7ab97745a60daec812a286f61b9680f4714e2c38
   1git-fast-import(1)
   2==================
   3
   4NAME
   5----
   6git-fast-import - Backend for fast Git data importers
   7
   8
   9SYNOPSIS
  10--------
  11[verse]
  12frontend | 'git fast-import' [<options>]
  13
  14DESCRIPTION
  15-----------
  16This program is usually not what the end user wants to run directly.
  17Most end users want to use one of the existing frontend programs,
  18which parses a specific type of foreign source and feeds the contents
  19stored there to 'git fast-import'.
  20
  21fast-import reads a mixed command/data stream from standard input and
  22writes one or more packfiles directly into the current repository.
  23When EOF is received on standard input, fast import writes out
  24updated branch and tag refs, fully updating the current repository
  25with the newly imported data.
  26
  27The fast-import backend itself can import into an empty repository (one that
  28has already been initialized by 'git init') or incrementally
  29update an existing populated repository.  Whether or not incremental
  30imports are supported from a particular foreign source depends on
  31the frontend program in use.
  32
  33
  34OPTIONS
  35-------
  36
  37--force::
  38        Force updating modified existing branches, even if doing
  39        so would cause commits to be lost (as the new commit does
  40        not contain the old commit).
  41
  42--quiet::
  43        Disable the output shown by --stats, making fast-import usually
  44        be silent when it is successful.  However, if the import stream
  45        has directives intended to show user output (e.g. `progress`
  46        directives), the corresponding messages will still be shown.
  47
  48--stats::
  49        Display some basic statistics about the objects fast-import has
  50        created, the packfiles they were stored into, and the
  51        memory used by fast-import during this run.  Showing this output
  52        is currently the default, but can be disabled with --quiet.
  53
  54Options for Frontends
  55~~~~~~~~~~~~~~~~~~~~~
  56
  57--cat-blob-fd=<fd>::
  58        Write responses to `get-mark`, `cat-blob`, and `ls` queries to the
  59        file descriptor <fd> instead of `stdout`.  Allows `progress`
  60        output intended for the end-user to be separated from other
  61        output.
  62
  63--date-format=<fmt>::
  64        Specify the type of dates the frontend will supply to
  65        fast-import within `author`, `committer` and `tagger` commands.
  66        See ``Date Formats'' below for details about which formats
  67        are supported, and their syntax.
  68
  69--done::
  70        Terminate with error if there is no `done` command at the end of
  71        the stream.  This option might be useful for detecting errors
  72        that cause the frontend to terminate before it has started to
  73        write a stream.
  74
  75Locations of Marks Files
  76~~~~~~~~~~~~~~~~~~~~~~~~
  77
  78--export-marks=<file>::
  79        Dumps the internal marks table to <file> when complete.
  80        Marks are written one per line as `:markid SHA-1`.
  81        Frontends can use this file to validate imports after they
  82        have been completed, or to save the marks table across
  83        incremental runs.  As <file> is only opened and truncated
  84        at checkpoint (or completion) the same path can also be
  85        safely given to --import-marks.
  86
  87--import-marks=<file>::
  88        Before processing any input, load the marks specified in
  89        <file>.  The input file must exist, must be readable, and
  90        must use the same format as produced by --export-marks.
  91        Multiple options may be supplied to import more than one
  92        set of marks.  If a mark is defined to different values,
  93        the last file wins.
  94
  95--import-marks-if-exists=<file>::
  96        Like --import-marks but instead of erroring out, silently
  97        skips the file if it does not exist.
  98
  99--[no-]relative-marks::
 100        After specifying --relative-marks the paths specified
 101        with --import-marks= and --export-marks= are relative
 102        to an internal directory in the current repository.
 103        In git-fast-import this means that the paths are relative
 104        to the .git/info/fast-import directory. However, other
 105        importers may use a different location.
 106+
 107Relative and non-relative marks may be combined by interweaving
 108--(no-)-relative-marks with the --(import|export)-marks= options.
 109
 110Performance and Compression Tuning
 111~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 112
 113--active-branches=<n>::
 114        Maximum number of branches to maintain active at once.
 115        See ``Memory Utilization'' below for details.  Default is 5.
 116
 117--big-file-threshold=<n>::
 118        Maximum size of a blob that fast-import will attempt to
 119        create a delta for, expressed in bytes.  The default is 512m
 120        (512 MiB).  Some importers may wish to lower this on systems
 121        with constrained memory.
 122
 123--depth=<n>::
 124        Maximum delta depth, for blob and tree deltification.
 125        Default is 50.
 126
 127--export-pack-edges=<file>::
 128        After creating a packfile, print a line of data to
 129        <file> listing the filename of the packfile and the last
 130        commit on each branch that was written to that packfile.
 131        This information may be useful after importing projects
 132        whose total object set exceeds the 4 GiB packfile limit,
 133        as these commits can be used as edge points during calls
 134        to 'git pack-objects'.
 135
 136--max-pack-size=<n>::
 137        Maximum size of each output packfile.
 138        The default is unlimited.
 139
 140fastimport.unpackLimit::
 141        See linkgit:git-config[1]
 142
 143PERFORMANCE
 144-----------
 145The design of fast-import allows it to import large projects in a minimum
 146amount of memory usage and processing time.  Assuming the frontend
 147is able to keep up with fast-import and feed it a constant stream of data,
 148import times for projects holding 10+ years of history and containing
 149100,000+ individual commits are generally completed in just 1-2
 150hours on quite modest (~$2,000 USD) hardware.
 151
 152Most bottlenecks appear to be in foreign source data access (the
 153source just cannot extract revisions fast enough) or disk IO (fast-import
 154writes as fast as the disk will take the data).  Imports will run
 155faster if the source data is stored on a different drive than the
 156destination Git repository (due to less IO contention).
 157
 158
 159DEVELOPMENT COST
 160----------------
 161A typical frontend for fast-import tends to weigh in at approximately 200
 162lines of Perl/Python/Ruby code.  Most developers have been able to
 163create working importers in just a couple of hours, even though it
 164is their first exposure to fast-import, and sometimes even to Git.  This is
 165an ideal situation, given that most conversion tools are throw-away
 166(use once, and never look back).
 167
 168
 169PARALLEL OPERATION
 170------------------
 171Like 'git push' or 'git fetch', imports handled by fast-import are safe to
 172run alongside parallel `git repack -a -d` or `git gc` invocations,
 173or any other Git operation (including 'git prune', as loose objects
 174are never used by fast-import).
 175
 176fast-import does not lock the branch or tag refs it is actively importing.
 177After the import, during its ref update phase, fast-import tests each
 178existing branch ref to verify the update will be a fast-forward
 179update (the commit stored in the ref is contained in the new
 180history of the commit to be written).  If the update is not a
 181fast-forward update, fast-import will skip updating that ref and instead
 182prints a warning message.  fast-import will always attempt to update all
 183branch refs, and does not stop on the first failure.
 184
 185Branch updates can be forced with --force, but it's recommended that
 186this only be used on an otherwise quiet repository.  Using --force
 187is not necessary for an initial import into an empty repository.
