Experimental log batching

[Adda0]

The experimental log batching in 0.8.8 works great. It really helps on the nixpkgs repo. The initial loading time after opening jjui still takes quite some time, but overall a wonderful feature, and the subsequent loadings are then fast enough to not be all that disrupting one's workflow.

Improving the startup time to show only the first like 10 changes might help a lot, too. Just so that we get something we can work on immediately, without unnecessary waiting times.

[idursun]

Hey thanks!

nixpkgs repo is a little bit weird because it is also slowing jj down quite a bit and there's always some noticeable snapshotting happening before every command (due to the sheer size of the repository). I also noticed a delay in jj while it is listing the revisions until it fills the first page.

I had set the batch count to 50 and it is fine in most of the large repos I have tested against with. However, in nixpkgs case, I think it is spending most of the time until it accumulates up to 50 revisions. I can experiment with lower the batch sizes but then they have their own set of problems (half filled pages etc).

My current priority is to iron out some weird bugs due to introduced concurrency.

[Adda0]

I am way too spoiled by Lazygit where the nixpkgs repo loads instantly, and I can quickly run the few commands I need. jjui is getting there, though.

If the issue is loading up the first few revisions, then loading just a few from the get-go, and have a loading ... text below which will be eventually replaced with the next few revisions up to the first full screen, for example, seems like a good temporary (or even a permanent) solution. Of course, nixpkgs is really a special case. Another one might be the Linux kernel, or some browser engines, etc. Other than that, most people will not really need anything like this and are already well served by the log batching of today.

It might be interesting to think about possible optimization on the jj side as well, then. Maybe there could be something done to better support getting information for the large repos such as these. That would ultimately improve the user experience for both the TUIs and the CLI.

All of this is of course not a priority, and just some musings regarding the user experience. Wonderful work overall ❤️