Re: [RFC] The best way to choose an "action" at point: context-menu-mode, transient, which-key or embark?

[Kierin Bell <fernseed@fernseed.me>]

Hi Ihor,

Ihor Radchenko <yantar92@posteo.net> writes:

> TL;DR: We are in the process of designing a more unified selection
> interface for Org mode and want to see if there is some way to unify
> context-menu-mode, transient, which-key and embark together. The idea
> is to (1) avoid too many customizations; (2) allow users to decide how
> to choose between multiple options - by mouse, keyboard, and using
> customizable UIs.

I think that the built-in Emacs thingatpt.el should not be overlooked
here.

Instead of implementing an entire system specific to Org, imagine a
generic action-at-point interface that works on "things" from
thingatpt.el. For the various targets, Org could add new "providers" to
`thing-at-point-provider-alist', `forward-thing-provider-alist', and
`bounds-of-thing-at-point-provider-alist'. [ Org actually does already
register its own 'url' provider for links. ]

Then, Org could implement a number of action selection interfaces that
act on the various classes of "thing". An exemplary package would be
Philip Kaludercic's great =do-at-point= package, which provides a simple
action selection menu for the thing-at-point using
`read-multiple-choice', which I find elegant and intuitive.[1]

I have gone as far as implementing a 'heading' provider for Org and
`outline-mode' (for use with =do-at-point.el=). I don't see any reason
why Org couldn't define a 'citation' provider, a 'source-block'
provider, etc.

The only downside that I have found with adding lots of thingatpt.el
providers is that it can be difficult to write providers efficient
enough for `forward-thing' in particular
(`forward-thing-provider-alist').

I may also be misunderstanding the proposed interface. For example,
instead of a generic interface for acting on a single thing at point,
maybe you are describing more of an interface for associating commands
with multiple potential targets that must be located (e.g., in a
subtree), which are then each associated with actions.

Even if that's the case, there is a good case for implementing
thingatpt.el providers for the targets, so that users could bring our
own action-at-point packages/interfaces. [ I would be willing to help
write some of those providers. ] And if thingatpt.el isn't generalized
or fast enough, then there is a case for creating a new, more flexible
/de facto/ library like this for Emacs.


[1] https://codeberg.org/pkal/do-at-point

Thanks,
Kierin