XML Schema translation w/ included schemas

elliottg · May 15th, 2008, 02:06 PM

Hello, new member here. I've searched the forum for questions related to this but haven't found much of use.
I have the unfortunate task of attempting to translate a large collection of XML Schema's to templates for further application processing. These schema's have several includes to other schema's (which in turn might include yet other schema's, with the unfortunate case of there being cyclic dependencies) and a sample primary schema might include up to 70 other schema's after all includes have been handled.
Since XML Schema's are themselves XML I figured XSLT's would be the natural way to go, especially since the template outputs are also pseudo-XML Schema's themselves, and certainly are XML-based. But it seems as though this might not be the case. The templates need to be one resulting document that contains transformed elements referenced by elements in the primary schema to secondary schema's.
ie
primary schema

Code:

<xs:include schemaLocation="datatypes.xsd" />
<xs:element name="foo" type="barDataType" />

secondary schema

Code:

<xs:complexType name="barDataType">
  ...
</xs:complexType>

with the result hopefully being something like:

Code:

<template>
  <complexType name="barDataType">
    ...
  </complexType>
  ...
  <element name="foo" type="barDataType" />
</template>

or possibly:

Code:

<template>
  <element name="foo">
    <complexType>
      ...(transformed barDataType in-lined)
    </complexType>
  </element>
</template>

Is this possible (with one level of dependancy? with multiple? with cyclic dependancies?) If so, do I need a schema-aware processor (if so, what would I need to do to accomplish this) or can I somehow store some list of all possible includes to look through when I encounter a non-local/non-base datatype (or xsl:copy-of, or some other method I haven't considered)? Namespace awareness isn't an issue, as all schema's don't use a namespace. Or should I forget XSLT's for this and write some program to create a temporary (and massively large) schema that contains all necessary types locally and translate that with my other transformations?

Thanks in advance for any assistance.

Grant

mhkay · May 15th, 2008, 03:13 PM

Processing raw schema documents using XSLT or anything else is not an easy task, because there are so many different ways users can write effectively the same thing.

You might find it easier to start with an XML dump of a compiled schema, which you can get from Saxon (option -scmout on the com.saxonica.Validate command line) - this removes many of the syntactic variations and avoids the need to process multiple input documents.

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

elliottg · May 16th, 2008, 04:14 PM

Thanks for your reply.
I'm certainly not keen on the idea of translating all these schema's but unfortunately that's what I'm tasked to do so here I am.
One plus is the files are all generated and aren't hand-crafted and there doesn't seem to be any real complexity to the files themselves, just the dependencies.
I might have figured something out, but I might come back with more questions. Thanks.