bug#61002: 28.2; Gnus - "Date" scoring scores all articles.

[Eric Abrahamsen]

Jakub Ječmínek <jecminek.k@gmail.com> writes:

> Hi, thanks for the feedback!
>
> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> Thanks for the ping. TBH I've never used scoring in Gnus, so this is new
>> territory for me. I tried the patch and it worked correctly, but...
>>
>> The patch is doing this:
>>
>> (gnus-date-get-time (gnus-date-iso8601 match))
>>
>> This is the definition of `gnus-date-iso8601':
>>
>> (defun gnus-date-iso8601 (date)
>>   "Convert the DATE to YYYYMMDDTHHMMSS."
>>   (condition-case ()
>>       (gnus-time-iso8601 (gnus-date-get-time date))
>>     (error "")))
>>
>> So error handling aside, the patch ends up doing the equivalent of:
>>
>> (let ((ds "Sun, 22 Jan 2023 09:30:17 +0100"))
>>   (gnus-date-get-time
>>    (gnus-time-iso8601
>>     (gnus-date-get-time ds))))
>>
>> which effectively round-trips the date string through two different
>> formats.
>
> You're right. I've replaced the offending form with this:
>
> (car (time-convert (gnus-date-get-time match) 1))
>
>> [...] I think we should just keep the inner
>> `gnus-date-get-time' call, and wrap the whole `int-to-string' form in a
>> `condition-case'. If anything at all goes wrong we really don't care, we
>> can just skip it and return a "1" or something.
>>
>> WDYT?
>
> I believe that error handling is now redundant, because even if we try to
> parse invalid date using `gnus-date-get-time' we still end up with valid
> value (0).
>
> (/ (car (time-convert (gnus-date-get-time "invalid date") 1)) 86400)
>
> But, there's one important thing I failed to realize - the code I wrote
> would break prompts for other headers. Therefore I propose we add `cond'
> and distinguish which header we're scoring on.
>
> Please let me know what you think, here's the patch:
>
>>From 4e2174503c1d69345536929a22483ae309048271 Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?Jakub=20Je=C4=8Dm=C3=ADnek?= <jecminek.k@gmail.com>
> Date: Mon, 12 Feb 2024 14:03:24 +0100
> Subject: [PATCH] Provide valid match string when scoring on date header
>  (Bug#61002)
>
> * lisp/gnus/gnus-score.el (gnus-summary-score-entry): Fix invalid
> default match string when creating score file interactively on date header.
> ---
>  lisp/gnus/gnus-score.el | 11 ++++++++---
>  1 file changed, 8 insertions(+), 3 deletions(-)
>
> diff --git a/lisp/gnus/gnus-score.el b/lisp/gnus/gnus-score.el
> index bd19e7d7cd7..479b7496cf1 100644
> --- a/lisp/gnus/gnus-score.el
> +++ b/lisp/gnus/gnus-score.el
> @@ -893,9 +893,14 @@ If optional argument `EXTRA' is non-nil, it's a 
> non-standard overview header."
>                                (t "permanent"))
>                          header
>                          (if (< score 0) "lower" "raise"))
> -                (if (numberp match)
> -                    (int-to-string match)
> -                  match))))
> +                   (cond ((numberp match) (int-to-string match))
> +                         ((string= header "date")
> +                          (int-to-string
> +                           (-
> +                            (/ (car (time-convert (current-time) 1)) 86400)
> +                            (/ (car (time-convert (gnus-date-get-time match) 
> 1))
> +                               86400))))
> +                         (t match)))))
>
>      ;; If this is an integer comparison, we transform from string to int.
>      (if (eq (nth 2 (assoc header gnus-header-index)) 'gnus-score-integer)
> --
> 2.39.3 (Apple Git-145)
>
> The patch should be correctly indented even though it doesn't look like
> it is. The reason is that previous lines contain mixture of tabs and
> spaces.

Looks good! I've pushed, and am closing this bug, thanks.