[mini.diff] API to refresh mini.diff for a buffer

[emmanueltouzery]

Contributing guidelines

• I have read CODE_OF_CONDUCT.md

Module(s)

mini.diff

Question

first off, thank you a lot for your plugins! i really appreciate their small size and extensibility!

now, i'm leveraging mini.diff's extensibility for my own needs:
https://github.com/emmanueltouzery/nvim_config/blob/main/lua/telescope_buffers_entry_maker.lua#L6

i find it useful to see in the telescope buffer picker whether i have changes in individual buffers. I can leverage mini.diff's cache for that, since obviously the buffers are loaded.

there is a downside: when i commit, the changes should clear, now everything is clean. but the changes remain displayed in the telescope buffers picker until i manually reopen each buffer.

i think i could fix this if i had access to mini.diff's update_buf_cache(buf_id) function. That function is not exported, and I understand why... normally a user should not need it. however due to me leveraging the api, it would come in handy for my case.

i can easily fork mini.diff for my own needs (i routinely do that for other plugins), i think just exporting the function would be enough. But i'm wondering whether you see another way i could handle this or whether you'd be prepared to accept a PR to export that function?

thank you again for your work!

[emmanueltouzery]

Another similar use case: I'm using mini.diff stats to display in lualine diff stats for each window, with the same concern: on commit, only the current window has its stats updated, I must focus each window in turn to update its display. But at least there are less windows than open buffers.

[echasnovski]

i find it useful to see in the telescope buffer picker whether i have changes in individual buffers. I can leverage mini.diff's cache for that, since obviously the buffers are loaded.

there is a downside: when i commit, the changes should clear, now everything is clean. but the changes remain displayed in the telescope buffers picker until i manually reopen each buffer.
...
Another similar use case: I'm using mini.diff stats to display in lualine diff stats for each window, with the same concern: on commit, only the current window has its stats updated, I must focus each window in turn to update its display.

I am not quite sure where the problem is exactly. With default Git source executing git add * outside of Neovim or in a built-in terminal will update buffer's reference text (and get_buf_data() output as a consequence). I've just checked it: both signs and statusline component (even in not-current windows) are updated (i.e. cleared as everything is staged).

This is the case thanks to the created file watchers which watch any change in Git index and react with MiniDiff.set_ref_text() on relevant changes.

My suspicion is that the observed behavior might be not due to 'mini.diff' not updating its data but from its consumers not updating (or redrawing) to show the relevant data. There is a :h MiniDiff-update-event that can be helpful here.

I am afraid I can not help any further without explicit reproduction steps.

---

To answer the specific question in the title about refreshing: I think either MiniDiff.set_ref_text(MiniDiff.get_buf_data(buf_id).ref_text) can work. If not - MiniDiff.disable() + MiniDiff.enable() should always work, but that will do much more than simple update.

Also, if all you need is diff summary, using vim.b.minidiff_summary is more concise and performant than get_buf_data().

---

first off, thank you a lot for your plugins! i really appreciate their small size and extensibility!

Thanks for using them!

[emmanueltouzery]

so i concluded that it's about checktime. meaning when i do a git stash, my git repo doesn't change. but my checkout changes. meaning the open buffers should be refreshed from disk but they aren't being refreshed, and so mini.diff still computes the same old diff. i thought that :checktime would force these hidden buffers to refresh from disk, but i now found it it doesn't operate on hidden buffers. i'm now doing for _, b in ipairs(vim.api.nvim_list_bufs()) do vim.api.nvim_buf_call(b, function() vim.cmd('checktime') end) end and things seem to work, so i guess the point is that this was never a mini.diff issue. So I guess there's nothing to check for you!

[echasnovski]

So I guess there's nothing to check for you!

Thanks for the follow up! I'll still take a quick look some time later, and this info will be very helpful.

[emmanueltouzery]

if you find out anything relevant, i'd be thankful if you update it here or cc me on any related issue. i won't know if i fixed all my issues for a while as i'll use the software so... :) thank for your work again!

[echasnovski]

but my checkout changes. meaning the open buffers should be refreshed from disk but they aren't being refreshed, and so mini.diff still computes the same old diff.

Yes, this is exactly what happens. The git stash changes the file on disk but the Neovim buffer used to show the file is not yet updated. As 'mini.diff' is designed around "compare buffer text with its reference text" and the Neovim's buffer watcher doesn't notify it about the changes (since there are none), the summary is the same (because the buffer is the same).

Using git add * instead of git stash here works because of how 'mini.diff' watches Git index. It can detect the change via file watcher and adjust reference text on its own.

So I'd indeed say that 'mini.diff' here works as expected. And I am not entirely sure if there is anything reasonably simple it can do to make this use case less painful. Probably explicitly reloading buffers is the way to go here. Although judging from :h :checktime the single :checktime should be enough, but it doesn't look like it. Also, Instead of vim.api.nvim_buf_call(b, function() vim.cmd('checktime') end, the vim.cmd('checktime ' .. b) should work.

[emmanueltouzery]

interesting, i had only briefly checked but i missed the file watchers. the set_ref_text is also a good suggestion that seems likely to work. i'll check this out further and update here with my findings, thank you for your answer so far 👍