Programming with org-element-cache -> short introduction?

Programming with org-element-cache -> short introduction?

[Thorsten Jolitz]

Hi List (or rather Hi Nicolas),

when programming with Org elements sometimes things seem to work and
then something strange happens - what smells like a cache problem. I
don't mean a cache bug, but a programmer (me) not taking the cache
into account the right way.

Is there a short introduction somewhere about the 'todos' and 'nogos' in
programming with elements wrt to the org-element-cache?

-- 
cheers,
Thorsten

Re: Programming with org-element-cache -> short introduction?

[Nicolas Goaziou]

Hello,

Thorsten Jolitz <tjolitz@gmail.com> writes:

> when programming with Org elements sometimes things seem to work and
> then something strange happens - what smells like a cache problem. I
> don't mean a cache bug, but a programmer (me) not taking the cache
> into account the right way.

It might also be a cache problem. Do not hesitate to report it.

> Is there a short introduction somewhere about the 'todos' and 'nogos' in
> programming with elements wrt to the org-element-cache?

The only "nogo" is to never modify (destructively) a value returned by
`org-element-at-point' or `org-element-context'. Consider their return
value as read-only, and possibly invalid as soon as you modify the
buffer.

These advices don't apply to `org-element-parse-buffer', which doesn't
use cache (if it did, it should copy cached elements beforehand anyway).


HTH,

-- 
Nicolas Goaziou

Re: Programming with org-element-cache -> short introduction?

[Thorsten Jolitz]

Nicolas Goaziou <mail@nicolasgoaziou.fr> writes:

Hello,

> Thorsten Jolitz <tjolitz@gmail.com> writes:
>
>> when programming with Org elements sometimes things seem to work and
>> then something strange happens - what smells like a cache problem. I
>> don't mean a cache bug, but a programmer (me) not taking the cache
>> into account the right way.
>
> It might also be a cache problem. Do not hesitate to report it.

Ok

>> Is there a short introduction somewhere about the 'todos' and 'nogos' in
>> programming with elements wrt to the org-element-cache?
>
> The only "nogo" is to never modify (destructively) a value returned by
> `org-element-at-point' or `org-element-context'. Consider their return
> value as read-only, and possibly invalid as soon as you modify the
> buffer.

This one directly applies to my use-case (wrt `org-element-at-point'). 
Is it ok to do these two things:

 - let-bind a value returned by `org-element-at-point' and modify it (with
   plist-put), and 

   1. then return the interpreted value or

   2. do (delete-region ...) on old element and then insert the
      (interpreted) new one?

 - globally set a value returned by `org-element-at-point', modify it
   (with plist-put) in a function call, but use the variable as a quoted
   (!) function arg and reset the original value with this trick:
   
   #+begin_src emacs-lisp
     (setq X (org-element-at-point))

     (defun foo (element &optional replace-p)
       (let ((elem (eval element)))
         [...destructively modify elem...]
         [...interpret elem ...]
         (when replace-p
           [...delete old-element ...]
           [...insert interpreted elem, goto beg ...]
           (set element (org-element-at-point)))))

     (foo 'X t)
   #+end_src

> These advices don't apply to `org-element-parse-buffer', which doesn't
> use cache (if it did, it should copy cached elements beforehand anyway).

Ok

-- 
cheers,
Thorsten

Re: Programming with org-element-cache -> short introduction?

[Nicolas Goaziou]

Thorsten Jolitz <tjolitz@gmail.com> writes:

> Is it ok to do these two things:
>
>  - let-bind a value returned by `org-element-at-point' and modify it (with
>    plist-put), and 

No, per advice given before. `plist-put' is destructive.

Instead, create a new element. You can use `org-combine-plists' to
generate its property list.


[...]

>  - globally set a value returned by `org-element-at-point', modify it
>    (with plist-put) in a function call, but use the variable as a quoted
>    (!) function arg and reset the original value with this trick:

Ditto.