 188
 189
 190TECHNICAL DISCUSSION
 191--------------------
 192fast-import tracks a set of branches in memory.  Any branch can be created
 193or modified at any point during the import process by sending a
 194`commit` command on the input stream.  This design allows a frontend
 195program to process an unlimited number of branches simultaneously,
 196generating commits in the order they are available from the source
 197data.  It also simplifies the frontend programs considerably.
 198
 199fast-import does not use or alter the current working directory, or any
 200file within it.  (It does however update the current Git repository,
 201as referenced by `GIT_DIR`.)  Therefore an import frontend may use
 202the working directory for its own purposes, such as extracting file
 203revisions from the foreign source.  This ignorance of the working
 204directory also allows fast-import to run very quickly, as it does not
 205need to perform any costly file update operations when switching
 206between branches.
 207
 208INPUT FORMAT
 209------------
 210With the exception of raw file data (which Git does not interpret)
 211the fast-import input format is text (ASCII) based.  This text based
 212format simplifies development and debugging of frontend programs,
 213especially when a higher level language such as Perl, Python or
 214Ruby is being used.
 215
 216fast-import is very strict about its input.  Where we say SP below we mean
 217*exactly* one space.  Likewise LF means one (and only one) linefeed
 218and HT one (and only one) horizontal tab.
 219Supplying additional whitespace characters will cause unexpected
 220results, such as branch names or file names with leading or trailing
 221spaces in their name, or early termination of fast-import when it encounters
 222unexpected input.
 223
 224Stream Comments
 225~~~~~~~~~~~~~~~
 226To aid in debugging frontends fast-import ignores any line that
 227begins with `#` (ASCII pound/hash) up to and including the line
 228ending `LF`.  A comment line may contain any sequence of bytes
 229that does not contain an LF and therefore may be used to include
 230any detailed debugging information that might be specific to the
 231frontend and useful when inspecting a fast-import data stream.
 232
 233Date Formats
 234~~~~~~~~~~~~
 235The following date formats are supported.  A frontend should select
 236the format it will use for this import by passing the format name
 237in the --date-format=<fmt> command-line option.
 238
 239`raw`::
 240        This is the Git native format and is `<time> SP <offutc>`.
 241        It is also fast-import's default format, if --date-format was
 242        not specified.
 243+
 244The time of the event is specified by `<time>` as the number of
 245seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is
 246written as an ASCII decimal integer.
 247+
 248The local offset is specified by `<offutc>` as a positive or negative
 249offset from UTC.  For example EST (which is 5 hours behind UTC)
 250would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''.
 251The local offset does not affect `<time>`; it is used only as an
 252advisement to help formatting routines display the timestamp.
 253+
 254If the local offset is not available in the source material, use
 255``+0000'', or the most common local offset.  For example many
 256organizations have a CVS repository which has only ever been accessed
 257by users who are located in the same location and time zone.  In this
 258case a reasonable offset from UTC could be assumed.
 259+
 260Unlike the `rfc2822` format, this format is very strict.  Any
 261variation in formatting will cause fast-import to reject the value.
 262
 263`rfc2822`::
 264        This is the standard email format as described by RFC 2822.
 265+
 266An example value is ``Tue Feb 6 11:22:18 2007 -0500''.  The Git
 267parser is accurate, but a little on the lenient side.  It is the
 268same parser used by 'git am' when applying patches
 269received from email.
 270+
 271Some malformed strings may be accepted as valid dates.  In some of
 272these cases Git will still be able to obtain the correct date from
 273the malformed string.  There are also some types of malformed
 274strings which Git will parse wrong, and yet consider valid.
 275Seriously malformed strings will be rejected.
 276+
 277Unlike the `raw` format above, the time zone/UTC offset information
 278contained in an RFC 2822 date string is used to adjust the date
 279value to UTC prior to storage.  Therefore it is important that
 280this information be as accurate as possible.
 281+
 282If the source material uses RFC 2822 style dates,
 283the frontend should let fast-import handle the parsing and conversion
 284(rather than attempting to do it itself) as the Git parser has
 285been well tested in the wild.
 286+
 287Frontends should prefer the `raw` format if the source material
 288already uses UNIX-epoch format, can be coaxed to give dates in that
 289format, or its format is easily convertible to it, as there is no
 290ambiguity in parsing.
 291
 292`now`::
 293        Always use the current time and time zone.  The literal
 294        `now` must always be supplied for `<when>`.
 295+
 296This is a toy format.  The current time and time zone of this system
 297is always copied into the identity string at the time it is being
 298created by fast-import.  There is no way to specify a different time or
 299time zone.
 300+
 301This particular format is supplied as it's short to implement and
 302may be useful to a process that wants to create a new commit
 303right now, without needing to use a working directory or
 304'git update-index'.
 305+
 306If separate `author` and `committer` commands are used in a `commit`
 307the timestamps may not match, as the system clock will be polled
 308twice (once for each command).  The only way to ensure that both
 309author and committer identity information has the same timestamp
 310is to omit `author` (thus copying from `committer`) or to use a
 311date format other than `now`.
 312
 313Commands
 314~~~~~~~~
 315fast-import accepts several commands to update the current repository
 316and control the current import process.  More detailed discussion
 317(with examples) of each command follows later.
 318
 319`commit`::
 320        Creates a new branch or updates an existing branch by
 321        creating a new commit and updating the branch to point at
 322        the newly created commit.
 323
 324`tag`::
 325        Creates an annotated tag object from an existing commit or
 326        branch.  Lightweight tags are not supported by this command,
 327        as they are not recommended for recording meaningful points
 328        in time.
 329
 330`reset`::
 331        Reset an existing branch (or a new branch) to a specific
 332        revision.  This command must be used to change a branch to
 333        a specific revision without making a commit on it.
 334
 335`blob`::
 336        Convert raw file data into a blob, for future use in a
 337        `commit` command.  This command is optional and is not
 338        needed to perform an import.
 339
 340`checkpoint`::
 341        Forces fast-import to close the current packfile, generate its
 342        unique SHA-1 checksum and index, and start a new packfile.
 343        This command is optional and is not needed to perform
 344        an import.
 345
 346`progress`::
 347        Causes fast-import to echo the entire line to its own
 348        standard output.  This command is optional and is not needed
 349        to perform an import.
 350
 351`done`::
 352        Marks the end of the stream. This command is optional
 353        unless the `done` feature was requested using the
 354        `--done` command-line option or `feature done` command.
 355
 356`get-mark`::
 357        Causes fast-import to print the SHA-1 corresponding to a mark
 358        to the file descriptor set with `--cat-blob-fd`, or `stdout` if
 359        unspecified.
 360
 361`cat-blob`::
 362        Causes fast-import to print a blob in 'cat-file --batch'
 363        format to the file descriptor set with `--cat-blob-fd` or
 364        `stdout` if unspecified.
 365
 366`ls`::
 367        Causes fast-import to print a line describing a directory
 368        entry in 'ls-tree' format to the file descriptor set with
 369        `--cat-blob-fd` or `stdout` if unspecified.
