(mini.misc) Add support for restoring the cursor

[cryptomilk]

What it does:

• When reopening a file, this will make sure the cursor is placed back to the position where you left before.
• This implements :help restore-cursor in a nicer way.
• Creates autocommand for |BufReadPre| and |FileType| event with |MiniMisc.restore_cursor|

• I have read CONTRIBUTING.md
• I have read CODE_OF_CONDUCT.md

This is still work in progress. However I wanted to ask if you would be fine integrating this.

[echasnovski]

Hi!

First of all, thanks for taking time to write this. It would be even better if you created an issue to discuss it first (as said in CONTRIBUTING.md).

Second, could you clarify the purpose and scope of this proposed module? Is it a slightly polished version of the autocommand from :h last-position-jump (like open fold, etc.)?

Third, it is a strong opinion of mine that no ignore_buftype or ignore_filetype will be used as options of 'mini.nvim' modules. It gets complicated to track and not easy to make future proof.

Edit: for reference, this was already discussed earlier

[cryptomilk]

I thought before opening an issue I could also create a MR. I wanted to play with mini.nvim anyway :-) So this was an exercise learning the code ...

About the plugin:

By default if you open a file with neovim, the cursor is positioned always at line 1 column 1 (1,1).

If you have the plugin loaded and lets assume you jump to line 60 and move the cursor to column 5 (60, 5). Now you close nvim and open the file again with nvim. With the plugin loaded the cursor will be positioned back on line 60 and column 5 (60, 5). It is last position were you left the file.

The information about the position where you left a file is stored in shada (see :help shada)!

The plugin is an enhanced version of the :h last-position-jump. I find this extremely useful. Try it.

[echasnovski]

I thought before opening an issue I could also create a MR. I wanted to play with mini.nvim anyway :-) So this was an exercise learning the code ...

Sure, I can understand this motivation :) Would be nice following the guidelines, though.

By default if you open a file with neovim, the cursor is positioned always at line 1 column 1 (1,1).

If you have the plugin loaded and lets assume you jump to line 60 and move the cursor to column 5 (60, 5). Now you close nvim and open the file again with nvim. With the plugin loaded the cursor will be positioned back on line 60 and column 5 (60, 5). It is last position were you left the file.

The information about the position where you left a file is stored in shada (see :help shada)!

OK, I see. Having in mind comments about no ignore_* options in modules and the common practice of trying to always open just enough folds when moving cursor, it seems too small of a functionality to be a whole module.

I am not opposed to having this as a part of 'mini.misc', though. Along the lines of MiniMisc.setup_auto_root() and MiniMisc.find_root(). Say setup_goto_last_cursor() and goto_last_cursor(). Might be tricky to write tests, though.

The plugin is an enhanced version of the :h last-position-jump. I find this extremely useful. Try it.

I now understand why I've never found this functionality useful... I am using sessions 99% of the time, and they restore this information automatically.

[cryptomilk]

You really do not want the 'cursor goto' for git commit or the quickfix window. You want to disable it for some buffer/file types. If you do not want a maintain a list, the variables could be empty. However a user might want to be able to set those. It is really annoying if you do a git commit and the cursor is in the middle of the comment section. Is it ok to have the options cursor_goto_ignore_filetype but an emtpy table as the default? Then the user could maintain the list ...

[echasnovski]

You really do not want the 'cursor goto' for git commit or the quickfix window. You want to disable it for some buffer/file types. If you do not want a maintain a list, the variables could be empty. However a user might want to be able to set those. It is really annoying if you do a git commit and the cursor is in the middle of the comment section. Is it ok to have the options cursor_goto_ignore_filetype but an emtpy table as the default? Then the user could maintain the list ...

If we are talking about a module - no, any filetype/buftype related disabling is done only through vim.b.minixxx_disable variables. User can then create autocommands to automatically set them up (possibly using a provided common recipes). It is also easier to implement.

If we are talking about 'mini.misc' setup_* type of function, then I think these lists can be an optional arguments.

[cryptomilk]

I've added it to MiniMisc.setup() but I can move it to setup_goto_last_cursor()

[cryptomilk]

I've fixed the last issues, added more code comments and improved documentation.

[cryptomilk]

Looks like we can't test the center option. I don't find a way to get the cursor postion in the window.

vim.api.nvim_win_get_position() gives you the buffer-relative postion, not the window :-(

[cryptomilk]

I already do.

You use os.tmpname(), which is not predetermined. With this approach it is easy to not detect problems with cleanup.

Please, use approach similar to 'test_sessions.lua': this, this, and this (but inside new_set() for restoring cursor).

Done. Thanks for the pointer :-)

[cryptomilk]

I think, I have covered everything with tests now.

[echasnovski]

Sorry for many comments. This does look very good, only many small optimizations needed.

And general request: please use more explicit test case names describing what should work correctly. Like:

• Instead of ['nocenter'] - ['respects `opts.center`'].
• Instead of ['file'] - ['works in general file'].

[cryptomilk]

I don't see auto_root() documented in readmes/mini-misc.md. If you could do it I could follow it and document restore_cursor().

[echasnovski]

I don't see auto_root() documented in readmes/mini-misc.md. If you could do it I could follow it and document restore_cursor().

I'll probably add some kind of summary for all interesting 'mini.misc' functions when this one is done.

[echasnovski]

I've taken a quick look. Seems to be nothing major to change. Still some small-ish code style issues, but I think I'll fix them myself during merge. Thanks!

I'll take closer look, probably, the day after tomorrow.

[cryptomilk]

Thanks for all the help!

[echasnovski]

Thank you for your work!

I'll merge this in separate branch and will polish the rest myself (like not modifying the commonly used set_cursor() helper). Of course, you'll be credited as initial author.

[echasnovski]

@cryptomilk, I've polished it and merged into main.

Couple of differences that I made:

• Removed ignore_buftype and made it work only for normal buffers. This seems to be a more reasonable approach. If not, can add those later.
• Decided to act zv always and not behind foldclosed(), as I didn't see a use case where this goes wrong.
• Rewritten tests to test setup_restore_cursor() directly, which required some tricky approach for restarting child process.
• Added note that it won't work if file doesn't have recognized file type. This is a limitation of FileType approach, but it doesn't seem to be a huge deal to me right now.

Thanks again for going forward with this!

[cryptomilk]

@echasnovski You're welcome.

I've rebased my branch which implements the autocmd using vim.api.nvim_create_augroup() here:

https://github.com/cryptomilk/mini.nvim/tree/mini.restore_cursor_lua