Le 28/11/2010 18:47, Heiko Oberdiek a écrit :
On Sun, Nov 28, 2010 at 04:51:55PM +0100, Paul Isambert wrote:
As you may know, PDF reads strings encoded in either its own scheme,
or in UTF-16.
My problem is I want accented characters in bookmarks, e.g:
\pdfoutline goto name {there}{Héhé}
I can't feed PDF-encoded strings directly (LuaTeX would complain),
but PDF also understands \nnn-encoded characters, where \nnn is a
number in base-8, so I actually do
\pdfoutline goto name {there}{\octal{Héhé}}
where \octal calls Lua to convert each character into \nnn, so that
LuaTeX actually reads (the backslash has catcode 12):
\pdfoutline goto name {there}{\110\351\150\351}
which works fine, except the number are Unicode, which the PDF
encoding doesn't follow exactly, so that I also need a mapping from
Unicode to PDF encoding, and then octal.
If hyperref is loaded, you can use \pdfstringdef.
Otherwise package `stringenc' might help, PDFDocEncoding
and UTF-16 are supported. But I think I haven't
implemented big char support (chars with character code> 255
for LuaTeX and XeTeX) yet (it's done in hyperref). If this is
true, then the package might still help you at some conversion
steps.
I do not load hyperref; as for stringenc, I might have used it, except
it doesn't work. strings more than one letter produce the non-utf8
character error (even if they don't contain any), as in:
\StringEncodingConvert\outlinetitle{aa}{utf8}{utf16}
and if I try with just one letter, bookmarks show nothing:
\StringEncodingConvert\outlinetitle{a}{utf8}{utf16}
\pdfoutline goto name {there}{\outlinetitle}
The other solution is UTF-16BE, which can't be fed directly to
LuaTeX (you can't use \nnn to represent bytes). I.e you can't have:
\pdfoutline goto name {there}{\sixteen{Héhé}}
where \sixteen would return a string encoded in UTF-16BE, because
LuaTeX will complain about non-utf-8 characters. You can do that
directly in Lua, though, but there's no pdf.outline primitive:
pdf.outline(to_utf16("Héhé"))
You can still use \nnn, because this notation is restricted
to plain ASCII ('\', '0', ... '9'). Of course the \nnn are
used for bytes only.
Meaning \nnn can denote bytes? [... TeXing ... TeXing ...] Oh yes! I
thought \nnn only denoted characters. So the solution is really simple
and works! Thanks Heiko.
Best,
Paul
