Re: Completion of links to man pages

[Eli Zaretskii <eliz@gnu.org>]

> From: Ihor Radchenko <yantar92@posteo.net>
> Cc: manikulin@gmail.com, emacs-orgmode@gnu.org, emacs-devel@gnu.org
> Date: Thu, 05 Oct 2023 16:53:57 +0000
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> > How is it different from the "M-x man" completion we already have?
> >> 
> >> M-x man will display the man page, while we just need `completing-read'
> >> from the same source M-x man or M-x woman use.
> >
> > Sorry, I don't understand: "M-x man" does provide completion.
> 
> Yes, but one cannot replicate the same completion dialogue
> programmatically in future-compatible way.

What do you mean by that?  "M-x man" does this:

  (interactive
   (list (let* ((default-entry (Man-default-man-entry))
		;; ignore case because that's friendly for bizarre
		;; caps things like the X11 function names and because
		;; "man" itself is case-insensitive on the command line
		;; so you're accustomed not to bother about the case
		;; ("man -k" is case-insensitive similarly, so the
		;; table has everything available to complete)
		(completion-ignore-case t)
		Man-completion-cache    ;Don't cache across calls.
		(input (completing-read
			(format-prompt "Manual entry"
                                       (and (not (equal default-entry ""))
                                            default-entry))
                        'Man-completion-table
			nil nil nil 'Man-topic-history default-entry)))

This uses completing-read, as I think you wanted.

> > And what do you mean by "`completing-read' from the same source M-x
> > man or M-x woman use"?
> >
> > IOW, I think I have no clue of what are you trying to accomplish,
> > sorry.
> 
> We aim to create Org links like [[man:ls]].
> To create a link in Org, the interface is C-c C-l (org-insert-link),
> which then prompts for link type (man:) and link path (ls).
> When querying for the path, we want to have the same completion
> COLLECTION as M-x man/woman has.

Why cannot you reuse Man-completion-table?

> For now, as you can see in the quoted code from my initial message, we
> have to partially replicate the code from man.el and woman.el:
> 
> (defun org-man--complete-man (prompt)
>    (require 'man)
>    (let (Man-completion-cache) ;; <- implementation detail in man.el
>      (completing-read
>       prompt
>       'Man-completion-table)))

And why is that a problem?

> However, `Man-completion-table' is not documented (no docstring),

If the only thing that's missing is its doc string, that is easy to
add.

> What I am asking here is to provide a stable Elisp API for the above use
> case. Currently, we have to rely on implementation details.

From where I stand, we have already a stable API tested by years of
use.  What is maybe missing is some documentation to allow its easier
use, that's all.