Anders Johansson <mejlaandersj@gmail.com> writes:
> Hi,
> I noticed that this commit:
> b1a570b3b  org-agenda: Fix regression when diary sexp timestamps are ignored
>
> made a call like (org-at-timestamp-p 'agenda) match against the regexp
> `org-element--timestamp-regexp`. However, this regexp doesn’t have the
> match groups setup as the others (`org-ts-regexp3` and `org-ts-regexp2`) so
> the last part of the docstring for org-at-timestamp-p doesn’t hold anymore
> (that the timestamp parts should now be put in match groups).

Confirmed.
Can you try the attached patch?

Also, unless I miss something, docstring was not fully obeyed even in
the past. Consider the following timestamp:

 <2022-04-15 Fri 11:40>

I ran the following code with point at the timestamp:
(progn (org-at-timestamp-p ) (mapcar #'substring-no-properties (list (match-string 0) (match-string 1) (match-string 2) (match-string 3) (match-string 4) (match-string 5) (match-string 6) (match-string 7))))

The output is:
("<2022-04-15 Fri 11:40>" "2022-04-15 Fri 11:40" "2022" "04" "15" "Fri" " 11:40" "11" "40")

There are two extra groups (group 1 and group 6), presumably due to
earlier changes in org-ts-regexp1.

> I was using this in some code of mine, but I also realize that this doesn’t
> feel all that clean anyway. I am not sure if this is used elsewhere in org
> though.

I am currently working on an alternative concept of analyzing arbitrary
org element components. Instead of named groups, I plan to provide named
match groups:

(progn (org-element-match 'timestamp) (org-element-match-data))

((:minute-start 314 316)
 (:hour-start 311 313)
 (:day-start 304 306)
 (:month-start 301 303)
 (:year-start 296 300)
 (:date-start 296 316)
 (:timestamp-start 295 317)
 (:end-marker 316 317)
 (:begin-marker 295 296)
 (:full 295 317)
 (:full-no-blank 295 317)
 (:full-no-affiliated 295 317))

The current WIP implementation is in
https://github.com/yantar92/org/blob/feature/org-font-lock-element/lisp/org-element-match.el

The match groups set by org-at-timestamp-p are currently being used by
the following functions: org-mouse-delete-timestamp,
org-follow-timestamp-link, org-time-stamp, org-toggle-timestamp-type

However, all but one of them are only using group 0 and
org-follow-timestamp-link is using group 1 as currently set by
org-at-timestamp-p against the docstring.

Best,
Ihor