Documentation / gitattributes.txton commit add tests for git diff --submodule (86140d5)
   1gitattributes(5)
   2================
   3
   4NAME
   5----
   6gitattributes - defining attributes per path
   7
   8SYNOPSIS
   9--------
  10$GIT_DIR/info/attributes, .gitattributes
  11
  12
  13DESCRIPTION
  14-----------
  15
  16A `gitattributes` file is a simple text file that gives
  17`attributes` to pathnames.
  18
  19Each line in `gitattributes` file is of form:
  20
  21        pattern attr1 attr2 ...
  22
  23That is, a pattern followed by an attributes list,
  24separated by whitespaces.  When the pattern matches the
  25path in question, the attributes listed on the line are given to
  26the path.
  27
  28Each attribute can be in one of these states for a given path:
  29
  30Set::
  31
  32        The path has the attribute with special value "true";
  33        this is specified by listing only the name of the
  34        attribute in the attribute list.
  35
  36Unset::
  37
  38        The path has the attribute with special value "false";
  39        this is specified by listing the name of the attribute
  40        prefixed with a dash `-` in the attribute list.
  41
  42Set to a value::
  43
  44        The path has the attribute with specified string value;
  45        this is specified by listing the name of the attribute
  46        followed by an equal sign `=` and its value in the
  47        attribute list.
  48
  49Unspecified::
  50
  51        No pattern matches the path, and nothing says if
  52        the path has or does not have the attribute, the
  53        attribute for the path is said to be Unspecified.
  54
  55When more than one pattern matches the path, a later line
  56overrides an earlier line.  This overriding is done per
  57attribute.  The rules how the pattern matches paths are the
  58same as in `.gitignore` files; see linkgit:gitignore[5].
  59
  60When deciding what attributes are assigned to a path, git
  61consults `$GIT_DIR/info/attributes` file (which has the highest
  62precedence), `.gitattributes` file in the same directory as the
  63path in question, and its parent directories up to the toplevel of the
  64work tree (the further the directory that contains `.gitattributes`
  65is from the path in question, the lower its precedence).
  66
  67If you wish to affect only a single repository (i.e., to assign
  68attributes to files that are particular to one user's workflow), then
  69attributes should be placed in the `$GIT_DIR/info/attributes` file.
  70Attributes which should be version-controlled and distributed to other
  71repositories (i.e., attributes of interest to all users) should go into
  72`.gitattributes` files.
  73
  74Sometimes you would need to override an setting of an attribute
  75for a path to `unspecified` state.  This can be done by listing
  76the name of the attribute prefixed with an exclamation point `!`.
  77
  78
  79EFFECTS
  80-------
  81
  82Certain operations by git can be influenced by assigning
  83particular attributes to a path.  Currently, the following
  84operations are attributes-aware.
  85
  86Checking-out and checking-in
  87~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  88
  89These attributes affect how the contents stored in the
  90repository are copied to the working tree files when commands
  91such as 'git-checkout' and 'git-merge' run.  They also affect how
  92git stores the contents you prepare in the working tree in the
  93repository upon 'git-add' and 'git-commit'.
  94
  95`crlf`
  96^^^^^^
  97
  98This attribute controls the line-ending convention.
  99
 100Set::
 101
 102        Setting the `crlf` attribute on a path is meant to mark
 103        the path as a "text" file.  'core.autocrlf' conversion
 104        takes place without guessing the content type by
 105        inspection.
 106
 107Unset::
 108
 109        Unsetting the `crlf` attribute on a path tells git not to
 110        attempt any end-of-line conversion upon checkin or checkout.
 111
 112Unspecified::
 113
 114        Unspecified `crlf` attribute tells git to apply the
 115        `core.autocrlf` conversion when the file content looks
 116        like text.
 117
 118Set to string value "input"::
 119
 120        This is similar to setting the attribute to `true`, but
 121        also forces git to act as if `core.autocrlf` is set to
 122        `input` for the path.
 123
 124Any other value set to `crlf` attribute is ignored and git acts
 125as if the attribute is left unspecified.
 126
 127
 128The `core.autocrlf` conversion
 129^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 130
 131If the configuration variable `core.autocrlf` is false, no
 132conversion is done.
 133
 134When `core.autocrlf` is true, it means that the platform wants
 135CRLF line endings for files in the working tree, and you want to
 136convert them back to the normal LF line endings when checking
 137in to the repository.
 138
 139When `core.autocrlf` is set to "input", line endings are
 140converted to LF upon checkin, but there is no conversion done
 141upon checkout.
