bug#39600: [PATCH] Fix handling of non-exclusive non-prefix completion functions

bug#39600: [PATCH] Fix handling of non-exclusive non-prefix completion functions

[Amai Kinono]

[-- Attachment #1.1: Type: text/plain, Size: 3025 bytes --]

# What does this fix?

Currently, with non-exclusive completion functions, Emacs will do
`try-completion` on the current text, and decide whether to try next
completion function based on that. This makes completion functions that
can do non-prefix completions fails when the current text only occurs in
the middle of the candidates. This is a problem I found in a FIXME in
the code.

# How does this work?

I use `completion-all-completions` instead. As far as I know, this
respects the `completion-styles`.

Here's a simple test. Eval this:

```
(require 'thingatpt)
(defun my-completion-at-point ()
  (let* ((symbol (thing-at-point 'symbol))
         (bounds (bounds-of-thing-at-point 'symbol)))
    (list (car bounds) (cdr bounds)
          '("gnuemacs" "xemacs" "uemacs")
          :exclusive 'no)))
(setq completion-at-point-functions
      '(my-completion-at-point elisp-completion-at-point))
(setq completion-styles '(substring))
```

Now type "emacs" and press `C-M-i`, the candidates defined in
`my-completion-at-point` will show up correctly. Now change
`completion-styles` to `(basic)`, and complete "emacs" again, it
fallbacks to `elisp-completion-at-point`.

# The patch

See the attachment. Since I don't know much about email, I'm not sure if
gmail's attachment is permanent, so I'll also paste the diff and commit
log here:

## Diff

--- a/lisp/minibuffer.el
+++ b/lisp/minibuffer.el
@@ -2244,18 +2244,11 @@ completion--capf-wrapper
           (unless (member fun completion--capf-safe-funs)
             (push fun completion--capf-safe-funs))
           (and (eq 'no (plist-get (nthcdr 3 res) :exclusive))
-               ;; FIXME: Here we'd need to decide whether there are
-               ;; valid completions against the current text.  But this
depends
-               ;; on the actual completion UI (e.g. with the default
completion
-               ;; it depends on completion-style) ;-(
-               ;; We approximate this result by checking whether prefix
-               ;; completion might work, which means that non-prefix
completion
-               ;; will not work (or not right) for completion functions
that
-               ;; are non-exclusive.
-               (null (try-completion (buffer-substring-no-properties
-                                      (car res) (point))
-                                     (nth 2 res)
-                                     (plist-get (nthcdr 3 res)
:predicate)))
+               (null (completion-all-completions
+                      (buffer-substring-no-properties (car res) (point))
+                      (nth 2 res)
+                      (plist-get (nthcdr 3 res) :predicate)
+                      (- (point) (car res))))
                (setq res nil)))
          ((not (or (listp res) (functionp res)))
           (unless (member fun completion--capf-misbehave-funs)

## Commit Log

Fix handling for non-exclusive non-prefix completion functions

* lisp/minibuffer.el (completion--capf-wrapper):
use completion-all-completions to do the test.

[-- Attachment #1.2: Type: text/html, Size: 3604 bytes --]

[-- Attachment #2: 0001-Fix-handling-for-non-exclusive-non-prefix-completion.patch --]
[-- Type: text/x-patch, Size: 2019 bytes --]

From 4dd4d58d56c111eb1ba482119c07134b85591747 Mon Sep 17 00:00:00 2001
From: AmaiKinono <amaikinono@gmail.com>
Date: Fri, 14 Feb 2020 14:52:55 +0800
Subject: [PATCH] Fix handling for non-exclusive non-prefix completion
 functions