 370
 371`feature`::
 372        Enable the specified feature. This requires that fast-import
 373        supports the specified feature, and aborts if it does not.
 374
 375`option`::
 376        Specify any of the options listed under OPTIONS that do not
 377        change stream semantic to suit the frontend's needs. This
 378        command is optional and is not needed to perform an import.
 379
 380`commit`
 381~~~~~~~~
 382Create or update a branch with a new commit, recording one logical
 383change to the project.
 384
 385....
 386        'commit' SP <ref> LF
 387        mark?
 388        ('author' (SP <name>)? SP LT <email> GT SP <when> LF)?
 389        'committer' (SP <name>)? SP LT <email> GT SP <when> LF
 390        data
 391        ('from' SP <commit-ish> LF)?
 392        ('merge' SP <commit-ish> LF)?
 393        (filemodify | filedelete | filecopy | filerename | filedeleteall | notemodify)*
 394        LF?
 395....
 396
 397where `<ref>` is the name of the branch to make the commit on.
 398Typically branch names are prefixed with `refs/heads/` in
 399Git, so importing the CVS branch symbol `RELENG-1_0` would use
 400`refs/heads/RELENG-1_0` for the value of `<ref>`.  The value of
 401`<ref>` must be a valid refname in Git.  As `LF` is not valid in
 402a Git refname, no quoting or escaping syntax is supported here.
 403
 404A `mark` command may optionally appear, requesting fast-import to save a
 405reference to the newly created commit for future use by the frontend
 406(see below for format).  It is very common for frontends to mark
 407every commit they create, thereby allowing future branch creation
 408from any imported commit.
 409
 410The `data` command following `committer` must supply the commit
 411message (see below for `data` command syntax).  To import an empty
 412commit message use a 0 length data.  Commit messages are free-form
 413and are not interpreted by Git.  Currently they must be encoded in
 414UTF-8, as fast-import does not permit other encodings to be specified.
 415
 416Zero or more `filemodify`, `filedelete`, `filecopy`, `filerename`,
 417`filedeleteall` and `notemodify` commands
 418may be included to update the contents of the branch prior to
 419creating the commit.  These commands may be supplied in any order.
 420However it is recommended that a `filedeleteall` command precede
 421all `filemodify`, `filecopy`, `filerename` and `notemodify` commands in
 422the same commit, as `filedeleteall` wipes the branch clean (see below).
 423
 424The `LF` after the command is optional (it used to be required).
 425
 426`author`
 427^^^^^^^^
 428An `author` command may optionally appear, if the author information
 429might differ from the committer information.  If `author` is omitted
 430then fast-import will automatically use the committer's information for
 431the author portion of the commit.  See below for a description of
 432the fields in `author`, as they are identical to `committer`.
 433
 434`committer`
 435^^^^^^^^^^^
 436The `committer` command indicates who made this commit, and when
 437they made it.
 438
 439Here `<name>` is the person's display name (for example
 440``Com M Itter'') and `<email>` is the person's email address
 441(``\cm@example.com'').  `LT` and `GT` are the literal less-than (\x3c)
 442and greater-than (\x3e) symbols.  These are required to delimit
 443the email address from the other fields in the line.  Note that
 444`<name>` and `<email>` are free-form and may contain any sequence
 445of bytes, except `LT`, `GT` and `LF`.  `<name>` is typically UTF-8 encoded.
 446
 447The time of the change is specified by `<when>` using the date format
 448that was selected by the --date-format=<fmt> command-line option.
 449See ``Date Formats'' above for the set of supported formats, and
 450their syntax.
 451
 452`from`
 453^^^^^^
 454The `from` command is used to specify the commit to initialize
 455this branch from.  This revision will be the first ancestor of the
 456new commit.  The state of the tree built at this commit will begin
 457with the state at the `from` commit, and be altered by the content
 458modifications in this commit.
 459
 460Omitting the `from` command in the first commit of a new branch
 461will cause fast-import to create that commit with no ancestor. This
 462tends to be desired only for the initial commit of a project.
 463If the frontend creates all files from scratch when making a new
 464branch, a `merge` command may be used instead of `from` to start
 465the commit with an empty tree.
 466Omitting the `from` command on existing branches is usually desired,
 467as the current commit on that branch is automatically assumed to
 468be the first ancestor of the new commit.
 469
 470As `LF` is not valid in a Git refname or SHA-1 expression, no
 471quoting or escaping syntax is supported within `<commit-ish>`.
 472
 473Here `<commit-ish>` is any of the following:
 474
 475* The name of an existing branch already in fast-import's internal branch
 476  table.  If fast-import doesn't know the name, it's treated as a SHA-1
 477  expression.
 478
 479* A mark reference, `:<idnum>`, where `<idnum>` is the mark number.
 480+
 481The reason fast-import uses `:` to denote a mark reference is this character
 482is not legal in a Git branch name.  The leading `:` makes it easy
 483to distinguish between the mark 42 (`:42`) and the branch 42 (`42`
 484or `refs/heads/42`), or an abbreviated SHA-1 which happened to
 485consist only of base-10 digits.
 486+
 487Marks must be declared (via `mark`) before they can be used.
 488
 489* A complete 40 byte or abbreviated commit SHA-1 in hex.
 490
 491* Any valid Git SHA-1 expression that resolves to a commit.  See
 492  ``SPECIFYING REVISIONS'' in linkgit:gitrevisions[7] for details.
 493
 494* The special null SHA-1 (40 zeros) specifies that the branch is to be
 495  removed.
 496
 497The special case of restarting an incremental import from the
 498current branch value should be written as:
 499----
 500        from refs/heads/branch^0
 501----
 502The `^0` suffix is necessary as fast-import does not permit a branch to
 503start from itself, and the branch is created in memory before the
 504`from` command is even read from the input.  Adding `^0` will force
 505fast-import to resolve the commit through Git's revision parsing library,
 506rather than its internal branch table, thereby loading in the
 507existing value of the branch.
 508
 509`merge`
 510^^^^^^^
 511Includes one additional ancestor commit.  The additional ancestry
 512link does not change the way the tree state is built at this commit.
 513If the `from` command is
 514omitted when creating a new branch, the first `merge` commit will be
 515the first ancestor of the current commit, and the branch will start
 516out with no files.  An unlimited number of `merge` commands per
 517commit are permitted by fast-import, thereby establishing an n-way merge.
 518
 519Here `<commit-ish>` is any of the commit specification expressions
 520also accepted by `from` (see above).
 521
 522`filemodify`
 523^^^^^^^^^^^^
 524Included in a `commit` command to add a new file or change the
 525content of an existing file.  This command has two different means
 526of specifying the content of the file.
 527
 528External data format::
 529        The data content for the file was already supplied by a prior
 530        `blob` command.  The frontend just needs to connect it.
 531+
 532....
 533        'M' SP <mode> SP <dataref> SP <path> LF
 534....
