Mimic `flyspell-autocorrect-previous-word`

[joostkremers]

One thing I really like about flyspell, which I've been using so far for spell checking, is the command flyspell-auto-correct-previous-word. Are you planning to add something similar to jinx?

In case you're not familiar with it: flyspell-auto-correct-previous-word corrects the last misspelled word before point. (Or, as the doc string says "the first misspelled word that occurs before point". 😄 Scanning backward from point, it's the first misspelled word, but it's likely the last word that you misspelled.)

This works very smoothly and unobtrusively. You don't need to move point back to the word you want to correct, and you don't need to select a correction, the most likely correction is applied automatically. The list of possible corrections is shown in the minibuffer, and you can repeat the command to select the next one if the first one isn't what you were looking for.

In most of the cases, though, the first option is the right one, so usually when I misspell a word, all I need to do is hit C-; and the error is corrected. That even works when I've already typed the next word (which happens quite a bit).

[minad]

This is quite a different UI than what I have now, also with the cycling. We could do something similar, where on the second key press the list is shown or even only cycle only between the possible suggestions. There is also a long UI discussion ongoing in #9. For now I will keep the current UI as is, until it is clear how the best UI looks like. To be clear, the main advantage of Jinx is how it works regarding highlighting of spelling mistakes - something where I was never satisfied with Flyspell. The correction UI is a whole different story.

[joostkremers]

Yeah, it's a very different UI, that's true. I wouldn't suggest that this should be the only UI, though. I do still use ispell-word, so to me, it's really about being able to use both.

[minad]

On 3/31/23 13:27, Joost Kremers wrote: Yeah, it's a very different UI, that's true. I wouldn't suggest that this should be the only UI, though.

I am not really interested in maintaining multiple UIs. Maybe two but not more (completion and transient for example).

I do still use |ispell-word|, so to me, it's really about being able to use both.

Note that Jinx replaces ispell/flyspell completely. They can coexist next to each other, but combining them is not a goal. Why you would want to use Ispell together with Jinx? If you miss something in Jinx, it can be added here.

[gusbrs]

@joostkremers minad pointed you to #9 , but, in particular, you may be interested in #9 (comment), which implements something akin to flyspell-correct (the package), which also has a "backward looking" behavior by default.

With that, assuming you use Vertico, the following could get you something similar (not equal) to flyspell-auto-correct-previous-word (at the expense of the easy skipping, though).

(add-hook 'minibuffer-setup-hook #'my/jinx-correct-previous-cycle-setup)
(defun my/jinx-correct-previous-cycle-setup ()
  (when (eq this-command 'gb/jinx-correct-previous)
    (setq-local vertico-cycle t)
    (keymap-local-set "C-;" #'vertico-next)))

Of course, at this point this is just experimenting and playing with ideas.

[agreselin]

The only correction command I use is one that scans backward. I think it makes sense: you correct what you just typed. Where it leaves the point or whether it asks which candidate to use should be consistent with the rest of the interface, but a command that scans backward would be a very useful addition in my opinion.

PS: What do you think about having jinx-correct move backward if called with negative argument? Alternatively you could expose a function that, like ispell-word, returns different values based on the user's action. Then one could build a wrapper around it to make his or her own correction command.

Also, after trying out jinx-correct some more I wish there was a quicker way to insert a candidate. I like the way good old ispell-word does it, with one-key shortcuts to candidates. This could be used by jinx-correct as well. (Did I miss something or the only way to insert a candidate is to type it (with completion)?)

[joostkremers]

With that, assuming you use Vertico, the following could get you something similar (not equal) to flyspell-auto-correct-previous-word (at the expense of the easy skipping, though).

@gusbrs thanks for pointing out #9 , I haven't had time to try out the code you point to, but the fact that you mention Vertico (which I do use, BTW), suggests that I still need to do more than just hit C-; once or twice to correct a misspelled word.

flyspell-auto-correct-previous-word does show the list of possible candidates in the minibuffer, but it does so horizontally, so the minibuffer doesn't expand. I usually don't even notice the candidates are there, my eyes stay on the misspelled word and when I see it's corrected, I continue typing.

[gusbrs]

@gusbrs thanks for pointing out #9 , I haven't had time to try out the code you point to, but the fact that you mention Vertico (which I do use, BTW), suggests that I still need to do more than just hit C-; once or twice to correct a misspelled word.

@joostkremers Yes, as I said, similar, but not equal. Technically, you need one C-; more at the start to call the command, and one RET more at the end to commit your selection. Other smaller differences, presumably positive ones, are that you get saving and accepting the word among the options.

