1GITATTRIBUTES(5) Git Manual GITATTRIBUTES(5)
2
6 gitattributes - Defining attributes per path
7
9 $GIT_DIR/info/attributes, .gitattributes
10
12 A gitattributes file is a simple text file that gives attributes to
13 pathnames.
14 Each line in gitattributes file is of form:
16 pattern attr1 attr2 ...
18 That is, a pattern followed by an attributes list, separated by
20 whitespaces. Leading and trailing whitespaces are ignored. Lines that
21 begin with # are ignored. Patterns that begin with a double quote are
22 quoted in C style. When the pattern matches the path in question, the
23 attributes listed on the line are given to the path.
24 Each attribute can be in one of these states for a given path:
26 Set
28 The path has the attribute with special value "true"; this is
29 specified by listing only the name of the attribute in the
30 attribute list.
31 Unset
33 The path has the attribute with special value "false"; this is
34 specified by listing the name of the attribute prefixed with a dash
35 - in the attribute list.
36 Set to a value
38 The path has the attribute with specified string value; this is
39 specified by listing the name of the attribute followed by an equal
40 sign = and its value in the attribute list.
41 Unspecified
43 No pattern matches the path, and nothing says if the path has or
44 does not have the attribute, the attribute for the path is said to
45 be Unspecified.
46 When more than one pattern matches the path, a later line overrides an
48 earlier line. This overriding is done per attribute.
49 The rules by which the pattern matches paths are the same as in
51 .gitignore files (see gitignore(5)), with a few exceptions:
52 • negative patterns are forbidden
54 • patterns that match a directory do not recursively match paths
56 inside that directory (so using the trailing-slash path/ syntax is
57 pointless in an attributes file; use path/** instead)
58 When deciding what attributes are assigned to a path, Git consults
60 $GIT_DIR/info/attributes file (which has the highest precedence),
61 .gitattributes file in the same directory as the path in question, and
62 its parent directories up to the toplevel of the work tree (the further
63 the directory that contains .gitattributes is from the path in
64 question, the lower its precedence). Finally global and system-wide
65 files are considered (they have the lowest precedence).
66 When the .gitattributes file is missing from the work tree, the path in
68 the index is used as a fall-back. During checkout process,
69 .gitattributes in the index is used and then the file in the working
70 tree is used as a fall-back.
71 If you wish to affect only a single repository (i.e., to assign
73 attributes to files that are particular to one user’s workflow for that
74 repository), then attributes should be placed in the
75 $GIT_DIR/info/attributes file. Attributes which should be
76 version-controlled and distributed to other repositories (i.e.,
77 attributes of interest to all users) should go into .gitattributes
78 files. Attributes that should affect all repositories for a single user
79 should be placed in a file specified by the core.attributesFile
80 configuration option (see git-config(1)). Its default value is
81 $XDG_CONFIG_HOME/git/attributes. If $XDG_CONFIG_HOME is either not set
82 or empty, $HOME/.config/git/attributes is used instead. Attributes for
83 all users on a system should be placed in the
84 $(prefix)/etc/gitattributes file.
85 Sometimes you would need to override a setting of an attribute for a
87 path to Unspecified state. This can be done by listing the name of the
88 attribute prefixed with an exclamation point !.
89
91 Certain operations by Git can be influenced by assigning particular
92 attributes to a path. Currently, the following operations are
93 attributes-aware.
94 Checking-out and checking-in
96 These attributes affect how the contents stored in the repository are
97 copied to the working tree files when commands such as git switch, git
98 checkout and git merge run. They also affect how Git stores the
99 contents you prepare in the working tree in the repository upon git add
100 and git commit.
101 text
103 This attribute marks the path as a text file, which enables
104 end-of-line conversion: When a matching file is added to the index,
105 the file’s line endings are normalized to LF in the index.
106 Conversely, when the file is copied from the index to the working
107 directory, its line endings may be converted from LF to CRLF
108 depending on the eol attribute, the Git config, and the platform
109 (see explanation of eol below).
110 Set
112 Setting the text attribute on a path enables end-of-line
113 conversion on checkin and checkout as described above. Line
114 endings are normalized to LF in the index every time the file
115 is checked in, even if the file was previously added to Git
116 with CRLF line endings.
117 Unset
119 Unsetting the text attribute on a path tells Git not to attempt
120 any end-of-line conversion upon checkin or checkout.
121 Set to string value "auto"
123 When text is set to "auto", Git decides by itself whether the
124 file is text or binary. If it is text and the file was not
125 already in Git with CRLF endings, line endings are converted on
126 checkin and checkout as described above. Otherwise, no
127 conversion is done on checkin or checkout.
128 Unspecified
130 If the text attribute is unspecified, Git uses the
131 core.autocrlf configuration variable to determine if the file
132 should be converted.
133 Any other value causes Git to act as if text has been left
135 unspecified.
136 eol
138 This attribute marks a path to use a specific line-ending style in
139 the working tree when it is checked out. It has effect only if text
140 or text=auto is set (see above), but specifying eol automatically
141 sets text if text was left unspecified.
142 Set to string value "crlf"
144 This setting converts the file’s line endings in the working
145 directory to CRLF when the file is checked out.