 535+
 536Here usually `<dataref>` must be either a mark reference (`:<idnum>`)
 537set by a prior `blob` command, or a full 40-byte SHA-1 of an
 538existing Git blob object.  If `<mode>` is `040000`` then
 539`<dataref>` must be the full 40-byte SHA-1 of an existing
 540Git tree object or a mark reference set with `--import-marks`.
 541
 542Inline data format::
 543        The data content for the file has not been supplied yet.
 544        The frontend wants to supply it as part of this modify
 545        command.
 546+
 547....
 548        'M' SP <mode> SP 'inline' SP <path> LF
 549        data
 550....
 551+
 552See below for a detailed description of the `data` command.
 553
 554In both formats `<mode>` is the type of file entry, specified
 555in octal.  Git only supports the following modes:
 556
 557* `100644` or `644`: A normal (not-executable) file.  The majority
 558  of files in most projects use this mode.  If in doubt, this is
 559  what you want.
 560* `100755` or `755`: A normal, but executable, file.
 561* `120000`: A symlink, the content of the file will be the link target.
 562* `160000`: A gitlink, SHA-1 of the object refers to a commit in
 563  another repository. Git links can only be specified by SHA or through
 564  a commit mark. They are used to implement submodules.
 565* `040000`: A subdirectory.  Subdirectories can only be specified by
 566  SHA or through a tree mark set with `--import-marks`.
 567
 568In both formats `<path>` is the complete path of the file to be added
 569(if not already existing) or modified (if already existing).
 570
 571A `<path>` string must use UNIX-style directory separators (forward
 572slash `/`), may contain any byte other than `LF`, and must not
 573start with double quote (`"`).
 574
 575A path can use C-style string quoting; this is accepted in all cases
 576and mandatory if the filename starts with double quote or contains
 577`LF`. In C-style quoting, the complete name should be surrounded with
 578double quotes, and any `LF`, backslash, or double quote characters
 579must be escaped by preceding them with a backslash (e.g.,
 580`"path/with\n, \\ and \" in it"`).
 581
 582The value of `<path>` must be in canonical form. That is it must not:
 583
 584* contain an empty directory component (e.g. `foo//bar` is invalid),
 585* end with a directory separator (e.g. `foo/` is invalid),
 586* start with a directory separator (e.g. `/foo` is invalid),
 587* contain the special component `.` or `..` (e.g. `foo/./bar` and
 588  `foo/../bar` are invalid).
 589
 590The root of the tree can be represented by an empty string as `<path>`.
 591
 592It is recommended that `<path>` always be encoded using UTF-8.
 593
 594`filedelete`
 595^^^^^^^^^^^^
 596Included in a `commit` command to remove a file or recursively
 597delete an entire directory from the branch.  If the file or directory
 598removal makes its parent directory empty, the parent directory will
 599be automatically removed too.  This cascades up the tree until the
 600first non-empty directory or the root is reached.
 601
 602....
 603        'D' SP <path> LF
 604....
 605
 606here `<path>` is the complete path of the file or subdirectory to
 607be removed from the branch.
 608See `filemodify` above for a detailed description of `<path>`.
 609
 610`filecopy`
 611^^^^^^^^^^
 612Recursively copies an existing file or subdirectory to a different
 613location within the branch.  The existing file or directory must
 614exist.  If the destination exists it will be completely replaced
 615by the content copied from the source.
 616
 617....
 618        'C' SP <path> SP <path> LF
 619....
 620
 621here the first `<path>` is the source location and the second
 622`<path>` is the destination.  See `filemodify` above for a detailed
 623description of what `<path>` may look like.  To use a source path
 624that contains SP the path must be quoted.
 625
 626A `filecopy` command takes effect immediately.  Once the source
 627location has been copied to the destination any future commands
 628applied to the source location will not impact the destination of
 629the copy.
 630
 631`filerename`
 632^^^^^^^^^^^^
 633Renames an existing file or subdirectory to a different location
 634within the branch.  The existing file or directory must exist. If
 635the destination exists it will be replaced by the source directory.
 636
 637....
 638        'R' SP <path> SP <path> LF
 639....
 640
 641here the first `<path>` is the source location and the second
 642`<path>` is the destination.  See `filemodify` above for a detailed
 643description of what `<path>` may look like.  To use a source path
 644that contains SP the path must be quoted.
 645
 646A `filerename` command takes effect immediately.  Once the source
 647location has been renamed to the destination any future commands
 648applied to the source location will create new files there and not
 649impact the destination of the rename.
 650
 651Note that a `filerename` is the same as a `filecopy` followed by a
 652`filedelete` of the source location.  There is a slight performance
 653advantage to using `filerename`, but the advantage is so small
 654that it is never worth trying to convert a delete/add pair in
 655source material into a rename for fast-import.  This `filerename`
 656command is provided just to simplify frontends that already have
 657rename information and don't want bother with decomposing it into a
 658`filecopy` followed by a `filedelete`.
 659
 660`filedeleteall`
 661^^^^^^^^^^^^^^^
 662Included in a `commit` command to remove all files (and also all
 663directories) from the branch.  This command resets the internal
 664branch structure to have no files in it, allowing the frontend
 665to subsequently add all interesting files from scratch.
 666
 667....
 668        'deleteall' LF
 669....
 670
 671This command is extremely useful if the frontend does not know
 672(or does not care to know) what files are currently on the branch,
 673and therefore cannot generate the proper `filedelete` commands to
 674update the content.
 675
 676Issuing a `filedeleteall` followed by the needed `filemodify`
 677commands to set the correct content will produce the same results
 678as sending only the needed `filemodify` and `filedelete` commands.
 679The `filedeleteall` approach may however require fast-import to use slightly
 680more memory per active branch (less than 1 MiB for even most large
 681projects); so frontends that can easily obtain only the affected
 682paths for a commit are encouraged to do so.
 683
 684`notemodify`
 685^^^^^^^^^^^^
 686Included in a `commit` `<notes_ref>` command to add a new note
 687annotating a `<commit-ish>` or change this annotation contents.
 688Internally it is similar to filemodify 100644 on `<commit-ish>`
 689path (maybe split into subdirectories). It's not advised to
 690use any other commands to write to the `<notes_ref>` tree except
 691`filedeleteall` to delete all existing notes in this tree.
 692This command has two different means of specifying the content
 693of the note.
 694
 695External data format::
 696        The data content for the note was already supplied by a prior
 697        `blob` command.  The frontend just needs to connect it to the
 698        commit that is to be annotated.
 699+
 700....
 701        'N' SP <dataref> SP <commit-ish> LF
 702....
 703+
 704Here `<dataref>` can be either a mark reference (`:<idnum>`)
 705set by a prior `blob` command, or a full 40-byte SHA-1 of an
 706existing Git blob object.
 707
 708Inline data format::
 709        The data content for the note has not been supplied yet.
 710        The frontend wants to supply it as part of this modify
 711        command.
 712+
 713....
 714        'N' SP 'inline' SP <commit-ish> LF
 715        data
 716....
 717+
 718See below for a detailed description of the `data` command.
