Re: [emacs-wiki-discuss] RE: New: planner-tasks-file-behavior

[fouvry+emacs-wiki]

[...]

| I finally got around to working on Frederik Fouvry's patch to
| planner-update-task, [snip] He wanted it to close
| files that it had to open itself.
| 
| planner--dev--1.0--patch-167 now includes the
| planner-tasks-file-behavior user option. By default, it closes all the
| extra buffers opened by plan, planner-copy-or-move-region,
| planner-copy-or-move-task, planner-replan-task, and
| planner-update-task. (plan) should add just one new buffer: the buffer
| for today.

The patch as I submitted it was actually too slow when many files
had to be opened (mainly due to font-locking).  I'm not using it
anymore, but Sacha's suggestion to collect the open files first,
then update them (without saving & closing after every task
update), and then restore them, seems a better and more viable
alternative.  My patch was just a fast way to get a "prototype"
working.

Perhaps it an even simpler technique is to keep track of the
files that have to be opened for saving/updating, and then to
save and kill these buffers once done.  The advantage: the names
of the buffers to be killed are immediately available, and it's
less likely to interfere with other things (like just a snapshot
of the buffer list is doing [1]).  Even better would be if these
files could be opened without font-locking: that would speed
things up a lot, and may make buffer bookkeeping unnecessary
(although there's nothing against it).

[Something else: The duplication of recording the live buffers
 (both in planner-copy-or-move-region and
 planner-copy-or-move-task, where the former calls the latter),
 makes me uneasy.]

Cheers,

Frederik


[1] A (possible) example of this going astray: suppose that a
    hook/function is called during the update that changes the
    major mode of an open buffer B to planner-mode.  The
    live-buffer list will not contain B, so the buffer, even
    though it may not have anything to do with the day page, will
    be killed.  IMO that is wrong: it should only kill buffers it
    had to open itself, NOT buffers that were opened/affected by
    other functions.  A realistic scenario: the update may
    contain a call the function to generate a task overview.  If
    so, it will be killed at the end.