What do people do to avoid tracking intermediate source files?

[halloleo]

Sometimes I make a local copy of a source file, work on the original and then look at both files.

E.g say, I have a file Main.elm. I copy it briefly to Main_sav.elm, so I have it handy to see both files next to each other for a little while. Then I will delete Main_sav.elm, continue development and eventutally commit my changes to Main.elm.

The intermediate file Main_sav.elm should never be in the git storage system. But from my understanding jj will write it to git when I do nothing.

So I have to untrack it, right? But it feels tedious to be forced to untrack a brief intermediate file... What do people do?:

• Do they have a *_sav.* exclusion in their .gitgignore?
• Or do they just live with the _sav files being tracked in git?

Thanks for pointers!

[CarlMitchellKT]

I use my editor's ability to show a diff between two commits. Set it to compare HEAD & the starting commit, and there's no need to make the _sav file.

[halloleo]

Totally understand. The mentioned case to take a copy of a source is only one example - Sometimes I have short-lived intermediate files like test data in CSVs, etc which I don't want to end up in in the storage system.

For all of these intermediate files, do I have to create .gitignore rules, or do I just live with the fact that they are recorded in the (ephemeral) jj tracking system, but not get committed to git?

[nertzy]

I personally make a tmp/ directory in the root of my project and add it to .gitignore - then any files placed there are treated as "scratch" and are untracked.

[jennings]

FYI: If you want to do this but don't want to add to the committed .gitignore, jj also respects ignore patterns in .git/info/exclude

[roteiro]

For all of these intermediate files, do I have to create .gitignore rules, or do I just live with the fact that they are recorded in the (ephemeral) jj tracking system, but not get committed to git?

@halloleo
Are you looking for the setting snapshot.auto-track=none()?
With that setting, new files won't be tracked unless you add them with jj file track

[halloleo]

@roteiro

Are you looking for for the setting snapshot.auto-track=none()?

Uh, that is really interesting! Maybe this will help! I will read up on it.

Do people use this setting?

[steveklabnik]

I would edit the file and then use jj diff to look at the difference between my changes and the original file. To me, the entire point of version control systems is to not need to keep extra copies of files around. But we work differently :)

That said:

But from my understanding jj will write it to git when I do nothing.

Sort of. It'll end up in your local repo, but when you end up moving it back to Main.elm, the contents of the commit you actually send to others isn't going to have Main_sav.elm anywhere.

[halloleo]

Thanks for this @steveklabnik.

But from my understanding jj will write it to git when I do nothing.

Sort of. It'll end up in your local repo, but when you end up moving it back to Main.elm, the contents of the commit you actually send to others isn't going to have Main_sav.elm anywhere.

I see. Interesting! So as long as I don't create a commit containing Main_sav.elm, it is only track in jj, but not in the git storage system, right?

[steveklabnik]

jj will create local commits that contain Main_sav.elm, but those aren't shared outside of your machine.