 719
 720In both formats `<commit-ish>` is any of the commit specification
 721expressions also accepted by `from` (see above).
 722
 723`mark`
 724~~~~~~
 725Arranges for fast-import to save a reference to the current object, allowing
 726the frontend to recall this object at a future point in time, without
 727knowing its SHA-1.  Here the current object is the object creation
 728command the `mark` command appears within.  This can be `commit`,
 729`tag`, and `blob`, but `commit` is the most common usage.
 730
 731....
 732        'mark' SP ':' <idnum> LF
 733....
 734
 735where `<idnum>` is the number assigned by the frontend to this mark.
 736The value of `<idnum>` is expressed as an ASCII decimal integer.
 737The value 0 is reserved and cannot be used as
 738a mark.  Only values greater than or equal to 1 may be used as marks.
 739
 740New marks are created automatically.  Existing marks can be moved
 741to another object simply by reusing the same `<idnum>` in another
 742`mark` command.
 743
 744`tag`
 745~~~~~
 746Creates an annotated tag referring to a specific commit.  To create
 747lightweight (non-annotated) tags see the `reset` command below.
 748
 749....
 750        'tag' SP <name> LF
 751        'from' SP <commit-ish> LF
 752        'tagger' (SP <name>)? SP LT <email> GT SP <when> LF
 753        data
 754....
 755
 756where `<name>` is the name of the tag to create.
 757
 758Tag names are automatically prefixed with `refs/tags/` when stored
 759in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would
 760use just `RELENG-1_0-FINAL` for `<name>`, and fast-import will write the
 761corresponding ref as `refs/tags/RELENG-1_0-FINAL`.
 762
 763The value of `<name>` must be a valid refname in Git and therefore
 764may contain forward slashes.  As `LF` is not valid in a Git refname,
 765no quoting or escaping syntax is supported here.
 766
 767The `from` command is the same as in the `commit` command; see
 768above for details.
 769
 770The `tagger` command uses the same format as `committer` within
 771`commit`; again see above for details.
 772
 773The `data` command following `tagger` must supply the annotated tag
 774message (see below for `data` command syntax).  To import an empty
 775tag message use a 0 length data.  Tag messages are free-form and are
 776not interpreted by Git.  Currently they must be encoded in UTF-8,
 777as fast-import does not permit other encodings to be specified.
 778
 779Signing annotated tags during import from within fast-import is not
 780supported.  Trying to include your own PGP/GPG signature is not
 781recommended, as the frontend does not (easily) have access to the
 782complete set of bytes which normally goes into such a signature.
 783If signing is required, create lightweight tags from within fast-import with
 784`reset`, then create the annotated versions of those tags offline
 785with the standard 'git tag' process.
 786
 787`reset`
 788~~~~~~~
 789Creates (or recreates) the named branch, optionally starting from
 790a specific revision.  The reset command allows a frontend to issue
 791a new `from` command for an existing branch, or to create a new
 792branch from an existing commit without creating a new commit.
 793
 794....
 795        'reset' SP <ref> LF
 796        ('from' SP <commit-ish> LF)?
 797        LF?
 798....
 799
 800For a detailed description of `<ref>` and `<commit-ish>` see above
 801under `commit` and `from`.
 802
 803The `LF` after the command is optional (it used to be required).
 804
 805The `reset` command can also be used to create lightweight
 806(non-annotated) tags.  For example:
 807
 808====
 809        reset refs/tags/938
 810        from :938
 811====
 812
 813would create the lightweight tag `refs/tags/938` referring to
 814whatever commit mark `:938` references.
 815
 816`blob`
 817~~~~~~
 818Requests writing one file revision to the packfile.  The revision
 819is not connected to any commit; this connection must be formed in
 820a subsequent `commit` command by referencing the blob through an
 821assigned mark.
 822
 823....
 824        'blob' LF
 825        mark?
 826        data
 827....
 828
 829The mark command is optional here as some frontends have chosen
 830to generate the Git SHA-1 for the blob on their own, and feed that
 831directly to `commit`.  This is typically more work than it's worth
 832however, as marks are inexpensive to store and easy to use.
 833
 834`data`
 835~~~~~~
 836Supplies raw data (for use as blob/file content, commit messages, or
 837annotated tag messages) to fast-import.  Data can be supplied using an exact
 838byte count or delimited with a terminating line.  Real frontends
 839intended for production-quality conversions should always use the
 840exact byte count format, as it is more robust and performs better.
 841The delimited format is intended primarily for testing fast-import.
 842
 843Comment lines appearing within the `<raw>` part of `data` commands
 844are always taken to be part of the body of the data and are therefore
 845never ignored by fast-import.  This makes it safe to import any
 846file/message content whose lines might start with `#`.
 847
 848Exact byte count format::
 849        The frontend must specify the number of bytes of data.
 850+
 851....
 852        'data' SP <count> LF
 853        <raw> LF?
 854....
 855+
 856where `<count>` is the exact number of bytes appearing within
 857`<raw>`.  The value of `<count>` is expressed as an ASCII decimal
 858integer.  The `LF` on either side of `<raw>` is not
 859included in `<count>` and will not be included in the imported data.
 860+
 861The `LF` after `<raw>` is optional (it used to be required) but
 862recommended.  Always including it makes debugging a fast-import
 863stream easier as the next command always starts in column 0
 864of the next line, even if `<raw>` did not end with an `LF`.
 865
 866Delimited format::
 867        A delimiter string is used to mark the end of the data.
 868        fast-import will compute the length by searching for the delimiter.
 869        This format is primarily useful for testing and is not
 870        recommended for real data.
 871+
 872....
 873        'data' SP '<<' <delim> LF
 874        <raw> LF
 875        <delim> LF
 876        LF?
 877....
 878+
 879where `<delim>` is the chosen delimiter string.  The string `<delim>`
 880must not appear on a line by itself within `<raw>`, as otherwise
 881fast-import will think the data ends earlier than it really does.  The `LF`
 882immediately trailing `<raw>` is part of `<raw>`.  This is one of
 883the limitations of the delimited format, it is impossible to supply
 884a data chunk which does not have an LF as its last byte.
 885+
 886The `LF` after `<delim> LF` is optional (it used to be required).
 887
 888`checkpoint`
 889~~~~~~~~~~~~
 890Forces fast-import to close the current packfile, start a new one, and to
 891save out all current branch refs, tags and marks.
 892
 893....
 894        'checkpoint' LF
 895        LF?
 896....
 897
 898Note that fast-import automatically switches packfiles when the current
 899packfile reaches --max-pack-size, or 4 GiB, whichever limit is
 900smaller.  During an automatic packfile switch fast-import does not update
 901the branch refs, tags or marks.
 902
 903As a `checkpoint` can require a significant amount of CPU time and
 904disk IO (to compute the overall pack SHA-1 checksum, generate the
 905corresponding index file, and update the refs) it can easily take
 906several minutes for a single `checkpoint` command to complete.
 907
 908Frontends may choose to issue checkpoints during extremely large
 909and long running imports, or when they need to allow another Git
 910process access to a branch.  However given that a 30 GiB Subversion
 911repository can be loaded into Git through fast-import in about 3 hours,
 912explicit checkpointing may not be necessary.
