fix handling multiple windows for a buffer - fix #11

[FelipeLema]

fixes #11

[FelipeLema]

it's working now, although it's missing tests... but you may want to add those in a separate PR

[FelipeLema]

cause I'm uncertain of when can I add them

[wookayin]

This looks good overall, but I'm not 100% sure about the change in the way the viewport is obtained. We get the viewport inside the autocmd handler, and would you please give more context/explanation about why this change is necessary and makes more sense? (the previous logic was written by the original author in 2019, so I'm not quite familiar with why @numiria had needed this -- see 3dc14aa for the context).

[FelipeLema]

oh, I see ... it seems fetching the viewports from python can be costly (not sure how bad)

the change to obtaining them from python was rather because I prefer to have as little vimscript as possible... I don't mind making the changes to fetch all viewports from vimscript. will do just that

[FelipeLema]

ready to go

I re-implemented the get_viewports() function in lua, but it could be written in vimscript in the future

[wookayin]

Thanks for the contribution again. Actually I'm having a hard time reproducing the bug. I tried this session but haven't had the bug. Can you make a more clean step please, possibly using neither mksession nor absolute (user-specific) paths, starting from nvim --clean? Or writing a unit test will be even better, but I can take care of it as long as it's easily reproducible.

[FelipeLema]

lemme try and add an unit test

It's probably not going to be ready this week. Hopefully this month

[FelipeLema]

oh, I see that we're not testing highlights. the to-be-implemented test points to neovim/neovim#6412, which has been solved

we may want to address that TODO first

[wookayin]

I see. I will work on refactoring to switch to extmarks from the legacy highlight API soon. Adding a test is not a big priority.

BTW using the classic vim API synIDattr(synID(line, col, 1), 'name') would still work.

[wookayin]

Actually I don't mind we don't have tests yet as we're lacking a infra for testing highlights, but I still would like to reproduce the issue.

[FelipeLema]

Will pick up the MVE-as-test, although I was shifting my focus into implementing an LSP server that only does semantic highlight using code from this repo

anyone reading this can chip in with such test in the meantime