Documentation / gitattributes.txton commit parse-options: multi-word argh should use dash to separate words (e703d71)
   1gitattributes(5)
   2================
   3
   4NAME
   5----
   6gitattributes - defining attributes per path
   7
   8SYNOPSIS
   9--------
  10$GIT_DIR/info/attributes, .gitattributes
  11
  12
  13DESCRIPTION
  14-----------
  15
  16A `gitattributes` file is a simple text file that gives
  17`attributes` to pathnames.
  18
  19Each line in `gitattributes` file is of form:
  20
  21        pattern attr1 attr2 ...
  22
  23That is, a pattern followed by an attributes list,
  24separated by whitespaces.  When the pattern matches the
  25path in question, the attributes listed on the line are given to
  26the path.
  27
  28Each attribute can be in one of these states for a given path:
  29
  30Set::
  31
  32        The path has the attribute with special value "true";
  33        this is specified by listing only the name of the
  34        attribute in the attribute list.
  35
  36Unset::
  37
  38        The path has the attribute with special value "false";
  39        this is specified by listing the name of the attribute
  40        prefixed with a dash `-` in the attribute list.
  41
  42Set to a value::
  43
  44        The path has the attribute with specified string value;
  45        this is specified by listing the name of the attribute
  46        followed by an equal sign `=` and its value in the
  47        attribute list.
  48
  49Unspecified::
  50
  51        No pattern matches the path, and nothing says if
  52        the path has or does not have the attribute, the
  53        attribute for the path is said to be Unspecified.
  54
  55When more than one pattern matches the path, a later line
  56overrides an earlier line.  This overriding is done per
  57attribute.  The rules how the pattern matches paths are the
  58same as in `.gitignore` files; see linkgit:gitignore[5].
  59Unlike `.gitignore`, negative patterns are forbidden.
  60
  61When deciding what attributes are assigned to a path, Git
  62consults `$GIT_DIR/info/attributes` file (which has the highest
  63precedence), `.gitattributes` file in the same directory as the
  64path in question, and its parent directories up to the toplevel of the
  65work tree (the further the directory that contains `.gitattributes`
  66is from the path in question, the lower its precedence). Finally
  67global and system-wide files are considered (they have the lowest
  68precedence).
  69
  70When the `.gitattributes` file is missing from the work tree, the
  71path in the index is used as a fall-back.  During checkout process,
  72`.gitattributes` in the index is used and then the file in the
  73working tree is used as a fall-back.
  74
  75If you wish to affect only a single repository (i.e., to assign
  76attributes to files that are particular to
  77one user's workflow for that repository), then
  78attributes should be placed in the `$GIT_DIR/info/attributes` file.
  79Attributes which should be version-controlled and distributed to other
  80repositories (i.e., attributes of interest to all users) should go into
  81`.gitattributes` files. Attributes that should affect all repositories
  82for a single user should be placed in a file specified by the
  83`core.attributesfile` configuration option (see linkgit:git-config[1]).
  84Its default value is $XDG_CONFIG_HOME/git/attributes. If $XDG_CONFIG_HOME
  85is either not set or empty, $HOME/.config/git/attributes is used instead.
  86Attributes for all users on a system should be placed in the
  87`$(prefix)/etc/gitattributes` file.
  88
  89Sometimes you would need to override an setting of an attribute
  90for a path to `Unspecified` state.  This can be done by listing
  91the name of the attribute prefixed with an exclamation point `!`.
  92
  93
  94EFFECTS
  95-------
  96
  97Certain operations by Git can be influenced by assigning
  98particular attributes to a path.  Currently, the following
  99operations are attributes-aware.
 100
 101Checking-out and checking-in
 102~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 103
 104These attributes affect how the contents stored in the
 105repository are copied to the working tree files when commands
 106such as 'git checkout' and 'git merge' run.  They also affect how
 107Git stores the contents you prepare in the working tree in the
 108repository upon 'git add' and 'git commit'.
 109
 110`text`
 111^^^^^^
 112
 113This attribute enables and controls end-of-line normalization.  When a
 114text file is normalized, its line endings are converted to LF in the
 115repository.  To control what line ending style is used in the working
 116directory, use the `eol` attribute for a single file and the
 117`core.eol` configuration variable for all text files.
 118
 119Set::
 120
 121        Setting the `text` attribute on a path enables end-of-line
 122        normalization and marks the path as a text file.  End-of-line
 123        conversion takes place without guessing the content type.
