gitattributes(5) Manual Page

This attribute marks the path as a text file, which enables end-of-line conversion: When a matching file is added to the index, the file’s line endings are normalized to LF in the index. Conversely, when the file is copied from the index to the working directory, its line endings may be converted from LF to CRLF depending on the eol attribute, the Git config, and the platform (see explanation of eol below).

Any other value causes Git to act as if text has been left unspecified.

This attribute marks a path to use a specific line-ending style in the working tree when it is checked out. It has effect only if text or text=auto is set (see above), but specifying eol automatically sets text if text was left unspecified.

For backwards compatibility, the crlf attribute is interpreted as follows:

While Git normally leaves file contents alone, it can be configured to normalize line endings to LF in the repository and, optionally, to convert them to CRLF when files are checked out.

If you simply want to have CRLF line endings in your working directory regardless of the repository you are working with, you can set the config variable "core.autocrlf" without using any attributes.

This does not force normalization of text files, but does ensure that text files that you introduce to the repository have their line endings normalized to LF when they are added, and that files that are already normalized in the repository stay normalized.

If you want to ensure that text files that any contributor introduces to the repository have their line endings normalized, you can set the text attribute to "auto" for all files.

The attributes allow a fine-grained control, how the line endings are converted. Here is an example that will make Git normalize .txt, .vcproj and .sh files, ensure that .vcproj files have CRLF and .sh files have LF in the working directory, and prevent .jpg files from being normalized regardless of their content.

From a clean working directory:

If any files that should not be normalized show up in git status, unset their text attribute before running git add -u.

Conversely, text files that Git does not detect can have normalization enabled manually.

If core.safecrlf is set to "true" or "warn", Git verifies if the conversion is reversible for the current setting of core.autocrlf. For "true", Git rejects irreversible conversions; for "warn", Git only prints a warning but accepts an irreversible conversion. The safety triggers to prevent such a conversion done to the files in the work tree, but there are a few exceptions. Even though…

Git recognizes files encoded in ASCII or one of its supersets (e.g. UTF-8, ISO-8859-1, …) as text files. Files encoded in certain other encodings (e.g. UTF-16) are interpreted as binary and consequently built-in Git text processing tools (e.g. git diff) as well as most Git web front ends do not visualize the contents of these files by default.

In these cases you can tell Git the encoding of a file in the working directory with the working-tree-encoding attribute. If a file with this attribute is added to Git, then Git re-encodes the content from the specified encoding to UTF-8. Finally, Git stores the UTF-8 encoded content in its internal data structure (called "the index"). On checkout the content is re-encoded back to the specified encoding.

Please note that using the working-tree-encoding attribute may have a number of pitfalls:

Use the working-tree-encoding attribute only if you cannot store a file in UTF-8 encoding and if you want Git to be able to process the content as text.

As an example, use the following attributes if your *.ps1 files are UTF-16 encoded with byte order mark (BOM) and you want Git to perform automatic line ending conversion based on your platform.

Use the following attributes if your *.ps1 files are UTF-16 little endian encoded without BOM and you want Git to use Windows line endings in the working directory (use UTF-16LE-BOM instead of UTF-16LE if you want UTF-16 little endian with BOM). Please note, it is highly recommended to explicitly define the line endings with eol if the working-tree-encoding attribute is used to avoid ambiguity.

You can get a list of all available encodings on your platform with the following command:

If you do not know the encoding of a file, then you can use the file command to guess the encoding:

When the attribute ident is set for a path, Git replaces $Id$ in the blob object with $Id:, followed by the 40-character hexadecimal blob object name, followed by a dollar sign $ upon checkout. Any byte sequence that begins with $Id: and ends with $ in the worktree file is replaced with $Id$ upon check-in.

A filter attribute can be set to a string value that names a filter driver specified in the configuration.

A filter driver consists of a clean command and a smudge command, either of which can be left unspecified. Upon checkout, when the smudge command is specified, the command is fed the blob object from its standard input, and its standard output is used to update the worktree file. Similarly, the clean command is used to convert the contents of worktree file upon checkin. By default these commands process only a single blob and terminate. If a long running process filter is used in place of clean and/or smudge filters, then Git can process all blobs with a single filter command invocation for the entire life of a single Git command, for example git add --all. If a long running process filter is configured then it always takes precedence over a configured single blob filter. See section below for the description of the protocol used to communicate with a process filter.

One use of the content filtering is to massage the content into a shape that is more convenient for the platform, filesystem, and the user to use. For this mode of operation, the key phrase here is "more convenient" and not "turning something unusable into usable". In other words, the intent is that if someone unsets the filter driver definition, or does not have the appropriate filter program, the project should still be usable.