146 Set to string value "lf"
148 This setting uses the same line endings in the working
149 directory as in the index when the file is checked out.
150 Unspecified
152 If the eol attribute is unspecified for a file, its line
153 endings in the working directory are determined by the
154 core.autocrlf or core.eol configuration variable (see the
155 definitions of those options in git-config(1)). If text is set
156 but neither of those variables is, the default is eol=crlf on
157 Windows and eol=lf on all other platforms.
158 Backwards compatibility with crlf attribute
160 For backwards compatibility, the crlf attribute is interpreted as
161 follows:
162 crlf text
164 -crlf -text
165 crlf=input eol=lf
166 End-of-line conversion
168 While Git normally leaves file contents alone, it can be configured
169 to normalize line endings to LF in the repository and, optionally,
170 to convert them to CRLF when files are checked out.
171 If you simply want to have CRLF line endings in your working
173 directory regardless of the repository you are working with, you
174 can set the config variable "core.autocrlf" without using any
175 attributes.
176 [core]
178 autocrlf = true
179 This does not force normalization of text files, but does ensure
181 that text files that you introduce to the repository have their
182 line endings normalized to LF when they are added, and that files
183 that are already normalized in the repository stay normalized.
184 If you want to ensure that text files that any contributor
186 introduces to the repository have their line endings normalized,
187 you can set the text attribute to "auto" for all files.
188 * text=auto
190 The attributes allow a fine-grained control, how the line endings
192 are converted. Here is an example that will make Git normalize
193 .txt, .vcproj and .sh files, ensure that .vcproj files have CRLF
194 and .sh files have LF in the working directory, and prevent .jpg
195 files from being normalized regardless of their content.
196 * text=auto
198 *.txt text
199 *.vcproj text eol=crlf
200 *.sh text eol=lf
201 *.jpg -text
202 Note
205 When text=auto conversion is enabled in a cross-platform
206 project using push and pull to a central repository the text
207 files containing CRLFs should be normalized.
208 From a clean working directory:
210 $ echo "* text=auto" >.gitattributes
212 $ git add --renormalize .
213 $ git status # Show files that will be normalized
214 $ git commit -m "Introduce end-of-line normalization"
215 If any files that should not be normalized show up in git status,
217 unset their text attribute before running git add -u.
218 manual.pdf -text
220 Conversely, text files that Git does not detect can have
222 normalization enabled manually.
223 weirdchars.txt text
225 If core.safecrlf is set to "true" or "warn", Git verifies if the
227 conversion is reversible for the current setting of core.autocrlf.
228 For "true", Git rejects irreversible conversions; for "warn", Git
229 only prints a warning but accepts an irreversible conversion. The
230 safety triggers to prevent such a conversion done to the files in
231 the work tree, but there are a few exceptions. Even though...
232 • git add itself does not touch the files in the work tree, the
234 next checkout would, so the safety triggers;
235 • git apply to update a text file with a patch does touch the
237 files in the work tree, but the operation is about text files
238 and CRLF conversion is about fixing the line ending
239 inconsistencies, so the safety does not trigger;
240 • git diff itself does not touch the files in the work tree, it
242 is often run to inspect the changes you intend to next git add.
243 To catch potential problems early, safety triggers.
244 working-tree-encoding
246 Git recognizes files encoded in ASCII or one of its supersets (e.g.
247 UTF-8, ISO-8859-1, ...) as text files. Files encoded in certain
248 other encodings (e.g. UTF-16) are interpreted as binary and
249 consequently built-in Git text processing tools (e.g. git diff) as
250 well as most Git web front ends do not visualize the contents of
251 these files by default.
252 In these cases you can tell Git the encoding of a file in the
254 working directory with the working-tree-encoding attribute. If a
255 file with this attribute is added to Git, then Git re-encodes the
256 content from the specified encoding to UTF-8. Finally, Git stores
257 the UTF-8 encoded content in its internal data structure (called
258 "the index"). On checkout the content is re-encoded back to the
259 specified encoding.
260 Please note that using the working-tree-encoding attribute may have
262 a number of pitfalls:
263 • Alternative Git implementations (e.g. JGit or libgit2) and
265 older Git versions (as of March 2018) do not support the
266 working-tree-encoding attribute. If you decide to use the
267 working-tree-encoding attribute in your repository, then it is
268 strongly recommended to ensure that all clients working with
269 the repository support it.
270 For example, Microsoft Visual Studio resources files (*.rc) or
272 PowerShell script files (*.ps1) are sometimes encoded in
273 UTF-16. If you declare *.ps1 as files as UTF-16 and you add
274 foo.ps1 with a working-tree-encoding enabled Git client, then
275 foo.ps1 will be stored as UTF-8 internally. A client without
276 working-tree-encoding support will checkout foo.ps1 as UTF-8
277 encoded file. This will typically cause trouble for the users
278 of this file.
279 If a Git client that does not support the working-tree-encoding
281 attribute adds a new file bar.ps1, then bar.ps1 will be stored
282 "as-is" internally (in this example probably as UTF-16). A
283 client with working-tree-encoding support will interpret the
284 internal contents as UTF-8 and try to convert it to UTF-16 on
285 checkout. That operation will fail and cause an error.
