feat: minimal diff

[will-lynas]

See #582

[yetone]

I encountered an error during my local testing:

[will-lynas]

I've added a check for that. Im interested, what do you do to get it to do that though?

[yetone]

I just pulled the latest version and tried it, but I can no longer trigger any diff effects.

minimize diff:

minimal-diff.mp4

original diff:

normal-diff.mp4

[aarnphm]

wondering if we can just generate the minimal diff instead of parsing the snippets

[will-lynas]

I just pulled the latest version and tried it, but I can no longer trigger any diff effects.

minimize diff:
minimal-diff.mp4

original diff:
normal-diff.mp4

Thats weird. Not sure why that's happening. The code that should be applied is definetly different from whats there already, right?

[will-lynas]

wondering if we can just generate the minimal diff instead of parsing the snippets

I've moved it to where the diff is created

[yetone]

wondering if we can just generate the minimal diff instead of parsing the snippets

LLMs are still unable to achieve this.

[aarnphm]

Quote reply

o1 can do this, but it is pretty expensive afaict

[will-lynas]

I think I'm actually running into a bug which is on main.

As far as I understand snippet.range is supposed to be start inclusive and end exclusive.

But there's this loop in insert_conflict_contents which treats the end as inclusive:

    for i = start_line, end_line do
      table.insert(result, lines[i])
    end

Heres a video of the ranges getting messed up:

Screen.Recording.2024-09-16.at.00.20.15.mov

But I also have one where it works somehow?

Screen.Recording.2024-09-16.at.00.22.51.mov

[aarnphm]

ok seems like a bug then :))