Allow multiple workspaces per buffer

[gagbo]

Hello again,

I hope this time I'll be clear enough, I tend to over-complicate things and I try to work on that :(

Rescoping the previous messy PR I made. The point is that here, we do not need non-consuming groups in order to allow multiple workspaces for a single buffer.

Feature summary

The workflow I want to cover is :

• Set the frame to a given workspace context (let's say "Bufler") with bufler-workspace-frame-set
• Whenever I find a file in the frame, I add "Bufler" to its list of workspaces. So find-file on "bufler.el" will (through hooks) call (bufler-workspace-buffer-name-workspace "Bufler")
• Change context (starting to work on my Emacs config instead)
• Set the frame to the new workspace ("Emacs Config") with bufler-workspace-frame-set
• Find the same "bufler.el" file. Now the hook will call (bufler-workspace-buffer-name-workspace "Emacs Config")

Then, if I list buffers with M-x bufler, I want to see the bufler.el buffer under 2 different named workspaces.

Change summary

• Rename bufler-workspace-name to bufler-workspace-names to show it
  can hold multiple values
• Add prefix to workspace names in the groups
• Make bufler-workspace-buffer-name-workspace add to list of
  buffer-local workspace value instead of replacing it
• Use another version of seq-group-by in bufler-group-tree that adds
  an element in multiple groups if the grouping function returns a list.

For the last point, it would be better to use cl-defmethod to override
the seq-group-by function only when grouping workspaces but currently
no version tried work. Last try :

(cl-defmethod seq-group-by ((function (eql 'bufler-group-auto-workspace)) sequence)
  "Specialization of `seq-group-by' that puts elements of SEQUENCES in each of
the workspaces returned by FUNCTION."
  (seq-reduce
   (lambda (acc elt)
     (let ((keys (funcall function elt))
           (add-to-groups (lambda (key)
                            (let ((cell (assoc key acc)))
                              (if cell
                                  (setcdr cell (push elt (cdr cell)))
                                (push (list key elt) acc))))))
       (mapc add-to-groups keys)
       acc))
   (seq-reverse sequence)
   nil))

[alphapapa]

FYI, I probably won't have time to look at this soon. Thanks for your patience.

[alphapapa]

FYI, due to some IRL issues and computer problems, I haven't had time to work on Emacs stuff. It will probably be a while longer before I can look at this carefully.

[gagbo]

Don't worry, living comes first.

Actually I think I'll move the PR back to draft as I'm only starting using the features I'd like to add and I'm seeing a few bugs/usability issues that I'm fixing

[alphapapa]

Ok, thanks. Let me know when you're ready for it to be looked at again.

[alphapapa]

For reference, this PR's ideas were previously discussed at #49.

Thinking about this a bit more, I think what's needed is:

1. Allow groups to be specified with a plist for options, e.g. a :no-consume t one.
2. The grouping function (i.e. the equivalent of seq-group-by, assuming we have to reimplement one like it with our changes) should use that plist to control what it does with the grouped values. So if non-consuming is specified, a buffer would not be consumed, and it would be eligible for also being matched by other predicates. It might be necessary to distinguish between "match into this group, but don't consume this, and allow other predicates in this chain to match it" and "match into this group, but don't consume this, but don't allow other predicates in this chain to match it, only ones in a parallel tree"...which starts to feel a little complicated, so we'll have to see.

[alphapapa]

We should probably discuss these ideas in an issue rather than a PR, so: #69