Re: Proposal: 'executable' org-capture-templaes

[Arthur Miller <arthur.miller@live.com>]

Max Nikulin <manikulin@gmail.com> writes:

> On 20/06/2022 19:10, Ihor Radchenko wrote:
>> Max Nikulin writes:
>> Sure. I was hinting about such functionality in the new library we are
>> discussing. Multiple capture menus are not possible now anyway, so it
>> will be a new feature if we decide that we need it (I am not 100% sure
>> if having multiple menus is not going to be confusing).
>
> In my opinion application logic should be consistent with UI. Do you think
> current behavior of help buffers has no problems? If so I have no chance to
> convince you that upgrade of menu should be accompanied by changes in
> org-capture.
>
> Consider the following case. A user has 2 virtual desktops dedicated to rather
> independent tasks and an Emacs frame on each desktop. On desktop 1 she opens
> help for some function with aim to evaluate if it is appropriate for some
> particular purpose. Next she has to switch to another task and applications for
> it reside on virtual desktop 2. To proceed further she has to read several help
> pages relevant for task 2. After completion of this task it is time to resume
> task 1 and to switch to virtual desktop 1. Does a window with the help buffer
> still display the same content as before switching to task 2? No, context on
> desktop 1 lost due to some actions in the Emacs frame on desktop 2.
>
> Such synchronization is against multitasking and I do not like the idea to get
> it for org-capture as well. Currently read-key and modal prompt is a kind of
> excuse, but the idea of new menu library is non-blocking keymap-based selection.

Yes, that is true, current Emacs help-buffer implementation is not
designed with your use-case in mind. If we leave out "virtual desktops",
it is unimportant detail. What you speak about is that user is working with
two different contexts at the same time and prefers to keep each context
in it's own frame. Emacs as a whole is not designed to work in the way I
percieve it has clean separation between contexts in each frame. Menu
buffer is "global" for entire Emacs process, and there are other
features of Emacs that does not work well in such scenarion. Some people
prefer to keep an Emacs process per project/task for that reason.

I can agree with you, that it would be nice if things were not designed
the way they are, but I don't have an opinion how org-capture should
behave in that regard, and how much work would it be to design it the
way you suggest. With other words, I say neither yes or no to anything
:). I would just like to remind that we are speaking of two things here;
one is menu buffer, that is replacing org-mks & co, as in org-agenda and
other places Ihor mentioned in earlier mails, and the other one is
org-capture itself.

>>> Currently several capture menu instances may be requested though
>>> org-protocol (or calling org-capture from emacsclient). The behavior is
>>> rather confusing. New menu may help to fix the issue, that is why I
>>> raised the question about parallel captures.
>> I am not sure which behavior you have in mind.
>
> try the following commands without selecting a template in an Emacs frame in
> between
>
>     emacsclient 'org-protocol:/capture?url=url-A&title=title-A&body=body=A'
>     emacsclient 'org-protocol:/capture?url=url-B&title=title-B&body=body=B'
>
>> What I was thinking as a conservative implementation that would not
>> introduce any new features is replacing the old menu with the new one
>> every time the same menu is called. So, every time the user calls menu
>> (e.g. capture menu), only the last capture environment is preserved. The
>> previous, potentially unfinished, capture menus will be destroyed.
>
> Causing loss of user data. Currently it is hard to start new capture before
> selecting a template.

Current org-capture is one at a time because of how org-mks works. There
is nothing that prevents org-capture to open enumerated buffers,
org-caputre<1>, org-capture<2> etc. User has to manually serialize data
anyway, via C-c C-c from withing capture buffer? So in principle it is
still one capture buffer at a time that manipulates the file on the disk
itself?

Problem would arise if serialization takes long time, i.e. there is lots
of data to write to file, and user switches to another org-capture
buffer and tries to save that one. So the serialization itself has to be
guarded somehow, but a mutex or file lock or something. At least I think
so, I am not sure if I am correct completely. Just trying to think loud
about it.