Re: org-fill-paragraph leaves point at end of table

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Carsten Dominik <carsten.dominik@gmail.com> writes:

> On 25 feb. 2013, at 10:10, Nicolas Goaziou <n.goaziou@gmail.com> wrote:
>
>> Hello,
>> 
>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>> 
>>> Calling `org-fill-paragraph' inside a table leaves point at the end of
>>> the table, for reasons that are totally unclear to me.
>>> 
>>> I've tested this with up-to-date org and emacs -Q, so I'm hoping it's
>>> reproducible. I edebugged org-fill-paragraph, and it appears to do the
>>> right thing, going from the save-excursion to the cond to the org-table
>>> cond statement, and there calling `org-table-align'. That works
>>> correctly, but stepping forward you come to the end of the enclosing
>>> `save-excursion', and emerging from `save-excursion' puts point at the
>>> end of the table -- precisely what it's not supposed to do!
>>> 
>>> I made a minimum sexp to reproduce the relevant bits of
>>> org-fill-paragraph:
>>> 
>>> #+BEGIN_SRC emacs-lisp
>>>  (save-excursion
>>>    (let ((element (org-element-at-point)))
>>>      (case (org-element-type element)
>>>        (table-row (org-table-align) t))))
>>> #+END_SRC
>>> 
>>> Putting point in a table and eval'ing that also leaves point at the end
>>> of the table. I tried using (call-interactively 'org-table-align) and it
>>> did the same thing.
>>> 
>>> I'm baffled, particularly as it doesn't do this for any other element
>>> type. Any clever ideas? M-q after a bit of typing is already stuck in my
>>> fingers, and this bit of strangeness doesn't set the mark, so editing
>>> long tables is a pain...
>> 
>> `org-table-align' inserts a whole new table and removes completely the
>> previous one. This confuses `save-excursion' which doesn't recognize any
>> familiar location anymore.
>> 
>> I've pushed a fix for it. All filling tests pass, but if you notice
>> anything suspicious, please signal it.
>
> I do not expect problems, because you now fall back onto
> org-table-align without any save-excursion around it, AFAICS.
> org-table-align remembers line number and table row number and
> restores them, so this should work.

And sure enough, it does!

Sorry for all the table-related stuff. I'm in one of those moments where
you throw caution to the wind and entrust a time-sensitive company
project to areas of git-master Orgmode you've never used before. Likely
very stupid, but it's too late to turn back now.

Thanks!
E