resurrect scala-indent:indent-on-parentheses

[hvesalai]

This commit removed our own scala-indent:indent-on-parentheses handling in favour of using smarparens but never added documentation on how to achieve the same functionality.

[hvesalai]

618f0fbef4d3d94156f834a66e36c79f07a54e02

[hvesalai]

The functionality removed was:

(defun scala-indent:indent-on-parentheses ()
  (when (and (= (char-syntax (char-before)) ?\))
             (= (save-excursion (back-to-indentation) (point)) (1- (point))))
    (scala-indent:indent-line)))

What in effect indents the closing parenthesis when inserted on an empty line. The added documentation:

  (sp-local-pair 'scala-mode "(" nil :post-handlers '(("||\n[i]" "RET")))
  (sp-local-pair 'scala-mode "{" nil :post-handlers '(("||\n[i]" "RET") ("| " "SPC")))

does something completely different.

[fommil]

Ok, I thought it did the same thing. Maybe I've got some other mode/setting active that does the indentation when I insert parens that I didn't document properly. I'm pretty sure smart parens can handle this.

If we need to add it back in, I'd like to avoid adding it to post-self-insert-hook as it causes performance problems and conflicts with other modes... it's easy enough for somebody to manually add it as part of their scala-mode-hooks.

[hvesalai]

I'm fine with just adding documentation on how to do this with smartparens. Just can't figure it out my self.

[fommil]

with smart parens, type { then RET and you should end up with

{
  <point>
}

all three lines are indented for me. Maybe not the first line, come to thing of it.

[hvesalai]

Yes, but I don't want that. I want it to be as it was before so that with

foo { RET I get

foo {
  <point>

And then if I continue with bar RET }

I get

foo {
  bar
}<point>

[hvesalai]

I.e. I don't want any of the automatic insertion of parens.

[hvesalai]

It seems to me smartparens can't do it so I'd say we have to revert to the code as it was before. I'd like to understand the performance problem though. I can't see any reason why it would affect performance since it is a very small piece of code (when compared to, for example, smartparens).

[fommil]

OK, let's add the function back in but please don't add it to the post-self-insert-hook... that seems like a personal preference and adding anything in there is a known performance problem as it gets called on every keystroke.

[fommil]

btw, isn't this a sort of general function that other modes could benefit from? if you replace scala-indent:indent-line with indent-line doesn't it work everywhere?

[hvesalai]

I still don't see the problem with post-self-insert-hook. If you are happy with smartparens having its self insert hook (it is really complex), then I don't see why scala-mode could have it's two-liner there. Yes, it is a general function, but I would like to see scala-mode working with "sane defaults" out of the box, without having to install tens of other modes to support basic code editing.

[fommil]

if that's what you want. I can always add a hook to remove it from my config. I already do this for many post-self-insert-hook functions.

[fommil]

incidentally, what does this functionality achieve that electric indent doesn't do? That's part of stock emacs https://www.emacswiki.org/emacs/AutoIndentation

I think most people would be surprised that they get some sort of electic indent without having enabled it.

[hvesalai]

The function we had only indented on } in the very specific case that it was inserted onto an empty line. As such it is much less interfering than any of various auto paren modes or electric.