Hi Sandy,
It's Great that you have lots of changes in mind & it's good that they are
coming out but in bits & pieces.Had we known about it before, we would have
saved time.
I feel, we can save coding time if everyone knows about complete
design/framework in advance.
Sandy, I just received your mail of SimpleType redesign... Thx
I will look into it tomorrow.
Cheers
Gopal
-----------------------
SUN Microsystems, India
> Hi Rahul,
>
> Thanks for looking into the XSDGroupTraverser and
> XSDAttributeGroupTraverser.
>
> Though you code basically looks fine to us, there are some points that I
> want to share with you before you proceed.
>
> 1. How attribute information will be stored in Xerces2 schema grammar.
>
> As the schema spec states, globally declared attributes result in
> "attribute declaration" in the grammar, while locally declared ones result
> in "attribute use" in the grammar. But in Xerces1, we don't distinguish
> these two concepts. So there were lots of redundant information in Xerces1
> about attribute declaration.
>
> In Xerces2, we plan to separate these two concepts. (Similar to the
> separation between element declaration and particle.) I'll add a new
> XSAttributeUse class in the v2 branch, which corresponds to the "attribute
> use" concept. And inside each XSAttributeUse instance, there will be a
> reference to an XSAttributeDecl (by index, not object reference).
>
> 2. How attribute group information will be stored in Xerces2 schema
> grammar.
>
> An attribute group contains two parts: attribute uses and attribute
> wildcard. I'm thinking of introducing an XSAttributeGroupDecl class to hold
> such information. Do you have any other suggestion, or any opinion on how
> to do it?
>
> 3. What should be returned from the traversal methods of
> XSDAttributeTraverser and XSDAttributeGroupTraverser.
>
> XSDAttributeTraverser.traverseGlobal: register an XSAttributeDecl, and
> return the index;
> XSDAttributeTraverser.traverseLocal: register an XSAttributeUse, and return
> the index;
> XSDAttributeGroupTraverser.traverseGlobal: register an
> XSAttributeGroupDecl, and return the index;
> XSDAttributeGroupTraverser.traverseLocal: resolve the "ref" attribute to an
> XSAttributeGroupDecl index, and return such index.
>
> What do you think?
>
> 4. What's the responsibility of XSAttributeChecker, and what's the
> responsibility of the traversers.
>
> As you noticed (and you already did), whenever we traverse an element in
> the schema document, we call XSAttributeChecker.checkAttributes to get the
> attribute values. This method is responsible for checking all attribute
> (and attribute value) errors except one case: it won't report an error if a
> required attribute is missing. The reasons are:
> a. If it's a required attribute, the caller will definitely need it and
> check whether it's present, so we don't need to duplicate the same effort
> in XSAttributeChecker;
> b. In some case, we don't know whether an attribute is required. For
> example, attribute "base" of <restriction> is required when it's in a
> complexType, but optional for a simpleType. So XSAttributeChecker doesn't
> know whether it's an error if "base" is missing.
>
> No it's clear that the caller (traverse methods) is only responsible for
> reporting errors about missing required attributes. But in your code, you
> also check for attributes that's not allowed for a certain element, which
> is not necessary. (As an example, "name" on local group/attributeGroup)
>
> 5. Some coding habit
>
> While this is completely up to you, just some thoughts:
> - In Xerces-J, we seldom use '_' in variable names. Andy Clark had an
> outline of coding convention in Xerces-J, which might be helpful to you.
> - In both XSDGroupTraverser and XSDAttributeGroupTraverser classes, both
> traverseLocal and traverseGlobal invoke a common method
> "traverseGroupDecl/traverseAttributeGroupDecl". But there is almost no code
> in common between the global and local cases. Do you have any reason doing
> so? I think we can get rid of the traverse???Decl method, and just put the
> right code into each of the appropriate methods. In fact, Lisa Martin is
> working on removing traverseGroupDecl from the group traverser.
>
> I'm so glad to see you guys actively involving in Xerces2 developing and
> design. And you further code contributions and design opinions are very
> welcome.
>
> Thanks,
> Sandy Gao
> Software Developer, IBM Canada
> (1-416) 448-3255
> [EMAIL PROTECTED]
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
