consecutive narrowing

[blaenk]

I love this package. Would it be possible to make it so that if I choose a prefix command from helm-descbinds, it'll close helm-descbinds and leave emacs in a state where it's waiting for the next key?

For example the scenario I'm imagining is:

1. Press C-x
2. Wait a bit, emacs shows "C-x -" at the bottom
3. Press C-h to trigger helm-descbinds, choose C-x 4 which is a prefix command
4. Descbinds should close and I should be back where I was
5. After a bit, emacs shows "C-x 4 -" at the bottom
6. I can press C-h again to trigger descbinds if necessary, now narrowing from the point of "C-x 4"

I saw which-key which looks pretty cool, and lets you do this (same with guide-key, which it's an improvement on AFAIK). I much prefer helm-descbinds though, by far, but this is the functionality I sorely miss.

[michael-heerdegen]

That sounds useful.

But is it really helpful to be able to leave the Helm session in between? Because I think it would be easier to implement that when you select a prefix key, you get the narrowed session instantly without quitting helm. Doing otherwise might also be confusing.

[blaenk]

Yeah I suppose that would be perfectly fine.

[michael-heerdegen]

Jorge Israel Peña notifications@github.com writes:

Yeah I suppose that would be perfectly fine.

Ok, good.

Currently, the behavior for the main action on a prefix key is not
useful at all (in my tests, just nothing happened).

So I'll try to just change that. Should not be hard, AFAICT, all we
need is already there.

[michael-heerdegen]

I did exactly that now, via action-transformer attribute, and bumped the version to 1.10.

Please give it some testing.

[blaenk]

Thanks.

One thing I've noticed is that when I go say 2 levels deep into a prefix, descbinds disappears and only reappears once I press a key, though sometimes (not always) it comes back on its own after a delay:

I believe this sequence would replicate it:

1. C-x C-h
2. pattern: 8 prefix
3. RET
4. pattern: ' prefix
5. RET
6. helm-descbinds disappears. it comes back if I press a key, but the key is also output to the file buffer
7. pattern: A
8. RET
9. properly inserts the character

I'll let you know if I find anything else.

[michael-heerdegen]

I can't reproduce - but you may have hit a problem I faced in other situations.

In 6., does Emacs display any text in the minibuffer ("RET -" maybe)?

You could also try to (if you know how to load lisp code) substitute the run-with-idle-timer in my commit with just run-with-timer.

[blaenk]

Nope the minibuffer doesn't show anything after step 5 actually, that's weird.

Ok I opened the helm-descbinds.el, made the change to run-with-timer, did M-x eval-sexp, and re-ran the steps I outlined. Indeed that seems to have resolved it; it seems to work perfectly. Any chance we can go with that function instead of the idle one? Or does this better narrow where my problem lies?

The character was properly inserted, but now I do see a "RET -" in the minibuffer once the character was inserted (at step 9 above) and descbinds exited. It seems like the final RET both did the descbinds action as well as passed RET over to emacs separately.

[blaenk]

I'm on 24.5.1 if that matters.

[michael-heerdegen]

I see no problem with run-with-timer, seemingly your Emacs is not always idle after leaving Helm and that caused the problem. I'll just make the change. I think it's a separate issue, different from the "RET - " thing.

I see the "RET -" thing here and there all the time when exiting Helm. I guess it's a subtle thing that has to with our usage of minor-mode-overriding-map-alist or something like that. discard-input doesn't seem to be able to break it.

I this case here, it's only a visual effect (the "second" RET does not have any effect), but in some situations it had (e.g. I hit "a" then and got "RET-a is undefined" or something like that.)

[michael-heerdegen]

Ok, please check if it's ok now.

[blaenk]

I think you forgot to bump the version? My emacs doesn't detect any updates for the package. Anyways, that's the change I used myself when you told me to try it, so I think it's safe to say it'll work again :)

[blaenk]

My mistake, I was on melpa anyways, so version doesn't matter. I noticed that it sees the update now. I tested it and it does work :) I'll let you know if anything comes up, but I suppose it's fine to close this now if you want!

[michael-heerdegen]

Closing. Please reopen if you should find a problem.

Thanks for the idea!