Re: [PHP5-DEV] Replacing expat with libxml

Shane,

As for BC: again, I agree 100%.  If it breaks BC in anyway, other than a
few error codes/error messages changing, then we can fix it, or revert
back to expat.  I've tested it with pear and pres2 (including my xml and
php slides), and everything seems to be in working order.  However, once
I commit it, other people can start testing.

PHP5 is going to be unstable, things will break, but we will have a very
long QA period, and a lot of people testing their apps against it to
make sure that there are no BC breaks.  Couple that with the fact that
libxml2 tries to be expat "compliant," I'm pretty confident we can
squash any and all BC breaks before a PHP5 release.

As for the other features, well, that's a discussion for a different
thread.  My main concern in this one is just switching the underlying
library, that way we're shipping with a robust XML library bundled as
default.  This will allow extension developers (hopefully) to work on
more XML related technologies in extension space.

-Sterling

On Mon, 2003-04-28 at 08:01, Shane Caraveo wrote:
> > After some discussions with various people at the PHP-Con, I decided it
> > was important that we (at least) have libxml integrated with PHP by
> > PHP5.  When it comes to XML processing, expat is a legacy library and it
> > doesn't support nearly what is required for processing XML by todays
> > standards.
> 
> > The current version of expat makes processing SOAP documents (for
> > example) very hard, because XML schema is not available.
> 
> I'd like to talk with you about what features will be implemented in 
> time for php5, as there are some specific features that would be good 
> for soap (ie. pull parsing).  I'm at the airport now, so later this week.
> 
> > - FTP and HTTP transports (as well as an IO wrapper library like
> > Streams)
> 
> Can the wrapper be made to work with the php streams?  It would be nice 
> for instance, to simply be able to pass php://stdin to the library and 
> have it handle input.  A call back to handle protocol headers, or 
> alternate encodings such as dime would be necessary.  Anyway, those are 
> some things I'd like to discuss.
> 
> > I've currently completed the first two steps of the integration.  I've
> > removed the expat library from ext/xml and replaced it with libxml. 
> > I've also ported the XML extension to use libxml as the underlying
> > processing engine.
> > 
> > The following incompatibilities exist:
> > 
> > 1) some XML_ERROR_ * constants are irrelevant (they are stilled defined,
> > but they have no meaning for libxml).
> 
> Are they mappable in any way, or simply library specific errors?
> 
> > 2) xml_error_get_string() just returns a blank string.  This can be
> > changed in the next couple of days, i just need to implement error
> > strings ontop of libxml error codes.
> > 
> > Having this library bundled internally will allow people to develop
> > other extensions which use libxml features, and will allow for future
> > extensions (for example, a fully compliant DOM extension) to easily be
> > added, without requiring extra bundling.  
> > 
> > Thoughts?
> 
> My concern is of course BC.  I think a little bit of breakage is ok, but 
> by little I'm thinking < 1%.  Otherwise, there needs to be some way to 
> load the expat xml extension.  Perhaps the expat extension functions 
> could be renamed to expat_xml_*, libxml would be libxml_*, and a utility 
> function such as xml_use_library(XML_EXPAT) would map xml_* functions to 
> the xpat library (or libxml).  The default mapping would be to libxml.
> 
> Shane
-- 
"Whether you think you can or think you can't -- you are right." 
    - Henry Ford