MathML-content with mtext afer postprocessingPostprocessing

[kohlhase]

[Originally Ticket 1393]

sometimes postprocessing and goes quite wrong, I get the following thing in http://arxmliv.kwarc.info/files/quant-ph/papers/0506062/determinism.xhtml

It seems that all the formulae in this paper have the same weird form.

```

P f , G = ∏ i ∈ O c > M i α i ⁢ K G ⁢ ( f ⁢ ( i ) ) s i ⁢ Z i s i ⁢ E G ⁢ N I c 0 PfG=∏>∈iOc⁢MiαiK⁢G⁢fisiZisiEGNIc0

```

[brucemiller]

“It seems that all the formulae in this paper have the same weird form. "
basically because all the TeX markup has the same “weird form”.
Namely, it looks like all displayed formula are typeset using either
eqnarray or an array with a dozen columns (most of which end up empty).

The immediate cause of what you’re seeing is that mtext is the fallback
for cmml conversion when latexml doesn’t know what else to generate.
When I wrote it, it wasn’t 100% clear that it was an error, but perhaps
merror would be more appropriate?

At the next level, I took some trouble to make at least a little sense
out of eqnarrays, but that leads to a dilemma: An eqnarray (or alignments of
equations in general) is a single presentational object (either an mtable in mml,
or a table at the html level containing mml), but contains several content
objects (ie. the equations, themselves, if I could heuristicate them).
I can, and do, markup such a beast in latexml, but there’s no direct
correspondence to an mml semantics structure, which is what you’re trying
to generate here. In other words, it’s not clear what “parallel markup”
should be in this case. Any suggestions?

Knowing what the author intended with the array markup, in principle
one could generate structures similar to eqnarray, but such a heuristic
certainly doesn’t apply to math arrays in general.

Finally, the author is using rather specialized notations, bra-ket etc.
One could do some interesting semantics here, but it needs human intervention.

In summary, there are some information architecture issues here
about what the desired representations should be. But, it also
seems unrealistic to expect automatic content extraction from
documents like this.

[brucemiller]

In the meantime, I have extended the grammar to handle some
quantum mechanical notations.

However, there are 2 basic issues:
- content markup (cmml, OM) cannot even realistically be attempted if
latexml has failed to parse the formula.
- even then, there are ambiguities that need to be resolved
(esp scripts) before a useful content can be generated.
Since the offending document fails on the first issue,
and may not even be reasonable to interpret, I’ll call this one “invalid”

Sorry, trac is bugging me to close tickets!