mjhjmtu

I want to compare lines in two files, but to minimize noise in the output, I want only the actual differences in the lines to be printed.

For instance, given the two files below:

a.txt

a b c d e f g h i j k l m n o p q r s t u v w x y z

b.txt

a B c d e f g h i j k l m n o p q r s t u v w x y z

(the difference between them is the case of letter b)

I want the output to be something like:

[-b-]+B+

Currently, the best approach I found was to use git diff --word-diff, but it outputs the whole line:

a [-b-]+B+ c d e f g h i j k l m n o p q r s t u v w x y z

Is there a more direct way to do it, other than manually parsing the output? Also, ideally I would prefer to use something more commonly available than git diff, e.g. a POSIX shell tool that would not require the user to install extra packages.

Using wdiff:

$ wdiff -3 a.txt b.txt

======================================================================
 [-b-] +B+
======================================================================

The -3 or ---no-common option will remove words that are common between the two files and only show the differences.

The ===... banner (and empty lines) may be removed with grep:

$ wdiff -3 a.txt b.txt | grep -vx '=*'
 [-b-] +B+

wdiff may also read unified diff data if you give it the -d or --diff-input option, for example from git:

git diff somefile | wdiff -d -3

Although wdiff is not a POSIX tool, it is commonly available.