286 • Reencoding content to non-UTF encodings can cause errors as the
288 conversion might not be UTF-8 round trip safe. If you suspect
289 your encoding to not be round trip safe, then add it to
290 core.checkRoundtripEncoding to make Git check the round trip
291 encoding (see git-config(1)). SHIFT-JIS (Japanese character
292 set) is known to have round trip issues with UTF-8 and is
293 checked by default.
294 • Reencoding content requires resources that might slow down
296 certain Git operations (e.g git checkout or git add).
297 Use the working-tree-encoding attribute only if you cannot store a
299 file in UTF-8 encoding and if you want Git to be able to process
300 the content as text.
301 As an example, use the following attributes if your *.ps1 files are
303 UTF-16 encoded with byte order mark (BOM) and you want Git to
304 perform automatic line ending conversion based on your platform.
305 *.ps1 text working-tree-encoding=UTF-16
307 Use the following attributes if your *.ps1 files are UTF-16 little
309 endian encoded without BOM and you want Git to use Windows line
310 endings in the working directory (use UTF-16LE-BOM instead of
311 UTF-16LE if you want UTF-16 little endian with BOM). Please note,
312 it is highly recommended to explicitly define the line endings with
313 eol if the working-tree-encoding attribute is used to avoid
314 ambiguity.
315 *.ps1 text working-tree-encoding=UTF-16LE eol=CRLF
317 You can get a list of all available encodings on your platform with
319 the following command:
320 iconv --list
322 If you do not know the encoding of a file, then you can use the
324 file command to guess the encoding:
325 file foo.ps1
327 ident
329 When the attribute ident is set for a path, Git replaces $Id$ in
330 the blob object with $Id:, followed by the 40-character hexadecimal
331 blob object name, followed by a dollar sign $ upon checkout. Any
332 byte sequence that begins with $Id: and ends with $ in the worktree
333 file is replaced with $Id$ upon check-in.
334 filter
336 A filter attribute can be set to a string value that names a filter
337 driver specified in the configuration.
338 A filter driver consists of a clean command and a smudge command,
340 either of which can be left unspecified. Upon checkout, when the
341 smudge command is specified, the command is fed the blob object
342 from its standard input, and its standard output is used to update
343 the worktree file. Similarly, the clean command is used to convert
344 the contents of worktree file upon checkin. By default these
345 commands process only a single blob and terminate. If a long
346 running process filter is used in place of clean and/or smudge
347 filters, then Git can process all blobs with a single filter
348 command invocation for the entire life of a single Git command, for
349 example git add --all. If a long running process filter is
350 configured then it always takes precedence over a configured single
351 blob filter. See section below for the description of the protocol
352 used to communicate with a process filter.
353 One use of the content filtering is to massage the content into a
355 shape that is more convenient for the platform, filesystem, and the
356 user to use. For this mode of operation, the key phrase here is
357 "more convenient" and not "turning something unusable into usable".
358 In other words, the intent is that if someone unsets the filter
359 driver definition, or does not have the appropriate filter program,
360 the project should still be usable.
361 Another use of the content filtering is to store the content that
363 cannot be directly used in the repository (e.g. a UUID that refers
364 to the true content stored outside Git, or an encrypted content)
365 and turn it into a usable form upon checkout (e.g. download the
366 external content, or decrypt the encrypted content).
367 These two filters behave differently, and by default, a filter is
369 taken as the former, massaging the contents into more convenient
370 shape. A missing filter driver definition in the config, or a
371 filter driver that exits with a non-zero status, is not an error
372 but makes the filter a no-op passthru.
373 You can declare that a filter turns a content that by itself is
375 unusable into a usable content by setting the
376 filter.<driver>.required configuration variable to true.
377 Note: Whenever the clean filter is changed, the repo should be
379 renormalized: $ git add --renormalize .
380 For example, in .gitattributes, you would assign the filter
382 attribute for paths.
383 *.c filter=indent
385 Then you would define a "filter.indent.clean" and
387 "filter.indent.smudge" configuration in your .git/config to specify
388 a pair of commands to modify the contents of C programs when the
389 source files are checked in ("clean" is run) and checked out (no
390 change is made because the command is "cat").
391 [filter "indent"]
393 clean = indent
394 smudge = cat
395 For best results, clean should not alter its output further if it
397 is run twice ("clean→clean" should be equivalent to "clean"), and
398 multiple smudge commands should not alter clean's output
399 ("smudge→smudge→clean" should be equivalent to "clean"). See the
400 section on merging below.
401 The "indent" filter is well-behaved in this regard: it will not
403 modify input that is already correctly indented. In this case, the
404 lack of a smudge filter means that the clean filter must accept its
405 own output without modifying it.
406 If a filter must succeed in order to make the stored contents
408 usable, you can declare that the filter is required, in the
409 configuration:
410 [filter "crypt"]
412 clean = openssl enc ...
413 smudge = openssl enc -d ...
414 required
415 Sequence "%f" on the filter command line is replaced with the name
417 of the file the filter is working on. A filter might use this in
418 keyword substitution. For example:
419 [filter "p4"]
421 clean = git-p4-filter --clean %f
422 smudge = git-p4-filter --smudge %f
423 Note that "%f" is the name of the path that is being worked on.
425 Depending on the version that is being filtered, the corresponding
426 file on disk may not exist, or may have different contents. So,
427 smudge and clean commands should not try to access the file on
428 disk, but only act as filters on the content provided to them on
429 standard input.
