haskell: use ghc-mod for "mhi" and "mht" bindings when enabled

[tko]

ghc-mod can get type / info without having to manually push the code to
REPL first. #3173

I'd think maybe the "needs repl" bindings should be moved under m s prefix to make the context more clear. Also the "mhT" function could probably be reimplemented in terms of ghc-mod.

[syl20bnr]

Can we disambiguate SPC m h t before merging this PR ? I can see that SPC m h t can be 3 different things.
Also the documentation has to be updated with the new key bindings.

[d12frosted]

Originally haskell-mode has 'magic' key binding C-c C-t which is used by different modes to set proper behaviour. I don't think we should really mimic it. Instead, we could map SPC m h t to C-c C-t.

There are several such key bindings. So similarly we could map SPC m h i to C-c C-i.

[PierreR]

I think we should go and ask the authors from haskell-mode what to do here ...

Right now ghc-show-type is doing more work compared to haskell-process-do-type so the PR is alright but what if in a near future the situation is opposite ...

That and the fact that in a near future we might want to integrate stack-ide which is yet another (quite promising) ghc-mod competitor.

[d12frosted]

Well, that's why I propose to map SPC m h t to C-c C-t, because haskell-mode itself maps this key binding to function that will ask you to setup interactive mode. But, if you are using interactive mode, C-c C-t will be bound to haskell-process-do-type. When you are not using interactive mode, but using ghc-mod, C-c C-t will be bound to ghc-show-type.

As for ghci-ng - it uses haskell-mode-show-type-at (defined here in haskell-mode), but official tutorial says to bind this function to C-c C-t. For ghc-ng we already set SPC m h t to haskell-mode-show-type-at here, but probably it's easier to set it to C-c C-t (so it works as expected in haskell-mode) and then bind SPC m h t to C-c C-t.

As for stack-ide - it's immature yet, but I think that they will also reuse C-c C-t, because it's common.

P. S. Probably I am using wrong terminology, when I say bind SPC m h t to C-c C-t, but I hope you understand me 😄

[PierreR]

bind SPC m h t to C-c C-t

That makes sense. How do we do this ?

[d12frosted]

The only way I know is defining function like

(defun spacemacs/haskell-show-type ()
      (interactive)
      (call-interactively (lookup-key haskell-mode-map (kbd "C-c C-t"))))

And just making regular bind from SPC m h t to spacemacs/haskell-show-type.

[d12frosted]

@tko What do you think about it?

[syl20bnr]

Thank you for the PR, I close it in favor of #3345.

[PierreR]

I have just tested the PR but it does not seem to work out. I am currently on develop e72dd8a.

Pressing c-c c-t will show me the type. On the other hand mht still requires loading haskell-process. It looks like mht is still bound to haskell-mod-show-type-at (or haskell-mode-enable-process-minor-mode if I don't use ghc-ng),

If I am correct the aim of the PR was to map mht to whatever c-c c-t is doing

[d12frosted]

If I am correct the aim of the PR was to map mht to whatever c-c c-t is doing.

Yes, it should be mapped to whatever C-c C-t is. But, if you have ghci-ng enabled - then C-c C-t should not work, while mht should work, because for ghci-ng we need to explicitly set key bindings. Since C-c C-t is not spacemacs keybinding - we don't setup it for ghci-ng users.

As for you problem - 3 days ago it was working for me like a charm. I'll try to reproduce it once I am back.

[d12frosted]

I'll take a look into problems with ghci-ng on weekends. Sorry for such a big delay!

[d12frosted]

Still working. Just make sure that you are really running ghci-ng - there are some caveats with it, which I described in #3544.