319Typically branch names are prefixed with `refs/heads/` in 320Git, so importing the CVS branch symbol `RELENG-1_0` would use 321`refs/heads/RELENG-1_0` for the value of `<ref>`. The value of 322`<ref>` must be a valid refname in Git. As `LF` is not valid in 323a Git refname, no quoting or escaping syntax is supported here. 324 325A `mark` command may optionally appear, requesting fast-import to save a 326reference to the newly created commit for future use by the frontend 327(see below for format). It is very common for frontends to mark 328every commit they create, thereby allowing future branch creation 329from any imported commit. 330 331The `data` command following `committer` must supply the commit 332message (see below for `data` command syntax). To import an empty 333commit message use a 0 length data. Commit messages are free-form 334and are not interpreted by Git. Currently they must be encoded in 335UTF-8, as fast-import does not permit other encodings to be specified. 336 337Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename` 338and `filedeleteall` commands 339may be included to update the contents of the branch prior to 340creating the commit. These commands may be supplied in any order. 341However it is recommended that a `filedeleteall` command preceed 342all `filemodify`, `filecopy` and `filerename` commands in the same 343commit, as `filedeleteall` 344wipes the branch clean (see below). 345 346`author` 347^^^^^^^^ 348An `author` command may optionally appear, if the author information 349might differ from the committer information. If `author` is omitted 350then fast-import will automatically use the committer's information for 351the author portion of the commit. See below for a description of 352the fields in `author`, as they are identical to `committer`. 353 354`committer` 355^^^^^^^^^^^ 356The `committer` command indicates who made this commit, and when 357they made it. 358 359Here `<name>` is the person's display name (for example 360``Com M Itter'') and `<email>` is the person's email address 361(``cm@example.com''). `LT` and `GT` are the literal less-than (\x3c) 362and greater-than (\x3e) symbols. These are required to delimit 363the email address from the other fields in the line. Note that 364`<name>` is free-form and may contain any sequence of bytes, except 365`LT` and `LF`. It is typically UTF-8 encoded. 366 367The time of the change is specified by `<when>` using the date format 368that was selected by the \--date-format=<fmt> command line option. 369See ``Date Formats'' above for the set of supported formats, and 370their syntax. 371 372`from` 373^^^^^^ 374The `from` command is used to specify the commit to initialize 375this branch from. This revision will be the first ancestor of the 376new commit. 377 378Omitting the `from` command in the first commit of a new branch 379will cause fast-import to create that commit with no ancestor. This 380tends to be desired only for the initial commit of a project. 381Omitting the `from` command on existing branches is usually desired, 382as the current commit on that branch is automatically assumed to 383be the first ancestor of the new commit. 384 385As `LF` is not valid in a Git refname or SHA-1 expression, no 386quoting or escaping syntax is supported within `<committish>`. 387 388Here `<committish>` is any of the following: 389 390* The name of an existing branch already in fast-import's internal branch 391 table. If fast-import doesn't know the name, its treated as a SHA-1 392 expression. 393 394* A mark reference, `:<idnum>`, where `<idnum>` is the mark number. 395+ 396The reason fast-import uses `:` to denote a mark reference is this character 397is not legal in a Git branch name. The leading `:` makes it easy 398to distingush between the mark 42 (`:42`) and the branch 42 (`42` 399or `refs/heads/42`), or an abbreviated SHA-1 which happened to 400consist only of base-10 digits. 401+ 402Marks must be declared (via `mark`) before they can be used. 403 404* A complete 40 byte or abbreviated commit SHA-1 in hex. 405 406* Any valid Git SHA-1 expression that resolves to a commit. See 407 ``SPECIFYING REVISIONS'' in gitlink:git-rev-parse[1] for details. 408 409The special case of restarting an incremental import from the 410current branch value should be written as: 411---- 412 from refs/heads/branch^0 413---- 414The `{caret}0` suffix is necessary as fast-import does not permit a branch to 415start from itself, and the branch is created in memory before the 416`from` command is even read from the input. Adding `{caret}0` will force 417fast-import to resolve the commit through Git's revision parsing library, 418rather than its internal branch table, thereby loading in the 419existing value of the branch. 