> Note that Jan-Ake Larsson stated that supporting postscript specials
> means embedding a postscript interpreter into dvipng, and he has no
> time for that. So, I doubt that you are on the right track.
I read that post too, but what I'm saying is not to make dvipng actually work
but just detect if it would.
DVI, as you'll know, is a sort of assembly language (but not Turing-complete,
more info [1]) made of directives (identified by opcodes). Now, opcodes from 239
to 242 [2] identifies specials, dvipng can interpret several of these (e.g. GIF,
PNG, EPS image inclusion) but not all of them, in particular doesn't work well
with PostScript literals. PostScript literals are identified by the "ps:"
prefix, so if we find in a DVI file a sequence of one of those special opcodes
followed by "ps:", we'll know that dvipng most probably will fail.
To be truth not all postscript literals are unsupported, but I think that using
the legacy method when postscript is needed is the safest solution. Math,
XY-pic, image rendering and so on will work without problems in dvipng. Moreover
this solution is extremly specific for each preview, instead the heuristic way
would make all the preview in the same document work with the legacy method, as
the "\usepackage{pstricks}" directive goes in the preamble which is included in
all the tex preview code.
I wrote to the author of dvipng (Jan-Ake Larsson [jan-ake.lars...@liu.se]) and
even him thinks this is the best solution:
> The imminent 1.14 release will also have a switch --norawps that will
> completely disable trying to render all raw PS specials, even those that
> dvipng (thinks it) can handle.
We can't wait for 1.14 diffusion, and in any case some processing would get done
in any case (e.g. a very long code without PS and a last PS special), so in the
next days I'm going to write the python code to detect PS specials. It should be
pretty straightforward, as DVI is a simple format.
What do you think?
venom00
P.S. Sorry for the multiple e-mails, but I've been experiencing problems
[1] http://www-users.math.umd.edu/~asnowden/comp-cont/dvi.html
[2] http://www-users.math.umd.edu/~asnowden/comp-cont/dvi.html#xxx