* lisp/minibuffer.el (completion--capf-wrapper):
use completion-all-completions to do the test.
---
 lisp/minibuffer.el | 17 +++++------------
 1 file changed, 5 insertions(+), 12 deletions(-)

diff --git a/lisp/minibuffer.el b/lisp/minibuffer.el
index 49daabc..f811a5f 100644
--- a/lisp/minibuffer.el
+++ b/lisp/minibuffer.el
@@ -2244,18 +2244,11 @@ completion--capf-wrapper
           (unless (member fun completion--capf-safe-funs)
             (push fun completion--capf-safe-funs))
           (and (eq 'no (plist-get (nthcdr 3 res) :exclusive))
-               ;; FIXME: Here we'd need to decide whether there are
-               ;; valid completions against the current text.  But this depends
-               ;; on the actual completion UI (e.g. with the default completion
-               ;; it depends on completion-style) ;-(
-               ;; We approximate this result by checking whether prefix
-               ;; completion might work, which means that non-prefix completion
-               ;; will not work (or not right) for completion functions that
-               ;; are non-exclusive.
-               (null (try-completion (buffer-substring-no-properties
-                                      (car res) (point))
-                                     (nth 2 res)
-                                     (plist-get (nthcdr 3 res) :predicate)))
+               (null (completion-all-completions
+                      (buffer-substring-no-properties (car res) (point))
+                      (nth 2 res)
+                      (plist-get (nthcdr 3 res) :predicate)
+                      (- (point) (car res))))
                (setq res nil)))
          ((not (or (listp res) (functionp res)))
           (unless (member fun completion--capf-misbehave-funs)
-- 
2.25.0

bug#39600: [PATCH] Fix handling of non-exclusive non-prefix completion functions

[Lars Ingebrigtsen]

Amai Kinono <amaikinono@gmail.com> writes:

> Currently, with non-exclusive completion functions, Emacs will do
> `try-completion` on the current text, and decide whether to try next
> completion function based on that. This makes completion functions that
> can do non-prefix completions fails when the current text only occurs in
> the middle of the candidates. This is a problem I found in a FIXME in
> the code.
>
> # How does this work?
>
> I use `completion-all-completions` instead. As far as I know, this
> respects the `completion-styles`.

[...]

Stefan, how does this patch look to you?

> -               ;; FIXME: Here we'd need to decide whether there are
> -               ;; valid completions against the current text.  But this depends
> -               ;; on the actual completion UI (e.g. with the default completion
> -               ;; it depends on completion-style) ;-(
> -               ;; We approximate this result by checking whether prefix
> -               ;; completion might work, which means that non-prefix completion
> -               ;; will not work (or not right) for completion functions that
> -               ;; are non-exclusive.
> -               (null (try-completion (buffer-substring-no-properties
> -                                      (car res) (point))
> -                                     (nth 2 res)
> -                                     (plist-get (nthcdr 3 res) :predicate)))
> +               (null (completion-all-completions
> +                      (buffer-substring-no-properties (car res) (point))
> +                      (nth 2 res)
> +                      (plist-get (nthcdr 3 res) :predicate)
> +                      (- (point) (car res))))


-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#39600: [PATCH] Fix handling of non-exclusive non-prefix completion functions

[Stefan Monnier]

>> Currently, with non-exclusive completion functions, Emacs will do
>> `try-completion` on the current text, and decide whether to try next
>> completion function based on that. This makes completion functions that
>> can do non-prefix completions fails when the current text only occurs in
>> the middle of the candidates. This is a problem I found in a FIXME in
>> the code.
>>
>> # How does this work?
>>
>> I use `completion-all-completions` instead. As far as I know, this
>> respects the `completion-styles`.
> [...]
> Stefan, how does this patch look to you?

I think it replaces one set of problems with another.

Within the completion-at-point-function we shouldn't call anything like
`try-completion`: we should only decide which completion table to use
but we shouldn't look inside those completion tables.

So a better solution is to build a new completion table which combines
(e.g. with `completion-table-in-turn`) the current one with that of "the
rest".  [ The tricky (but still doable) case being when when the current
completion table and that of the rest don't use the same BEG and END.  ]


        Stefan