bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eshel Yaron via Bug reports for GNU Emacs, the Swiss army knife of text editors]

Hi,

1. emacs -Q
2. Insert "(cond (current-p
3. Hit M-TAB to complete

We're completing a symbol in the condition of a cond clause, so it is
expected to be completed as a variable, to current-prefix-argument.
However, Emacs mistakenly thinks that we're completing a function name,
and completes to current-bidi-paragraph-direction.


Thanks,

Eshel

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eli Zaretskii]

> Date: Wed, 18 Sep 2024 14:36:18 +0200
> From:  Eshel Yaron via "Bug reports for GNU Emacs,
>  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
> 
> 1. emacs -Q
> 2. Insert "(cond (current-p
> 3. Hit M-TAB to complete
> 
> We're completing a symbol in the condition of a cond clause, so it is
> expected to be completed as a variable, to current-prefix-argument.
> However, Emacs mistakenly thinks that we're completing a function name,
> and completes to current-bidi-paragraph-direction.

Thanks.  Does the patch below give good results?

diff --git a/lisp/progmodes/elisp-mode.el b/lisp/progmodes/elisp-mode.el
index 9bf6f92..2471296 100644
--- a/lisp/progmodes/elisp-mode.el
+++ b/lisp/progmodes/elisp-mode.el
@@ -784,13 +784,14 @@ elisp-completion-at-point
                         (list t (elisp--completion-local-symbols)
                               :predicate (lambda (sym)
                                            (get sym 'error-conditions))))
-                       ((and (or ?\( 'let 'let*)
+                       ((and (or ?\( 'let 'let* 'cond 'cond* 'bind*)
                              (guard (save-excursion
                                       (goto-char (1- beg))
                                       (when (eq parent ?\()
                                         (up-list -1))
                                       (forward-symbol -1)
-                                      (looking-at "\\_<let\\*?\\_>"))))
+                                      (looking-at
+                                       "\\_<\\(let\\|cond\\|bind\\*\\)\\*?\\_>"))))
                         (list t (elisp--completion-local-symbols)
                               :predicate #'elisp--shorthand-aware-boundp
                               :company-kind (lambda (_) 'variable)

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eshel Yaron via Bug reports for GNU Emacs, the Swiss army knife of text editors]

Eli Zaretskii <eliz@gnu.org> writes:

>> Date: Wed, 18 Sep 2024 14:36:18 +0200
>> From:  Eshel Yaron via "Bug reports for GNU Emacs,
>>  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
>> 
>> 1. emacs -Q
>> 2. Insert "(cond (current-p
>> 3. Hit M-TAB to complete
>> 
>> We're completing a symbol in the condition of a cond clause, so it is
>> expected to be completed as a variable, to current-prefix-argument.
>> However, Emacs mistakenly thinks that we're completing a function name,
>> and completes to current-bidi-paragraph-direction.
>
> Thanks.  Does the patch below give good results?

It helps, but also seems to cause a regression.  With this change I get:

1. emacs -Q
2. Insert "(cond ((current-p"
                  ^note the additional parenthesis
3. Hit M-TAB to complete

This shows an opposite bug: Emacs should complete function names (and
without the change, it does), but with the change we now get variable
name completion.


Thanks,

Eshel

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eli Zaretskii]

> From: Eshel Yaron <me@eshelyaron.com>
> Cc: 73330@debbugs.gnu.org
> Date: Thu, 19 Sep 2024 14:25:13 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> Date: Wed, 18 Sep 2024 14:36:18 +0200
> >> From:  Eshel Yaron via "Bug reports for GNU Emacs,
> >>  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
> >> 
> >> 1. emacs -Q
> >> 2. Insert "(cond (current-p
> >> 3. Hit M-TAB to complete
> >> 
> >> We're completing a symbol in the condition of a cond clause, so it is
> >> expected to be completed as a variable, to current-prefix-argument.
> >> However, Emacs mistakenly thinks that we're completing a function name,
> >> and completes to current-bidi-paragraph-direction.
> >
> > Thanks.  Does the patch below give good results?
> 
> It helps

Does "helps" mean the original scenario is fixed? or are there any
leftovers in that scenario?

> but also seems to cause a regression.  With this change I get:
> 
> 1. emacs -Q
> 2. Insert "(cond ((current-p"
>                   ^note the additional parenthesis
> 3. Hit M-TAB to complete
> 
> This shows an opposite bug: Emacs should complete function names (and
> without the change, it does), but with the change we now get variable
> name completion.

