Also record labels as metadata for longtables, when requested in lxRDFa

[dginev]

This PR is a follow-up to #596 and I would appreciate a quick review & merge, if possible.

It turns out that my generic recording of \label as metadata did not extend to packages redefining it, such as longtable.

The problem there is that longtable redefines (via Let) the \label macro on each environment start, which makes it impossible to maintain the RDFa hook I introduced globally in the preamble. However, I can hook into the lower level \@longtable@label macro, and it all works great again!

Example test being:

\documentclass{article}
\usepackage{longtable}
\usepackage[labels]{lxRDFa}
\begin{document}

\begin{longtable}[]{@{}lccc@{}}
1 & 2 & 3 & 4 \\

\caption{\label{table:long} caption}
\end{longtable}

I should be able to refer to Table \ref{table:long}.

\end{document}

with the expected snippet of metadata produced:

<figcaption class="ltx_caption" property="dcterms:alternative" content="table:long">

Tested locally and all works well.

[brucemiller]

Hmm... I guess the original option "labels" was a patch from you, and I accepted it? But...

It kinda seems the wrong way to go; since we'd conceivably need to mung every package that plays around with \label. Shouldn't it be more correct to use Tag (or something similar) to check for any labels attributes, and add RDF when appropriate?

[dginev]

But then I would need to add that rule to all heading-like tags? Or attach it to * which would hurt performance badly.

I double-checked the bindings, i think this is the only case one needs to make a special exception. All other table packages, such as tabularx or tabulary keep using the standard label.

Happy to backtrack when it becomes unmanageable (which I hope will never happen)?

[brucemiller]

Have you actually tried whether it really hurts performance? I'd bet it's not as bas as you think (but might be).

[dginev]

I mean I have a perfectly safe 3-line PR that does everything one needs with no issue...

[dginev]

Maybe a compromise - if we merge here as-is, I promise to rework this using Tag the next time I find a missing label macro that isn't supported.

[brucemiller]

On 08/24/2017 10:32 AM, Deyan Ginev wrote: Maybe a compromise - if we merge here as-is, I promise to rework this using |Tag| the next time I find a missing label macro that isn't supported.

Better trade: I merge, you swallow a few "i told you so"'s next time I start whining about XS ?

[dginev]

XS? I love XS! I'm sure it will work out in the end :) Thanks for merging 👍

[brucemiller]

OK, I'm convinced!