Question about completion mode

[rpgoldman]

I wonder if you can help me debug a problem I am having. sly completions are not working as expected (I think).

Here's what I do:

1. M-TAB to request symbol creation
2. I get a minibuffer (*sly-completions*) with a list of completions and the prompt to use <right> and <left> to navigate completions.
3. I press <right>
4. The cursor moves right in the original buffer and the minibuffer manages.

My best guess is that this prompt reflects an expectation that the cursor will be in the *sly-completions* minibuffer, but for me it is, instead, left in the source buffer. Is that correct?

If so, I have no idea what I could be doing to break the cursor placement. But before I go crazy trying to figure this out, I wanted to make sure that I am right in my guess above. Thanks!

[joaotavora]

There's osmething odd with that prompt. My prompt doesn't suggest those keys:

Can you reproduce this exact situation from Emacs -Q?

[rpgoldman]

Sorry to be so long getting to this. I did reproduce this on an un-customized emacs. I still see this:

The one confound is that I'm using Aquamacs, and maybe that is interfering. I will check further.

Meanwhile, when you are in this situation, is the cursor in the minibuffer, or does it stay in "thingy.lisp"? It's hard for me to tell from your image, so I am left wondering -- does completion override the bindings for C-n and C-p in "thingy.lisp", or does the cursor get left in the minibuffer so you use C-n and C-p there to navigate and select?

For me, navigating the set of completions works, but annoyingly I have to do C-x b to get to the completion buffer before the keystrokes work (which I often forget to do). It seems like this is just my problem, and given the results on emacs w/o customizations, it is most likely an Aquamacs thing.

P.S. the contents of the *scratch* buffer in my figure should be the only customizations in that emacs session.

[joaotavora]

The one confound is that I'm using Aquamacs, and maybe that is interfering. I will check further.

I'm very inclined to believe that. Aquamacs is extremely heavily customized last time I used it. I didn't even know it was still a thing, and am suprised SLY works on it.

[rpgoldman]

@joaotavora Sorry to bother you again, but would you just let me know where the cursor is positioned after the completion key for you? Is it in the completion minibuffer (as I suspect), or in the buffer where the completion command was typed?

If it is in the original buffer (not the minibuffer), do C-n and C-p move up and down in the minibuffer or are they interpreted in the source buffer?

[joaotavora]

When SLY's built in completion UI kicks in, like in the screenshot above, the active window is the source-editing window (which can be the REPL). Also in this active window, the point (which is the correct term to mean "cursor" in Emacs parlance) is unchanged (i.e. it is where you last left it).

When one presses C-n and C-p the temporary and inactive *sly-completions* window which is showing below the source window is updated to highlight the completion candidate that will be inserted "at point" if you press RET (Enter) or TAB.

[rpgoldman]

Thank you, @joaotavora ! I will look and see why that is not happening for me. Seems like you are saying that the C-n and C-p characters are read in the main window, but move the subsidiary window.

That's a very helpful clue and I will try to figure out what is breaking this!

[rpgoldman]

I looked further and now I see that the next and previous completion keys seem to work, it's just that for some reason Aquamacs (I think it's Aquamacs!) is rewriting the keystrokes in here:

"Use \\[sly-next-completion] and \\[sly-prev-completion] to navigate completions."
          " \\[sly-choose-completion] or [mouse-1] selects a completion."
          "\n\nAnnotation flags: (b)oundp (f)boundp (g)eneric-function (c)lass (m)acro (s)pecial-operator\n\n"

to be wrong. I wouldn't know how to do this even if I wanted to!

[rpgoldman]

For anyone else who encounters this problem, I offer a proposed fix from Win Treese, the Aquamacs maintainer:

1. Aquamacs redefines C-n and C-p to be aquamacs-next-line and aquamacs-previous-line, respectively. I think this was originally done to make things more intuitive with certain configurations different than default Emacs, but I can’t reconstruct the details at the moment.

2. In defining the keymap for the Sly completions buffer, Sly remaps certain keys to next/prev. This makes sense, but doesn’t have the desired effect in Aquamacs.

To clean up the bindings and give the right hint about the keys to use, I think you can use the following code after Sly is loaded:

(define-key sly--completion-display-mode-map [right] nil)
(define-key sly--completion-display-mode-map [left] nil)
(define-key sly--completion-display-mode-map [up] nil)
(define-key sly--completion-display-mode-map [down] nil)
(define-key sly--completion-display-mode-map [remap previous-line] nil)
(define-key sly--completion-display-mode-map [remap next-line] nil)
(define-key sly--completion-display-mode-map [backtab] nil)
(define-key sly--completion-display-mode-map [remap aquamacs-next-line] #'sly-next-completion)
(define-key sly--completion-display-mode-map [remap aquamacs-previous-line] #'sly-prev-completion)

I will update this issue after testing and close it if this provides the solution.