420 421`merge` 422^^^^^^^ 423Includes one additional ancestor commit, and makes the current 424commit a merge commit. An unlimited number of `merge` commands per 425commit are permitted by fast-import, thereby establishing an n-way merge. 426However Git's other tools never create commits with more than 15 427additional ancestors (forming a 16-way merge). For this reason 428it is suggested that frontends do not use more than 15 `merge` 429commands per commit. 430 431Here `<committish>` is any of the commit specification expressions 432also accepted by `from` (see above). 433 434`filemodify` 435^^^^^^^^^^^^ 436Included in a `commit` command to add a new file or change the 437content of an existing file. This command has two different means 438of specifying the content of the file. 439 440External data format:: 441 The data content for the file was already supplied by a prior 442 `blob` command. The frontend just needs to connect it. 443+ 444.... 445 'M' SP <mode> SP <dataref> SP <path> LF 446.... 447+ 448Here `<dataref>` can be either a mark reference (`:<idnum>`) 449set by a prior `blob` command, or a full 40-byte SHA-1 of an 450existing Git blob object. 451 452Inline data format:: 453 The data content for the file has not been supplied yet. 454 The frontend wants to supply it as part of this modify 455 command. 456+ 457.... 458 'M' SP <mode> SP 'inline' SP <path> LF 459 data 460.... 461+ 462See below for a detailed description of the `data` command. 463 464In both formats `<mode>` is the type of file entry, specified 465in octal. Git only supports the following modes: 466 467* `100644` or `644`: A normal (not-executable) file. The majority 468 of files in most projects use this mode. If in doubt, this is 469 what you want. 470* `100755` or `755`: A normal, but executable, file. 471* `120000`: A symlink, the content of the file will be the link target. 472 473In both formats `<path>` is the complete path of the file to be added 474(if not already existing) or modified (if already existing). 475 476A `<path>` string must use UNIX-style directory separators (forward 477slash `/`), may contain any byte other than `LF`, and must not 478start with double quote (`"`). 479 480If an `LF` or double quote must be encoded into `<path>` shell-style 481quoting should be used, e.g. `"path/with\n and \" in it"`. 482 483The value of `<path>` must be in canoncial form. That is it must not: 484 485* contain an empty directory component (e.g. `foo//bar` is invalid), 486* end with a directory separator (e.g. `foo/` is invalid), 487* start with a directory separator (e.g. `/foo` is invalid), 488* contain the special component `.` or `..` (e.g. `foo/./bar` and 489 `foo/../bar` are invalid). 490 491It is recommended that `<path>` always be encoded using UTF-8. 492 493`filedelete` 494^^^^^^^^^^^^ 495Included in a `commit` command to remove a file or recursively 496delete an entire directory from the branch. If the file or directory 497removal makes its parent directory empty, the parent directory will 498be automatically removed too. This cascades up the tree until the 499first non-empty directory or the root is reached. 500 501.... 502 'D' SP <path> LF 503.... 504 505here `<path>` is the complete path of the file or subdirectory to 506be removed from the branch. 507See `filemodify` above for a detailed description of `<path>`. 508 509`filecopy` 510^^^^^^^^^^^^ 511Recursively copies an existing file or subdirectory to a different 512location within the branch. The existing file or directory must 513exist. If the destination exists it will be completely replaced 514by the content copied from the source. 515 516.... 517 'C' SP <path> SP <path> LF 518.... 519 520here the first `<path>` is the source location and the second 521`<path>` is the destination. See `filemodify` above for a detailed 522description of what `<path>` may look like. To use a source path 523that contains SP the path must be quoted. 524 525A `filecopy` command takes effect immediately. Once the source 526location has been copied to the destination any future commands 527applied to the source location will not impact the destination of 528the copy. 529 530`filerename` 531^^^^^^^^^^^^ 532Renames an existing file or subdirectory to a different location 533within the branch. The existing file or directory must exist. If 534the destination exists it will be replaced by the source directory. 535 536.... 537 'R' SP <path> SP <path> LF 538.... 539 540here the first `<path>` is the source location and the second 541`<path>` is the destination. See `filemodify` above for a detailed 542description of what `<path>` may look like. To use a source path 543that contains SP the path must be quoted. 