430 Long Running Filter Process
432 If the filter command (a string value) is defined via
433 filter.<driver>.process then Git can process all blobs with a
434 single filter invocation for the entire life of a single Git
435 command. This is achieved by using the long-running process
436 protocol (described in
437 technical/long-running-process-protocol.txt).
438 When Git encounters the first file that needs to be cleaned or
440 smudged, it starts the filter and performs the handshake. In the
441 handshake, the welcome message sent by Git is "git-filter-client",
442 only version 2 is supported, and the supported capabilities are
443 "clean", "smudge", and "delay".
444 Afterwards Git sends a list of "key=value" pairs terminated with a
446 flush packet. The list will contain at least the filter command
447 (based on the supported capabilities) and the pathname of the file
448 to filter relative to the repository root. Right after the flush
449 packet Git sends the content split in zero or more pkt-line packets
450 and a flush packet to terminate content. Please note, that the
451 filter must not send any response before it received the content
452 and the final flush packet. Also note that the "value" of a
453 "key=value" pair can contain the "=" character whereas the key
454 would never contain that character.
455 packet: git> command=smudge
457 packet: git> pathname=path/testfile.dat
458 packet: git> 0000
459 packet: git> CONTENT
460 packet: git> 0000
461 The filter is expected to respond with a list of "key=value" pairs
463 terminated with a flush packet. If the filter does not experience
464 problems then the list must contain a "success" status. Right after
465 these packets the filter is expected to send the content in zero or
466 more pkt-line packets and a flush packet at the end. Finally, a
467 second list of "key=value" pairs terminated with a flush packet is
468 expected. The filter can change the status in the second list or
469 keep the status as is with an empty list. Please note that the
470 empty list must be terminated with a flush packet regardless.
471 packet: git< status=success
473 packet: git< 0000
474 packet: git< SMUDGED_CONTENT
475 packet: git< 0000
476 packet: git< 0000 # empty list, keep "status=success" unchanged!
477 If the result content is empty then the filter is expected to
479 respond with a "success" status and a flush packet to signal the
480 empty content.
481 packet: git< status=success
483 packet: git< 0000
484 packet: git< 0000 # empty content!
485 packet: git< 0000 # empty list, keep "status=success" unchanged!
486 In case the filter cannot or does not want to process the content,
488 it is expected to respond with an "error" status.
489 packet: git< status=error
491 packet: git< 0000
492 If the filter experiences an error during processing, then it can
494 send the status "error" after the content was (partially or
495 completely) sent.
496 packet: git< status=success
498 packet: git< 0000
499 packet: git< HALF_WRITTEN_ERRONEOUS_CONTENT
500 packet: git< 0000
501 packet: git< status=error
502 packet: git< 0000
503 In case the filter cannot or does not want to process the content
505 as well as any future content for the lifetime of the Git process,
506 then it is expected to respond with an "abort" status at any point
507 in the protocol.
508 packet: git< status=abort
510 packet: git< 0000
511 Git neither stops nor restarts the filter process in case the
513 "error"/"abort" status is set. However, Git sets its exit code
514 according to the filter.<driver>.required flag, mimicking the
515 behavior of the filter.<driver>.clean / filter.<driver>.smudge
516 mechanism.
517 If the filter dies during the communication or does not adhere to
519 the protocol then Git will stop the filter process and restart it
520 with the next file that needs to be processed. Depending on the
521 filter.<driver>.required flag Git will interpret that as error.
522 Delay
524 If the filter supports the "delay" capability, then Git can send
525 the flag "can-delay" after the filter command and pathname. This
526 flag denotes that the filter can delay filtering the current blob
527 (e.g. to compensate network latencies) by responding with no
528 content but with the status "delayed" and a flush packet.
529 packet: git> command=smudge
531 packet: git> pathname=path/testfile.dat
532 packet: git> can-delay=1
533 packet: git> 0000
534 packet: git> CONTENT
535 packet: git> 0000
536 packet: git< status=delayed
537 packet: git< 0000
538 If the filter supports the "delay" capability then it must support
540 the "list_available_blobs" command. If Git sends this command, then
541 the filter is expected to return a list of pathnames representing
542 blobs that have been delayed earlier and are now available. The
543 list must be terminated with a flush packet followed by a "success"
544 status that is also terminated with a flush packet. If no blobs for
545 the delayed paths are available, yet, then the filter is expected
546 to block the response until at least one blob becomes available.
547 The filter can tell Git that it has no more delayed blobs by
548 sending an empty list. As soon as the filter responds with an empty
549 list, Git stops asking. All blobs that Git has not received at this
550 point are considered missing and will result in an error.
551 packet: git> command=list_available_blobs
553 packet: git> 0000
554 packet: git< pathname=path/testfile.dat
555 packet: git< pathname=path/otherfile.dat
556 packet: git< 0000
557 packet: git< status=success
558 packet: git< 0000
559 After Git received the pathnames, it will request the corresponding
561 blobs again. These requests contain a pathname and an empty content
562 section. The filter is expected to respond with the smudged content
563 in the usual way as explained above.
