fix high cpu

[jinzhongjia]

Recently, my main development environment was changed to windows. I found that the event used by git-blame is cursormoved, and this event is triggered very frequently.

Git itself on windows runs slower, which causes the entire neovim to be blocked

I replaced cursormoved with cursorhold

[jinzhongjia]

for some users, they may like using cursormoved to get an immediate response, we can implement one debounce function

[jinzhongjia]

this pr can solved there issue:

• uncommited lines issues (slow navigation, refresh issues) #83
• Scrolling quickly throws errors | too many open files: "/bin/zsh" #119
• Performance on big git projects #52

[jinzhongjia]

I decided to implement additional debounce functionality, although CursorHold is enough for me

[jinzhongjia]

This pr add three config options:

• gitblame_schedule_event
• gitblame_clear_event
  Instructions have been added to the readme

[f-person]

Hi! Thanks for the PR! I'll check it in the near future!
In the meantime, is the debounce functionality different from https://github.com/f-person/git-blame.nvim?tab=readme-ov-file#visual-delay-for-displaying-the-blame-info?

[jinzhongjia]

Oh, I didn't notice this before
I implemented debounce, schedule and clear virtual text for both

[jinzhongjia]

Now, I think every thing is ok, current code only enhances the original functions, adds event configurations for schedule_show_info_display and clear_virtual_text, and debounce time is 250 milliseconds by default.

[bossley9]

This is great! I left a few comments mostly on pull request polishing.

[bossley9]

Other than that, I'm ok with this as long as @f-person is ok with it! 🚀

[jinzhongjia]

ok, thx

[bossley9]

@f-person have you had any time to take a look at this?

[f-person]

thank you, this is really great! i approve this pr, but left a comment; let's discuss that matter a bit further and maybe merge it or decide to do something else. looking forward from hearing back from you! thanks! <3

[bossley9]

I linked the issues you mentioned since it is very likely this PR will fix them. The best way to test a big repository is to clone https://github.com/nixos/nixpkgs and try to open pkgs/top-level/all-packages.nix 😄

[jinzhongjia]

I tried it(nixpkgs pkgs/top-level/all-packages.nix) briefly. At least for now, gitblame will not prevent me from scrolling, but when git gets the commit message in the background, there will still be a high CPU usage.
Both windows and linux are

[jinzhongjia]

By the way, it is recommended to close all treesitter related plugins during testing, including itself, because the performance of treesitter is really poor.
Only turning on treesitter related plugins will prevent me from scrolling. . .

[f-person]

Okay, let's merge this! Thank you very much for the effort on this, @jinzhongjia & @bossley9!! <3
Ping me if you're ever in the Netherlands, and I'll get you a coffee or a beer :)

[f-person]

Maybe clearing the existing message should be done immediately on cursor move and not whenever we set the next one?

[jinzhongjia]

I think that's ok, let me create a new pr
Well, the amount of code is a bit too much, so I won’t createa new PR for the time being.