 124
 125Unset::
 126
 127        Unsetting the `text` attribute on a path tells Git not to
 128        attempt any end-of-line conversion upon checkin or checkout.
 129
 130Set to string value "auto"::
 131
 132        When `text` is set to "auto", the path is marked for automatic
 133        end-of-line normalization.  If Git decides that the content is
 134        text, its line endings are normalized to LF on checkin.
 135
 136Unspecified::
 137
 138        If the `text` attribute is unspecified, Git uses the
 139        `core.autocrlf` configuration variable to determine if the
 140        file should be converted.
 141
 142Any other value causes Git to act as if `text` has been left
 143unspecified.
 144
 145`eol`
 146^^^^^
 147
 148This attribute sets a specific line-ending style to be used in the
 149working directory.  It enables end-of-line normalization without any
 150content checks, effectively setting the `text` attribute.
 151
 152Set to string value "crlf"::
 153
 154        This setting forces Git to normalize line endings for this
 155        file on checkin and convert them to CRLF when the file is
 156        checked out.
 157
 158Set to string value "lf"::
 159
 160        This setting forces Git to normalize line endings to LF on
 161        checkin and prevents conversion to CRLF when the file is
 162        checked out.
 163
 164Backwards compatibility with `crlf` attribute
 165^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 166
 167For backwards compatibility, the `crlf` attribute is interpreted as
 168follows:
 169
 170------------------------
 171crlf            text
 172-crlf           -text
 173crlf=input      eol=lf
 174------------------------
 175
 176End-of-line conversion
 177^^^^^^^^^^^^^^^^^^^^^^
 178
 179While Git normally leaves file contents alone, it can be configured to
 180normalize line endings to LF in the repository and, optionally, to
 181convert them to CRLF when files are checked out.
 182
 183Here is an example that will make Git normalize .txt, .vcproj and .sh
 184files, ensure that .vcproj files have CRLF and .sh files have LF in
 185the working directory, and prevent .jpg files from being normalized
 186regardless of their content.
 187
 188------------------------
 189*.txt           text
 190*.vcproj        eol=crlf
 191*.sh            eol=lf
 192*.jpg           -text
 193------------------------
 194
 195Other source code management systems normalize all text files in their
 196repositories, and there are two ways to enable similar automatic
 197normalization in Git.
 198
 199If you simply want to have CRLF line endings in your working directory
 200regardless of the repository you are working with, you can set the
 201config variable "core.autocrlf" without changing any attributes.
 202
 203------------------------
 204[core]
 205        autocrlf = true
 206------------------------
 207
 208This does not force normalization of all text files, but does ensure
 209that text files that you introduce to the repository have their line
 210endings normalized to LF when they are added, and that files that are
 211already normalized in the repository stay normalized.
 212
 213If you want to interoperate with a source code management system that
 214enforces end-of-line normalization, or you simply want all text files
 215in your repository to be normalized, you should instead set the `text`
 216attribute to "auto" for _all_ files.
 217
 218------------------------
 219*       text=auto
 220------------------------
 221
 222This ensures that all files that Git considers to be text will have
 223normalized (LF) line endings in the repository.  The `core.eol`
 224configuration variable controls which line endings Git will use for
 225normalized files in your working directory; the default is to use the
 226native line ending for your platform, or CRLF if `core.autocrlf` is
 227set.
 228
 229NOTE: When `text=auto` normalization is enabled in an existing
 230repository, any text files containing CRLFs should be normalized.  If
 231they are not they will be normalized the next time someone tries to
 232change them, causing unfortunate misattribution.  From a clean working
 233directory:
 234
 235-------------------------------------------------
 236$ echo "* text=auto" >>.gitattributes
 237$ rm .git/index     # Remove the index to force Git to
 238$ git reset         # re-scan the working directory
 239$ git status        # Show files that will be normalized
 240$ git add -u
 241$ git add .gitattributes
 242$ git commit -m "Introduce end-of-line normalization"
 243-------------------------------------------------
 244
 245If any files that should not be normalized show up in 'git status',
 246unset their `text` attribute before running 'git add -u'.
 247
 248------------------------
 249manual.pdf      -text
 250------------------------
 251
 252Conversely, text files that Git does not detect can have normalization
 253enabled manually.