564 packet: git> command=smudge
566 packet: git> pathname=path/testfile.dat
567 packet: git> 0000
568 packet: git> 0000 # empty content!
569 packet: git< status=success
570 packet: git< 0000
571 packet: git< SMUDGED_CONTENT
572 packet: git< 0000
573 packet: git< 0000 # empty list, keep "status=success" unchanged!
574 Example
576 A long running filter demo implementation can be found in
577 contrib/long-running-filter/example.pl located in the Git core
578 repository. If you develop your own long running filter process
579 then the GIT_TRACE_PACKET environment variables can be very helpful
580 for debugging (see git(1)).
581 Please note that you cannot use an existing filter.<driver>.clean
583 or filter.<driver>.smudge command with filter.<driver>.process
584 because the former two use a different inter process communication
585 protocol than the latter one.
586 Interaction between checkin/checkout attributes
588 In the check-in codepath, the worktree file is first converted with
589 filter driver (if specified and corresponding driver defined), then
590 the result is processed with ident (if specified), and then finally
591 with text (again, if specified and applicable).
592 In the check-out codepath, the blob content is first converted with
594 text, and then ident and fed to filter.
595 Merging branches with differing checkin/checkout attributes
597 If you have added attributes to a file that cause the canonical
598 repository format for that file to change, such as adding a
599 clean/smudge filter or text/eol/ident attributes, merging anything
600 where the attribute is not in place would normally cause merge
601 conflicts.
602 To prevent these unnecessary merge conflicts, Git can be told to
604 run a virtual check-out and check-in of all three stages of a file
605 when resolving a three-way merge by setting the merge.renormalize
606 configuration variable. This prevents changes caused by check-in
607 conversion from causing spurious merge conflicts when a converted
608 file is merged with an unconverted file.
609 As long as a "smudge→clean" results in the same output as a "clean"
611 even on files that are already smudged, this strategy will
612 automatically resolve all filter-related conflicts. Filters that do
613 not act in this way may cause additional merge conflicts that must
614 be resolved manually.
615 Generating diff text
617 diff
618 The attribute diff affects how Git generates diffs for particular
619 files. It can tell Git whether to generate a textual patch for the
620 path or to treat the path as a binary file. It can also affect what
621 line is shown on the hunk header @@ -k,l +n,m @@ line, tell Git to
622 use an external command to generate the diff, or ask Git to convert
623 binary files to a text format before generating the diff.
624 Set
626 A path to which the diff attribute is set is treated as text,
627 even when they contain byte values that normally never appear
628 in text files, such as NUL.
629 Unset
631 A path to which the diff attribute is unset will generate
632 Binary files differ (or a binary patch, if binary patches are
633 enabled).
634 Unspecified
636 A path to which the diff attribute is unspecified first gets
637 its contents inspected, and if it looks like text and is
638 smaller than core.bigFileThreshold, it is treated as text.
639 Otherwise it would generate Binary files differ.
640 String
642 Diff is shown using the specified diff driver. Each driver may
643 specify one or more options, as described in the following
644 section. The options for the diff driver "foo" are defined by
645 the configuration variables in the "diff.foo" section of the
646 Git config file.
647 Defining an external diff driver
649 The definition of a diff driver is done in gitconfig, not
650 gitattributes file, so strictly speaking this manual page is a
651 wrong place to talk about it. However...
652 To define an external diff driver jcdiff, add a section to your
654 $GIT_DIR/config file (or $HOME/.gitconfig file) like this:
655 [diff "jcdiff"]
657 command = j-c-diff
658 When Git needs to show you a diff for the path with diff attribute
660 set to jcdiff, it calls the command you specified with the above
661 configuration, i.e. j-c-diff, with 7 parameters, just like
662 GIT_EXTERNAL_DIFF program is called. See git(1) for details.
663 Setting the internal diff algorithm
665 The diff algorithm can be set through the diff.algorithm config
666 key, but sometimes it may be helpful to set the diff algorithm per
667 path. For example, one may want to use the minimal diff algorithm
668 for .json files, and the histogram for .c files, and so on without
669 having to pass in the algorithm through the command line each time.
670 First, in .gitattributes, assign the diff attribute for paths.
672 *.json diff=<name>
674 Then, define a "diff.<name>.algorithm" configuration to specify the
676 diff algorithm, choosing from myers, patience, minimal, or
677 histogram.
678 [diff "<name>"]
680 algorithm = histogram
681 This diff algorithm applies to user facing diff output like
683 git-diff(1), git-show(1) and is used for the --stat output as well.
684 The merge machinery will not use the diff algorithm set through
685 this method.
686 Note
688 If diff.<name>.command is defined for path with the diff=<name>
689 attribute, it is executed as an external diff driver (see
690 above), and adding diff.<name>.algorithm has no effect, as the
691 algorithm is not passed to the external diff driver.
692 Defining a custom hunk-header
694 Each group of changes (called a "hunk") in the textual diff output
695 is prefixed with a line of the form:
696 @@ -k,l +n,m @@ TEXT
698 This is called a hunk header. The "TEXT" portion is by default a
700 line that begins with an alphabet, an underscore or a dollar sign;
701 this matches what GNU diff -p output uses. This default selection
702 however is not suited for some contents, and you can use a
703 customized pattern to make a selection.
