[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: Type-assignment in one pass

• From: Eric van der Vlist <vdv@d...>
• To: "K.Kawaguchi" <k-kawa@b...>
• Date: Fri, 02 Feb 2001 09:32:40 +0100

"K.Kawaguchi" wrote:
> 
[Detailed and interesting example and algorithm skipped]

This is getting even more interesting.

Your example makes it clear that TREX pattern is a specialized syntax to
express rules.

This is good news since we should be able to easily couple TREX with
"generic rules".

It's also showing more similarities that I would have expected between
TREX and XSLT even if there is a major difference: XSLT only evaluates
the "matches" to select the best template and will never select another
one for the same context node while a validation process using TREX
needs to evaluate "matches" for every nodes and try other matches if one
fails.

This is making TREX more similar to Prolog than to XSLT since in Prolog,
you can "fail" at any level of any statement and the process will
continue by selecting the next possible choice of the last test.

This similarity between TREX and rules based language might be used to
solve the issue of ambiguity like other rule based languages have done.

What we would need to do would "only" be to specify the behavior to
remove the ambiguity.

The possible algorithm would include a priority mechanism (ala XLST)
based on a hierarchy between datatypes and document order.

If we take the (modified) example from James Clark:

<choice>
  <element name="e">
    <attribute name="a">
      <data type="xsd:string"/>
    </attribute>
  </element>
  <element name="e">
    <attribute name="a">
      <data type="xsd:integer"/>
    </attribute>
  </element>
</choice>

and the following instance

<e a="7"/>

A system based only on the document order would evaluate the first
choice, see that it's OK and stop there, reporting the datatype as
string.

A system based on a hierarchy between datatypes would evaluate both, see
that integer is more restrictive than string and that both are matching
and report the datatype as an integer.

A combination of both with an attribute to override the priority like
XSLT does it is also possible.

If both case, if the algorithm is documented, there is no ambiguity any
longer...

Eric
-- 
------------------------------------------------------------------------
Eric van der Vlist       Dyomedea                    http://dyomedea.com
http://xmlfr.org         http://4xt.org              http://ducotede.com
------------------------------------------------------------------------

• Follow-Ups:
  • Re: Type-assignment in one pass
    • From: "K.Kawaguchi" <k-kawa@b...>

• References:
  • Re: Are we losing out because of grammars?
    • From: James Clark <jjc@j...>
  • Re: Are we losing out because of grammars?
    • From: Norman Walsh <ndw@n...>
  • Type-assignment in one pass
    • From: "K.Kawaguchi" <k-kawa@b...>

• Prev by Date: Re: XML versus Relational Database
• Next by Date: Re: Are we losing out because of grammars?
• Previous by thread: Type-assignment in one pass
• Next by thread: Re: Type-assignment in one pass
• Index(es):
  • Date
  • Thread