[help wanted] How to keep extmark from removed after commenting code

[shellRaining]

hello, very thanks for your nice work! mini.indentscope has cool effect, and provide many inspiration when I wrote my plugin. I am writing a indent (chunk) plugin, and encounter a bug, when comment a line which has extmark, the extmark was remove, detail can be seen here shellRaining/hlchunk.nvim#109

I found that mini.indentscope also has lazy feat and can keep extmark after commenting a line, But I feel very confused about how he achieved this effect.

I originally wanted to post on Reddit, but it seems that my account and IP have been banned. I'm sorry if this has caused any inconvenience.

[echasnovski]

Hi, @shellRaining! I remember seeing screenshots of that chunk highlighting and thinking that it was beautiful. Nice work!

Applying nvim_buf_set_lines() indeed "squashes" all extmarks to be on either side of set lines. This is essentially close to manually deleting those lines (which moves extmarks) and pasting text (which also might move extmarks).

Extmarks from 'mini.indentscope' are not exceptions. The way 'mini.indentscope' makes it look like extmarks don't move is by redrawing on every 'TextChanged' event immediately before Neovim redraws the screen.

In turn, 'mini.indentscope' decides whether it should redraw scope immediately or not is by comparing currently drawn scope with target. If those scopes intersect, then it draws with delay 0 the whole scope right away.

So in short: it re-adds extmarks very quickly after every text change.

---

For the future, this type of question is better be posted in Discussions directly. I've transferred this one myself.

[shellRaining]

Thank you very much for your help. I understand how to solve it, and I noticed that you mentioned lockmarks in the official repositories of Neovim and comment.nvim. Can we use this to ensure that calling nvim_buf_set_lines() not clear extmark?

I have searched a lot of information online, but they are all used in conjunction with visualMode, and I still don't quite understand how to use it. Thanks!

[echasnovski]

... I noticed that you mentioned lockmarks in the official repositories of Neovim and comment.nvim. Can we use this to ensure that calling nvim_buf_set_lines() not clear extmark?

Unfortunately, no. The lockmarks command modifier only preserves regular marks (like </> for latest visual selection, etc.), not extended marks (which are used for all sorts of things).

The only way (right now) to preserve extmarks is to precisely use nvim_buf_set_text() to edit only those parts of lines which need to be changed.

But (as always) it is not that simple: preserving extmarks is not a universally good thing. I spent quite a bit of time contemplating this in 'mini.comment' (and built-in commenting by extension). In the end, using nvim_buf_set_lines() and not preserving extmarks seems like the best thing to do for commenting. Here is a full PhD size comment with motivation.

[shellRaining]

Thank you for your clarification. I saw that you mentioned a set of data. When commenting four thousand lines, using nvim_buf_set_text() will consume about 400ms. There are two points of doubt:

1. Will we often comment out so many lines of code? Are there other reasons why we choose nvim_buf_set_lines()?
2. Can we divide the four thousand lines of code into blocks (such as one win_height as a block), and then send them to the task queue to avoid lagging as much as possible?

[echasnovski]

1. Are there other reasons why we choose nvim_buf_set_lines()?

Yes, see the next point.

2. Can we divide the four thousand lines of code into blocks (such as one win_height as a block), and then send them to the task queue to avoid lagging as much as possible?

Sure, but that is a more complex approach and won't solve other issues with preserving extmarks.