704 First, in .gitattributes, you would assign the diff attribute for
706 paths.
707 *.tex diff=tex
709 Then, you would define a "diff.tex.xfuncname" configuration to
711 specify a regular expression that matches a line that you would
712 want to appear as the hunk header "TEXT". Add a section to your
713 $GIT_DIR/config file (or $HOME/.gitconfig file) like this:
714 [diff "tex"]
716 xfuncname = "^(\\\\(sub)*section\\{.*)$"
717 Note. A single level of backslashes are eaten by the configuration
719 file parser, so you would need to double the backslashes; the
720 pattern above picks a line that begins with a backslash, and zero
721 or more occurrences of sub followed by section followed by open
722 brace, to the end of line.
723 There are a few built-in patterns to make this easier, and tex is
725 one of them, so you do not have to write the above in your
726 configuration file (you still need to enable this with the
727 attribute mechanism, via .gitattributes). The following built in
728 patterns are available:
729 • ada suitable for source code in the Ada language.
731 • bash suitable for source code in the Bourne-Again SHell
733 language. Covers a superset of POSIX shell function
734 definitions.
735 • bibtex suitable for files with BibTeX coded references.
737 • cpp suitable for source code in the C and C++ languages.
739 • csharp suitable for source code in the C# language.
741 • css suitable for cascading style sheets.
743 • dts suitable for devicetree (DTS) files.
745 • elixir suitable for source code in the Elixir language.
747 • fortran suitable for source code in the Fortran language.
749 • fountain suitable for Fountain documents.
751 • golang suitable for source code in the Go language.
753 • html suitable for HTML/XHTML documents.
755 • java suitable for source code in the Java language.
757 • kotlin suitable for source code in the Kotlin language.
759 • markdown suitable for Markdown documents.
761 • matlab suitable for source code in the MATLAB and Octave
763 languages.
764 • objc suitable for source code in the Objective-C language.
766 • pascal suitable for source code in the Pascal/Delphi language.
768 • perl suitable for source code in the Perl language.
770 • php suitable for source code in the PHP language.
772 • python suitable for source code in the Python language.
774 • ruby suitable for source code in the Ruby language.
776 • rust suitable for source code in the Rust language.
778 • scheme suitable for source code in the Scheme language.
780 • tex suitable for source code for LaTeX documents.
782 Customizing word diff
784 You can customize the rules that git diff --word-diff uses to split
785 words in a line, by specifying an appropriate regular expression in
786 the "diff.*.wordRegex" configuration variable. For example, in TeX
787 a backslash followed by a sequence of letters forms a command, but
788 several such commands can be run together without intervening
789 whitespace. To separate them, use a regular expression in your
790 $GIT_DIR/config file (or $HOME/.gitconfig file) like this:
791 [diff "tex"]
793 wordRegex = "\\\\[a-zA-Z]+|[{}]|\\\\.|[^\\{}[:space:]]+"
794 A built-in pattern is provided for all languages listed in the
796 previous section.
797 Performing text diffs of binary files
799 Sometimes it is desirable to see the diff of a text-converted
800 version of some binary files. For example, a word processor
801 document can be converted to an ASCII text representation, and the
802 diff of the text shown. Even though this conversion loses some
803 information, the resulting diff is useful for human viewing (but
804 cannot be applied directly).
805 The textconv config option is used to define a program for
807 performing such a conversion. The program should take a single
808 argument, the name of a file to convert, and produce the resulting
809 text on stdout.
810 For example, to show the diff of the exif information of a file
812 instead of the binary information (assuming you have the exif tool
813 installed), add the following section to your $GIT_DIR/config file
814 (or $HOME/.gitconfig file):
815 [diff "jpg"]
817 textconv = exif
818 Note
821 The text conversion is generally a one-way conversion; in this
822 example, we lose the actual image contents and focus just on
823 the text data. This means that diffs generated by textconv are
824 not suitable for applying. For this reason, only git diff and
825 the git log family of commands (i.e., log, whatchanged, show)
826 will perform text conversion. git format-patch will never
827 generate this output. If you want to send somebody a
828 text-converted diff of a binary file (e.g., because it quickly
829 conveys the changes you have made), you should generate it
830 separately and send it as a comment in addition to the usual
831 binary diff that you might send.
832 Because text conversion can be slow, especially when doing a large
834 number of them with git log -p, Git provides a mechanism to cache
835 the output and use it in future diffs. To enable caching, set the
836 "cachetextconv" variable in your diff driver’s config. For example:
837 [diff "jpg"]
839 textconv = exif
840 cachetextconv = true
841 This will cache the result of running "exif" on each blob
843 indefinitely. If you change the textconv config variable for a diff
844 driver, Git will automatically invalidate the cache entries and
845 re-run the textconv filter. If you want to invalidate the cache
846 manually (e.g., because your version of "exif" was updated and now
847 produces better output), you can remove the cache manually with git
848 update-ref -d refs/notes/textconv/jpg (where "jpg" is the name of
849 the diff driver, as in the example above).
850 Choosing textconv versus external diff
852 If you want to show differences between binary or
853 specially-formatted blobs in your repository, you can choose to use
854 either an external diff command, or to use textconv to convert them
855 to a diff-able text format. Which method you choose depends on your
856 exact situation.
