Re: [Xmlunit-general] Schema Validation patch development

[James A. <jam...@gm...>]

On 05/01/07, Stefan Bodewig <bo...@ap...> wrote:
> On Tue, 2 Jan 2007, James Abley <jam...@gm...> wrote:
> > On 22/12/06, Stefan Bodewig <bo...@ap...> wrote:
> >> On Thu, 21 Dec 2006, James Abley <jam...@gm...> wrote:
> >>
> >> > I'm writing a patch for consideration that will handle schema
> >> > validation.
> >>
> >> [please use CVS diff to create patches, it is easier to see what
> >> has changed that way]
> >
> > Will do - I'm just TDDing the feature at the moment and wanted to
> > give a little context.
>
> That's fine.  Thank you.
>
> >> > In developing the tests for this patch though, I seem to have hit
> >> > a bug, and was hoping someone could help me see what I'm doing
> >> > wrong.  The below tests fail, although I don't think they
> >> > should. I can't see what properties I should be setting that
> >> > would cause it to pass.
> >>
> >> >     private String expected = "<root
> >> >     xmlns='http://www.example.com/xmlns/example'" + "
> >> >         + xmlns:xsi='http://exampe.com/xmlns/not/schema-instance/namespace'/>";;
> >> >     private String actual =  "<root xmlns='http://www.example.com/xmlns/example' />";
> >>
> >> Why do you expect them to be different?  The
> >> http://exampe.com/xmlns/not/schema-instance/namespace is never
> >> used.
> >>
> >
> > Well, because it has the namespace declaration and to my current way
> > of thinking, that's just an attribute with special semantics.
>
> The way XMLUnit treats it it is not an attribute at all.  This has a
> few benefits like allowing to documents to be equal if they use the
> same namespaces but chose different prefixes for example.
>
> If you declare a namespace but never use it, this will never be seen
> by XMLUnit since (as you've seen) this doesn't affect the DOM tree at
> all.
>
> Do you think this is a problem?

It is a little annoying when I'm trying to write comprehensive tests.
I've got around it by adding a flag to XMLUnit that prevents any xmlns
attributes from being ignored, but I would probably want to do that in
a more robust fashion as you suggest. If namespaces are being
compared, then  (probably) the namespace name is the important thing
to compare, rather than the prefixes. I'll write some tests around
this area and see how it goes (in a separate patch though).

>
> > I just realised that the mails on this list don't have the
> > Reply-To header set to the mailing list, hence the reason I've been
> > replying to individuals rather than the list. Might it be possible
> > to alter that?
>
> I'll note that setting ot not setting Reply-To headers is the subject
> of heated discussions (just do a search for "reply-to considered
> harmful" with your favorite search engine).  Fortunately I use a MUA
> that deals with either way so I'm pretty indifferent.
>
> I don't have karma to change that setting, though.  Jeff?

I'm using GMail for this list at the moment. I'll have a look for
Greasemonkey stuff to work around my habit of just hitting reply to
list emails, otherwise, I'll just have to be more careful!

>
> Cheers
>
>         Stefan
>

An initial draft of the patch is done; I'm just writing more tests
around some serialization, so I'll create a patch on sourceforge
hopefully this weekend.

Cheers,

James