 254
 255------------------------
 256weirdchars.txt  text
 257------------------------
 258
 259If `core.safecrlf` is set to "true" or "warn", Git verifies if
 260the conversion is reversible for the current setting of
 261`core.autocrlf`.  For "true", Git rejects irreversible
 262conversions; for "warn", Git only prints a warning but accepts
 263an irreversible conversion.  The safety triggers to prevent such
 264a conversion done to the files in the work tree, but there are a
 265few exceptions.  Even though...
 266
 267- 'git add' itself does not touch the files in the work tree, the
 268  next checkout would, so the safety triggers;
 269
 270- 'git apply' to update a text file with a patch does touch the files
 271  in the work tree, but the operation is about text files and CRLF
 272  conversion is about fixing the line ending inconsistencies, so the
 273  safety does not trigger;
 274
 275- 'git diff' itself does not touch the files in the work tree, it is
 276  often run to inspect the changes you intend to next 'git add'.  To
 277  catch potential problems early, safety triggers.
 278
 279
 280`ident`
 281^^^^^^^
 282
 283When the attribute `ident` is set for a path, Git replaces
 284`$Id$` in the blob object with `$Id:`, followed by the
 28540-character hexadecimal blob object name, followed by a dollar
 286sign `$` upon checkout.  Any byte sequence that begins with
 287`$Id:` and ends with `$` in the worktree file is replaced
 288with `$Id$` upon check-in.
 289
 290
 291`filter`
 292^^^^^^^^
 293
 294A `filter` attribute can be set to a string value that names a
 295filter driver specified in the configuration.
 296
 297A filter driver consists of a `clean` command and a `smudge`
 298command, either of which can be left unspecified.  Upon
 299checkout, when the `smudge` command is specified, the command is
 300fed the blob object from its standard input, and its standard
 301output is used to update the worktree file.  Similarly, the
 302`clean` command is used to convert the contents of worktree file
 303upon checkin.
 304
 305One use of the content filtering is to massage the content into a shape
 306that is more convenient for the platform, filesystem, and the user to use.
 307For this mode of operation, the key phrase here is "more convenient" and
 308not "turning something unusable into usable".  In other words, the intent
 309is that if someone unsets the filter driver definition, or does not have
 310the appropriate filter program, the project should still be usable.
 311
 312Another use of the content filtering is to store the content that cannot
 313be directly used in the repository (e.g. a UUID that refers to the true
 314content stored outside Git, or an encrypted content) and turn it into a
 315usable form upon checkout (e.g. download the external content, or decrypt
 316the encrypted content).
 317
 318These two filters behave differently, and by default, a filter is taken as
 319the former, massaging the contents into more convenient shape.  A missing
 320filter driver definition in the config, or a filter driver that exits with
 321a non-zero status, is not an error but makes the filter a no-op passthru.
 322
 323You can declare that a filter turns a content that by itself is unusable
 324into a usable content by setting the filter.<driver>.required configuration
 325variable to `true`.
 326
 327For example, in .gitattributes, you would assign the `filter`
 328attribute for paths.
 329
 330------------------------
 331*.c     filter=indent
 332------------------------
 333
 334Then you would define a "filter.indent.clean" and "filter.indent.smudge"
 335configuration in your .git/config to specify a pair of commands to
 336modify the contents of C programs when the source files are checked
 337in ("clean" is run) and checked out (no change is made because the
 338command is "cat").
 339
 340------------------------
 341[filter "indent"]
 342        clean = indent
 343        smudge = cat
 344------------------------
 345
 346For best results, `clean` should not alter its output further if it is
 347run twice ("clean->clean" should be equivalent to "clean"), and
 348multiple `smudge` commands should not alter `clean`'s output
 349("smudge->smudge->clean" should be equivalent to "clean").  See the
 350section on merging below.
 351
 352The "indent" filter is well-behaved in this regard: it will not modify
 353input that is already correctly indented.  In this case, the lack of a
 354smudge filter means that the clean filter _must_ accept its own output
 355without modifying it.
 356
 357If a filter _must_ succeed in order to make the stored contents usable,
 358you can declare that the filter is `required`, in the configuration:
 359
 360------------------------
 361[filter "crypt"]
 362        clean = openssl enc ...
 363        smudge = openssl enc -d ...
