Re: [O] RFC: inheritance of export attributes

[Aaron Ecay]

Hi Nicolas,

2013ko martxoak 21an, Nicolas Goaziou-ek idatzi zuen:
> Aaron Ecay <address@hidden> writes:
> 
>> I have noticed a difficulty with :results table drawer babel blocks.  It
>> isn’t possible to put ATTR_LATEX keywords on the table in that case.  If
>> they are placed outside of the drawer, they apply to the drawer and not
>> the table.  If they are placed inside it, they will be deleted when the
>> block is reevaluated.
> 
> What about evaluating it, inserting the attr_latex keyword, and then
> disabling evaluation for that block?

This isn’t a solution, generally speaking.  I may be working on a graph
in R; I know that I want it to be inserted in the LaTeX output with
width 6in, even as I go through several iterations of changing the
graph.  So, each time I do, either I have to manually re-enter the
width, or be resigned to the LaTeX export being “broken” until I am
done.  (and the same goes for a table, of course)

> I understand the interest for the problem at hand, but, generally
> speaking, I tend to think it could lead to confusion.
> 
> Attributes inheritance is but a hack used to parametrize inline images,
> until we agree on a proper link syntax including its own attributes.

Well, this is one specific current use.  But the fact that it is a hack
doesn’t mean the inheritance approach is inherently hackish.

> 
> Another way to solve the problem would be to let Babel generate
> attributes from source code with a specific keyword, e.g.:
> 
>   #+begin_src :results table :attr-latex ":align lll"
>   ...
>   #+end_src

This would lead to long begin_src lines, especially if one wants to
export to multiple backends.  I think it is much more pleasant to be
able to edit several lines:

#+ATTR_LaTeX: foo
#+ATTR_HTML: bar

rather than packing them all into one begin_src line.  (But I think I
could live with this as a solution, if I had to.)

What if attribute inheritance was only implemented for RESULTS drawers?
I guess that would look more like a “hack,” but it would avoid
interfering with other areas where attribute inheritance might cause
confusion.

Thanks,

-- 
Aaron Ecay