Hi,
I recently found a bug in reLyX that I think is related to th infamous
underscore bug in labels: reLyX simply crash if you have two inline maths
not separed by any character, like $a$$b$ (wich is somewat common in .tex
files generated by scientific workplace). Latex understands this as two
inline maths, but reLyX tries to understand $$ as \[.
The problem seems to be that reLyX has the "Cleaning..." phase, in which
it scans the .tex file replacing latex tokens by "canonical equivalents",
so, for example, $$ -> \[ and a_b -> a_{b}, no matter the context these
tokens appear.
Sure enough, these "canonical substitutions" must be performed, but making
them in a separate phase before parsing (as relyx does) seems to be a
totaly wrong way. It looks that "cleaning" and parsing should be
integrated, but maybe this would involve a lot of reLyX rewriting.
Considering that now there is support for regular expressions in C++,
complete rewrite of reLyX in C++ is also a good option. Is there someone
working on this? Or at least planning to work? If so, I would like to
participate in such a project! I thik it would be a long term project,
but as long as we have a good latex parser, it realy would worth, and if
it must be done, the sooner we start, the better.
Regards,
João.
