Confusing doc string of list-or-string at point

Confusing doc string of list-or-string at point

[Eli Zaretskii]

The doc string says:

  Like `thing-at-point-bounds-of-list-at-point', but also
  prefer to find of any enclosing string."

"Prefer to find of" is clearly a typo of sorts, but I don't have a
clear idea what you wanted to say there.  The same problem is with the
NEWS entry.

Thanks.

Re: Confusing doc string of list-or-string at point

[Juri Linkov]

> The doc string says:
>
>   Like `thing-at-point-bounds-of-list-at-point', but also
>   prefer to find of any enclosing string."
>
> "Prefer to find of" is clearly a typo of sorts, but I don't have a
> clear idea what you wanted to say there.  The same problem is with the
> NEWS entry.

I tried to copy the meaning of the arg NO-SYNTAX-CROSSING of 'up-list'
that uses such words in the docstring:

  If NO-SYNTAX-CROSSING is non-nil (as it is interactively), prefer
  to break out of any enclosing string instead of moving to the
  end of a list broken across multiple strings.

The idea is that when this arg is non-nil, and point is inside a string,
then 'up-list' prefers to move out of string, instead of moving out
of parentheses.

Re: Confusing doc string of list-or-string at point

[Eli Zaretskii]

> From: Juri Linkov <juri@linkov.net>
> Cc: emacs-devel@gnu.org
> Date: Thu, 23 Sep 2021 20:11:45 +0300
> 
> >   Like `thing-at-point-bounds-of-list-at-point', but also
> >   prefer to find of any enclosing string."
> >
> > "Prefer to find of" is clearly a typo of sorts, but I don't have a
> > clear idea what you wanted to say there.  The same problem is with the
> > NEWS entry.
> 
> I tried to copy the meaning of the arg NO-SYNTAX-CROSSING of 'up-list'
> that uses such words in the docstring:
> 
>   If NO-SYNTAX-CROSSING is non-nil (as it is interactively), prefer
>   to break out of any enclosing string instead of moving to the
>   end of a list broken across multiple strings.
> 
> The idea is that when this arg is non-nil, and point is inside a string,
> then 'up-list' prefers to move out of string, instead of moving out
> of parentheses.

I understand the documentation of NO-SYNTAX-CROSSING, which talks
about "breaking out of string".  But I still don't understand what you
meant by "prefer to find of string".

If the list is enclosed in a string, does it mean this will return the
string, not the list?  IOW, what will it return in these two cases:

  (foo "bar baz" quux)
  "foo (bar baz) guux"

Re: Confusing doc string of list-or-string at point

[Juri Linkov]

> I understand the documentation of NO-SYNTAX-CROSSING, which talks
> about "breaking out of string".  But I still don't understand what you
> meant by "prefer to find of string".

I meant that if point is inside a string that is inside a list
then it will prefer to find boundaries of the string, not the list.

> If the list is enclosed in a string, does it mean this will return the
> string, not the list?

For a string enclosed in a list it returns the string.  For a list
enclosed in a string it returns the list.

> IOW, what will it return in these two cases:
>
>   (foo "bar baz" quux)
>   "foo (bar baz) guux"

You can type 'C-M-u' (backward-up-list) on different positions
(in a prog mode where strings are highlighted) to see its effect.

Re: Confusing doc string of list-or-string at point

[Eli Zaretskii]

> Date: Thu, 23 Sep 2021 20:33:28 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> If the list is enclosed in a string, does it mean this will return the
> string, not the list?  IOW, what will it return in these two cases:
> 
>   (foo "bar baz" quux)
>   "foo (bar baz) guux"

It looks like it prefers the innermost thing, whether it's a list or a
string.  Is that correct?  If so, why/when/where is this kind of
"thing" useful in practice?

Re: Confusing doc string of list-or-string at point

[Juri Linkov]

>> If the list is enclosed in a string, does it mean this will return the
>> string, not the list?  IOW, what will it return in these two cases:
>>
>>   (foo "bar baz" quux)
>>   "foo (bar baz) guux"
>
> It looks like it prefers the innermost thing, whether it's a list or a
> string.  Is that correct?  If so, why/when/where is this kind of
> "thing" useful in practice?