857 The advantage of using an external diff command is flexibility. You
859 are not bound to find line-oriented changes, nor is it necessary
860 for the output to resemble unified diff. You are free to locate and
861 report changes in the most appropriate way for your data format.
862 A textconv, by comparison, is much more limiting. You provide a
864 transformation of the data into a line-oriented text format, and
865 Git uses its regular diff tools to generate the output. There are
866 several advantages to choosing this method:
867 1. Ease of use. It is often much simpler to write a binary to text
869 transformation than it is to perform your own diff. In many
870 cases, existing programs can be used as textconv filters (e.g.,
871 exif, odt2txt).
872 2. Git diff features. By performing only the transformation step
874 yourself, you can still utilize many of Git’s diff features,
875 including colorization, word-diff, and combined diffs for
876 merges.
877 3. Caching. Textconv caching can speed up repeated diffs, such as
879 those you might trigger by running git log -p.
880 Marking files as binary
882 Git usually guesses correctly whether a blob contains text or
883 binary data by examining the beginning of the contents. However,
884 sometimes you may want to override its decision, either because a
885 blob contains binary data later in the file, or because the
886 content, while technically composed of text characters, is opaque
887 to a human reader. For example, many postscript files contain only
888 ASCII characters, but produce noisy and meaningless diffs.
889 The simplest way to mark a file as binary is to unset the diff
891 attribute in the .gitattributes file:
892 *.ps -diff
894 This will cause Git to generate Binary files differ (or a binary
896 patch, if binary patches are enabled) instead of a regular diff.
897 However, one may also want to specify other diff driver attributes.
899 For example, you might want to use textconv to convert postscript
900 files to an ASCII representation for human viewing, but otherwise
901 treat them as binary files. You cannot specify both -diff and
902 diff=ps attributes. The solution is to use the diff.*.binary config
903 option:
904 [diff "ps"]
906 textconv = ps2ascii
907 binary = true
908 Performing a three-way merge
910 merge
911 The attribute merge affects how three versions of a file are merged
912 when a file-level merge is necessary during git merge, and other
913 commands such as git revert and git cherry-pick.
914 Set
916 Built-in 3-way merge driver is used to merge the contents in a
917 way similar to merge command of RCS suite. This is suitable for
918 ordinary text files.
919 Unset
921 Take the version from the current branch as the tentative merge
922 result, and declare that the merge has conflicts. This is
923 suitable for binary files that do not have a well-defined merge
924 semantics.
925 Unspecified
927 By default, this uses the same built-in 3-way merge driver as
928 is the case when the merge attribute is set. However, the
929 merge.default configuration variable can name different merge
930 driver to be used with paths for which the merge attribute is
931 unspecified.
932 String
934 3-way merge is performed using the specified custom merge
935 driver. The built-in 3-way merge driver can be explicitly
936 specified by asking for "text" driver; the built-in "take the
937 current branch" driver can be requested with "binary".
938 Built-in merge drivers
940 There are a few built-in low-level merge drivers defined that can
941 be asked for via the merge attribute.
942 text
944 Usual 3-way file level merge for text files. Conflicted regions
945 are marked with conflict markers <<<<<<<, ======= and >>>>>>>.
946 The version from your branch appears before the ======= marker,
947 and the version from the merged branch appears after the
948 ======= marker.
949 binary
951 Keep the version from your branch in the work tree, but leave
952 the path in the conflicted state for the user to sort out.
953 union
955 Run 3-way file level merge for text files, but take lines from
956 both versions, instead of leaving conflict markers. This tends
957 to leave the added lines in the resulting file in random order
958 and the user should verify the result. Do not use this if you
959 do not understand the implications.
960 Defining a custom merge driver
962 The definition of a merge driver is done in the .git/config file,
963 not in the gitattributes file, so strictly speaking this manual
964 page is a wrong place to talk about it. However...
965 To define a custom merge driver filfre, add a section to your
967 $GIT_DIR/config file (or $HOME/.gitconfig file) like this:
968 [merge "filfre"]
970 name = feel-free merge driver
971 driver = filfre %O %A %B %L %P
972 recursive = binary
973 The merge.*.name variable gives the driver a human-readable name.
975 The merge.*.driver variable’s value is used to construct a command
977 to run to merge ancestor’s version (%O), current version (%A) and
978 the other branches' version (%B). These three tokens are replaced
979 with the names of temporary files that hold the contents of these
980 versions when the command line is built. Additionally, %L will be
981 replaced with the conflict marker size (see below).
982 The merge driver is expected to leave the result of the merge in
984 the file named with %A by overwriting it, and exit with zero status
985 if it managed to merge them cleanly, or non-zero if there were
986 conflicts. When the driver crashes (e.g. killed by SEGV), it is
987 expected to exit with non-zero status that are higher than 128, and
988 in such a case, the merge results in a failure (which is different
989 from producing a conflict).
990 The merge.*.recursive variable specifies what other merge driver to
992 use when the merge driver is called for an internal merge between
993 common ancestors, when there are more than one. When left
994 unspecified, the driver itself is used for both internal merge and
995 the final merge.
996 The merge driver can learn the pathname in which the merged result
998 will be stored via placeholder %P.
