feat!: expand options for LSP formatting

[stevearc]

This deprecates the lsp_fallback option (gracefully; it should continue to work) and replaces it with the lsp_format enum which can be several values. Three of the values produce behavior that was possible with lsp_fallback, and two of them are new behaviors. The new behaviors fix #316

lsp_format	lsp_fallback	description
nil or "never"	nil or false	Do not use LSP formatting
"fallback"	true	Use LSP formatting if there are no formatters configured for this filetype
"prefer"	-	Use LSP formatting when it is available, otherwise fall back to the formatters configured for this filetype
"last"	"always"	Run configured formatters, then run LSP formatting
"first"	-	Run LSP formatting, then run configured formatters

[folke]

Hi!

I'm getting multiple reports from users that the lsp fallback doesn't work anymore.
Haven't looked into it myself.

• bug: breaking change in conform.nvim LazyVim/LazyVim#3706
• bug:Prettier not working again.  LazyVim/LazyVim#3525

[dpetka2001]

I don't know if this is an intended change or not, but the previous fallback behavior didn't check if there was a formatters_by_ft defined for the current buffer, so it just used lsp formatting as a last resort.

Now, with this line introduced in the latest update, it seems that it also checks if there's no filetype formatter defined in formatters_by_ft, which is clearly different from before. This leads to confusion to users who were accustomed to previous behavior, that used LSP formatting if no other formatting was used before. Again, I don't know if this is intended behavior or not in the latest update.

[folke]

@dpetka2001 nice find!

The LSP should definitely be used as a fallback.
@stevearc I assume this change was not intentional?

[dpetka2001]

Also, this function M.will_fallback_lsp does not have any references after the last update. So, it's not being used. Previously, it was used here. So, I believe more changes would have to be introduced other than the line that checks for the filetype formatters. Unfortunately, I don't understand the codebase that well to really understand how this should be solved and can't submit a PR. Sorry about that.

[dpetka2001]

Did my best and submitted a PR that fixes this behavior if it's not intentional. Please feel free to disregard it if you think a better approach is possible, as I'm not that familiar with the codebase myself.

[stevearc]

This behavior change is not intentional, it was supposed to be a transparent API change with no effect on behavior. I'll take a look at the PRs and make sure a fix gets in soon

[stevearc]

Should be fixed by bde3bee