Re: multitable widths problem

[Werner LEMBERG]

>> Regardless of improvements, the documentation should be improved so
>> that people like me are not suprised by unexpected results of
>> `@columnfractions`.
>
> I added a note to the manual.

Thanks!

> I think you can already do something with a macro [...]  which does
> work (although you never know with macros and conditionals).  This
> approach would allow you to mix fraction values and prototypes for
> TeX output only.  Perhaps your use case is rare enough that this is
> tolerable?

Alas, it's not really tolerable, since we have multitables
everywhere.  Our problem is that in most of those tables at least one
column contains PDFs with LilyPond graphics, for example

  @item
  @code{\clef G}
    @tab
    @lilypond[line-width=3\cm,notime,ragged-right,relative=1]
    \clef G c1
    @end lilypond
  ...

Those 'lilypond' environments are handled by the `lilypond-book`
script, which processes the LilyPond code and replaces it with PDF
files for TeX output and PNG images for everything else:

  @item
  @code{\clef G}
    @tab
    @noindent
    @ifinfo
    @image{lilypond/3f/lily-2af16f63,,,[image of music],}
    @end ifinfo
    @html
    <p>
     <a href="3f/lily-2af16f63.ly">
      <img align="middle"
           border="0"
           src="3f/lily-2af16f63.png"
           alt="[image of music]">
     </a>
    </p>
    @end html

    @iftex
    @include 3f/lily-2af16f63-systems.texi
    @end iftex

  ...

File `3f/lily-2af16f63-systems.texi` looks like this:

  @image{./3f/lily-2af16f63-1}

In other words, we can't get the real width of the PDFs easily within
the texinfo framework.  A variant of `@columnfractions` that accepts
both fractions and sample strings would be thus ideal.


    Werner