544 545A `filerename` command takes effect immediately. Once the source 546location has been renamed to the destination any future commands 547applied to the source location will create new files there and not 548impact the destination of the rename. 549 550Note that a `filerename` is the same as a `filecopy` followed by a 551`filedelete` of the source location. There is a slight performance 552advantage to using `filerename`, but the advantage is so small 553that it is never worth trying to convert a delete/add pair in 554source material into a rename for fast-import. This `filerename` 555command is provided just to simplify frontends that already have 556rename information and don't want bother with decomposing it into a 557`filecopy` followed by a `filedelete`. 558 559`filedeleteall` 560^^^^^^^^^^^^^^^ 561Included in a `commit` command to remove all files (and also all 562directories) from the branch. This command resets the internal 563branch structure to have no files in it, allowing the frontend 564to subsequently add all interesting files from scratch. 565 566.... 567 'deleteall' LF 568.... 569 570This command is extremely useful if the frontend does not know 571(or does not care to know) what files are currently on the branch, 572and therefore cannot generate the proper `filedelete` commands to 573update the content. 574 575Issuing a `filedeleteall` followed by the needed `filemodify` 576commands to set the correct content will produce the same results 577as sending only the needed `filemodify` and `filedelete` commands. 578The `filedeleteall` approach may however require fast-import to use slightly 579more memory per active branch (less than 1 MiB for even most large 580projects); so frontends that can easily obtain only the affected 581paths for a commit are encouraged to do so. 582 583`mark` 584~~~~~~ 585Arranges for fast-import to save a reference to the current object, allowing 586the frontend to recall this object at a future point in time, without 587knowing its SHA-1. Here the current object is the object creation 588command the `mark` command appears within. This can be `commit`, 589`tag`, and `blob`, but `commit` is the most common usage. 590 591.... 592 'mark' SP ':' <idnum> LF 593.... 594 595where `<idnum>` is the number assigned by the frontend to this mark. 596The value of `<idnum>` is expressed as an ASCII decimal integer. 597The value 0 is reserved and cannot be used as 598a mark. Only values greater than or equal to 1 may be used as marks. 599 600New marks are created automatically. Existing marks can be moved 601to another object simply by reusing the same `<idnum>` in another 602`mark` command. 603 604`tag` 605~~~~~ 606Creates an annotated tag referring to a specific commit. To create 607lightweight (non-annotated) tags see the `reset` command below. 608 609.... 610 'tag' SP <name> LF 611 'from' SP <committish> LF 612 'tagger' SP <name> SP LT <email> GT SP <when> LF 613 data 614.... 615 616where `<name>` is the name of the tag to create. 617 618Tag names are automatically prefixed with `refs/tags/` when stored 619in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would 620use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the 621corresponding ref as `refs/tags/RELENG-1_0-FINAL`. 622 623The value of `<name>` must be a valid refname in Git and therefore 624may contain forward slashes. As `LF` is not valid in a Git refname, 625no quoting or escaping syntax is supported here. 626 627The `from` command is the same as in the `commit` command; see 628above for details. 629 630The `tagger` command uses the same format as `committer` within 631`commit`; again see above for details. 632 633The `data` command following `tagger` must supply the annotated tag 634message (see below for `data` command syntax). To import an empty 635tag message use a 0 length data. Tag messages are free-form and are 636not interpreted by Git. Currently they must be encoded in UTF-8, 637as fast-import does not permit other encodings to be specified. 638 639Signing annotated tags during import from within fast-import is not 640supported. Trying to include your own PGP/GPG signature is not 641recommended, as the frontend does not (easily) have access to the 642complete set of bytes which normally goes into such a signature. 643If signing is required, create lightweight tags from within fast-import with 644`reset`, then create the annotated versions of those tags offline 645with the standard gitlink:git-tag[1] process. 646 647`reset` 648~~~~~~~ 649Creates (or recreates) the named branch, optionally starting from 650a specific revision. The reset command allows a frontend to issue 651a new `from` command for an existing branch, or to create a new 652branch from an existing commit without creating a new commit. 