Thanks for implementing this!

I've added some tests, and they didn't pass initially. Turns out, diffy lines already contain newlines character in them, so things like .join("\n") and writeln!() would add extra newlines, breaking the patch. I've made a fix, the tests are green now.

It's too bad that diffy doesn't support mutating hunks. This makes the code overly complex and less efficient, since we're reformatting the patch on every calculate_size() call. I know I was the one who pointed you to a reference implementation that uses diffy, but now I think it's not a good fit.

I think we can get rid of the dependency on diffy. Instead, let's just do simple diff string parsing. Essentially, we need two levels of parsing:

• Split by file (we already do this)
• Split by hunk headers (which always start with @@ )

Then we can keep a list of hunks as a Vec[String]. This way we don't have to reconstruct the patch, and computing lengths becomes much easier.

Are you willing to implement this change or should I go ahead? We can also pair on it, if you like.

Are you willing to implement this change or should I go ahead? We can also pair on it, if you like.

Yes, I am willing to implement this, I would request for a pair session if I face any difficulties.
Thank you for your time : )

@osyvokon would appreciate your review!
Thank you : )

Looks much better now, thanks! I've made a few nit renames while reviewing this. Ready to merge.