 142
 143If `core.safecrlf` is set to "true" or "warn", git verifies if
 144the conversion is reversible for the current setting of
 145`core.autocrlf`.  For "true", git rejects irreversible
 146conversions; for "warn", git only prints a warning but accepts
 147an irreversible conversion.  The safety triggers to prevent such
 148a conversion done to the files in the work tree, but there are a
 149few exceptions.  Even though...
 150
 151- 'git-add' itself does not touch the files in the work tree, the
 152  next checkout would, so the safety triggers;
 153
 154- 'git-apply' to update a text file with a patch does touch the files
 155  in the work tree, but the operation is about text files and CRLF
 156  conversion is about fixing the line ending inconsistencies, so the
 157  safety does not trigger;
 158
 159- 'git-diff' itself does not touch the files in the work tree, it is
 160  often run to inspect the changes you intend to next 'git-add'.  To
 161  catch potential problems early, safety triggers.
 162
 163
 164`ident`
 165^^^^^^^
 166
 167When the attribute `ident` is set for a path, git replaces
 168`$Id$` in the blob object with `$Id:`, followed by the
 16940-character hexadecimal blob object name, followed by a dollar
 170sign `$` upon checkout.  Any byte sequence that begins with
 171`$Id:` and ends with `$` in the worktree file is replaced
 172with `$Id$` upon check-in.
 173
 174
 175`filter`
 176^^^^^^^^
 177
 178A `filter` attribute can be set to a string value that names a
 179filter driver specified in the configuration.
 180
 181A filter driver consists of a `clean` command and a `smudge`
 182command, either of which can be left unspecified.  Upon
 183checkout, when the `smudge` command is specified, the command is
 184fed the blob object from its standard input, and its standard
 185output is used to update the worktree file.  Similarly, the
 186`clean` command is used to convert the contents of worktree file
 187upon checkin.
 188
 189A missing filter driver definition in the config is not an error
 190but makes the filter a no-op passthru.
 191
 192The content filtering is done to massage the content into a
 193shape that is more convenient for the platform, filesystem, and
 194the user to use.  The key phrase here is "more convenient" and not
 195"turning something unusable into usable".  In other words, the
 196intent is that if someone unsets the filter driver definition,
 197or does not have the appropriate filter program, the project
 198should still be usable.
 199
 200
 201Interaction between checkin/checkout attributes
 202^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 203
 204In the check-in codepath, the worktree file is first converted
 205with `filter` driver (if specified and corresponding driver
 206defined), then the result is processed with `ident` (if
 207specified), and then finally with `crlf` (again, if specified
 208and applicable).
 209
 210In the check-out codepath, the blob content is first converted
 211with `crlf`, and then `ident` and fed to `filter`.
 212
 213
 214Generating diff text
 215~~~~~~~~~~~~~~~~~~~~
 216
 217`diff`
 218^^^^^^
 219
 220The attribute `diff` affects how 'git' generates diffs for particular
 221files. It can tell git whether to generate a textual patch for the path
 222or to treat the path as a binary file.  It can also affect what line is
 223shown on the hunk header `@@ -k,l +n,m @@` line, tell git to use an
 224external command to generate the diff, or ask git to convert binary
 225files to a text format before generating the diff.
 226
 227Set::
 228
 229        A path to which the `diff` attribute is set is treated
 230        as text, even when they contain byte values that
 231        normally never appear in text files, such as NUL.
 232
 233Unset::
 234
 235        A path to which the `diff` attribute is unset will
 236        generate `Binary files differ` (or a binary patch, if
 237        binary patches are enabled).
 238
 239Unspecified::
 240
 241        A path to which the `diff` attribute is unspecified
 242        first gets its contents inspected, and if it looks like
 243        text, it is treated as text.  Otherwise it would
 244        generate `Binary files differ`.
 245
 246String::
 247
 248        Diff is shown using the specified diff driver.  Each driver may
 249        specify one or more options, as described in the following
 250        section. The options for the diff driver "foo" are defined
 251        by the configuration variables in the "diff.foo" section of the
 252        git config file.
 253
 254
 255Defining an external diff driver
 256^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 257
 258The definition of a diff driver is done in `gitconfig`, not
 259`gitattributes` file, so strictly speaking this manual page is a
 260wrong place to talk about it.  However...
