OUTDATED: Support staged vs. unstaged changes, to support splitting hunks

[jeyj0]

This is mostly a placeholder for a longer text that I hope I'll get around to writing, but the core of it can be summarized with these points:

• Splitting hunks can be done using git add -p, then e.
• GitButler makes no distinction between staged and unstaged changes.
• If distinction was made, we could split hunks using git add -p, then e.

[Byron]

Thanks a lot for sharing!

Even at the early stage of this idea I'd like to chime in and share what I know.

First of all, I think it's fair to say that this would be a tough sell - not hunk-splitting and all kinds of workflows around it, of course, but the re-introduction of the index that GitButler tries very hard to hide away.

GitButler would certainly prefer to become so good at hunk splitting/hunk-setup that you wouldn't want to use the command-line workflow anymore.

Admittedly, without the index, it won't be super easy to persist multiple hunk selections, but maybe GitButler could manage to use the index under the hood somehow. And if it would use it, then of course git add -p would be more of the same. Maybe something clever is possible in that regard eventually?

[jeyj0]

Yeah I didn't necessarily mean that the index is exposed in GitButler as a "manageable" status.
I was mostly surprised that when I did stage something, I didn't see any difference in GitButler.
To enable a super simple way of hunk-splitting, if GitButler was at least able to split hunks that are partially staged and partially unstaged, that would already satisfy my needs for now. Currently, it takes a lot of gymnastics with additional stashing to get things where I want them to:

1. git add -p
2. git stash -S
3. git stash
4. git stash apply 1
5. do the commit with GitButler
6. git stash apply 0 (often resulting in having to resolve conflicts)
7. git stash clear

If after git add -p the hunk simply showed as being split (no further information or features about the index necessary in GitButler), the workflow would already be greatly simplified:

1. git add -p
2. do the commit with GitButler

But I wouldn't be surprised if that would take a significant amount of effort in GitButler to support...

[jeyj0]

Damn, by accident I just realized that I can do line-based commits in GitButler already 🙃
Nvm this idea then! 😄
I could've avoided a lot of work yesterday...

[Byron]

I my, so sorry to hear that! I also didn't catch up on it either and thought you really wanted to use the index, knowing well enough that GitButler can already do hunk selections, albeit rather differently from Git.

[jeyj0]

All good; I'm happy to have learned!