Readme Intro about consult

[minad]

Turns out, the implementation is also much better than before! :)

[minad]

Good idea. Note that I haven't done sorting via the completion metadata yet. I have to find out how the metadata callback works. If you have time it would be great if you go a bit over the stuff I added so far. There are a few places marked with TODOs, which should be resolved but I haven't done yet or have to find the right solution first.

[minad]

Added a note in 8d0a860. I will polish the readme more and more when this thing becomes more mature.

[clemera]

The manual describes the callback at (info "(elisp) Programmed Completion"). The action argument is used to query for metadata. The following example shows how you can add some metadata to completions. Emacs comes with the little helper function complete-with-action which you can use to handle the rest of the API so you only need to care about handling the metadata action:

(let* ((alist '(("GNU" . "GNU is not Unix")
                ("Emacs" . "Eight Megabytes And Constantly Swapping")))
       (annotf (lambda (str)
                 (format " (%s)" (cdr (assoc str alist))))))
  (completing-read "Candidates: "
                   (lambda (str pred action)
                     (if (eq action 'metadata)
                         `(metadata
                           (annotation-function . ,annotf)
                           (cycle-sort-function . identity)
                           (display-sort-function . identity))
                       (complete-with-action action alist str pred)))))

[clemera]

I will look through the code on the weekend

[minad]

Thank you for the example. I will play around a bit and see how that works. One major downside of the standard api I observed is that completing-read does not yet preserve properties (only in 28). This complicates everything unnecessarily. And I think we want to support at least Emacs 26.

I will look through the code on the weekend

Looking forward to it!

[clemera]

One major downside of the standard api I observed is that completing-read does not yet preserve properties (only in 28). This complicates everything unnecessarily. And I think we want to support at least Emacs 26.

Given that Emacs will eventually gain support via minibuffer-allow-text-properties and helm and ivy return text with properties we should think about adding this back to selectrum, as well. Most users will use a completion framework and it will work with any of those except icomplete or default completion before Emacs 28. One approach could be to let commands relying on this to bail out if there isn't any framework in use (eq completing-read-function #'completing-read-default) and minibuffer-allow-text-properties isn't defined/set.

[clemera]

Here is a corresponding issue for selectrum.

[minad]

Yes, consult could detect if there is either emacs 28 or a completion framwork and refuse to work otherwise.

Another thing I wonder about are annotations. The Emacs completion system only supports suffix annotations and this is too weak for many things. What do you propose in that case? Simply use the selectrum property keys? For now I introduce "defcustom consult-property-prefix" etc, but seems to be more like a fig leaf...

[clemera]

I just discovered that we could add prefixes using the display property. It works with helm,ivy, default completion. In selectrum the highlighting is currently broken but we should be able to fix that:

(completing-read "Candidates: "
                     `(,(propertize "GNU" 'display "some-prefix: GNU")
                       ,(propertize "Emacs" 'display "otherprefix: Emacs")))

[clemera]

Oh, the other frameworks also get the match highlighting wrong but at least the selection works as usual.

[clemera]

I don't think there is a way to highlight the matches in this case...(I mean highlighting the parts matched by your input, the selection issue can get fixed).

[minad]

Hmm, this seems to be some kind of abuse of the display attribute? I wouldn't try to bend the api too much, the same holds regarding returning propertized text. I only think that's ok since you say that in emacs 28 this is proper api use. Otherwise I would rather fallback to the selectrum api. This is cleaner than bending an api, we don't really own. I think using these prefix selectrum properties is ok, since this degrades gracefully. Things don't look so nice but will work at least.

[clemera]

Yes, the better way would be to propose adding the prefix feature to the metadata in Emacs. As you said there is no problem using the selectrum properties as long this isn't supported.

[clemera]

I will look through the code on the weekend

I don't have time for an extensive review but looks all pretty good to me, I don't have much to add for now. For a quicker way for computing line numbers see here. From a user perspective I'm not a big fan of making the preview behaviour a default and think it would be good to have a single setting to disable/enable it for all but that is just me ;)

[minad]

I don't have time for an extensive review but looks all pretty good to me, I don't have much to add for now.

Thank you for your help! I feel pretty confident right now regarding my elisp skills. Your help and excessively hacking a few days helped a lot ;) There are only very few TODOs in the consult.el. If you have time at some point, it would be great if you just read over those and give your opinion (I also made github issues for some of the TODOs).

For a quicker way for computing line numbers see here.

Yes, I've seen that before. But my quick tests didn't instill confidence. This is not a robust approach. I rather go with some algorithmic changes. E.g. consult--outline-candidates should now be O(n).

From a user perspective I'm not a big fan of making the preview behaviour a default and think it would be good to have a single setting to disable/enable it for all but that is just me ;)

I generally agree that packages should stay out of the way. But in this case I enable everything which I consider in a "working state" and which I would like to use myself. This way the user gets the full power right away. And you have to bind the commands explicitly yourself after all, so my criterion regarding going out of the way still holds :)

[clemera]

I generally agree that packages should stay out of the way. But in this case I enable everything which I consider in a "working state" and which I would like to use myself. This way the user gets the full power right away. And you have to bind the commands explicitly yourself after all, so my criterion regarding going out of the way still holds :)

Sure, it's a matter of taste and I can disable those manually but as I said it would be nice to be able to disable this by one general variable because every time you add a new command users (I ;)) would need to update the config.

One more thing regarding the review, in consult--buffer you let bind selectrum-should-sort-p which binds it for all recursive minibuffer sessions, too. To avoid that you can set the variable locally via minibuffer-setup-hook (see this PR. This relies on the nice behaviour that local variable bindings in the minibuffer are automatically killed when you exit the minibuffer.

[minad]

Sure, it's a matter of taste and I can disable those manually but as I said it would be nice to be able to disable this by one general variable because every time you add a new command users (I ;)) would need to update the config.

Ok, I misread you! Maybe it makes sense to add some override variable? But I am not sure if it is worth the effort. I prefer to have fine grained settings here since I may want to disable some previews. For example the buffer preview can be very annoying if you are not accustomed to it. But the consult-line preview is essential I think in order to allow jumping between search and buffer for recursive editing. So I would want to enable that.

One more thing regarding the review, in consult--buffer you let bind selectrum-should-sort-p which binds it for all recursive minibuffer sessions, too. To avoid that you can set the variable locally via minibuffer-setup-hook (see this PR. This relies on the nice behaviour that local variable bindings in the minibuffer are automatically killed when you exit the minibuffer.

Ok, this is good to know. I can probably use this too for other places where I am using a stack which I push/pop in setup/exit.
It makes sense since the minibuffer is a buffer after all.

[minad]

@clemera You will like this 970a84e

[clemera]

Yes :) Thanks for implementing this!