 913
 914The `LF` after the command is optional (it used to be required).
 915
 916`progress`
 917~~~~~~~~~~
 918Causes fast-import to print the entire `progress` line unmodified to
 919its standard output channel (file descriptor 1) when the command is
 920processed from the input stream.  The command otherwise has no impact
 921on the current import, or on any of fast-import's internal state.
 922
 923....
 924        'progress' SP <any> LF
 925        LF?
 926....
 927
 928The `<any>` part of the command may contain any sequence of bytes
 929that does not contain `LF`.  The `LF` after the command is optional.
 930Callers may wish to process the output through a tool such as sed to
 931remove the leading part of the line, for example:
 932
 933====
 934        frontend | git fast-import | sed 's/^progress //'
 935====
 936
 937Placing a `progress` command immediately after a `checkpoint` will
 938inform the reader when the `checkpoint` has been completed and it
 939can safely access the refs that fast-import updated.
 940
 941`get-mark`
 942~~~~~~~~~~
 943Causes fast-import to print the SHA-1 corresponding to a mark to
 944stdout or to the file descriptor previously arranged with the
 945`--cat-blob-fd` argument. The command otherwise has no impact on the
 946current import; its purpose is to retrieve SHA-1s that later commits
 947might want to refer to in their commit messages.
 948
 949....
 950        'get-mark' SP ':' <idnum> LF
 951....
 952
 953This command can be used anywhere in the stream that comments are
 954accepted.  In particular, the `get-mark` command can be used in the
 955middle of a commit but not in the middle of a `data` command.
 956
 957See ``Responses To Commands'' below for details about how to read
 958this output safely.
 959
 960`cat-blob`
 961~~~~~~~~~~
 962Causes fast-import to print a blob to a file descriptor previously
 963arranged with the `--cat-blob-fd` argument.  The command otherwise
 964has no impact on the current import; its main purpose is to
 965retrieve blobs that may be in fast-import's memory but not
 966accessible from the target repository.
 967
 968....
 969        'cat-blob' SP <dataref> LF
 970....
 971
 972The `<dataref>` can be either a mark reference (`:<idnum>`)
 973set previously or a full 40-byte SHA-1 of a Git blob, preexisting or
 974ready to be written.
 975
 976Output uses the same format as `git cat-file --batch`:
 977
 978====
 979        <sha1> SP 'blob' SP <size> LF
 980        <contents> LF
 981====
 982
 983This command can be used anywhere in the stream that comments are
 984accepted.  In particular, the `cat-blob` command can be used in the
 985middle of a commit but not in the middle of a `data` command.
 986
 987See ``Responses To Commands'' below for details about how to read
 988this output safely.
 989
 990`ls`
 991~~~~
 992Prints information about the object at a path to a file descriptor
 993previously arranged with the `--cat-blob-fd` argument.  This allows
 994printing a blob from the active commit (with `cat-blob`) or copying a
 995blob or tree from a previous commit for use in the current one (with
 996`filemodify`).
 997
 998The `ls` command can be used anywhere in the stream that comments are
 999accepted, including the middle of a commit.
1000
1001Reading from the active commit::
1002        This form can only be used in the middle of a `commit`.
1003        The path names a directory entry within fast-import's
1004        active commit.  The path must be quoted in this case.
1005+
1006....
1007        'ls' SP <path> LF
1008....
1009
1010Reading from a named tree::
1011        The `<dataref>` can be a mark reference (`:<idnum>`) or the
1012        full 40-byte SHA-1 of a Git tag, commit, or tree object,
1013        preexisting or waiting to be written.
1014        The path is relative to the top level of the tree
1015        named by `<dataref>`.
1016+
1017....
1018        'ls' SP <dataref> SP <path> LF
1019....
1020
1021See `filemodify` above for a detailed description of `<path>`.
1022
1023Output uses the same format as `git ls-tree <tree> -- <path>`:
1024
1025====
1026        <mode> SP ('blob' | 'tree' | 'commit') SP <dataref> HT <path> LF
1027====
1028
1029The <dataref> represents the blob, tree, or commit object at <path>
1030and can be used in later 'get-mark', 'cat-blob', 'filemodify', or
1031'ls' commands.
1032
1033If there is no file or subtree at that path, 'git fast-import' will
1034instead report
1035
1036====
1037        missing SP <path> LF
1038====
1039
1040See ``Responses To Commands'' below for details about how to read
1041this output safely.
1042
1043`feature`
1044~~~~~~~~~
1045Require that fast-import supports the specified feature, or abort if
1046it does not.
1047
1048....
1049        'feature' SP <feature> ('=' <argument>)? LF
1050....
1051
1052The <feature> part of the command may be any one of the following:
1053
1054date-format::
1055export-marks::
1056relative-marks::
1057no-relative-marks::
1058force::
1059        Act as though the corresponding command-line option with
1060        a leading `--` was passed on the command line
1061        (see OPTIONS, above).
1062
1063import-marks::
1064import-marks-if-exists::
1065        Like --import-marks except in two respects: first, only one
1066        "feature import-marks" or "feature import-marks-if-exists"
1067        command is allowed per stream; second, an --import-marks=
1068        or --import-marks-if-exists command-line option overrides
1069        any of these "feature" commands in the stream; third,
1070        "feature import-marks-if-exists" like a corresponding
1071        command-line option silently skips a nonexistent file.
1072
1073get-mark::
1074cat-blob::
1075ls::
1076        Require that the backend support the 'get-mark', 'cat-blob',
1077        or 'ls' command respectively.
1078        Versions of fast-import not supporting the specified command
1079        will exit with a message indicating so.
1080        This lets the import error out early with a clear message,
1081        rather than wasting time on the early part of an import
1082        before the unsupported command is detected.
1083
1084notes::
1085        Require that the backend support the 'notemodify' (N)
1086        subcommand to the 'commit' command.
1087        Versions of fast-import not supporting notes will exit
1088        with a message indicating so.
1089
1090done::
1091        Error out if the stream ends without a 'done' command.
1092        Without this feature, errors causing the frontend to end
1093        abruptly at a convenient point in the stream can go
1094        undetected.  This may occur, for example, if an import
1095        front end dies in mid-operation without emitting SIGTERM
1096        or SIGKILL at its subordinate git fast-import instance.
1097
1098`option`
1099~~~~~~~~
1100Processes the specified option so that git fast-import behaves in a
1101way that suits the frontend's needs.
1102Note that options specified by the frontend are overridden by any
1103options the user may specify to git fast-import itself.
1104
1105....
1106    'option' SP <option> LF
1107....
1108
1109The `<option>` part of the command may contain any of the options
1110listed in the OPTIONS section that do not change import semantics,
1111without the leading `--` and is treated in the same way.
1112
1113Option commands must be the first commands on the input (not counting
1114feature commands), to give an option command after any non-option
1115command is an error.