Another use of the content filtering is to store the content that cannot be directly used in the repository (e.g. a UUID that refers to the true content stored outside Git, or an encrypted content) and turn it into a usable form upon checkout (e.g. download the external content, or decrypt the encrypted content).

These two filters behave differently, and by default, a filter is taken as the former, massaging the contents into more convenient shape. A missing filter driver definition in the config, or a filter driver that exits with a non-zero status, is not an error but makes the filter a no-op passthru.

You can declare that a filter turns a content that by itself is unusable into a usable content by setting the filter.<driver>.required configuration variable to true.

Note: Whenever the clean filter is changed, the repo should be renormalized: $ git add --renormalize .

For example, in .gitattributes, you would assign the filter attribute for paths.

Then you would define a "filter.indent.clean" and "filter.indent.smudge" configuration in your .git/config to specify a pair of commands to modify the contents of C programs when the source files are checked in ("clean" is run) and checked out (no change is made because the command is "cat").

For best results, clean should not alter its output further if it is run twice ("clean→clean" should be equivalent to "clean"), and multiple smudge commands should not alter clean's output ("smudge→smudge→clean" should be equivalent to "clean"). See the section on merging below.

The "indent" filter is well-behaved in this regard: it will not modify input that is already correctly indented. In this case, the lack of a smudge filter means that the clean filter must accept its own output without modifying it.

If a filter must succeed in order to make the stored contents usable, you can declare that the filter is required, in the configuration:

Sequence "%f" on the filter command line is replaced with the name of the file the filter is working on. A filter might use this in keyword substitution. For example:

Note that "%f" is the name of the path that is being worked on. Depending on the version that is being filtered, the corresponding file on disk may not exist, or may have different contents. So, smudge and clean commands should not try to access the file on disk, but only act as filters on the content provided to them on standard input.

If the filter command (a string value) is defined via filter.<driver>.process then Git can process all blobs with a single filter invocation for the entire life of a single Git command. This is achieved by using the long-running process protocol (described in technical/long-running-process-protocol.txt).

When Git encounters the first file that needs to be cleaned or smudged, it starts the filter and performs the handshake. In the handshake, the welcome message sent by Git is "git-filter-client", only version 2 is supported, and the supported capabilities are "clean", "smudge", and "delay".

Afterwards Git sends a list of "key=value" pairs terminated with a flush packet. The list will contain at least the filter command (based on the supported capabilities) and the pathname of the file to filter relative to the repository root. Right after the flush packet Git sends the content split in zero or more pkt-line packets and a flush packet to terminate content. Please note, that the filter must not send any response before it received the content and the final flush packet. Also note that the "value" of a "key=value" pair can contain the "=" character whereas the key would never contain that character.

The filter is expected to respond with a list of "key=value" pairs terminated with a flush packet. If the filter does not experience problems then the list must contain a "success" status. Right after these packets the filter is expected to send the content in zero or more pkt-line packets and a flush packet at the end. Finally, a second list of "key=value" pairs terminated with a flush packet is expected. The filter can change the status in the second list or keep the status as is with an empty list. Please note that the empty list must be terminated with a flush packet regardless.

If the result content is empty then the filter is expected to respond with a "success" status and a flush packet to signal the empty content.

In case the filter cannot or does not want to process the content, it is expected to respond with an "error" status.

If the filter experiences an error during processing, then it can send the status "error" after the content was (partially or completely) sent.

In case the filter cannot or does not want to process the content as well as any future content for the lifetime of the Git process, then it is expected to respond with an "abort" status at any point in the protocol.

Git neither stops nor restarts the filter process in case the "error"/"abort" status is set. However, Git sets its exit code according to the filter.<driver>.required flag, mimicking the behavior of the filter.<driver>.clean / filter.<driver>.smudge mechanism.

If the filter dies during the communication or does not adhere to the protocol then Git will stop the filter process and restart it with the next file that needs to be processed. Depending on the filter.<driver>.required flag Git will interpret that as error.

If the filter supports the "delay" capability, then Git can send the flag "can-delay" after the filter command and pathname. This flag denotes that the filter can delay filtering the current blob (e.g. to compensate network latencies) by responding with no content but with the status "delayed" and a flush packet.

If the filter supports the "delay" capability then it must support the "list_available_blobs" command. If Git sends this command, then the filter is expected to return a list of pathnames representing blobs that have been delayed earlier and are now available. The list must be terminated with a flush packet followed by a "success" status that is also terminated with a flush packet. If no blobs for the delayed paths are available, yet, then the filter is expected to block the response until at least one blob becomes available. The filter can tell Git that it has no more delayed blobs by sending an empty list. As soon as the filter responds with an empty list, Git stops asking. All blobs that Git has not received at this point are considered missing and will result in an error.

After Git received the pathnames, it will request the corresponding blobs again. These requests contain a pathname and an empty content section. The filter is expected to respond with the smudged content in the usual way as explained above.