It's useful to just click inside either a list or string,
and choose "Select>List" or "Select>String" from the context menu,
and it will select the right thing, either a list when clicked inside list,
or string when clicked inside a string.

Re: Confusing doc string of list-or-string at point

[Eli Zaretskii]

> From: Juri Linkov <juri@linkov.net>
> Cc: emacs-devel@gnu.org
> Date: Thu, 23 Sep 2021 20:59:18 +0300
> 
> > It looks like it prefers the innermost thing, whether it's a list or a
> > string.  Is that correct?  If so, why/when/where is this kind of
> > "thing" useful in practice?
> 
> It's useful to just click inside either a list or string,
> and choose "Select>List" or "Select>String" from the context menu,
> and it will select the right thing, either a list when clicked inside list,
> or string when clicked inside a string.

But the same can happen with a slit/string enclosed within a defun, or
a word enclosed withing a string or a list.  Do we also need
defun-or-string or list-or-string-or-word or any other combinations?

Is this really the correct and useful analysis of this kind of
problems?

Re: Confusing doc string of list-or-string at point

[Eli Zaretskii]

> Date: Thu, 23 Sep 2021 21:43:29 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> But the same can happen with a slit/string enclosed within a defun, or
                                 ^^^^
Should be "list", sorry.

Re: Confusing doc string of list-or-string at point

[Juri Linkov]

>> > It looks like it prefers the innermost thing, whether it's a list or a
>> > string.  Is that correct?  If so, why/when/where is this kind of
>> > "thing" useful in practice?
>> 
>> It's useful to just click inside either a list or string,
>> and choose "Select>List" or "Select>String" from the context menu,
>> and it will select the right thing, either a list when clicked inside list,
>> or string when clicked inside a string.
>
> But the same can happen with a list/string enclosed within a defun, or
> a word enclosed withing a string or a list.  Do we also need
> defun-or-string or list-or-string-or-word or any other combinations?

An alternative would be to add a new thing-at-point target 'string'
instead of 'list-or-string'.  But since there is no such function as
'up-string' that would work like 'up-list' but only to go out of the
enclosing string and not out of a list, so what remains to do is
to use 'up-list' with the non-nil arg NO-SYNTAX-CROSSING, therefore
the new target 'list-or-string'.

Re: Confusing doc string of list-or-string at point

[Juri Linkov]

>> But the same can happen with a list/string enclosed within a defun, or
>> a word enclosed withing a string or a list.  Do we also need
>> defun-or-string or list-or-string-or-word or any other combinations?
>
> An alternative would be to add a new thing-at-point target 'string'
> instead of 'list-or-string'.  But since there is no such function as
> 'up-string' that would work like 'up-list' but only to go out of the
> enclosing string and not out of a list, so what remains to do is
> to use 'up-list' with the non-nil arg NO-SYNTAX-CROSSING, therefore
> the new target 'list-or-string'.

So now I reverted 'list-or-string', and added 'string' instead.

BTW, it seems there is a bug in thing-at-point--end-of-sexp,
but I'm not sure if this is a correct fix:

diff --git a/lisp/thingatpt.el b/lisp/thingatpt.el
index efe33982c3..3d2ea9dedf 100644
--- a/lisp/thingatpt.el
+++ b/lisp/thingatpt.el
@@ -241,7 +261,7 @@ in-string-p
 
 (defun thing-at-point--end-of-sexp ()
   "Move point to the end of the current sexp."
-  (let ((char-syntax (syntax-after (point))))
+  (let ((char-syntax (char-syntax (char-after))))
     (if (or (eq char-syntax ?\))
 	    (and (eq char-syntax ?\") (nth 3 (syntax-ppss))))
 	(forward-char 1)
-- 

Re: Confusing doc string of list-or-string at point

[Eli Zaretskii]

