stage changes in git based on regex

Question

A lot of my debug code is wrapped in a header/footer. Usually I run a script that just removes all of it from source before I stage and commit the changes. If the debug code needs to persist a commit I'll run git add -p and stage the chunks individually.

My question is if it's possible to stage changes based on a regex? For example, given this snippet of javascript:

function littleNinja( sword ) {
    /* debug:start */
    console.log( sword );
    /* debug:stop */
    // do stuff
}

I don't want git to stage the lines between and including debug:start and debug:stop, and my hope is that this process could be automated.

Answer 1

A good way of doing this is to use clean and smudge filters to filter out the marked lines as they are staged. (The description below is essentially the same as the procedure described in that section of "Pro Git", and which is very similar to that in the git attributes documentation.) First of all, we need a command that removes the lines between (and including) each debug:start and debug:end — fortunately, this is a simple job with sed:

sed '/debug:start/,/debug:stop/d' < foo.js

Now create a .gitattributes file in the top level of your git repository which contains the following line:

*.js filter=strip-debug

And now define what the strip-debug filter does both when 'cleaning' (on staging) or 'smudging' (on checkout):

git config filter.strip-debug.clean "sed '/debug:start/,/debug:stop/d'"
git config filter.strip-debug.smudge cat

I've (very briefly) tested this and it seems to do what you want.