Giters

oantolin / orderless

Emacs completion style that matches multiple regexps in any order

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Completion variables recommended configuration

Andre0991 opened this issue · comments

commented

Hi.

I found this comment in Eglot issue tracker.

It seems that using (setq completion-category-overrides) instead of add-to-list was an issue, since that replaced some Eglot configuration.

I haven’t faced this issue, but I’m wondering if it would make sense for orderless to recommend using add-to-list in its README.

WDYT?

commented

completion-category-overrides is for user-specific overrides. Eglot should probably rather use completion-category-defaults? However this will still leave us with the problem. Orderless recommends to reset these variables to ensure that you get a consistent Orderless UI everywhere. Unfortunately completion-category-defaults comes with many presets which are not really necessary and which lead to an inconsistent UI. On Emacs 28:

(defvar completion-category-defaults
  '((buffer (styles . (basic substring)))
    (unicode-name (styles . (basic substring)))
    ;; A new style that combines substring and pcm might be better,
    ;; e.g. one that does not anchor to bos.
    (project-file (styles . (substring)))
    (xref-location (styles . (substring)))
    (info-menu (styles . (basic substring)))
    (symbol-help (styles . (basic shorthand substring))))

We could recommend instead in the README that the user should remove the undesired entries from the list, e.g.,

(setq completion-category-defaults
      (assq-delete-all 'buffer completion-category-defaults))
(setq completion-category-defaults
      (assq-delete-all 'unicode-name completion-category-defaults))
...

At least a warning in the readme would be appropriate.

commented

The same issue applies to the Corfu and Vertico READMEs. Also some packages modify the completion-category-defaults variable afterwards, e.g., message.el adds (email (styles substring)) which is unnecessary given Orderless. My configuration even contains this:

(with-eval-after-load 'message
   (setq completion-category-defaults
           (delete '(email (styles substring))
                   completion-category-defaults)))

The impact of all these overrides is quite minimal (positively and negatively), but at least I've read reports before where people were confused that Orderless didn't set in properly due to the unexpected overrides.

commented

Indeed, having a consistent experience seems like the best default recommended configuration. Maybe the README could warn users about the caveats. But maybe not. Thanks for the explanation!