Den 2019-03-15 kl. 13:51, skrev BPJ:
Den 2019-03-15 kl. 12:31, skrev Zdenek Wagner:
I am also interested how you do it. I have tried with one of my
documents (I do not need this conversion, it was just a test). The
document contains 5 tables and 50 math equations. The first
equation
is OK, the remainng equations are total garbage, they will have
to be
entered manually from scratch. The tables are total garbage as
well,
they even do not look like tables. The table of contents is garbage
but this is not a major issue. The problem is that in the middle of
the first page, probably as an effect of math, the text becomes
garbage as well. In this situation copy&paste and manual conversion
will be faster unless there is a special (hidden) trick which I
do not
know.
As I said in my howto just posted your best bet if you have the
original LaTeX file is to redefine commands etc. in *TeX so that
the results become less garbagey and easier to correct by hand.
I don't know about math because I don't do math, so for me
not-so-simple tables are the biggest problem. If you or anyone
else comes up with a *TeX hack which makes column boundaries
"visible",
as in inserting pipe characters or some such, it will be much
easier to tidy things up after conversion to a text format with
Pandoc.
You may also want to try Pandoc's direct LaTeX-->Anything
conversion, although it is rather lossy for more advanced stuff
it does lists, tables, small caps and surely math quite OK.
I only use this PDF-->DOCX trick for PDFs I get from my clients
where the source is not included or may not exist.
I'm still to encounter a client handing me a *TeX file... :-(
BTW you can "tame" recalcitrant LaTeX commands when converting
with Pandoc by including `\renewcommand`s restating them in terms
of simpler LaTeX constructs which Pandoc can handle and Pandoc
will use them. IIRC there is a feature request out (or I'll make
one!) for getting some/all LaTeX commands "unknown" to Pandoc as
Pandoc's native Div/Span syntax with `custom-style` attributes
which you then could hook into when converting to DOCX with
Pandoc. It will probably become reality sooner than later. The
problem is how to handle commands/environments with multiple
arguments (which argument is "the" text?) You can already have
Pandoc preserve "unknown" LaTeX as raw LaTeX, which you then can
use a Pandoc filter (written in Lua, Python, Perl, your language
of choice) to massage into something suitable UTF-8 is no problem
as Pandoc uses it natively. It also understands standard
babel/polyglossia commands, giving you a native span or div with a
`lang` attribute which it then understands to handle correctly
when converting to other formats. There are some warts like
`\textgreek` giving `lang="el"` rather than "grc" but that can be
fixed with a Pandoc filter. DOCX's (lack of) math capabilities
may be another story though, but Pandoc surely does its best.
/bpj
