Re: [PATCH] Query when exiting with running clock

[Kyle Meyer]

Allen Li writes:

> This is a patch adding a query function when exiting Emacs, warning the
> user if there is a running clock.  This is useful for preventing the
> user from accidentally leaving dangling clocks.  I have had success
> using a modified personal version of this code.

Thanks.  I'd find this useful as well.

> My personal version instead prompts the user to clock out and save the
> buffer.  I didn't use it for the patch because it seems a little hacky
> to me; e.g., kill-emacs-query-functions doesn't mention whether
> functions can have side effects, and the UI is inconsistent with
> save-buffers-kill-emacs.  The code for my personal version:

Fwiw that's what Emacs's lisp/calendar/timeclock.el does:

  (defun timeclock-query-out ()
    "Ask the user whether to clock out.
  This is a useful function for adding to `kill-emacs-query-functions'."
    (and (equal (car timeclock-last-event) "i")
         (y-or-n-p "You're currently clocking time, clock out? ")
         (timeclock-out))
    ;; Unconditionally return t for `kill-emacs-query-functions'.
    t)

> Subject: [PATCH 1/2] org-clock: Replace org-clocking-buffer with
>  org-clock-is-active
>
> org-clocking-buffer and org-clock-is-active have the same definition.
> org-clocking-buffer is defined in org-clock.el while
> org-clock-is-active is defined in org.el.  org-clock.el requires
> org.el.
>
> org-clocking-buffer is kept as an alias to preserve backward
> compatibility with any user code.

Dropping the duplicate definitions using an alias sounds good, though as
I mention below I'd prefer the spots that are specifically concerned
with a buffer to keep using the org-clocking-buffer name.

> * lisp/org-clock.el (org-clocking-buffer): Changed to alias.
> (org-clocking-p): Changed reference to org-clocking-buffer.
> (org-clock-out): Changed reference to org-clocking-buffer.
> (org-clock-cancel): Changed reference to org-clocking-buffer.
> (org-clock-sum): Changed reference to org-clocking-buffer.
> (org-clock-out-if-current): Changed reference to org-clocking-buffer.
> (org-clock-save): Changed reference to org-clocking-buffer.
> ---
>  lisp/org-clock.el | 20 +++++++++-----------
>  1 file changed, 9 insertions(+), 11 deletions(-)
>
> diff --git a/lisp/org-clock.el b/lisp/org-clock.el
> index 892ae1810..37459580b 100644

Hmm, this patch doesn't apply to the current master (041459727).  The
preimage blob is quite old:

  $ git describe 892ae1810
  release_8.2.7b-7-gbacfe5b4f:lisp/org-clock.el
  $ git log --oneline --format="%h %cd" --find-object=892ae1810
  ca21b7b86 Mon Jan 12 12:35:10 2015 +0100
  bacfe5b4f Fri Jul 25 11:02:55 2014 +0200

> --- a/lisp/org-clock.el
> +++ b/lisp/org-clock.el
> @@ -503,13 +503,11 @@ of a different task.")
>    (mapc (lambda (m) (org-check-and-save-marker m beg end))
>       org-clock-history))
>  
> -(defun org-clocking-buffer ()
> -  "Return the clocking buffer if we are currently clocking a task or nil."
> -  (marker-buffer org-clock-marker))
> +(defalias 'org-clocking-buffer #'org-clock-is-active)
>  
>  (defun org-clocking-p ()
>    "Return t when clocking a task."
> -  (not (equal (org-clocking-buffer) nil)))
> +  (not (equal (org-clock-is-active) nil)))

Eh, so this too could just be an alias to org-clock-is-active, or, if it
looks like any callers really depend on this being t rather than just
non-nil (but why would they?), `(and (org-clock-is-active) t)' would be
better.  Anyway, I know you're just doing the plain substitution here,
so I'm fine keeping it at that.

>  
>  (defvar org-clock-before-select-task-hook nil
>    "Hook called in task selection just before prompting the user.")
> @@ -1501,7 +1499,7 @@ to, overriding the existing value of 
> `org-clock-out-switch-to-state'."
>         ts te s h m remove)
>        (setq org-clock-out-time now)
>        (save-excursion ; Do not replace this with `with-current-buffer'.
> -     (org-no-warnings (set-buffer (org-clocking-buffer)))
> +     (org-no-warnings (set-buffer (org-clock-is-active)))

This is an example of a spot that I think should continue using the
org-clocking-buffer variant for readability.  And scanning the other
spots in this patch, I think they actually all fall into this category.

> Subject: [PATCH 2/2] org-clock: Query when exiting with running clock
>
> It's annoying to accidentally quit Emacs with a running clock, then
> resolve the clock the next time when Emacs is started.
>
> * lisp/org-clock.el (org-clock-kill-emacs-query): New function.
> ---
>  lisp/org-clock.el | 9 +++++++++
>  1 file changed, 9 insertions(+)
>
> diff --git a/lisp/org-clock.el b/lisp/org-clock.el
> index 37459580b..bc1762f63 100644
> --- a/lisp/org-clock.el
> +++ b/lisp/org-clock.el
> @@ -2886,6 +2886,15 @@ The details of what will be saved are regulated by the 
> variable
>               (if (outline-invisible-p)
>                   (org-show-context))))))))))
>  
> +(defun org-clock-kill-emacs-query ()
> +  "Query user when killing Emacs.
> +This function is added to `kill-emacs-query-functions'."
> +  (if (org-clocking-buffer)

Hmm, why use org-clocking-buffer rather than org-clock-is-active?  After
a patch that introduced org-clock-is-active and switched a bunch of
spots over to it, I would have expected you to use it here.

> +      (y-or-n-p "Org clock is running; exit anyway? ")
> +    t))
> +
> +(add-hook 'kill-emacs-query-functions #'org-clock-kill-emacs-query)

lisp/calendar/timeclock.el has the following option:

  (defcustom timeclock-ask-before-exiting t
    "If non-nil, ask if the user wants to clock out before exiting Emacs.
  This variable only has effect if set with \\[customize]."
    :set (lambda (symbol value)
         (if value
             (add-hook 'kill-emacs-query-functions #'timeclock-query-out)
           (remove-hook 'kill-emacs-query-functions #'timeclock-query-out))
         (set symbol value))
    :type 'boolean)

I wonder if we should do something similar.