1116
1117The following command-line options change import semantics and may therefore
1118not be passed as option:
1119
1120* date-format
1121* import-marks
1122* export-marks
1123* cat-blob-fd
1124* force
1125
1126`done`
1127~~~~~~
1128If the `done` feature is not in use, treated as if EOF was read.
1129This can be used to tell fast-import to finish early.
1130
1131If the `--done` command-line option or `feature done` command is
1132in use, the `done` command is mandatory and marks the end of the
1133stream.
1134
1135RESPONSES TO COMMANDS
1136---------------------
1137New objects written by fast-import are not available immediately.
1138Most fast-import commands have no visible effect until the next
1139checkpoint (or completion).  The frontend can send commands to
1140fill fast-import's input pipe without worrying about how quickly
1141they will take effect, which improves performance by simplifying
1142scheduling.
1143
1144For some frontends, though, it is useful to be able to read back
1145data from the current repository as it is being updated (for
1146example when the source material describes objects in terms of
1147patches to be applied to previously imported objects).  This can
1148be accomplished by connecting the frontend and fast-import via
1149bidirectional pipes:
1150
1151====
1152        mkfifo fast-import-output
1153        frontend <fast-import-output |
1154        git fast-import >fast-import-output
1155====
1156
1157A frontend set up this way can use `progress`, `get-mark`, `ls`, and
1158`cat-blob` commands to read information from the import in progress.
1159
1160To avoid deadlock, such frontends must completely consume any
1161pending output from `progress`, `ls`, `get-mark`, and `cat-blob` before
1162performing writes to fast-import that might block.
1163
1164CRASH REPORTS
1165-------------
1166If fast-import is supplied invalid input it will terminate with a
1167non-zero exit status and create a crash report in the top level of
1168the Git repository it was importing into.  Crash reports contain
1169a snapshot of the internal fast-import state as well as the most
1170recent commands that lead up to the crash.
1171
1172All recent commands (including stream comments, file changes and
1173progress commands) are shown in the command history within the crash
1174report, but raw file data and commit messages are excluded from the
1175crash report.  This exclusion saves space within the report file
1176and reduces the amount of buffering that fast-import must perform
1177during execution.
1178
1179After writing a crash report fast-import will close the current
1180packfile and export the marks table.  This allows the frontend
1181developer to inspect the repository state and resume the import from
1182the point where it crashed.  The modified branches and tags are not
1183updated during a crash, as the import did not complete successfully.
1184Branch and tag information can be found in the crash report and
1185must be applied manually if the update is needed.
1186
1187An example crash:
1188
1189====
1190        $ cat >in <<END_OF_INPUT
1191        # my very first test commit
1192        commit refs/heads/master
1193        committer Shawn O. Pearce <spearce> 19283 -0400
1194        # who is that guy anyway?
1195        data <<EOF
1196        this is my commit
1197        EOF
1198        M 644 inline .gitignore
1199        data <<EOF
1200        .gitignore
1201        EOF
1202        M 777 inline bob
1203        END_OF_INPUT
1204
1205        $ git fast-import <in
1206        fatal: Corrupt mode: M 777 inline bob
1207        fast-import: dumping crash report to .git/fast_import_crash_8434
1208
1209        $ cat .git/fast_import_crash_8434
1210        fast-import crash report:
1211            fast-import process: 8434
1212            parent process     : 1391
1213            at Sat Sep 1 00:58:12 2007
1214
1215        fatal: Corrupt mode: M 777 inline bob
1216
1217        Most Recent Commands Before Crash
1218        ---------------------------------
1219          # my very first test commit
1220          commit refs/heads/master
1221          committer Shawn O. Pearce <spearce> 19283 -0400
1222          # who is that guy anyway?
1223          data <<EOF
1224          M 644 inline .gitignore
1225          data <<EOF
1226        * M 777 inline bob
1227
1228        Active Branch LRU
1229        -----------------
1230            active_branches = 1 cur, 5 max
1231
1232          pos  clock name
1233          ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1234           1)      0 refs/heads/master
1235
1236        Inactive Branches
1237        -----------------
1238        refs/heads/master:
1239          status      : active loaded dirty
1240          tip commit  : 0000000000000000000000000000000000000000
1241          old tree    : 0000000000000000000000000000000000000000
1242          cur tree    : 0000000000000000000000000000000000000000
1243          commit clock: 0
1244          last pack   :
1245
1246
1247        -------------------
1248        END OF CRASH REPORT
1249====
1250
1251TIPS AND TRICKS
1252---------------
1253The following tips and tricks have been collected from various
1254users of fast-import, and are offered here as suggestions.
1255
1256Use One Mark Per Commit
1257~~~~~~~~~~~~~~~~~~~~~~~
1258When doing a repository conversion, use a unique mark per commit
1259(`mark :<n>`) and supply the --export-marks option on the command
1260line.  fast-import will dump a file which lists every mark and the Git
1261object SHA-1 that corresponds to it.  If the frontend can tie
1262the marks back to the source repository, it is easy to verify the
1263accuracy and completeness of the import by comparing each Git
1264commit to the corresponding source revision.
1265
1266Coming from a system such as Perforce or Subversion this should be
1267quite simple, as the fast-import mark can also be the Perforce changeset
1268number or the Subversion revision number.
1269
1270Freely Skip Around Branches
1271~~~~~~~~~~~~~~~~~~~~~~~~~~~
1272Don't bother trying to optimize the frontend to stick to one branch
1273at a time during an import.  Although doing so might be slightly
1274faster for fast-import, it tends to increase the complexity of the frontend
1275code considerably.
1276
1277The branch LRU builtin to fast-import tends to behave very well, and the
1278cost of activating an inactive branch is so low that bouncing around
1279between branches has virtually no impact on import performance.
1280
1281Handling Renames
1282~~~~~~~~~~~~~~~~
1283When importing a renamed file or directory, simply delete the old
1284name(s) and modify the new name(s) during the corresponding commit.
1285Git performs rename detection after-the-fact, rather than explicitly
1286during a commit.
1287
1288Use Tag Fixup Branches
1289~~~~~~~~~~~~~~~~~~~~~~
1290Some other SCM systems let the user create a tag from multiple
1291files which are not from the same commit/changeset.  Or to create
1292tags which are a subset of the files available in the repository.
1293
1294Importing these tags as-is in Git is impossible without making at
1295least one commit which ``fixes up'' the files to match the content
1296of the tag.  Use fast-import's `reset` command to reset a dummy branch
1297outside of your normal branch space to the base commit for the tag,
1298then commit one or more file fixup commits, and finally tag the
1299dummy branch.
1300
1301For example since all normal branches are stored under `refs/heads/`
1302name the tag fixup branch `TAG_FIXUP`.  This way it is impossible for
1303the fixup branch used by the importer to have namespace conflicts
1304with real branches imported from the source (the name `TAG_FIXUP`
1305is not `refs/heads/TAG_FIXUP`).
1306
1307When committing fixups, consider using `merge` to connect the
1308commit(s) which are supplying file revisions to the fixup branch.
1309Doing so will allow tools such as 'git blame' to track
1310through the real commit history and properly annotate the source
1311files.
