Re: [gdome]Feedback on using gdome2 in C++ via gmetadom (and a problem with charsets)

[Enrico Zini <zinie cs unibo it>]

On Mon, Jun 24, 2002 at 12:03:03AM -0400, Luca Padovani wrote:

> >  DOMString lacks a constructor, and assignment operator that reads a
> >  std::string value, and an assignment operation for std::string that
> >  reads a DOMString value.  It is not essential, but it would be greatly
> >  appreciated.
> and it wouldn't be difficult at all. The only thing I'm concerned about is
> that using std::string inside DOMString.hh will cause the inclusion of a
> not-so-small .h file of the standard library into many DOM files, and
> consequently many user files.

No, you can avoid it: just use a forward declaration like this:

	namespace std {
		class string;
	};

And you should do fine #including <string> only in non-header sources.

Uhm... no, you cannot, since string is a damn typedef.  Uhm... is there
a way to forward declare a typedef?  After all, if you just use
references, the compiler has no use of knowing the structure size when
compiling the headers...


> The alternative possibility is to have helper functions (say in
> DOMStringAux), but they wouldn't be as natural as a constructor and an
> assignment operator.

I quite agree...


[double adding a node fails silently]
> I suspect it might be a gdome2 bug. I would suggest to write the smallest
> gdome2 program reproducing the problem, and posting the source on the
> mailing list.

Attached as test1.tar.gz


> [snip, about utf-8 encodings]
> Again, a small .xml file and a complete (small!) program with the problem
> woule be a great help. It might be a problem in the conversion
> UTF-8/UTF-16 done by gmetadom, or a low-level problem in libxml2, but I
> think the latter is less likely (libxml2 is quite well tested).

I tried to reproduce this with a program that just reads and saves an
XML file, but I had no luck.  I saw that Galeon stores accented chars
like this:

	<title>Universit&#195;&#160;</title>

...and gdome2 rewrites it like this:

	<title>Universit&#xC3;&#xA0;</title>

But this is not bad and I didn't get that warning.  Umpf.  I never alter
strings in the main application, I just move and copy elements around.
Tomorrow I'll try and do some more investigations...  It would help me
to know in what cases the libxml2 warning I've reported gets printed;
the warning was: "xmlEncodeEntitiesReentrant : char out of range".

Another curious note: when that error is given, the DOM tree is saved in
what seems to be latin1: accented chars are iso-latin1 accented chars
and other accented elements are as other binary data.  This seems to be
why libxml2 or libgdome2 complain when reloading those saved files.

Thanks!

				Bye, Enrico

--
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini <enrico debian org>

Attachment: test1.tar.gz
Description: Binary data