999 conflict-marker-size
1001 This attribute controls the length of conflict markers left in the
1002 work tree file during a conflicted merge. Only a positive integer
1003 has a meaningful effect.
1004 For example, this line in .gitattributes can be used to tell the
1006 merge machinery to leave much longer (instead of the usual
1007 7-character-long) conflict markers when merging the file
1008 Documentation/git-merge.txt results in a conflict.
1009 Documentation/git-merge.txt conflict-marker-size=32
1011 Checking whitespace errors
1013 whitespace
1014 The core.whitespace configuration variable allows you to define
1015 what diff and apply should consider whitespace errors for all paths
1016 in the project (See git-config(1)). This attribute gives you finer
1017 control per path.
1018 Set
1020 Notice all types of potential whitespace errors known to Git.
1021 The tab width is taken from the value of the core.whitespace
1022 configuration variable.
1023 Unset
1025 Do not notice anything as error.
1026 Unspecified
1028 Use the value of the core.whitespace configuration variable to
1029 decide what to notice as error.
1030 String
1032 Specify a comma separated list of common whitespace problems to
1033 notice in the same format as the core.whitespace configuration
1034 variable.
1035 Creating an archive
1037 export-ignore
1038 Files and directories with the attribute export-ignore won’t be
1039 added to archive files.
1040 export-subst
1042 If the attribute export-subst is set for a file then Git will
1043 expand several placeholders when adding this file to an archive.
1044 The expansion depends on the availability of a commit ID, i.e., if
1045 git-archive(1) has been given a tree instead of a commit or a tag
1046 then no replacement will be done. The placeholders are the same as
1047 those for the option --pretty=format: of git-log(1), except that
1048 they need to be wrapped like this: $Format:PLACEHOLDERS$ in the
1049 file. E.g. the string $Format:%H$ will be replaced by the commit
1050 hash. However, only one %(describe) placeholder is expanded per
1051 archive to avoid denial-of-service attacks.
1052 Packing objects
1054 delta
1055 Delta compression will not be attempted for blobs for paths with
1056 the attribute delta set to false.
1057 Viewing files in GUI tools
1059 encoding
1060 The value of this attribute specifies the character encoding that
1061 should be used by GUI tools (e.g. gitk(1) and git-gui(1)) to
1062 display the contents of the relevant file. Note that due to
1063 performance considerations gitk(1) does not use this attribute
1064 unless you manually enable per-file encodings in its options.
1065 If this attribute is not set or has an invalid value, the value of
1067 the gui.encoding configuration variable is used instead (See git-
1068 config(1)).
1069
1071 You do not want any end-of-line conversions applied to, nor textual
1072 diffs produced for, any binary file you track. You would need to
1073 specify e.g.
1074 *.jpg -text -diff
1076 but that may become cumbersome, when you have many attributes. Using
1078 macro attributes, you can define an attribute that, when set, also sets
1079 or unsets a number of other attributes at the same time. The system
1080 knows a built-in macro attribute, binary:
1081 *.jpg binary
1083 Setting the "binary" attribute also unsets the "text" and "diff"
1085 attributes as above. Note that macro attributes can only be "Set",
1086 though setting one might have the effect of setting or unsetting other
1087 attributes or even returning other attributes to the "Unspecified"
1088 state.
1089
1091 Custom macro attributes can be defined only in top-level gitattributes
1092 files ($GIT_DIR/info/attributes, the .gitattributes file at the top
1093 level of the working tree, or the global or system-wide gitattributes
1094 files), not in .gitattributes files in working tree subdirectories. The
1095 built-in macro attribute "binary" is equivalent to:
1096 [attr]binary -diff -merge -text
1098
1100 Git does not follow symbolic links when accessing a .gitattributes file
1101 in the working tree. This keeps behavior consistent when the file is
1102 accessed from the index or a tree versus from the filesystem.
1103
1105 If you have these three gitattributes file:
1106 (in $GIT_DIR/info/attributes)
1108 a* foo !bar -baz
1110 (in .gitattributes)
1112 abc foo bar baz
1113 (in t/.gitattributes)
1115 ab* merge=filfre
1116 abc -foo -bar
1117 *.c frotz
1118 the attributes given to path t/abc are computed as follows:
1120 1. By examining t/.gitattributes (which is in the same directory as
1122 the path in question), Git finds that the first line matches.
1123 merge attribute is set. It also finds that the second line matches,
1124 and attributes foo and bar are unset.
1125 2. Then it examines .gitattributes (which is in the parent directory),
1127 and finds that the first line matches, but t/.gitattributes file
1128 already decided how merge, foo and bar attributes should be given
1129 to this path, so it leaves foo and bar unset. Attribute baz is set.
1130 3. Finally it examines $GIT_DIR/info/attributes. This file is used to
1132 override the in-tree settings. The first line is a match, and foo
1133 is set, bar is reverted to unspecified state, and baz is unset.
1134 As the result, the attributes assignment to t/abc becomes:
1136 foo set to true
1138 bar unspecified
1139 baz set to false
1140 merge set to string value "filfre"
1141 frotz unspecified
1142
1144 git-check-attr(1).
1145
1147 Part of the git(1) suite
1148Git 2.43.0 11/20/2023 GITATTRIBUTES(5)