The formatting notification is too distracting

[banacorn]

I think there's really nothing worth notifying when formatting the document, especially when the formatting command has to be triggered by user by default. It becomes even more distracting when you enable, say, format on save.

There should be an option in the settings for disabling these notifications.

[konn]

Well, with some projects with large code base, Formatting only a single file can take much, much time and indicator spins for a while. If the codebase is relatively small, notification is not shown so regularly.
I think the cause of this delay is that HLS is waiting for the AST to be parsed, and compiler took much time to reach there..

My altenative suggestion is to add an option that "fallback text-based formatting when no progress has been made after N secs"; taking more than a few seconds for formatting code is not so desirable, so we need some escaping hatch for it, I think.

[banacorn]

reason-vscode only shows the notification when the formatting takes too much time. Perhaps we can have a threshold of some kind.

I wonder how much time it takes to format a Haskell file (of different code base sizes).

[lukel97]

I noticed this as well, formatting doesn't usually seem to be that time intensive of an operation so it would make sense to debounce it first.

[pepeiborra]

Formatting can block for other reasons though, like consulting a cradle. Debouncing is always good, even better if done on the LSP client side since it is a crosscutting concern across all progress notifications.

[georgefst]

Worth noting, currently the progress notification only shows for Ormolu/Fourmolu, presumably because the others aren't calling withIndefiniteProgress.

The notification doesn't annoy me per se, but what's odd is that it hangs around for almost a second after formatting appears to have finished, even on a trivial file. This could be because we call Ormolu with the idempotency check, which means that everything is actually formatted twice, but it's longer than I'd expect even accounting for that. Anyway, having a threshold of even a second would make this a non-issue for me.

[jneira]

I am gonna use withIndefiniteProgress in the hlint plugin, so this would affect it

[Profpatsch]

I get the problem that with format-on-save, formatting will block saving the file when hls is working on something in the background (especially when it crashes/hangs, which happens if the file has syntax errors on hls start for example).

Requiring me to manually stop hls and interrupt the formatting before I can save the file.

[michaelpj]

Possibly if we didn't declare them as cancellable the client would not show a UI element (which I think it might be doing just to show the cancel button). I think it's not really necessary to say they're cancelable.