A long running filter demo implementation can be found in contrib/long-running-filter/example.pl located in the Git core repository. If you develop your own long running filter process then the GIT_TRACE_PACKET environment variables can be very helpful for debugging (see git(1)).

Please note that you cannot use an existing filter.<driver>.clean or filter.<driver>.smudge command with filter.<driver>.process because the former two use a different inter process communication protocol than the latter one.

In the check-in codepath, the worktree file is first converted with filter driver (if specified and corresponding driver defined), then the result is processed with ident (if specified), and then finally with text (again, if specified and applicable).

In the check-out codepath, the blob content is first converted with text, and then ident and fed to filter.

If you have added attributes to a file that cause the canonical repository format for that file to change, such as adding a clean/smudge filter or text/eol/ident attributes, merging anything where the attribute is not in place would normally cause merge conflicts.

To prevent these unnecessary merge conflicts, Git can be told to run a virtual check-out and check-in of all three stages of a file when resolving a three-way merge by setting the merge.renormalize configuration variable. This prevents changes caused by check-in conversion from causing spurious merge conflicts when a converted file is merged with an unconverted file.

As long as a "smudge→clean" results in the same output as a "clean" even on files that are already smudged, this strategy will automatically resolve all filter-related conflicts. Filters that do not act in this way may cause additional merge conflicts that must be resolved manually.

The attribute diff affects how Git generates diffs for particular files. It can tell Git whether to generate a textual patch for the path or to treat the path as a binary file. It can also affect what line is shown on the hunk header @@ -k,l +n,m @@ line, tell Git to use an external command to generate the diff, or ask Git to convert binary files to a text format before generating the diff.

The definition of a diff driver is done in gitconfig, not gitattributes file, so strictly speaking this manual page is a wrong place to talk about it. However…

To define an external diff driver jcdiff, add a section to your $GIT_DIR/config file (or $HOME/.gitconfig file) like this:

When Git needs to show you a diff for the path with diff attribute set to jcdiff, it calls the command you specified with the above configuration, i.e. j-c-diff, with 7 parameters, just like GIT_EXTERNAL_DIFF program is called. See git(1) for details.

The diff algorithm can be set through the diff.algorithm config key, but sometimes it may be helpful to set the diff algorithm per path. For example, one may want to use the minimal diff algorithm for .json files, and the histogram for .c files, and so on without having to pass in the algorithm through the command line each time.

First, in .gitattributes, assign the diff attribute for paths.

Then, define a "diff.<name>.algorithm" configuration to specify the diff algorithm, choosing from myers, patience, minimal, or histogram.

This diff algorithm applies to user facing diff output like git-diff(1), git-show(1) and is used for the --stat output as well. The merge machinery will not use the diff algorithm set through this method.

Each group of changes (called a "hunk") in the textual diff output is prefixed with a line of the form:

This is called a hunk header. The "TEXT" portion is by default a line that begins with an alphabet, an underscore or a dollar sign; this matches what GNU diff -p output uses. This default selection however is not suited for some contents, and you can use a customized pattern to make a selection.

First, in .gitattributes, you would assign the diff attribute for paths.

Then, you would define a "diff.tex.xfuncname" configuration to specify a regular expression that matches a line that you would want to appear as the hunk header "TEXT". Add a section to your $GIT_DIR/config file (or $HOME/.gitconfig file) like this:

Note. A single level of backslashes are eaten by the configuration file parser, so you would need to double the backslashes; the pattern above picks a line that begins with a backslash, and zero or more occurrences of sub followed by section followed by open brace, to the end of line.

There are a few built-in patterns to make this easier, and tex is one of them, so you do not have to write the above in your configuration file (you still need to enable this with the attribute mechanism, via .gitattributes). The following built in patterns are available:

You can customize the rules that git diff --word-diff uses to split words in a line, by specifying an appropriate regular expression in the "diff.*.wordRegex" configuration variable. For example, in TeX a backslash followed by a sequence of letters forms a command, but several such commands can be run together without intervening whitespace. To separate them, use a regular expression in your $GIT_DIR/config file (or $HOME/.gitconfig file) like this:

A built-in pattern is provided for all languages listed in the previous section.

Sometimes it is desirable to see the diff of a text-converted version of some binary files. For example, a word processor document can be converted to an ASCII text representation, and the diff of the text shown. Even though this conversion loses some information, the resulting diff is useful for human viewing (but cannot be applied directly).

The textconv config option is used to define a program for performing such a conversion. The program should take a single argument, the name of a file to convert, and produce the resulting text on stdout.

