Re: [Po4a-devel]Sgml and generic file inclusion
On Fri, May 27, 2005 at 01:43:38PM +0200, Nicolas François wrote:
Hello,
With Martin having no time, but spending a lot of time on po4a, it seems
the right moment to ask this.
Arf, crétin ;)
You're right, I should speak less of working for real life and do it more :-D
First, extracting the file inclusion part from the TeX module to put
it in
the Transtractor works.
It has mostly one drawback: the file inclusion has to be done in one line.
For example:
.so foo.1
or:
\include{foo.tex}\input{bar}
should work, but:
<!ENTITY foo SYSTEM
foo_foo_foo_foo_foo_foo_foo_foo_foo_foo_foo_foo_foo_foo.sgml>
will not work (at least currently; if it is needed, it will work)
Actually, it shouldn't be an issue. At least not for ?ml since it's not the
lines you selected which should be changed to the file content, but &foo;
To use this file inclusion mechanism, a module will have to implement
one
function, which will find the filename and return the piece of text to put
before and after the inclusion of the given file. (it is a littel bit more
complicated, but basically works this way).
For example, the Man module could return "", "foo.1", ""
when it receive
".so foo.1"
Funny actually. I'd have done this the other way with a unshift_file()
function provided by transtractor, and all modules using it. Something like
$macro{'so'}=sub {
my $self=shift;
$self->unshiftfile($_[1]);
};
sub unshiftfile {
my $self=shift;
my $filename=shift
or croak wrap_msg(dgettext("po4a", "Can't read from unshift_file
without having a filename"));
my $linenum=0;
my @array;
# Get all the file into @array
open INPUT,"<$filename"
or croak wrap_msg(dgettext("po4a", "Can't read from %s: %s"),
$filename, $!);
while (defined (my $textline = <INPUT>)) {
$linenum++;
my $ref="$filename:$linenum";
my @entry=($textline,$ref);
push @array, @entry;
# Detect if this file has non-ascii characters
if($self->{TT}{ascii_input}) {
my $decoder = guess_encoding($textline);
if (!ref($decoder) or $decoder !~ /Encode::XS=/) {
# We have detected a non-ascii line
$self->{TT}{ascii_input} = 0;
# Save the reference for future error message
$self->{TT}{non_ascii_ref} ||= $ref;
}
}
}
close INPUT
or croak wrap_msg(dgettext("po4a", "Can't close %s after reading:
%s"), $filename, $!);
# unshift all the file content (in reverse order so that endoffile is unshifted
first)
map {$self->unshiftline(@_[0],@_[1])} reverse @array;
}
But I may have understand wrong what you mean?
Other facilities are also provided (for example, the TeX module needs
find
a file wether the .tex extension is given or not).
So here are some questions regarding the Sgml module:
Is there any problem with the current implementation of the inclusion?
No, as long as nsgml keeps on our way, I don't think we should touch at sgml
file inclusion (which is already implemented despite the complexity induced
by nsgml cruft).
The Sgml module is quite different from the other modules (the
lines are
not shifted one by one, but the file is given in its whole to nsgmls. So
is this module a good choice for testing the file inclusion? Maybe I can
try to implement it in order to be able to use the whole file.
Is there another module which could need inclusion, and could be used to
test if this mechanism is sufficient and generic enough
Man and Xml are good candidate for this feature, IMHO.
I will submit the patch to the list, so you can have a look at it.
But I'm
a little bit reluctant at committing it before it is tested by at least
two modules.
When have a plenty of time, I think.
Bye, Mt.
Attachments:
- signature.asc (application/pgp-signature — 189 bytes)