git-diff: not taking line order into account

Question

I have a file in my repository that is generated by a software program.

This program sometimes reorder the lines on that file, which is not really important because line order doesn't matter. The problem is that when performing git-diff it's very difficult to see whether anything has actually changed or not.

Is there anyway to perform a diff in which line order is not taken into account? Or, if not possible with git-diff, any shell command you might think of?

Thanks!

Answer 1

At the end of the day, I'm manually running this command before committing to know whether the file actually changed or it's just a line reorder. Maybe I'll set up a git hook for it.

diff -wB <(sort file.txt) <(git show HEAD:file.txt | sort -)

This command compares the file in the working directory to the file in the last commit of your branch without taking the line order into account.

In my case, I use -w and -B, to ignore blank spaces and lines which the program also adds. From man diff:

   -w, --ignore-all-space
          ignore all white space
   -B, --ignore-blank-lines
          ignore changes whose lines are all blank

Answer 2

here's a pretty nasty hack using DIFF_EXTERNAL_TOOL:

#!/bin/bash

# this will be called with path old-file old-hex old-mode new-file new-hex new-mode
echo $*
diff -uw <(tr -s '[[:space:]]' < "$2" | sort -fbi) <(tr -s '[[:space:]]' < "$5" | sort -fbi) || true

this can be called with, assuming you throw the above script in the wdiff-sort program in your $PATH:

GIT_EXTERNAL_DIFF=wdiff-sort git diff

Here, for example, how I used this to compare a bunch of changes to bind zone files:

#!/bin/bash

#                path old-file old-hex old-mode new-file new-hex new-mode
#echo $*
atomize_zone="sed 's/^@//;/^\;/d;s/[[:space:]][[:space:]]*/ /g' | egrep -v '^.TTL (1D|86400)' | grep -v '; serial' | grep -v 'RR TTL' | grep -v '^.TTL 1H' | egrep -v 'IN SOA (ns1|bureau).koumbit.net' | sort -fbi" 

output=$(diff -uw <(sh -c "$atomize_zone" <"$2") <(sh -c "$atomize_zone" < "$5") || true)

if [ ! -z "$output" ]; then
        echo $1
        cat <<EOF
$output
EOF

This must look totally insane, but it actually works great and allows for cleaning up the files in various ways before diffing.