RE: [xml-dev] patterns and restrictions

Your analysis is correct.  Pattern is indeed unusual in the manner you 
describe.  It's inclusion was debated quite a bit during the design of XML 
schema, because having constraints at two levels can complicate the use of 
XML schema for certain purpose.  For example, it's straightforward to 
validate a string using your "01" pattern for integer.  What's not so 
straightforward is to serialize an "int" stored in your program into a 
string that matches the pattern.  Few systems will (or should) even try, 
except perhaps for the simplest patterns. 

On balance, there were so many users who wanted to the capabilities 
offered by patterns that we decided to live with the added architectural 
complexity.  I still have some doubts that it was the right thing to do, 
but that's how we got to where we are. I think there are good arguments on 
both sides of the design question. 

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------








"Alessandro Triglia" <sandro@mclink.it>
Sent by: xmlschema-dev-request@w3.org
12/17/03 11:41 AM

 
        To:     "'David Tolpin'" <dvd@davidashen.net>, "[Public XML Schema-DEV]" 
<xmlschema-dev@w3.org>
        cc:     (bcc: Noah Mendelsohn/Cambridge/IBM)
        Subject:        RE: [xml-dev] patterns and restrictions





David Tolpin wrote:
> 
> [ Charset ISO-8859-1 unsupported, converting... ]
> > Hi,
> > 
> > Does the following allow 'abc' as a valid value?
> > 
> > <xs:simpleType name="MyDouble">
> >         <xs:restriction base="xs:double">
> >             <xs:pattern value="[^N].*"/>
> >         </xs:restriction>
> > </xs:simpleType>
> > 
> > We had been working under the belief that, via restriction 
> - patterns
> > from both the new datatype and the original one are 'And'ed 
> together. 
> > i.e - the above datatype would only allow values valid for 
> a double - 
> > with the exception of NaN.
> > 
> > However, we've read somewhere today that whenever a pattern facet is
> > evaluated - it is evaluated against a string.  Though 
> XMLSpy and Xerces 
> 
> The 'XML Schema Part 2: Datatypes' says
> 
> NOTE: It is a consequence of the schema representation 
> constraint Multiple patterns (?4.3.4.3) and of the rules for 
> ·restriction· that ·pattern· facets specified on the same 
> step in a type derivation are ORed together, while ·pattern· 
> facets specified on different steps of a type derivation are 
> ANDed together.
> 
> 'pattern' is a constraining facet, and a type defined by 
> applying a constraining facet to a primitive type is a derived type. 
> 
> Therefore, I would think that 'pattern' defines a subset of 
> the lexical space of the type it is applied to.


My understanding is that the "pattern" facet is a very special facet, in
that it constrains two different things:

1) the value space of the type;

2) the lexical representation of the values in the value space resulting
from 1, possibly beyond the extent implied by 1.

The text says that "pattern" constrains the value space "by constraining 
the
lexical space" (4.3.4), but also says that the literals must match the
pattern regexp (4.3.4.4).  Thus both (1) and (2) follow.

For example, in the very simple case of a pattern value "01" applied to
xsd:integer, the pattern facet has the following effects:

1) it restricts the value space of the datatype (derived from xsd:integer)
to include only the integer value 1;

2) it requires that the integer value 1 be lexically represented precisely
as "01", and not, say, as "1" or "001".

"Normal" facets (except "pattern") don't care about the lexical
representation.  Any lexical form that is legal for the values in the 
value
space may be used in instances, and will "represent" one of those values.
"Pattern" is different.

Alessandro Triglia


> 
> David Tolpin
> 
> -----------------------------------------------------------------
> The xml-dev list is sponsored by XML.org 
> <http://www.xml.org>, an initiative of OASIS 
<http://www.oasis-open.org>

The list archives are at http://lists.xml.org/archives/xml-dev/

To subscribe or unsubscribe from this list use the subscription
manager: <http://lists.xml.org/ob/adm.pl>

Received on Tuesday, 23 December 2003 10:57:10 UTC