Optimize and simplify haskell-ds-create-imenu-index with hashtable

[Hi-Angel]

The function has implemented a poor man's hash table by manually enlisting keys and then using cl-case on values. Just replace this with a hashtable and thus simplify the code considerably. This makes it faster as well because the cl-case + setq is O(n) whereas puthash is O(1).

The output of the function was tested on Pandoc's
Writers/Powerpoint/Output.hs file of 2841 lines and comparing the hashsum of the output before the changes and after. It didn't change.

The commit is also basically a backport of a similar change on purescript-mode that has been forked form haskell-mode at some point: purescript-emacs/purescript-mode#27

[Hi-Angel]

Just in case: CC: @cocreature as the last person who touched the function

[purcell]

Thanks for this. I approved the CI and it's complaining about part of the change, not immediately obvious to me why. :)

[Hi-Angel]

Hmm, bizarre indeed… The snippet (curr-alist (gethash (car type) imenu)) (as well as the line, and the whole "added" side of the diff for that matter as far as I compare) is exactly the same as in the purescript-mode PR, however there byte-compilation passed just fine everywhere starting with 25.1. Very strange.

Okay, thank you, I'll look at it later

[purcell]

It's a warning that is promoted to an error, but the cause is when-let* not necessarily being defined. Adding (require 'subr-x) fixes it for me, at least in Emacs 28.x.

[Hi-Angel]

Oh, thank you! I was comparing the require lines on the purescript and haskell "decl-scan" files, but they are the same, so I dismissed that thought. But perhaps in "purescript-mode" the (require 'subr-x) line comes from another import (both haskell and purescript "decl-scan" files include the main mode file).

I added the line, let's see if CI will be okay with it

[Hi-Angel]

Welp… 😊 I'll dig into it later, don't bother for now, thank you!

[purcell]

But perhaps in "purescript-mode" the (require 'subr-x) line comes from another import (both haskell and purescript "decl-scan" files include the main mode file).

Yeah, I don't really like that pattern of importing, partly for this reason!

The error with Emacs 25.1 is because when-let* was only added to subr-x in Emacs 26.1. Probably best to avoid that macro — when-let was available in Emacs 25.1, so you can just use that instead.

[Hi-Angel]

The error with Emacs 25.1 is because when-let* was only added to subr-x in Emacs 26.1. Probably best to avoid that macro — when-let was available in Emacs 25.1, so you can just use that instead.

Well, when-let* compiles just fine on 25.1 with the same change on purescript-mode

[purcell]

Huh! 🤔

[Hi-Angel]

Well, when-let* compiles just fine on 25.1 with the same change on purescript-mode

Oh, actually, scratch that. I just looked at the job log, and apparently the purescript-mode CI, for some reason, doesn't compile that specific file at all. It compiles other files, but not this one. So most likely this is the reason I don't see the failure there.

[Hi-Angel]

Okay, in this case, another attempt… I changed when-let* to when-let 😊

The reason I attempted when-let* is because when-let was declared obsolete in preference of when-let* in current Emacs master…

[purcell]

The reason I attempted when-let* is because when-let was declared obsolete in preference of when-let* in current Emacs master…

Yeah, always tricky juggling these things when aiming for compatibility with a wide range of Emacsen. It's possible that the compat package helps here, but I'd rather not add it as a dependency just for this case.

Merging now, thanks!