Skip to main content

18.10 Diff Mode

Diff mode is a major mode used for the output of M-x diff and other similar commands. This kind of output is called a patch, because it can be passed to the patch command to automatically apply the specified changes. To select Diff mode manually, type M-x diff-mode.

The changes specified in a patch are grouped into hunks, which are contiguous chunks of text that contain one or more changed lines. Hunks usually also include unchanged lines to provide context for the changes. Each hunk is preceded by a hunk header, which specifies the old and new line numbers where the hunk’s changes occur. Diff mode highlights each hunk header, to distinguish it from the actual contents of the hunk.

The first hunk in a patch is preceded by a file header, which shows the names of the new and the old versions of the file, and their time stamps. If a patch shows changes for more than one file, each file has such a header before the first hunk of that file’s changes.

You can edit a Diff mode buffer like any other buffer. (If it is read-only, you need to make it writable first; see Misc Buffer.) Whenever you edit a hunk, Diff mode attempts to automatically correct the line numbers in the hunk headers, to ensure that the patch remains correct, and could still be applied by patch. To disable automatic line number correction, change the variable diff-update-on-the-fly to nil.

Diff mode arranges for hunks to be treated as compiler error messages by M-g M-n and other commands that handle error messages (see Compilation Mode). Thus, you can use the compilation-mode commands to visit the corresponding source locations.

In addition, Diff mode provides the following commands to navigate, manipulate and apply parts of patches:

M-n​

Move to the next hunk-start (diff-hunk-next). With prefix argument n, move forward to the nth next hunk.

By default, Diff mode refines hunks as Emacs displays them, highlighting their changes with better granularity. Alternatively, if you set diff-refine to the symbol navigation, Diff mode only refines the hunk you move to with this command or with diff-hunk-prev.

M-p​

Move to the previous hunk-start (diff-hunk-prev). With prefix argument n, move back to the nth previous hunk. Like M-n, this command refines the hunk you move to if you set diff-refine to the symbol navigation.

M-}​

Move to the next file-start, in a multi-file patch (diff-file-next). With prefix argument n, move forward to the start of the nth next file.

M-{​

Move to the previous file-start, in a multi-file patch (diff-file-prev). With prefix argument n, move back to the start of the nth previous file.

M-k​

Kill the hunk at point (diff-hunk-kill).

M-K​

In a multi-file patch, kill the current file part. (diff-file-kill).

C-c C-a​

Apply this hunk to its target file (diff-apply-hunk). With a prefix argument of C-u, revert this hunk, i.e. apply the reverse of the hunk, which changes the “new" version into the “old" version. If diff-jump-to-old-file is non-nil, apply the hunk to the “old" version of the file instead.

C-c C-b​

Highlight the changes of the hunk at point with a finer granularity (diff-refine-hunk). This allows you to see exactly which parts of each changed line were actually changed.

By default, Diff mode refines hunks as Emacs displays them, so you may find this command useful if you customize diff-refine to a non-default value.

C-c C-c​

Go to the source file and line corresponding to this hunk (diff-goto-source). By default, this jumps to the “new" version of the file, the one shown first on the file header. With a prefix argument, jump to the “old" version instead. If diff-jump-to-old-file is non-nil, this command by default jumps to the “old" file, and the meaning of the prefix argument is reversed. If the prefix argument is a number greater than 8 (e.g., if you type C-u C-u C-c C-c), then this command also sets diff-jump-to-old-file for the next invocation. If the source file is under version control (see Version Control), this jumps to the work file by default. With a prefix argument, jump to the “old" revision of the file (see Old Revisions), when point is on the old line, or otherwise jump to the “new" revision.

C-c C-e​

Start an Ediff session with the patch (diff-ediff-patch). See Ediff in The Ediff Manual.

C-c C-n​

Restrict the view to the current hunk (diff-restrict-view). See Narrowing. With a prefix argument, restrict the view to the current file of a multiple-file patch. To widen again, use C-x n w (widen).

C-c C-r​

Reverse the direction of comparison for the entire buffer (diff-reverse-direction). With a prefix argument, reverse the direction only inside the current region (see Mark). Reversing the direction means changing the hunks and the file-start headers to produce a patch that would change the “new" version into the “old" one.

C-c C-s​

Split the hunk at point (diff-split-hunk) into two separate hunks. This inserts a hunk header and modifies the header of the current hunk. This command is useful for manually editing patches, and only works with the unified diff format produced by the -u or --unified options to the diff program. If you need to split a hunk in the context diff format produced by the -c or --context options to diff, first convert the buffer to the unified diff format with C-c C-u.

C-c C-d​

Convert the entire buffer to the context diff format (diff-unified->context). With a prefix argument, convert only the hunks within the region.

C-c C-u​

Convert the entire buffer to unified diff format (diff-context->unified). With a prefix argument, convert unified format to context format. When the mark is active, convert only the hunks within the region.

C-c C-w​

Re-generate the current hunk, disregarding changes in whitespace (diff-ignore-whitespace-hunk).

C-x 4 A​

Generate a ChangeLog entry, like C-x 4 a does (see Change Log), for each one of the hunks (diff-add-change-log-entries-other-window). This creates a skeleton of the log of changes that you can later fill with the actual descriptions of the changes. C-x 4 a itself in Diff mode operates on behalf of the current hunk’s file, but gets the function name from the patch itself. This is useful for making log entries for functions that are deleted by the patch.

Patches sometimes include trailing whitespace on modified lines, as an unintentional and undesired change. There are two ways to deal with this problem. Firstly, if you enable Whitespace mode in a Diff buffer (see Useless Whitespace), it automatically highlights trailing whitespace in modified lines. Secondly, you can use the command M-x diff-delete-trailing-whitespace, which searches for trailing whitespace in the lines modified by the patch, and removes that whitespace in both the patch and the patched source file(s). This command does not save the modifications that it makes, so you can decide whether to save the changes (the list of modified files is displayed in the echo area). With a prefix argument, it tries to modify the original (“old") source files rather than the patched (“new") source files.

If diff-font-lock-syntax is non-nil, fragments of source in hunks are highlighted according to the appropriate major mode.