For example, to show the diff of the exif information of a file instead of the binary information (assuming you have the exif tool installed), add the following section to your $GIT_DIR/config file (or $HOME/.gitconfig file):

Because text conversion can be slow, especially when doing a large number of them with git log -p, Git provides a mechanism to cache the output and use it in future diffs. To enable caching, set the "cachetextconv" variable in your diff driver’s config. For example:

This will cache the result of running "exif" on each blob indefinitely. If you change the textconv config variable for a diff driver, Git will automatically invalidate the cache entries and re-run the textconv filter. If you want to invalidate the cache manually (e.g., because your version of "exif" was updated and now produces better output), you can remove the cache manually with git update-ref -d refs/notes/textconv/jpg (where "jpg" is the name of the diff driver, as in the example above).

If you want to show differences between binary or specially-formatted blobs in your repository, you can choose to use either an external diff command, or to use textconv to convert them to a diff-able text format. Which method you choose depends on your exact situation.

The advantage of using an external diff command is flexibility. You are not bound to find line-oriented changes, nor is it necessary for the output to resemble unified diff. You are free to locate and report changes in the most appropriate way for your data format.

A textconv, by comparison, is much more limiting. You provide a transformation of the data into a line-oriented text format, and Git uses its regular diff tools to generate the output. There are several advantages to choosing this method:

Git usually guesses correctly whether a blob contains text or binary data by examining the beginning of the contents. However, sometimes you may want to override its decision, either because a blob contains binary data later in the file, or because the content, while technically composed of text characters, is opaque to a human reader. For example, many postscript files contain only ASCII characters, but produce noisy and meaningless diffs.

The simplest way to mark a file as binary is to unset the diff attribute in the .gitattributes file:

This will cause Git to generate Binary files differ (or a binary patch, if binary patches are enabled) instead of a regular diff.

However, one may also want to specify other diff driver attributes. For example, you might want to use textconv to convert postscript files to an ASCII representation for human viewing, but otherwise treat them as binary files. You cannot specify both -diff and diff=ps attributes. The solution is to use the diff.*.binary config option:

The attribute merge affects how three versions of a file are merged when a file-level merge is necessary during git merge, and other commands such as git revert and git cherry-pick.

There are a few built-in low-level merge drivers defined that can be asked for via the merge attribute.

The definition of a merge driver is done in the .git/config file, not in the gitattributes file, so strictly speaking this manual page is a wrong place to talk about it. However…

To define a custom merge driver filfre, add a section to your $GIT_DIR/config file (or $HOME/.gitconfig file) like this:

The merge.*.name variable gives the driver a human-readable name.

The ‘merge.*.driver` variable’s value is used to construct a command to run to merge ancestor’s version (%O), current version (%A) and the other branches’ version (%B). These three tokens are replaced with the names of temporary files that hold the contents of these versions when the command line is built. Additionally, %L will be replaced with the conflict marker size (see below).

The merge driver is expected to leave the result of the merge in the file named with %A by overwriting it, and exit with zero status if it managed to merge them cleanly, or non-zero if there were conflicts. When the driver crashes (e.g. killed by SEGV), it is expected to exit with non-zero status that are higher than 128, and in such a case, the merge results in a failure (which is different from producing a conflict).

The merge.*.recursive variable specifies what other merge driver to use when the merge driver is called for an internal merge between common ancestors, when there are more than one. When left unspecified, the driver itself is used for both internal merge and the final merge.

The merge driver can learn the pathname in which the merged result will be stored via placeholder %P.

This attribute controls the length of conflict markers left in the work tree file during a conflicted merge. Only a positive integer has a meaningful effect.

For example, this line in .gitattributes can be used to tell the merge machinery to leave much longer (instead of the usual 7-character-long) conflict markers when merging the file Documentation/git-merge.txt results in a conflict.

The core.whitespace configuration variable allows you to define what diff and apply should consider whitespace errors for all paths in the project (See git-config(1)). This attribute gives you finer control per path.

Files and directories with the attribute export-ignore won’t be added to archive files.

If the attribute export-subst is set for a file then Git will expand several placeholders when adding this file to an archive. The expansion depends on the availability of a commit ID, i.e., if git-archive(1) has been given a tree instead of a commit or a tag then no replacement will be done. The placeholders are the same as those for the option --pretty=format: of git-log(1), except that they need to be wrapped like this: $Format:PLACEHOLDERS$ in the file. E.g. the string $Format:%H$ will be replaced by the commit hash. However, only one %(describe) placeholder is expanded per archive to avoid denial-of-service attacks.

Delta compression will not be attempted for blobs for paths with the attribute delta set to false.

The value of this attribute specifies the character encoding that should be used by GUI tools (e.g. gitk(1) and git-gui(1)) to display the contents of the relevant file. Note that due to performance considerations gitk(1) does not use this attribute unless you manually enable per-file encodings in its options.

If this attribute is not set or has an invalid value, the value of the gui.encoding configuration variable is used instead (See git-config(1)).