 261
 262To define an external diff driver `jcdiff`, add a section to your
 263`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 264
 265----------------------------------------------------------------
 266[diff "jcdiff"]
 267        command = j-c-diff
 268----------------------------------------------------------------
 269
 270When git needs to show you a diff for the path with `diff`
 271attribute set to `jcdiff`, it calls the command you specified
 272with the above configuration, i.e. `j-c-diff`, with 7
 273parameters, just like `GIT_EXTERNAL_DIFF` program is called.
 274See linkgit:git[1] for details.
 275
 276
 277Defining a custom hunk-header
 278^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 279
 280Each group of changes (called a "hunk") in the textual diff output
 281is prefixed with a line of the form:
 282
 283        @@ -k,l +n,m @@ TEXT
 284
 285This is called a 'hunk header'.  The "TEXT" portion is by default a line
 286that begins with an alphabet, an underscore or a dollar sign; this
 287matches what GNU 'diff -p' output uses.  This default selection however
 288is not suited for some contents, and you can use a customized pattern
 289to make a selection.
 290
 291First, in .gitattributes, you would assign the `diff` attribute
 292for paths.
 293
 294------------------------
 295*.tex   diff=tex
 296------------------------
 297
 298Then, you would define a "diff.tex.xfuncname" configuration to
 299specify a regular expression that matches a line that you would
 300want to appear as the hunk header "TEXT". Add a section to your
 301`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 302
 303------------------------
 304[diff "tex"]
 305        xfuncname = "^(\\\\(sub)*section\\{.*)$"
 306------------------------
 307
 308Note.  A single level of backslashes are eaten by the
 309configuration file parser, so you would need to double the
 310backslashes; the pattern above picks a line that begins with a
 311backslash, and zero or more occurrences of `sub` followed by
 312`section` followed by open brace, to the end of line.
 313
 314There are a few built-in patterns to make this easier, and `tex`
 315is one of them, so you do not have to write the above in your
 316configuration file (you still need to enable this with the
 317attribute mechanism, via `.gitattributes`).  The following built in
 318patterns are available:
 319
 320- `bibtex` suitable for files with BibTeX coded references.
 321
 322- `cpp` suitable for source code in the C and C++ languages.
 323
 324- `html` suitable for HTML/XHTML documents.
 325
 326- `java` suitable for source code in the Java language.
 327
 328- `objc` suitable for source code in the Objective-C language.
 329
 330- `pascal` suitable for source code in the Pascal/Delphi language.
 331
 332- `php` suitable for source code in the PHP language.
 333
 334- `python` suitable for source code in the Python language.
 335
 336- `ruby` suitable for source code in the Ruby language.
 337
 338- `tex` suitable for source code for LaTeX documents.
 339
 340
 341Customizing word diff
 342^^^^^^^^^^^^^^^^^^^^^
 343
 344You can customize the rules that `git diff --color-words` uses to
 345split words in a line, by specifying an appropriate regular expression
 346in the "diff.*.wordRegex" configuration variable.  For example, in TeX
 347a backslash followed by a sequence of letters forms a command, but
 348several such commands can be run together without intervening
 349whitespace.  To separate them, use a regular expression in your
 350`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 351
 352------------------------
 353[diff "tex"]
 354        wordRegex = "\\\\[a-zA-Z]+|[{}]|\\\\.|[^\\{}[:space:]]+"
 355------------------------
 356
 357A built-in pattern is provided for all languages listed in the
 358previous section.
 359
 360
 361Performing text diffs of binary files
 362^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 363
 364Sometimes it is desirable to see the diff of a text-converted
 365version of some binary files. For example, a word processor
 366document can be converted to an ASCII text representation, and
 367the diff of the text shown. Even though this conversion loses
 368some information, the resulting diff is useful for human
 369viewing (but cannot be applied directly).
 370
 371The `textconv` config option is used to define a program for
 372performing such a conversion. The program should take a single
 373argument, the name of a file to convert, and produce the
 374resulting text on stdout.
 375
 376For example, to show the diff of the exif information of a
 377file instead of the binary information (assuming you have the
 378exif tool installed), add the following section to your
 379`$GIT_DIR/config` file (or `$HOME/.gitconfig` file):
 380
 381------------------------
 382[diff "jpg"]
 383        textconv = exif
 384------------------------
 385
 386NOTE: The text conversion is generally a one-way conversion;
 387in this example, we lose the actual image contents and focus
 388just on the text data. This means that diffs generated by
 389textconv are _not_ suitable for applying. For this reason,
 390only `git diff` and the `git log` family of commands (i.e.,
 391log, whatchanged, show) will perform text conversion. `git
 392format-patch` will never generate this output. If you want to
 393send somebody a text-converted diff of a binary file (e.g.,
 394because it quickly conveys the changes you have made), you
 395should generate it separately and send it as a comment _in
 396addition to_ the usual binary diff that you might send.