653 654.... 655 'reset' SP <ref> LF 656 ('from' SP <committish> LF)? 657 LF 658.... 659 660For a detailed description of `<ref>` and `<committish>` see above 661under `commit` and `from`. 662 663The `reset` command can also be used to create lightweight 664(non-annotated) tags. For example: 665 666==== 667 reset refs/tags/938 668 from :938 669==== 670 671would create the lightweight tag `refs/tags/938` referring to 672whatever commit mark `:938` references. 673 674`blob` 675~~~~~~ 676Requests writing one file revision to the packfile. The revision 677is not connected to any commit; this connection must be formed in 678a subsequent `commit` command by referencing the blob through an 679assigned mark. 680 681.... 682 'blob' LF 683 mark? 684 data 685.... 686 687The mark command is optional here as some frontends have chosen 688to generate the Git SHA-1 for the blob on their own, and feed that 689directly to `commit`. This is typically more work than its worth 690however, as marks are inexpensive to store and easy to use. 691 692`data` 693~~~~~~ 694Supplies raw data (for use as blob/file content, commit messages, or 695annotated tag messages) to fast-import. Data can be supplied using an exact 696byte count or delimited with a terminating line. Real frontends 697intended for production-quality conversions should always use the 698exact byte count format, as it is more robust and performs better. 699The delimited format is intended primarily for testing fast-import. 700 701Comment lines appearing within the `<raw>` part of `data` commands 702are always taken to be part of the body of the data and are therefore 703never ignored by fast-import. This makes it safe to import any 704file/message content whose lines might start with `#`. 705 706Exact byte count format:: 707 The frontend must specify the number of bytes of data. 708+ 709.... 710 'data' SP <count> LF 711 <raw> LF 712.... 713+ 714where `<count>` is the exact number of bytes appearing within 715`<raw>`. The value of `<count>` is expressed as an ASCII decimal 716integer. The `LF` on either side of `<raw>` is not 717included in `<count>` and will not be included in the imported data. 718 719Delimited format:: 720 A delimiter string is used to mark the end of the data. 721 fast-import will compute the length by searching for the delimiter. 722 This format is primarly useful for testing and is not 723 recommended for real data. 724+ 725.... 726 'data' SP '<<' <delim> LF 727 <raw> LF 728 <delim> LF 729.... 730+ 731where `<delim>` is the chosen delimiter string. The string `<delim>` 732must not appear on a line by itself within `<raw>`, as otherwise 733fast-import will think the data ends earlier than it really does. The `LF` 734immediately trailing `<raw>` is part of `<raw>`. This is one of 735the limitations of the delimited format, it is impossible to supply 736a data chunk which does not have an LF as its last byte. 737 738`checkpoint` 739~~~~~~~~~~~~ 740Forces fast-import to close the current packfile, start a new one, and to 741save out all current branch refs, tags and marks. 742 743.... 744 'checkpoint' LF 745 LF 746.... 747 748Note that fast-import automatically switches packfiles when the current 749packfile reaches \--max-pack-size, or 4 GiB, whichever limit is 750smaller. During an automatic packfile switch fast-import does not update 751the branch refs, tags or marks. 752 753As a `checkpoint` can require a significant amount of CPU time and 754disk IO (to compute the overall pack SHA-1 checksum, generate the 755corresponding index file, and update the refs) it can easily take 756several minutes for a single `checkpoint` command to complete. 757 758Frontends may choose to issue checkpoints during extremely large 759and long running imports, or when they need to allow another Git 760process access to a branch. However given that a 30 GiB Subversion 761repository can be loaded into Git through fast-import in about 3 hours, 762explicit checkpointing may not be necessary. 763 764 765Tips and Tricks 766--------------- 767The following tips and tricks have been collected from various 768users of fast-import, and are offered here as suggestions. 769 770Use One Mark Per Commit 771~~~~~~~~~~~~~~~~~~~~~~~ 772When doing a repository conversion, use a unique mark per commit 773(`mark :<n>`) and supply the \--export-marks option on the command 774line. fast-import will dump a file which lists every mark and the Git 775object SHA-1 that corresponds to it. If the frontend can tie 776the marks back to the source repository, it is easy to verify the 777accuracy and completeness of the import by comparing each Git 778commit to the corresponding source revision. 