Hello,
I'd like to set org-todo-keywords and org-todo-keyword-faces through
directory-local variables, to get rid of duplicate #+SEQ_TODO lines in
my Org files[1].
Right now I see two obstacles for this to work:
(1) org-set-regexps-and-options, which sets up a bunch of TODO-related
machinery, insists on using (default-value 'org-todo-keywords),
(2) this function is called in the major mode function, which IIUC means
that directory-local values have not been applied yet.
The first obstacle looks like it can be easily removed[2]; the second
obstacle looks more substantial. It is trivially side-stepped by
sticking (hack-local-variables) at the top of org-mode; to my untrained
eye, it looks like TRT would rather be for Org to add
org-set-regexps-and-options to hack-local-variables-hook.
This sounds like a risky change though: I imagine that a lot of what
happens in the major mode function depends on what
org-set-regexps-and-options sets up, and would therefore need to be
moved to this hook as well. Figuring which parts should be moved seems
like a non-trivial task that might introduce some regressions…
Can anyone confirm that this would (in principle) be the way forward, or
tell me if I am missing something[3]?
Thank you for your time.
[1] For example:
#+begin_src elisp
((org-mode
. ((org-todo-keywords
. ((sequence "REPORT" "REPORTED" "WAITING" "FIXED")
(sequence "CANCELED")))
(org-todo-keyword-faces
. (("REPORT" . org-todo)
("REPORTED" . warning)
("WAITING" . warning)
("FIXED" . org-done)
("CANCELED" . shadow))))))
#+end_src
I'd like that so much that I went through the trouble of writing
safe-local-variable predicates for these variables:
#+begin_src elisp
(put 'org-todo-keywords
'safe-local-variable
(lambda (x)
(cl-every
(lambda (seq)
(and (memq (car seq) '(sequence type))
(cl-every (lambda (i) (stringp i)) (cdr seq))))
x)))
(put 'org-todo-keyword-faces
'safe-local-variable
(lambda (x)
(cl-every
(lambda (pair)
(pcase pair
(`(,keyword . ,face)
(and (stringp keyword)
(or (facep face) (listp face))))))
x)))
#end_src
[2] I tried to go through org.el's history, but I could not find a
rationale for using default-value.
[3] Alternatively, maybe the answer is as simple as "Org documents
should be self-sufficient; keywords should be explicitly set in
#+SEQ_TODO lines"; that wouldn't sound right though, since
org-todo-keywords is customizable.