 397
 398
 399Performing a three-way merge
 400~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 401
 402`merge`
 403^^^^^^^
 404
 405The attribute `merge` affects how three versions of a file is
 406merged when a file-level merge is necessary during `git merge`,
 407and other commands such as `git revert` and `git cherry-pick`.
 408
 409Set::
 410
 411        Built-in 3-way merge driver is used to merge the
 412        contents in a way similar to 'merge' command of `RCS`
 413        suite.  This is suitable for ordinary text files.
 414
 415Unset::
 416
 417        Take the version from the current branch as the
 418        tentative merge result, and declare that the merge has
 419        conflicts.  This is suitable for binary files that does
 420        not have a well-defined merge semantics.
 421
 422Unspecified::
 423
 424        By default, this uses the same built-in 3-way merge
 425        driver as is the case the `merge` attribute is set.
 426        However, `merge.default` configuration variable can name
 427        different merge driver to be used for paths to which the
 428        `merge` attribute is unspecified.
 429
 430String::
 431
 432        3-way merge is performed using the specified custom
 433        merge driver.  The built-in 3-way merge driver can be
 434        explicitly specified by asking for "text" driver; the
 435        built-in "take the current branch" driver can be
 436        requested with "binary".
 437
 438
 439Built-in merge drivers
 440^^^^^^^^^^^^^^^^^^^^^^
 441
 442There are a few built-in low-level merge drivers defined that
 443can be asked for via the `merge` attribute.
 444
 445text::
 446
 447        Usual 3-way file level merge for text files.  Conflicted
 448        regions are marked with conflict markers `<<<<<<<`,
 449        `=======` and `>>>>>>>`.  The version from your branch
 450        appears before the `=======` marker, and the version
 451        from the merged branch appears after the `=======`
 452        marker.
 453
 454binary::
 455
 456        Keep the version from your branch in the work tree, but
 457        leave the path in the conflicted state for the user to
 458        sort out.
 459
 460union::
 461
 462        Run 3-way file level merge for text files, but take
 463        lines from both versions, instead of leaving conflict
 464        markers.  This tends to leave the added lines in the
 465        resulting file in random order and the user should
 466        verify the result. Do not use this if you do not
 467        understand the implications.
 468
 469
 470Defining a custom merge driver
 471^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 472
 473The definition of a merge driver is done in the `.git/config`
 474file, not in the `gitattributes` file, so strictly speaking this
 475manual page is a wrong place to talk about it.  However...
 476
 477To define a custom merge driver `filfre`, add a section to your
 478`$GIT_DIR/config` file (or `$HOME/.gitconfig` file) like this:
 479
 480----------------------------------------------------------------
 481[merge "filfre"]
 482        name = feel-free merge driver
 483        driver = filfre %O %A %B
 484        recursive = binary
 485----------------------------------------------------------------
 486
 487The `merge.*.name` variable gives the driver a human-readable
 488name.
 489
 490The `merge.*.driver` variable's value is used to construct a
 491command to run to merge ancestor's version (`%O`), current
 492version (`%A`) and the other branches' version (`%B`).  These
 493three tokens are replaced with the names of temporary files that
 494hold the contents of these versions when the command line is
 495built.
 496
 497The merge driver is expected to leave the result of the merge in
 498the file named with `%A` by overwriting it, and exit with zero
 499status if it managed to merge them cleanly, or non-zero if there
 500were conflicts.
 501
 502The `merge.*.recursive` variable specifies what other merge
 503driver to use when the merge driver is called for an internal
 504merge between common ancestors, when there are more than one.
 505When left unspecified, the driver itself is used for both
 506internal merge and the final merge.
 507
 508
 509Checking whitespace errors
 510~~~~~~~~~~~~~~~~~~~~~~~~~~
 511
 512`whitespace`
 513^^^^^^^^^^^^
 514
 515The `core.whitespace` configuration variable allows you to define what
 516'diff' and 'apply' should consider whitespace errors for all paths in
 517the project (See linkgit:git-config[1]).  This attribute gives you finer
 518control per path.
 519
 520Set::
 521
 522        Notice all types of potential whitespace errors known to git.
 523
 524Unset::
 525
 526        Do not notice anything as error.
 527
 528Unspecified::
 529
 530        Use the value of `core.whitespace` configuration variable to
 531        decide what to notice as error.