1312
1313After fast-import terminates the frontend will need to do `rm .git/TAG_FIXUP`
1314to remove the dummy branch.
1315
1316Import Now, Repack Later
1317~~~~~~~~~~~~~~~~~~~~~~~~
1318As soon as fast-import completes the Git repository is completely valid
1319and ready for use.  Typically this takes only a very short time,
1320even for considerably large projects (100,000+ commits).
1321
1322However repacking the repository is necessary to improve data
1323locality and access performance.  It can also take hours on extremely
1324large projects (especially if -f and a large --window parameter is
1325used).  Since repacking is safe to run alongside readers and writers,
1326run the repack in the background and let it finish when it finishes.
1327There is no reason to wait to explore your new Git project!
1328
1329If you choose to wait for the repack, don't try to run benchmarks
1330or performance tests until repacking is completed.  fast-import outputs
1331suboptimal packfiles that are simply never seen in real use
1332situations.
1333
1334Repacking Historical Data
1335~~~~~~~~~~~~~~~~~~~~~~~~~
1336If you are repacking very old imported data (e.g. older than the
1337last year), consider expending some extra CPU time and supplying
1338--window=50 (or higher) when you run 'git repack'.
1339This will take longer, but will also produce a smaller packfile.
1340You only need to expend the effort once, and everyone using your
1341project will benefit from the smaller repository.
1342
1343Include Some Progress Messages
1344~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1345Every once in a while have your frontend emit a `progress` message
1346to fast-import.  The contents of the messages are entirely free-form,
1347so one suggestion would be to output the current month and year
1348each time the current commit date moves into the next month.
1349Your users will feel better knowing how much of the data stream
1350has been processed.
1351
1352
1353PACKFILE OPTIMIZATION
1354---------------------
1355When packing a blob fast-import always attempts to deltify against the last
1356blob written.  Unless specifically arranged for by the frontend,
1357this will probably not be a prior version of the same file, so the
1358generated delta will not be the smallest possible.  The resulting
1359packfile will be compressed, but will not be optimal.
1360
1361Frontends which have efficient access to all revisions of a
1362single file (for example reading an RCS/CVS ,v file) can choose
1363to supply all revisions of that file as a sequence of consecutive
1364`blob` commands.  This allows fast-import to deltify the different file
1365revisions against each other, saving space in the final packfile.
1366Marks can be used to later identify individual file revisions during
1367a sequence of `commit` commands.
1368
1369The packfile(s) created by fast-import do not encourage good disk access
1370patterns.  This is caused by fast-import writing the data in the order
1371it is received on standard input, while Git typically organizes
1372data within packfiles to make the most recent (current tip) data
1373appear before historical data.  Git also clusters commits together,
1374speeding up revision traversal through better cache locality.
1375
1376For this reason it is strongly recommended that users repack the
1377repository with `git repack -a -d` after fast-import completes, allowing
1378Git to reorganize the packfiles for faster data access.  If blob
1379deltas are suboptimal (see above) then also adding the `-f` option
1380to force recomputation of all deltas can significantly reduce the
1381final packfile size (30-50% smaller can be quite typical).
1382
1383
1384MEMORY UTILIZATION
1385------------------
1386There are a number of factors which affect how much memory fast-import
1387requires to perform an import.  Like critical sections of core
1388Git, fast-import uses its own memory allocators to amortize any overheads
1389associated with malloc.  In practice fast-import tends to amortize any
1390malloc overheads to 0, due to its use of large block allocations.
1391
1392per object
1393~~~~~~~~~~
1394fast-import maintains an in-memory structure for every object written in
1395this execution.  On a 32 bit system the structure is 32 bytes,
1396on a 64 bit system the structure is 40 bytes (due to the larger
1397pointer sizes).  Objects in the table are not deallocated until
1398fast-import terminates.  Importing 2 million objects on a 32 bit system
1399will require approximately 64 MiB of memory.
1400
1401The object table is actually a hashtable keyed on the object name
1402(the unique SHA-1).  This storage configuration allows fast-import to reuse
1403an existing or already written object and avoid writing duplicates
1404to the output packfile.  Duplicate blobs are surprisingly common
1405in an import, typically due to branch merges in the source.
1406
1407per mark
1408~~~~~~~~
1409Marks are stored in a sparse array, using 1 pointer (4 bytes or 8
1410bytes, depending on pointer size) per mark.  Although the array
1411is sparse, frontends are still strongly encouraged to use marks
1412between 1 and n, where n is the total number of marks required for
1413this import.
1414
1415per branch
1416~~~~~~~~~~
1417Branches are classified as active and inactive.  The memory usage
1418of the two classes is significantly different.
1419
1420Inactive branches are stored in a structure which uses 96 or 120
1421bytes (32 bit or 64 bit systems, respectively), plus the length of
1422the branch name (typically under 200 bytes), per branch.  fast-import will
1423easily handle as many as 10,000 inactive branches in under 2 MiB
1424of memory.
1425
1426Active branches have the same overhead as inactive branches, but
1427also contain copies of every tree that has been recently modified on
1428that branch.  If subtree `include` has not been modified since the
1429branch became active, its contents will not be loaded into memory,
1430but if subtree `src` has been modified by a commit since the branch
1431became active, then its contents will be loaded in memory.
1432
1433As active branches store metadata about the files contained on that
1434branch, their in-memory storage size can grow to a considerable size
1435(see below).
1436
1437fast-import automatically moves active branches to inactive status based on
1438a simple least-recently-used algorithm.  The LRU chain is updated on
1439each `commit` command.  The maximum number of active branches can be
1440increased or decreased on the command line with --active-branches=.
1441
1442per active tree
1443~~~~~~~~~~~~~~~
1444Trees (aka directories) use just 12 bytes of memory on top of the
1445memory required for their entries (see ``per active file'' below).
1446The cost of a tree is virtually 0, as its overhead amortizes out
1447over the individual file entries.
1448
1449per active file entry
1450~~~~~~~~~~~~~~~~~~~~~
1451Files (and pointers to subtrees) within active trees require 52 or 64
1452bytes (32/64 bit platforms) per entry.  To conserve space, file and
1453tree names are pooled in a common string table, allowing the filename
1454``Makefile'' to use just 16 bytes (after including the string header
1455overhead) no matter how many times it occurs within the project.
1456
1457The active branch LRU, when coupled with the filename string pool
1458and lazy loading of subtrees, allows fast-import to efficiently import
1459projects with 2,000+ branches and 45,114+ files in a very limited
1460memory footprint (less than 2.7 MiB per active branch).
1461
1462SIGNALS
1463-------
1464Sending *SIGUSR1* to the 'git fast-import' process ends the current
1465packfile early, simulating a `checkpoint` command.  The impatient
1466operator can use this facility to peek at the objects and refs from an
1467import in progress, at the cost of some added running time and worse
1468compression.
1469
1470SEE ALSO
1471--------
1472linkgit:git-fast-export[1]
1473
1474GIT
1475---
1476Part of the linkgit:git[1] suite