`org-alert-time-match-string` is too permissive

[jwiegley]

Hello, I've found that org-alert-time-match-string is too lax in what it parses. I have lots of date stamps elsewhere in my properties that it thinks are associated with the scheduling time. What works better for me is to constrain it as follows:

"\\(?:SCHEDULED\\|DEADLINE\\):\\s-+<[^>]+\\([0-9]\\{2\\}:[0-9]\\{2\\}\\).*>"

This way it only checks the time that immediately follows SCHEDULED or DEADLINE, and only the time within that timestamp.

[ntBre]

Hi John, thanks for the report! (and for the great alert package that we depend on!)

We already expose org-alert-time-match-string as a customizable variable; are you suggesting that this would be a better default? I think this is similar to #29 and #42 (and possibly others), so it seems like a common issue that probably warrants an update to the default.

[jwiegley]

I have a feeling it may be! I'm still having one problem even with this string: I get alerts for the end of time-ranged appointments (09:00-10:00) rather than the beginning. Still tracking that one down...

[gnomon-]

I have a feeling it may be! I'm still having one problem even with this string: I get alerts for the end of time-ranged appointments (09:00-10:00) rather than the beginning. Still tracking that one down...

I believe this may be because of the <[^>]+\\(...\\).*> portion of the regex which picks out the timestamp. Since the + in the [^>]+ range is greedy, it will begin by matching all the way to the very end of the <yyyy-mm-dd 01:23--34:56> timestamp range, then back up to surrender as little as possible for the captured \\([0-9\\{2\\}:[0-9]\\{2\\}\\) matcher, which will match 34:56. However if it is converted to the non-greedy pattern <[^>]+?\\(...\\).*> then I believe it should only consume as much of the input space as necessary rather than as much as possible before matching the desired 01:23.

And indeed, if we check the greedy version..:

(save-match-data
  (string-match "\\(?:SCHEDULED\\|DEADLINE\\):\\s-+<[^>]+\\([0-9]\\{2\\}:[0-9]\\{2\\}\\).*>" "foo DEADLINE: <0123-45-67 01:23-45:67> bar")
  (match-data))

...we get the result..:

(4 38 32 37)

...where 32-37 corresponds to 45:67.

But if we switch to the non-greedy version..:

(save-match-data
  (string-match "\\(?:SCHEDULED\\|DEADLINE\\):\\s-+<[^>]+?\\([0-9]\\{2\\}:[0-9]\\{2\\}\\).*>" "foo DEADLINE: <0123-45-67 01:23-45:67> bar")
  (match-data))

...then we get..:

(4 38 26 31)

...where 26-31 corresponds to 01:23, as desired.