779 780Coming from a system such as Perforce or Subversion this should be 781quite simple, as the fast-import mark can also be the Perforce changeset 782number or the Subversion revision number. 783 784Freely Skip Around Branches 785~~~~~~~~~~~~~~~~~~~~~~~~~~~ 786Don't bother trying to optimize the frontend to stick to one branch 787at a time during an import. Although doing so might be slightly 788faster for fast-import, it tends to increase the complexity of the frontend 789code considerably. 790 791The branch LRU builtin to fast-import tends to behave very well, and the 792cost of activating an inactive branch is so low that bouncing around 793between branches has virtually no impact on import performance. 794 795Handling Renames 796~~~~~~~~~~~~~~~~ 797When importing a renamed file or directory, simply delete the old 798name(s) and modify the new name(s) during the corresponding commit. 799Git performs rename detection after-the-fact, rather than explicitly 800during a commit. 801 802Use Tag Fixup Branches 803~~~~~~~~~~~~~~~~~~~~~~ 804Some other SCM systems let the user create a tag from multiple 805files which are not from the same commit/changeset. Or to create 806tags which are a subset of the files available in the repository. 807 808Importing these tags as-is in Git is impossible without making at 809least one commit which ``fixes up'' the files to match the content 810of the tag. Use fast-import's `reset` command to reset a dummy branch 811outside of your normal branch space to the base commit for the tag, 812then commit one or more file fixup commits, and finally tag the 813dummy branch. 814 815For example since all normal branches are stored under `refs/heads/` 816name the tag fixup branch `TAG_FIXUP`. This way it is impossible for 817the fixup branch used by the importer to have namespace conflicts 818with real branches imported from the source (the name `TAG_FIXUP` 819is not `refs/heads/TAG_FIXUP`). 820 821When committing fixups, consider using `merge` to connect the 822commit(s) which are supplying file revisions to the fixup branch. 823Doing so will allow tools such as gitlink:git-blame[1] to track 824through the real commit history and properly annotate the source 825files. 826 827After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP` 828to remove the dummy branch. 829 830Import Now, Repack Later 831~~~~~~~~~~~~~~~~~~~~~~~~ 832As soon as fast-import completes the Git repository is completely valid 833and ready for use. Typicallly this takes only a very short time, 834even for considerably large projects (100,000+ commits). 835 836However repacking the repository is necessary to improve data 837locality and access performance. It can also take hours on extremely 838large projects (especially if -f and a large \--window parameter is 839used). Since repacking is safe to run alongside readers and writers, 840run the repack in the background and let it finish when it finishes. 841There is no reason to wait to explore your new Git project! 842 843If you choose to wait for the repack, don't try to run benchmarks 844or performance tests until repacking is completed. fast-import outputs 845suboptimal packfiles that are simply never seen in real use 846situations. 847 848Repacking Historical Data 849~~~~~~~~~~~~~~~~~~~~~~~~~ 850If you are repacking very old imported data (e.g. older than the 851last year), consider expending some extra CPU time and supplying 852\--window=50 (or higher) when you run gitlink:git-repack[1]. 853This will take longer, but will also produce a smaller packfile. 854You only need to expend the effort once, and everyone using your 855project will benefit from the smaller repository. 856 857 858Packfile Optimization 859--------------------- 860When packing a blob fast-import always attempts to deltify against the last 861blob written. Unless specifically arranged for by the frontend, 862this will probably not be a prior version of the same file, so the 863generated delta will not be the smallest possible. The resulting 864packfile will be compressed, but will not be optimal. 865 866Frontends which have efficient access to all revisions of a 867single file (for example reading an RCS/CVS ,v file) can choose 868to supply all revisions of that file as a sequence of consecutive 869`blob` commands. This allows fast-import to deltify the different file 870revisions against each other, saving space in the final packfile. 871Marks can be used to later identify individual file revisions during 872a sequence of `commit` commands. 