 364        required
 365------------------------
 366
 367Sequence "%f" on the filter command line is replaced with the name of
 368the file the filter is working on.  A filter might use this in keyword
 369substitution.  For example:
 370
 371------------------------
 372[filter "p4"]
 373        clean = git-p4-filter --clean %f
 374        smudge = git-p4-filter --smudge %f
 375------------------------
 376
 377
 378Interaction between checkin/checkout attributes
 379^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 380
 381In the check-in codepath, the worktree file is first converted
 382with `filter` driver (if specified and corresponding driver
 383defined), then the result is processed with `ident` (if
 384specified), and then finally with `text` (again, if specified
 385and applicable).
 386
 387In the check-out codepath, the blob content is first converted
 388with `text`, and then `ident` and fed to `filter`.
 389
 390
 391Merging branches with differing checkin/checkout attributes
 392^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 393
 394If you have added attributes to a file that cause the canonical
 395repository format for that file to change, such as adding a
 396clean/smudge filter or text/eol/ident attributes, merging anything
 397where the attribute is not in place would normally cause merge
 398conflicts.
 399
 400To prevent these unnecessary merge conflicts, Git can be told to run a
 401virtual check-out and check-in of all three stages of a file when
 402resolving a three-way merge by setting the `merge.renormalize`
 403configuration variable.  This prevents changes caused by check-in
 404conversion from causing spurious merge conflicts when a converted file
 405is merged with an unconverted file.
 406
 407As long as a "smudge->clean" results in the same output as a "clean"
 408even on files that are already smudged, this strategy will
 409automatically resolve all filter-related conflicts.  Filters that do
 410not act in this way may cause additional merge conflicts that must be
 411resolved manually.
 412
 413
 414Generating diff text
 415~~~~~~~~~~~~~~~~~~~~
 416
 417`diff`
 418^^^^^^
 419
 420The attribute `diff` affects how Git generates diffs for particular
 421files. It can tell Git whether to generate a textual patch for the path
 422or to treat the path as a binary file.  It can also affect what line is
 423shown on the hunk header `@@ -k,l +n,m @@` line, tell Git to use an
 424external command to generate the diff, or ask Git to convert binary
 425files to a text format before generating the diff.
 426
 427Set::
 428
 429        A path to which the `diff` attribute is set is treated
 430        as text, even when they contain byte values that
 431        normally never appear in text files, such as NUL.
 432
 433Unset::
 434
 435        A path to which the `diff` attribute is unset will
 436        generate `Binary files differ` (or a binary patch, if
 437        binary patches are enabled).
 438
 439Unspecified::
 440
 441        A path to which the `diff` attribute is unspecified
 442        first gets its contents inspected, and if it looks like
 443        text, it is treated as text.  Otherwise it would
 444        generate `Binary files differ`.
 445
 446String::
 447
 448        Diff is shown using the specified diff driver.  Each driver may
 449        specify one or more options, as described in the following
 450        section. The options for the diff driver "foo" are defined
 451        by the configuration variables in the "diff.foo" section of the
 452        Git config file.
 453
 454
 455Defining an external diff driver
 456^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 457
 458The definition of a diff driver is done in `gitconfig`, not
 459`gitattributes` file, so strictly speaking this manual page is a
 460wrong place to talk about it.  However...
 461
 462To define an external diff driver `jcdiff`, add a section to your
 463`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 464
 465----------------------------------------------------------------
 466[diff "jcdiff"]
 467        command = j-c-diff
 468----------------------------------------------------------------
 469
 470When Git needs to show you a diff for the path with `diff`
 471attribute set to `jcdiff`, it calls the command you specified
 472with the above configuration, i.e. `j-c-diff`, with 7
 473parameters, just like `GIT_EXTERNAL_DIFF` program is called.
 474See linkgit:git[1] for details.
 475
 476
 477Defining a custom hunk-header
 478^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 479
 480Each group of changes (called a "hunk") in the textual diff output
 481is prefixed with a line of the form:
 482
 483        @@ -k,l +n,m @@ TEXT
 484
 485This is called a 'hunk header'.  The "TEXT" portion is by default a line
 486that begins with an alphabet, an underscore or a dollar sign; this
 487matches what GNU 'diff -p' output uses.  This default selection however
 488is not suited for some contents, and you can use a customized pattern
 489to make a selection.
 490
 491First, in .gitattributes, you would assign the `diff` attribute
 492for paths.
