fix: don't ignore at_edge for floating windows

[thenbe]

trafficstars

fixes #217

Before this PR, we never invoke the user's custom at_edge behavior for floating windows (see minimal repro). Instead, we return early.

This PR ensures that, for floating windows, we call handle_at_edge() just before returning early.

And since we now invoke that logic in two different places, I extracted it into its own function handle_at_edge(). So really, the only real of this PR is the additional logic introduced on line 410.

Off-topic

The noisy diff is largely due to the indentation change. Block out the noise using:

git config --global diff.colormoved dimmed-zebra
git config --global diff.colormovedws allow-indentation-change

[mrjones2014]

Sorry for the delay. Been crazy at my day job.

I am pretty sure that ignoring at_edge on floating windows is intended behavior. There is a separate option float_win_behavior that controls the intended behavior for floating windows, since most users want them to be treated differently from normal windows.

See: #201 #202

I would be open to adding support for float_win_behavior = 'at-edge' to make the floating window behavior mimic the regular at-edge behavior.

[thenbe]

I would be open to adding support for float_win_behavior = 'at-edge' to make the floating window behavior mimic the regular at-edge behavior.

That would be even better. :)

[jkellz-dev]

Did anything happen with this? I am running into it with LazyVim's Snacks.Explorer plugin. Since the file explorer is technically a floating window, it breaks navigation. Works great with NeoTree. I might be missing configuration to make it work, but it seems like its the floating window thats mucking it all up.

[mrjones2014]

I proposed an alternative approach here but the PR hasn't moved since then.

Happy for anyone to pick this up, it should be relatively straightforward to implement.

[jkellz-dev]

happy to submit a PR. Not completely familiar with the inner workings here. Looking at the related changes, adding a universal handling of floating windows at_edge may not be desired. Since the edges may be different depending on configuration. Would adding an exception list for buftype/filetype be amenable?

I'm thinking about how you could have 2 projects open side by side. smart-splits needs to know the pane is in a position.

Maybe this is better fixed from the Snacks.Explorer side.

[mrjones2014]

Would adding an exception list for buftype/filetype be amenable?

That sounds fine. There's already global tables for that in config, but if we need a more context-specific config entry, that's also fine.