How about the patch below?

diff --git a/lisp/progmodes/elisp-mode.el b/lisp/progmodes/elisp-mode.el
index 9bf6f92..63bd685 100644
--- a/lisp/progmodes/elisp-mode.el
+++ b/lisp/progmodes/elisp-mode.el
@@ -784,13 +784,18 @@ elisp-completion-at-point
                         (list t (elisp--completion-local-symbols)
                               :predicate (lambda (sym)
                                            (get sym 'error-conditions))))
-                       ((and (or ?\( 'let 'let*)
+                       ((and (or ?\( 'let 'let* 'cond 'cond* 'bind*)
                              (guard (save-excursion
                                       (goto-char (1- beg))
                                       (when (eq parent ?\()
                                         (up-list -1))
                                       (forward-symbol -1)
-                                      (looking-at "\\_<let\\*?\\_>"))))
+                                      (or
+                                       (looking-at
+                                        "\\_<\\(let\\*?\\|bind\\*\\)\\_>")
+                                       (and (not (eq parent ?\())
+                                            (looking-at
+                                             "\\_<cond\\*?\\_>"))))))
                         (list t (elisp--completion-local-symbols)
                               :predicate #'elisp--shorthand-aware-boundp
                               :company-kind (lambda (_) 'variable)

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eshel Yaron via Bug reports for GNU Emacs, the Swiss army knife of text editors]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Eshel Yaron <me@eshelyaron.com>
>> Cc: 73330@debbugs.gnu.org
>> Date: Thu, 19 Sep 2024 14:25:13 +0200
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> >> Date: Wed, 18 Sep 2024 14:36:18 +0200
>> >> From:  Eshel Yaron via "Bug reports for GNU Emacs,
>> >>  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
>> >> 
>> >> 1. emacs -Q
>> >> 2. Insert "(cond (current-p
>> >> 3. Hit M-TAB to complete
>> >> 
>> >> We're completing a symbol in the condition of a cond clause, so it is
>> >> expected to be completed as a variable, to current-prefix-argument.
>> >> However, Emacs mistakenly thinks that we're completing a function name,
>> >> and completes to current-bidi-paragraph-direction.
>> >
>> > Thanks.  Does the patch below give good results?
>> 
>> It helps
>
> Does "helps" mean the original scenario is fixed? or are there any
> leftovers in that scenario?

Yes, it fixed the original scenario.

>> but also seems to cause a regression.  With this change I get:
>> 
>> 1. emacs -Q
>> 2. Insert "(cond ((current-p"
>>                   ^note the additional parenthesis
>> 3. Hit M-TAB to complete
>> 
>> This shows an opposite bug: Emacs should complete function names (and
>> without the change, it does), but with the change we now get variable
>> name completion.
>
> How about the patch below?

That works well in both the original scenario and the modified one.


Thank!

Eshel

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eshel Yaron via Bug reports for GNU Emacs, the Swiss army knife of text editors]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Eshel Yaron <me@eshelyaron.com>
>> Cc: 73330@debbugs.gnu.org
>> Date: Thu, 19 Sep 2024 14:25:13 +0200
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> >> Date: Wed, 18 Sep 2024 14:36:18 +0200
>> >> From:  Eshel Yaron via "Bug reports for GNU Emacs,
>> >>  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
>> >> 
>> >> 1. emacs -Q
>> >> 2. Insert "(cond (current-p
>> >> 3. Hit M-TAB to complete
>> >> 
>> >> We're completing a symbol in the condition of a cond clause, so it is
>> >> expected to be completed as a variable, to current-prefix-argument.
>> >> However, Emacs mistakenly thinks that we're completing a function name,
>> >> and completes to current-bidi-paragraph-direction.
>> >
>> > Thanks.  Does the patch below give good results?
>> 
>> It helps
>
> Does "helps" mean the original scenario is fixed? or are there any
> leftovers in that scenario?

Yes, it fixed the original scenario.

>> but also seems to cause a regression.  With this change I get:
>> 
>> 1. emacs -Q
>> 2. Insert "(cond ((current-p"
>>                   ^note the additional parenthesis
>> 3. Hit M-TAB to complete
>> 
>> This shows an opposite bug: Emacs should complete function names (and
>> without the change, it does), but with the change we now get variable
>> name completion.
>
> How about the patch below?

That works well in both the original scenario and the modified one.


Thanks!

Eshel

bug#73330: 31.0.50; Incorrect completions for 'cond' clauses

[Eli Zaretskii]

> From: Eshel Yaron <me@eshelyaron.com>
> Cc: 73330@debbugs.gnu.org
> Date: Thu, 19 Sep 2024 16:33:11 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Does "helps" mean the original scenario is fixed? or are there any
> > leftovers in that scenario?
> 
> Yes, it fixed the original scenario.
> 
> >> but also seems to cause a regression.  With this change I get:
> >> 
> >> 1. emacs -Q
> >> 2. Insert "(cond ((current-p"
> >>                   ^note the additional parenthesis
> >> 3. Hit M-TAB to complete
> >> 
> >> This shows an opposite bug: Emacs should complete function names (and
> >> without the change, it does), but with the change we now get variable
> >> name completion.
> >
> > How about the patch below?
> 
> That works well in both the original scenario and the modified one.

Thanks, installed on master, and closing the bug.