Hi all,

I'd like to remove them in favor of using `org-icompleting-read'
everywhere (or better yet, `completing-read').

`org-completing-read-no-i' doesn't do much, is called twice and can be
replaced with a let binding wrapper.

`org-completing-read' could be updated this way:

    (defmacro with-org-minibuffer-keys (&rest body)
      "Minibuffer read with SPACE being a normal character."
      `(let ((enable-recursive-minibuffers t)
             (minibuffer-local-completion-map
              (copy-keymap minibuffer-local-completion-map)))
         (org-defkey minibuffer-local-completion-map " " 'self-insert-command)
         (org-defkey minibuffer-local-completion-map "?" 'self-insert-command)
         (org-defkey minibuffer-local-completion-map (kbd "C-c !") 
'org-time-stamp-inactive)
         ,@body))

    (with-org-minibuffer-keys
        (org-icompleting-read args))

This change will simplify the code nicely.  One example that bothers me
is `org-tags-completion-function': it's essentially a hacky closure that
relies on several dynamic variables being set:
`org-last-tags-completion-table' and
`org-add-colon-after-tag-completion' are implicit arguments to this
function.

With the proposed change, each use of `org-tags-completion-function'
will be in conjunction with `org-icompleting-read'. That means a new
function can be written:

(defun org-read-tags (tag-list &optional colon)
  (let ((org-last-tags-completion-table tag-list)
        (org-add-colon-after-tag-completion colon))
    (org-icompleting-read
     "Tag: " 'org-tags-completion-function
     ;; ...
     )))

I think it's a better interface, since the coupling is now made
explicit, instead of having to remember to set
`org-last-tags-completion-table' and
`org-add-colon-after-tag-completion' each time
`org-tags-completion-function' is to be used.

An even better thing to do would be to use "lexical-binding: t", remove
`org-tags-completion-function' and have it be a real closure inside
`org-read-tags'. Is there any wish or effort to move Org to
lexical-binding? Adding it would allow us to get rid of those dynamic
variables and `org-tags-completion-function' altogether and have the
lambda enclose on tag-list and colon instead.

regards,
Oleh

Reply via email to