 493
 494------------------------
 495*.tex   diff=tex
 496------------------------
 497
 498Then, you would define a "diff.tex.xfuncname" configuration to
 499specify a regular expression that matches a line that you would
 500want to appear as the hunk header "TEXT". Add a section to your
 501`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 502
 503------------------------
 504[diff "tex"]
 505        xfuncname = "^(\\\\(sub)*section\\{.*)$"
 506------------------------
 507
 508Note.  A single level of backslashes are eaten by the
 509configuration file parser, so you would need to double the
 510backslashes; the pattern above picks a line that begins with a
 511backslash, and zero or more occurrences of `sub` followed by
 512`section` followed by open brace, to the end of line.
 513
 514There are a few built-in patterns to make this easier, and `tex`
 515is one of them, so you do not have to write the above in your
 516configuration file (you still need to enable this with the
 517attribute mechanism, via `.gitattributes`).  The following built in
 518patterns are available:
 519
 520- `ada` suitable for source code in the Ada language.
 521
 522- `bibtex` suitable for files with BibTeX coded references.
 523
 524- `cpp` suitable for source code in the C and C++ languages.
 525
 526- `csharp` suitable for source code in the C# language.
 527
 528- `fortran` suitable for source code in the Fortran language.
 529
 530- `html` suitable for HTML/XHTML documents.
 531
 532- `java` suitable for source code in the Java language.
 533
 534- `matlab` suitable for source code in the MATLAB language.
 535
 536- `objc` suitable for source code in the Objective-C language.
 537
 538- `pascal` suitable for source code in the Pascal/Delphi language.
 539
 540- `perl` suitable for source code in the Perl language.
 541
 542- `php` suitable for source code in the PHP language.
 543
 544- `python` suitable for source code in the Python language.
 545
 546- `ruby` suitable for source code in the Ruby language.
 547
 548- `tex` suitable for source code for LaTeX documents.
 549
 550
 551Customizing word diff
 552^^^^^^^^^^^^^^^^^^^^^
 553
 554You can customize the rules that `git diff --word-diff` uses to
 555split words in a line, by specifying an appropriate regular expression
 556in the "diff.*.wordRegex" configuration variable.  For example, in TeX
 557a backslash followed by a sequence of letters forms a command, but
 558several such commands can be run together without intervening
 559whitespace.  To separate them, use a regular expression in your
 560`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 561
 562------------------------
 563[diff "tex"]
 564        wordRegex = "\\\\[a-zA-Z]+|[{}]|\\\\.|[^\\{}[:space:]]+"
 565------------------------
 566
 567A built-in pattern is provided for all languages listed in the
 568previous section.
 569
 570
 571Performing text diffs of binary files
 572^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 573
 574Sometimes it is desirable to see the diff of a text-converted
 575version of some binary files. For example, a word processor
 576document can be converted to an ASCII text representation, and
 577the diff of the text shown. Even though this conversion loses
 578some information, the resulting diff is useful for human
 579viewing (but cannot be applied directly).
 580
 581The `textconv` config option is used to define a program for
 582performing such a conversion. The program should take a single
 583argument, the name of a file to convert, and produce the
 584resulting text on stdout.
 585
 586For example, to show the diff of the exif information of a
 587file instead of the binary information (assuming you have the
 588exif tool installed), add the following section to your
 589`$GIT_DIR/config` file (or `$HOME/.gitconfig` file):
 590
 591------------------------
 592[diff "jpg"]
 593        textconv = exif
 594------------------------
 595
 596NOTE: The text conversion is generally a one-way conversion;
 597in this example, we lose the actual image contents and focus
 598just on the text data. This means that diffs generated by
 599textconv are _not_ suitable for applying. For this reason,
 600only `git diff` and the `git log` family of commands (i.e.,
 601log, whatchanged, show) will perform text conversion. `git
 602format-patch` will never generate this output. If you want to
 603send somebody a text-converted diff of a binary file (e.g.,
 604because it quickly conveys the changes you have made), you
 605should generate it separately and send it as a comment _in
 606addition to_ the usual binary diff that you might send.