> From: Juri Linkov <juri@linkov.net>
> Cc: emacs-devel@gnu.org
> Date: Fri, 24 Sep 2021 09:32:18 +0300
> 
> >> But the same can happen with a list/string enclosed within a defun, or
> >> a word enclosed withing a string or a list.  Do we also need
> >> defun-or-string or list-or-string-or-word or any other combinations?
> >
> > An alternative would be to add a new thing-at-point target 'string'
> > instead of 'list-or-string'.  But since there is no such function as
> > 'up-string' that would work like 'up-list' but only to go out of the
> > enclosing string and not out of a list, so what remains to do is
> > to use 'up-list' with the non-nil arg NO-SYNTAX-CROSSING, therefore
> > the new target 'list-or-string'.
> 
> So now I reverted 'list-or-string', and added 'string' instead.

Thanks, I think this is a cleaner solution.

RE: [External] : Re: Confusing doc string of list-or-string at point

[Drew Adams]

> BTW, it seems there is a bug in thing-at-point--end-of-sexp,
> but I'm not sure if this is a correct fix:
> 
> diff --git a/lisp/thingatpt.el b/lisp/thingatpt.el
> index efe33982c3..3d2ea9dedf 100644
> --- a/lisp/thingatpt.el
> +++ b/lisp/thingatpt.el
> @@ -241,7 +261,7 @@ in-string-p
> 
>  (defun thing-at-point--end-of-sexp ()
>    "Move point to the end of the current sexp."
> -  (let ((char-syntax (syntax-after (point))))
> +  (let ((char-syntax (char-syntax (char-after))))
>      (if (or (eq char-syntax ?\))
>  	    (and (eq char-syntax ?\") (nth 3 (syntax-ppss))))
>  	(forward-char 1)

That looks right to me.

BTW2:
Why is function `in-string-p' even defined?
It doesn't seem to be used anywhere (but I'm
using Emacs 26.3, so ignore if it's used now).

RE: [External] : Re: Confusing doc string of list-or-string at point

[Stefan Kangas]

Drew Adams <drew.adams@oracle.com> writes:

> Why is function `in-string-p' even defined?
> It doesn't seem to be used anywhere (but I'm
> using Emacs 26.3, so ignore if it's used now).

describe-function tells me that it's "obsolete since 25.1; use (nth 3
(syntax-ppss)) instead".

Re: Confusing doc string of list-or-string at point

[Stefan Monnier]

>  (defun thing-at-point--end-of-sexp ()
>    "Move point to the end of the current sexp."
> -  (let ((char-syntax (syntax-after (point))))
> +  (let ((char-syntax (char-syntax (char-after))))

Splitting the work between `char-after` and `char-syntax` means that we
can't obey any `syntax-table` text properties placed on that char.
So I think the above change would be a regression in this sense (tho
I understand it fixes or circumvents a bug elsewhere).


        Stefan

Re: Confusing doc string of list-or-string at point

[Juri Linkov]

>>  (defun thing-at-point--end-of-sexp ()
>>    "Move point to the end of the current sexp."
>> -  (let ((char-syntax (syntax-after (point))))
>> +  (let ((char-syntax (char-syntax (char-after))))
>
> Splitting the work between `char-after` and `char-syntax` means that we
> can't obey any `syntax-table` text properties placed on that char.
> So I think the above change would be a regression in this sense (tho
> I understand it fixes or circumvents a bug elsewhere).

No bugs prompted this patch, just noticed the difference between
this function and its counterpart thing-at-point--beginning-of-sexp
that uses `(char-syntax (char-before))`.

Re: Confusing doc string of list-or-string at point

[Stefan Monnier]

>> Splitting the work between `char-after` and `char-syntax` means that we
>> can't obey any `syntax-table` text properties placed on that char.
>> So I think the above change would be a regression in this sense (tho
>> I understand it fixes or circumvents a bug elsewhere).
>
> No bugs prompted this patch, just noticed the difference between
> this function and its counterpart thing-at-point--beginning-of-sexp
> that uses `(char-syntax (char-before))`.

My point was just that `syntax-after` is almost always the better choice
over (char-syntax (char-after)) because it has access to more
information, so the better fix to unify the two would be to use
`syntax-after` in both cases ;-)


        Stefan