And the thing is only tied to Vertico in this cycling behavior I shared here, otherwise completing-read, since that's what jinx uses. But, since you do use it, you can get a similar visual with:

(add-to-list 'vertico-multiform-categories '(jinx flat))

But, of course, still not the same thing. And, as mentioned, this is just experimentation anyway.

flyspell-auto-correct-previous-word does show the list of possible candidates in the minibuffer, but it does so horizontally, so the minibuffer doesn't expand. I usually don't even notice the candidates are there, my eyes stay on the misspelled word and when I see it's corrected, I continue typing.

I personally never got used to it, but I know it well, so I understand what you are saying. I'd bet @minad does too.

[joostkremers]

I am not really interested in maintaining multiple UIs. Maybe two but not more (completion and transient for example).

Your call, obviously. It's just that both of these (completion and transient) would force me to mentally context-switch to the spell-check UI, because I need to see what exactly I need to type to correct the word. flyspell-auto-correct-previous-word does not require such a context-switch. I just hit C-; to correct the word and I can continue writing.

Note that Jinx replaces ispell/flyspell completely. They can coexist next to each other, but combining them is not a goal. Why you would want to use Ispell together with Jinx?

I didn't mean using flyspell and jink together, I meant using both ispell-word and flyspell-auto-correct-previous-word.

If you miss something in Jinx, it can be added here.

Yeah, I miss flyspell-auto-correct-previous-word and was wondering if you'd consider adding such a feature. 😉

[gusbrs]

It's just that both of these (completion and transient) would force me to mentally context-switch to the spell-check UI, because I need to see what exactly I need to type to correct the word. flyspell-auto-correct-previous-word does not require such a context-switch. I just hit C-; to correct the word and I can continue writing.

@minad I'm not sure where this would belong, but "Live preview" for jinx-correct would address this, at least partially.

[joostkremers]

@joostkremers Yes, as I said, similar, but not equal. Technically, you need one C-; more at the start to call the command, and one RET more at the end to commit your selection.

Yeah, too many key presses. 😄 Also, I want the change to be done in the buffer right away, without me having to commit the selection. When I use C-;, my eyes are close to the incorrect word, but possibly far away from the minibuffer.

Other smaller differences, presumably positive ones, are that you get saving and accepting the word among the options.

I don't care for those in this particular UI. I type fairly fast, but fat-finger a lot. flyspell-auto-correct-previous-word allows me to fix such mistakes quickly.

Note that for me, flyspell-auto-correct-previous-word is not an alternative to ispell-word. It's an alternative to moving point backward and correcting the word manually.

[minad]

On 4/3/23 13:36, Gustavo Barros wrote: @minad <https://github.com/minad> I'm not sure where this would belong, but "Live preview" for |jinx-correct| would address this, at least partially.

Yes, I know. But is unlikely to be added since it requires interaction with the completion UI ala Consult (see consult--completion-candidate-hook).

[paaguti]

And what about providing a list of words that could be fed to a completion-at-point function that can then be used in cape, company and friends?

I've been able to setup my language jungle with jinx, I've understood how to control the languages libenchant uses, I've seen that my current enchant-2 config still works and if I had a nicer UI, I'd switch from flyspell to jinx.

[minad]

@paaguti

And what about providing a list of words that could be fed to a completion-at-point function that can then be used in cape, company and friends?

You can use cape-dict if you want to complete words using a dictionary. Besides that, completion-at-point is not a good fit for a spell-checking UI, since it serves an entirely different purpose.

[paaguti]

Now that you say it... You are completely right... And I'm getting used to the jinx UI /PA

…

On Mon, 17 Apr 2023 at 23:24, Daniel Mendler ***@***.***> wrote: @paaguti <https://github.com/paaguti> And what about providing a list of words that could be fed to a completion-at-point function that can then be used in cape, company and friends? You can use cape-dict if you want to complete words using a dictionary. Besides that, completion-at-point is not a good fit for a spell-checking UI, since it serves an entirely different purpose. — Reply to this email directly, view it on GitHub <#30 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAZKU4DLWW5ZSYNLI22SYHTXBWYKNANCNFSM6AAAAAAWOOWEO4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

-- Fragen sind nicht da, um beantwortet zu werden, Fragen sind da um gestellt zu werden Georg Kreisler Headaches with a Juju log: unit-basic-16: 09:17:36 WARNING juju.worker.uniter.operation we should run a leader-deposed hook here, but we can't yet