 532
 533String::
 534
 535        Specify a comma separate list of common whitespace problems to
 536        notice in the same format as `core.whitespace` configuration
 537        variable.
 538
 539
 540Creating an archive
 541~~~~~~~~~~~~~~~~~~~
 542
 543`export-ignore`
 544^^^^^^^^^^^^^^^
 545
 546Files and directories with the attribute `export-ignore` won't be added to
 547archive files.
 548
 549`export-subst`
 550^^^^^^^^^^^^^^
 551
 552If the attribute `export-subst` is set for a file then git will expand
 553several placeholders when adding this file to an archive.  The
 554expansion depends on the availability of a commit ID, i.e., if
 555linkgit:git-archive[1] has been given a tree instead of a commit or a
 556tag then no replacement will be done.  The placeholders are the same
 557as those for the option `--pretty=format:` of linkgit:git-log[1],
 558except that they need to be wrapped like this: `$Format:PLACEHOLDERS$`
 559in the file.  E.g. the string `$Format:%H$` will be replaced by the
 560commit hash.
 561
 562
 563Viewing files in GUI tools
 564~~~~~~~~~~~~~~~~~~~~~~~~~~
 565
 566`encoding`
 567^^^^^^^^^^
 568
 569The value of this attribute specifies the character encoding that should
 570be used by GUI tools (e.g. linkgit:gitk[1] and linkgit:git-gui[1]) to
 571display the contents of the relevant file. Note that due to performance
 572considerations linkgit:gitk[1] does not use this attribute unless you
 573manually enable per-file encodings in its options.
 574
 575If this attribute is not set or has an invalid value, the value of the
 576`gui.encoding` configuration variable is used instead
 577(See linkgit:git-config[1]).
 578
 579
 580USING ATTRIBUTE MACROS
 581----------------------
 582
 583You do not want any end-of-line conversions applied to, nor textual diffs
 584produced for, any binary file you track.  You would need to specify e.g.
 585
 586------------
 587*.jpg -crlf -diff
 588------------
 589
 590but that may become cumbersome, when you have many attributes.  Using
 591attribute macros, you can specify groups of attributes set or unset at
 592the same time.  The system knows a built-in attribute macro, `binary`:
 593
 594------------
 595*.jpg binary
 596------------
 597
 598which is equivalent to the above.  Note that the attribute macros can only
 599be "Set" (see the above example that sets "binary" macro as if it were an
 600ordinary attribute --- setting it in turn unsets "crlf" and "diff").
 601
 602
 603DEFINING ATTRIBUTE MACROS
 604-------------------------
 605
 606Custom attribute macros can be defined only in the `.gitattributes` file
 607at the toplevel (i.e. not in any subdirectory).  The built-in attribute
 608macro "binary" is equivalent to:
 609
 610------------
 611[attr]binary -diff -crlf
 612------------
 613
 614
 615EXAMPLE
 616-------
 617
 618If you have these three `gitattributes` file:
 619
 620----------------------------------------------------------------
 621(in $GIT_DIR/info/attributes)
 622
 623a*      foo !bar -baz
 624
 625(in .gitattributes)
 626abc     foo bar baz
 627
 628(in t/.gitattributes)
 629ab*     merge=filfre
 630abc     -foo -bar
 631*.c     frotz
 632----------------------------------------------------------------
 633
 634the attributes given to path `t/abc` are computed as follows:
 635
 6361. By examining `t/.gitattributes` (which is in the same
 637   directory as the path in question), git finds that the first
 638   line matches.  `merge` attribute is set.  It also finds that
 639   the second line matches, and attributes `foo` and `bar`
 640   are unset.
 641
 6422. Then it examines `.gitattributes` (which is in the parent
 643   directory), and finds that the first line matches, but
 644   `t/.gitattributes` file already decided how `merge`, `foo`
 645   and `bar` attributes should be given to this path, so it
 646   leaves `foo` and `bar` unset.  Attribute `baz` is set.
 647
 6483. Finally it examines `$GIT_DIR/info/attributes`.  This file
 649   is used to override the in-tree settings.  The first line is
 650   a match, and `foo` is set, `bar` is reverted to unspecified
 651   state, and `baz` is unset.
 652
 653As the result, the attributes assignment to `t/abc` becomes:
 654
 655----------------------------------------------------------------
 656foo     set to true
 657bar     unspecified
 658baz     set to false
 659merge   set to string value "filfre"
 660frotz   unspecified
 661----------------------------------------------------------------
 662
 663
 664
 665GIT
 666---
 667Part of the linkgit:git[1] suite