 607
 608Because text conversion can be slow, especially when doing a
 609large number of them with `git log -p`, Git provides a mechanism
 610to cache the output and use it in future diffs.  To enable
 611caching, set the "cachetextconv" variable in your diff driver's
 612config. For example:
 613
 614------------------------
 615[diff "jpg"]
 616        textconv = exif
 617        cachetextconv = true
 618------------------------
 619
 620This will cache the result of running "exif" on each blob
 621indefinitely. If you change the textconv config variable for a
 622diff driver, Git will automatically invalidate the cache entries
 623and re-run the textconv filter. If you want to invalidate the
 624cache manually (e.g., because your version of "exif" was updated
 625and now produces better output), you can remove the cache
 626manually with `git update-ref -d refs/notes/textconv/jpg` (where
 627"jpg" is the name of the diff driver, as in the example above).
 628
 629Choosing textconv versus external diff
 630^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 631
 632If you want to show differences between binary or specially-formatted
 633blobs in your repository, you can choose to use either an external diff
 634command, or to use textconv to convert them to a diff-able text format.
 635Which method you choose depends on your exact situation.
 636
 637The advantage of using an external diff command is flexibility. You are
 638not bound to find line-oriented changes, nor is it necessary for the
 639output to resemble unified diff. You are free to locate and report
 640changes in the most appropriate way for your data format.
 641
 642A textconv, by comparison, is much more limiting. You provide a
 643transformation of the data into a line-oriented text format, and Git
 644uses its regular diff tools to generate the output. There are several
 645advantages to choosing this method:
 646
 6471. Ease of use. It is often much simpler to write a binary to text
 648   transformation than it is to perform your own diff. In many cases,
 649   existing programs can be used as textconv filters (e.g., exif,
 650   odt2txt).
 651
 6522. Git diff features. By performing only the transformation step
 653   yourself, you can still utilize many of Git's diff features,
 654   including colorization, word-diff, and combined diffs for merges.
 655
 6563. Caching. Textconv caching can speed up repeated diffs, such as those
 657   you might trigger by running `git log -p`.
 658
 659
 660Marking files as binary
 661^^^^^^^^^^^^^^^^^^^^^^^
 662
 663Git usually guesses correctly whether a blob contains text or binary
 664data by examining the beginning of the contents. However, sometimes you
 665may want to override its decision, either because a blob contains binary
 666data later in the file, or because the content, while technically
 667composed of text characters, is opaque to a human reader. For example,
 668many postscript files contain only ascii characters, but produce noisy
 669and meaningless diffs.
 670
 671The simplest way to mark a file as binary is to unset the diff
 672attribute in the `.gitattributes` file:
 673
 674------------------------
 675*.ps -diff
 676------------------------
 677
 678This will cause Git to generate `Binary files differ` (or a binary
 679patch, if binary patches are enabled) instead of a regular diff.
 680
 681However, one may also want to specify other diff driver attributes. For
 682example, you might want to use `textconv` to convert postscript files to
 683an ascii representation for human viewing, but otherwise treat them as
 684binary files. You cannot specify both `-diff` and `diff=ps` attributes.
 685The solution is to use the `diff.*.binary` config option:
 686
 687------------------------
 688[diff "ps"]
 689  textconv = ps2ascii
 690  binary = true
 691------------------------
 692
 693Performing a three-way merge
 694~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 695
 696`merge`
 697^^^^^^^
 698
 699The attribute `merge` affects how three versions of a file are
 700merged when a file-level merge is necessary during `git merge`,
 701and other commands such as `git revert` and `git cherry-pick`.
 702
 703Set::
 704
 705        Built-in 3-way merge driver is used to merge the
 706        contents in a way similar to 'merge' command of `RCS`
 707        suite.  This is suitable for ordinary text files.
 708
 709Unset::
 710
 711        Take the version from the current branch as the
 712        tentative merge result, and declare that the merge has
 713        conflicts.  This is suitable for binary files that do
 714        not have a well-defined merge semantics.
 715
 716Unspecified::
 717
 718        By default, this uses the same built-in 3-way merge
 719        driver as is the case when the `merge` attribute is set.
 720        However, the `merge.default` configuration variable can name
 721        different merge driver to be used with paths for which the
 722        `merge` attribute is unspecified.
 723
 724String::
 725
 726        3-way merge is performed using the specified custom
 727        merge driver.  The built-in 3-way merge driver can be
 728        explicitly specified by asking for "text" driver; the
 729        built-in "take the current branch" driver can be
 730        requested with "binary".
 731
 732
 733Built-in merge drivers
 734^^^^^^^^^^^^^^^^^^^^^^
 735
 736There are a few built-in low-level merge drivers defined that
 737can be asked for via the `merge` attribute.
