How to grep the git diff?

Question

Is there a way to show the git-diff filtered by a given pattern.

Something like

git grepdiff pattern

changed file
+++ some sentence with pattern
changed file 2
--- some other pattern

Unfortunately the simplest solution is not good enough

git diff | grep pattern 

+++ some sentence with pattern
--- some other pattern
# not an option as doesn't put the filename close to the match

I came with a workaround using awk

git diff | awk "/\+\+\+/{f = \$2}; /PATTERN/ {print f \$0} "

But would love to find out that there is a command for this.

Answer 1

Not sure but isn't git diff -G <regex> flag OK?

-G < regex>

Look for differences whose added or removed line matches the given <regex>.

Answer 2

Have you tried git diff -S<string> or git diff -G".*string.*"? Note that they are not equivalent, see the documentation about pickaxe for what -S does.

Answer 3

Another possibility would be to view the whole diff and search the output using the normal less commands (type / and then the pattern).

When you have less configured to show some lines before the match using --jump-target=N, this is pretty useful. Try it like this:

PAGER="/usr/bin/less --jump-target=10" git diff

This means that the match should be shown on line 10 (shows 9 lines of context above), which may be enough to also see the file name.

You can also use e.g. --jump-target=.5 to make it position the match in the middle of the screen.

Answer 4

I use git log -p, which opens less (configurable, though), which in turn can be searched for with /. There's also git log -S <searchword>.

Answer 5

I think your approach to "grep" diff output is the best workaround.

You may improve your awk script by using sed:

colored="(^[\[[0-9;]*[a-zA-Z])"
marker="^$colored+diff"
pattern="^$colored+.*(\+|\-).*PATTERN"
git diff --color | sed -rn -e "/$marker/! H; /$marker/ ba; $ ba; b; :a; x; /$pattern/ p"

• colored: regex to match terminal colored lines
• marker: marker to match division from differents diff hunks, lines starting with colored "diff"
• pattern: pattern to search for, lines starting with colored "+" or "-" and containing "PATTERN"

This will print full diff hunks, with added or removed PATTERN, also maintaining useful colored output.

Note that ^[ in colored should be actual, literal ^[. You can type them in bash by pressing Ctrl + V, Ctrl + [

Answer 6

Here is a custom diff tool that allows grepping inside changes (but not the context):

Usage

GIT_EXTERNAL_DIFF="mydiff --grep foo" git diff

This will output those lines in your changes that contain foo (including lines where foo disappeared because of your changes). Any grep pattern can be used instead of foo.

Each output line starts with the following prefix:

filename: oldlinenum: newlinenum|

The script can also be used without the --grep option, in which case it simply formats the full diff (i.e. providing full context) as described above.

mydiff

#!/bin/bash

my_diff()
{
    diff --old-line-format="$1"':%6dn:      |-%L'     \
         --new-line-format="$1"':      :%6dn|+%L'     \
         --unchanged-line-format="$1"':%6dn:%6dn| %L' \
         $2 $3
}

if [[ $1 == '--grep' ]]
then
    pattern="$2"
    shift 2
    my_diff "$1" "$2" "$5"|grep --color=never '^[^|]\+|[-+].\+'"$pattern"'.*'
else
    my_diff "$1" "$2" "$5"
fi

exit 0

Answer 7

I have been using this with great satisfaction :)

grep -ri <MY_PATTERN> $(git diff 790e26393d --name-only) 

Answer 8

On Windows, a simple solution is:

git diff -U0 | findstr string

If you want grouping by filename, use this

FOR /F "usebackq delims==" %i IN (`git diff --name-only`) do git diff -U0 %~fi | findstr string

Answer 9

This did the job for me, I hope it will help someone:

git diff | grep  -P '^\+|^\-'

Answer 10

The solutions offered didn't exactly fit my needs, this solved my issue.

(
    START_DIFF=abc123
    END_DIFF=123dcf 

    # loop over all the files that have changed inside the diff
    # you can add a `| grep '<ext>$'` to the end of `--name-only`
    # if you need to be more aggresive with the filtering / or 
    # make it go faster...
    for file in $(git diff $START_DIFF $END_DIFF --name-only); do
        # loop over every line of the diff FOR that file.
        while IFS= read -r line; do
            # prepend the file name to every line
            echo "$file:$line"
        done < <(git diff $START_DIFF $END_DIFF $file)
    done
) | grep what-youre-looking-for

I could not get the line numbers working, but I didn't really need them to get them to work. The prepended filename was enough for me.

---

My exact issue:

Find all the files that added either a from __future__ import .. or a -*- coding: utf-8 -*- out of 70+ files.

(
    START_DIFF=branch-a
    END_DIFF=HEAD
    for file in $(git diff $START_DIFF $END_DIFF --name-only); do
        while IFS= read -r line; do
            echo "$file:$line"
        done < <(git diff $START_DIFF $END_DIFF $file)
    done
) | grep ':+' | awk '(/import/ && /__future/) || (/coding/)'

The output looks like this:

....
app/tests/test_views.py:+# -*- coding: utf-8 -*-
app/tests/test_views.py:+from __future__ import absolute_import
app/tests/test_views.py:+from __future__ import division
app2/tests/test_views.py:+from __future__ import division
...

Answer 11

Ugly, potentially inaccurate, but might be enough to get by if you're just eyeballing it:

git diff origin/master | grep -Eo "(^diff.*|some-keyword)"

Where file1.txt contains some-keyword once, and file3.txt contains some-keyword 3 times:

Outputs:

diff --git a/path/to/file1.txt b/path/to/file1.txt
some-keyword
diff --git a/path/to/file2.txt b/path/to/file2.txt
diff --git a/path/to/file3.txt b/path/to/file3.txt
some-keyword
some-keyword
some-keyword
diff --git a/path/to/file4.txt b/path/to/file4.txt

Answer 12

Just another hacked-together script that produces slightly different formatting of the results, combining https://stackoverflow.com/a/50569950/1836776 and the original question. This will print filenames and lines that have changed in the diff, with the filename on the first line, and changes for the file on subsequent lines.

#!/usr/bin/env bash
# Adapted from https://stackoverflow.com/a/50569950/1836776

if [[ $1 == '--help' ]]; then
  script=`basename "$0"`
  echo "USAGE"
  echo "  $script term [git diff parameters]"
  echo
  echo "DESCRIPTION"
  echo "  Searches git diff results for <term>, returning only lines that have changed"
  echo
  echo "EXAMPLES"
  echo "  - Search all changes from master in current branch for 'TODO':"
  echo "    $script TODO master..."
  echo
  echo "  - As above, with pagination (color support):"
  echo "    $script TODO master... | less -r"
  exit 1
fi

colored=$'(\e\[[0-9;]*[a-zA-Z])'
marker="^$colored+diff"
pattern="^$colored+.*(\+|\-).*$1"
shift
# Yeah, okay, both sed and awk for maximum dork here, feel free to combine!
git diff --color $* | sed -rn -e "/$marker/! H; /$marker/ ba; $ ba; b; :a; x; /$pattern/ p" | awk "/\+\+\+/{f = \$2 \":\\n\"}; /$pattern/ {print f \$0; f = \"\"}"

Searches look something like this (you'll have to imagine the beautiful red/green colors for the changes):

# ./git-diff-grep.sh TODO main...
b/tsconfig.json:
+ /* TODO: don't forget to update this */
b/README.md:
-# TODO: update instructions
-TODO: license