873 874The packfile(s) created by fast-import do not encourage good disk access 875patterns. This is caused by fast-import writing the data in the order 876it is received on standard input, while Git typically organizes 877data within packfiles to make the most recent (current tip) data 878appear before historical data. Git also clusters commits together, 879speeding up revision traversal through better cache locality. 880 881For this reason it is strongly recommended that users repack the 882repository with `git repack -a -d` after fast-import completes, allowing 883Git to reorganize the packfiles for faster data access. If blob 884deltas are suboptimal (see above) then also adding the `-f` option 885to force recomputation of all deltas can significantly reduce the 886final packfile size (30-50% smaller can be quite typical). 887 888 889Memory Utilization 890------------------ 891There are a number of factors which affect how much memory fast-import 892requires to perform an import. Like critical sections of core 893Git, fast-import uses its own memory allocators to ammortize any overheads 894associated with malloc. In practice fast-import tends to ammoritize any 895malloc overheads to 0, due to its use of large block allocations. 896 897per object 898~~~~~~~~~~ 899fast-import maintains an in-memory structure for every object written in 900this execution. On a 32 bit system the structure is 32 bytes, 901on a 64 bit system the structure is 40 bytes (due to the larger 902pointer sizes). Objects in the table are not deallocated until 903fast-import terminates. Importing 2 million objects on a 32 bit system 904will require approximately 64 MiB of memory. 905 906The object table is actually a hashtable keyed on the object name 907(the unique SHA-1). This storage configuration allows fast-import to reuse 908an existing or already written object and avoid writing duplicates 909to the output packfile. Duplicate blobs are surprisingly common 910in an import, typically due to branch merges in the source. 911 912per mark 913~~~~~~~~ 914Marks are stored in a sparse array, using 1 pointer (4 bytes or 8 915bytes, depending on pointer size) per mark. Although the array 916is sparse, frontends are still strongly encouraged to use marks 917between 1 and n, where n is the total number of marks required for 918this import. 919 920per branch 921~~~~~~~~~~ 922Branches are classified as active and inactive. The memory usage 923of the two classes is significantly different. 924 925Inactive branches are stored in a structure which uses 96 or 120 926bytes (32 bit or 64 bit systems, respectively), plus the length of 927the branch name (typically under 200 bytes), per branch. fast-import will 928easily handle as many as 10,000 inactive branches in under 2 MiB 929of memory. 930 931Active branches have the same overhead as inactive branches, but 932also contain copies of every tree that has been recently modified on 933that branch. If subtree `include` has not been modified since the 934branch became active, its contents will not be loaded into memory, 935but if subtree `src` has been modified by a commit since the branch 936became active, then its contents will be loaded in memory. 937 938As active branches store metadata about the files contained on that 939branch, their in-memory storage size can grow to a considerable size 940(see below). 941 942fast-import automatically moves active branches to inactive status based on 943a simple least-recently-used algorithm. The LRU chain is updated on 944each `commit` command. The maximum number of active branches can be 945increased or decreased on the command line with \--active-branches=. 946 947per active tree 948~~~~~~~~~~~~~~~ 949Trees (aka directories) use just 12 bytes of memory on top of the 950memory required for their entries (see ``per active file'' below). 951The cost of a tree is virtually 0, as its overhead ammortizes out 952over the individual file entries. 953 954per active file entry 955~~~~~~~~~~~~~~~~~~~~~ 956Files (and pointers to subtrees) within active trees require 52 or 64 957bytes (32/64 bit platforms) per entry. To conserve space, file and 958tree names are pooled in a common string table, allowing the filename 959``Makefile'' to use just 16 bytes (after including the string header 960overhead) no matter how many times it occurs within the project. 961 962The active branch LRU, when coupled with the filename string pool 963and lazy loading of subtrees, allows fast-import to efficiently import 964projects with 2,000+ branches and 45,114+ files in a very limited 965memory footprint (less than 2.7 MiB per active branch). 966 967 968Author 969------ 970Written by Shawn O. Pearce <spearce@spearce.org>. 971 972Documentation 973-------------- 974Documentation by Shawn O. Pearce <spearce@spearce.org>. 975 976GIT 977--- 978Part of the gitlink:git[7] suite