 738
 739text::
 740
 741        Usual 3-way file level merge for text files.  Conflicted
 742        regions are marked with conflict markers `<<<<<<<`,
 743        `=======` and `>>>>>>>`.  The version from your branch
 744        appears before the `=======` marker, and the version
 745        from the merged branch appears after the `=======`
 746        marker.
 747
 748binary::
 749
 750        Keep the version from your branch in the work tree, but
 751        leave the path in the conflicted state for the user to
 752        sort out.
 753
 754union::
 755
 756        Run 3-way file level merge for text files, but take
 757        lines from both versions, instead of leaving conflict
 758        markers.  This tends to leave the added lines in the
 759        resulting file in random order and the user should
 760        verify the result. Do not use this if you do not
 761        understand the implications.
 762
 763
 764Defining a custom merge driver
 765^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 766
 767The definition of a merge driver is done in the `.git/config`
 768file, not in the `gitattributes` file, so strictly speaking this
 769manual page is a wrong place to talk about it.  However...
 770
 771To define a custom merge driver `filfre`, add a section to your
 772`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 773
 774----------------------------------------------------------------
 775[merge "filfre"]
 776        name = feel-free merge driver
 777        driver = filfre %O %A %B
 778        recursive = binary
 779----------------------------------------------------------------
 780
 781The `merge.*.name` variable gives the driver a human-readable
 782name.
 783
 784The `merge.*.driver` variable's value is used to construct a
 785command to run to merge ancestor's version (`%O`), current
 786version (`%A`) and the other branches' version (`%B`).  These
 787three tokens are replaced with the names of temporary files that
 788hold the contents of these versions when the command line is
 789built. Additionally, %L will be replaced with the conflict marker
 790size (see below).
 791
 792The merge driver is expected to leave the result of the merge in
 793the file named with `%A` by overwriting it, and exit with zero
 794status if it managed to merge them cleanly, or non-zero if there
 795were conflicts.
 796
 797The `merge.*.recursive` variable specifies what other merge
 798driver to use when the merge driver is called for an internal
 799merge between common ancestors, when there are more than one.
 800When left unspecified, the driver itself is used for both
 801internal merge and the final merge.
 802
 803
 804`conflict-marker-size`
 805^^^^^^^^^^^^^^^^^^^^^^
 806
 807This attribute controls the length of conflict markers left in
 808the work tree file during a conflicted merge.  Only setting to
 809the value to a positive integer has any meaningful effect.
 810
 811For example, this line in `.gitattributes` can be used to tell the merge
 812machinery to leave much longer (instead of the usual 7-character-long)
 813conflict markers when merging the file `Documentation/git-merge.txt`
 814results in a conflict.
 815
 816------------------------
 817Documentation/git-merge.txt     conflict-marker-size=32
 818------------------------
 819
 820
 821Checking whitespace errors
 822~~~~~~~~~~~~~~~~~~~~~~~~~~
 823
 824`whitespace`
 825^^^^^^^^^^^^
 826
 827The `core.whitespace` configuration variable allows you to define what
 828'diff' and 'apply' should consider whitespace errors for all paths in
 829the project (See linkgit:git-config[1]).  This attribute gives you finer
 830control per path.
 831
 832Set::
 833
 834        Notice all types of potential whitespace errors known to Git.
 835        The tab width is taken from the value of the `core.whitespace`
 836        configuration variable.
 837
 838Unset::
 839
 840        Do not notice anything as error.
 841
 842Unspecified::
 843
 844        Use the value of the `core.whitespace` configuration variable to
 845        decide what to notice as error.
 846
 847String::
 848
 849        Specify a comma separate list of common whitespace problems to
 850        notice in the same format as the `core.whitespace` configuration
 851        variable.
 852
 853
 854Creating an archive
 855~~~~~~~~~~~~~~~~~~~
 856
 857`export-ignore`
 858^^^^^^^^^^^^^^^
 859
 860Files and directories with the attribute `export-ignore` won't be added to
 861archive files.
 862
 863`export-subst`
 864^^^^^^^^^^^^^^
 865
 866If the attribute `export-subst` is set for a file then Git will expand
 867several placeholders when adding this file to an archive.  The
 868expansion depends on the availability of a commit ID, i.e., if
 869linkgit:git-archive[1] has been given a tree instead of a commit or a
 870tag then no replacement will be done.  The placeholders are the same
 871as those for the option `--pretty=format:` of linkgit:git-log[1],
 872except that they need to be wrapped like this: `$Format:PLACEHOLDERS$`
 873in the file.  E.g. the string `$Format:%H$` will be replaced by the
 874commit hash.
