Re: Chinese text in plots

[Allin C. <cot...@wf...>]

On Sat, 24 Jul 2010, sfeam (Ethan Merritt) wrote:

> On Tuesday 20 July 2010, Allin Cottrell wrote:
> > On Tue, 20 Jul 2010, Allin Cottrell wrote:
> >
> > > [W]hat if you encoded Chinese strings in a plot file as CP950? I
> > > notice that the symbol CHINESEBIG5_CHARSET (=136) seems to be
> > > defined in connection with EMF files. Gnuplot could support that
> > > quite easily, in the minimal sense of offering a "cp950"
> > > encoding option and, in case that is in force, setting the
> > > charset byte in EMF output to CHINESEBIG5_CHARSET.
> >
> > This seems to work. I made an EMF file with this recipe, encoding
> > to cp950 and using a Big5 font:
> >
> > set term emf font "AR PL Mingti2L Big5,12"
> > set encoding cp950
>
> That's really interesting.  I had interpreted the EMF documentation
> differently; wherever it refers to "unicode", I had understood that
> to mean "any multibyte encoding" and thus concluded that the wide
> character versions of all the text routines were necessary.
> But hey, if it works - that's great.

The docs are not very clear, but I did come across some references
to double-byte characters in the context of ExtTextOutA. I think
that ExtTextOutW is strictly for use with UTF-16LE.

> > The Chinese characters display correctly using the freeware EMF
> > viewer eview.exe under Wine. (Unfortunately the EMF crashes
> > OpenOffice on Linux; I'm not sure whose fault that is.)
>
> Seems only fair.  Until recently the EMF output was interpreted
> fine under linux but crashed the Windows viewer :-)
> Could you send me a sample EMF file that OpenOffice has troubles
> with?  I have a couple of other EMF viewers that I run under linux
> and I'm curious what they think about it.

Sure, I'm attaching a sample. I tried it with a different Windows
viewer (irfanview, under wine) and again it was OK. I've also
inspected the file with the program "printemf" (from the libEMF
package) and I can't see anything wrong with it.

> > In case it's of interest I'm attaching my patch.
>
> Do you know if the various other terminal types do anything
> reasonable with CP950 fonts?  That is, should we add this to the
> list of supported encodings in the "set encoding" command?

Hmm, I'm not aware of another terminal type that could handle
Big5-encoded input, but then my knowledge is mostly restricted to
png, pdf, postscript and emf.

Allin Cottrell