 875
 876
 877Packing objects
 878~~~~~~~~~~~~~~~
 879
 880`delta`
 881^^^^^^^
 882
 883Delta compression will not be attempted for blobs for paths with the
 884attribute `delta` set to false.
 885
 886
 887Viewing files in GUI tools
 888~~~~~~~~~~~~~~~~~~~~~~~~~~
 889
 890`encoding`
 891^^^^^^^^^^
 892
 893The value of this attribute specifies the character encoding that should
 894be used by GUI tools (e.g. linkgit:gitk[1] and linkgit:git-gui[1]) to
 895display the contents of the relevant file. Note that due to performance
 896considerations linkgit:gitk[1] does not use this attribute unless you
 897manually enable per-file encodings in its options.
 898
 899If this attribute is not set or has an invalid value, the value of the
 900`gui.encoding` configuration variable is used instead
 901(See linkgit:git-config[1]).
 902
 903
 904USING MACRO ATTRIBUTES
 905----------------------
 906
 907You do not want any end-of-line conversions applied to, nor textual diffs
 908produced for, any binary file you track.  You would need to specify e.g.
 909
 910------------
 911*.jpg -text -diff
 912------------
 913
 914but that may become cumbersome, when you have many attributes.  Using
 915macro attributes, you can define an attribute that, when set, also
 916sets or unsets a number of other attributes at the same time.  The
 917system knows a built-in macro attribute, `binary`:
 918
 919------------
 920*.jpg binary
 921------------
 922
 923Setting the "binary" attribute also unsets the "text" and "diff"
 924attributes as above.  Note that macro attributes can only be "Set",
 925though setting one might have the effect of setting or unsetting other
 926attributes or even returning other attributes to the "Unspecified"
 927state.
 928
 929
 930DEFINING MACRO ATTRIBUTES
 931-------------------------
 932
 933Custom macro attributes can be defined only in top-level gitattributes
 934files (`$GIT_DIR/info/attributes`, the `.gitattributes` file at the
 935top level of the working tree, or the global or system-wide
 936gitattributes files), not in `.gitattributes` files in working tree
 937subdirectories.  The built-in macro attribute "binary" is equivalent
 938to:
 939
 940------------
 941[attr]binary -diff -merge -text
 942------------
 943
 944
 945EXAMPLE
 946-------
 947
 948If you have these three `gitattributes` file:
 949
 950----------------------------------------------------------------
 951(in $GIT_DIR/info/attributes)
 952
 953a*      foo !bar -baz
 954
 955(in .gitattributes)
 956abc     foo bar baz
 957
 958(in t/.gitattributes)
 959ab*     merge=filfre
 960abc     -foo -bar
 961*.c     frotz
 962----------------------------------------------------------------
 963
 964the attributes given to path `t/abc` are computed as follows:
 965
 9661. By examining `t/.gitattributes` (which is in the same
 967   directory as the path in question), Git finds that the first
 968   line matches.  `merge` attribute is set.  It also finds that
 969   the second line matches, and attributes `foo` and `bar`
 970   are unset.
 971
 9722. Then it examines `.gitattributes` (which is in the parent
 973   directory), and finds that the first line matches, but
 974   `t/.gitattributes` file already decided how `merge`, `foo`
 975   and `bar` attributes should be given to this path, so it
 976   leaves `foo` and `bar` unset.  Attribute `baz` is set.
 977
 9783. Finally it examines `$GIT_DIR/info/attributes`.  This file
 979   is used to override the in-tree settings.  The first line is
 980   a match, and `foo` is set, `bar` is reverted to unspecified
 981   state, and `baz` is unset.
 982
 983As the result, the attributes assignment to `t/abc` becomes:
 984
 985----------------------------------------------------------------
 986foo     set to true
 987bar     unspecified
 988baz     set to false
 989merge   set to string value "filfre"
 990frotz   unspecified
 991----------------------------------------------------------------
 992
 993
 994SEE ALSO
 995--------
 996linkgit:git-check-attr[1].
 997
 998GIT
 999---
1000Part of the linkgit:git[1] suite