Resolved. Errata updated:

2.167. Section 23.5.3. Clarify that the dfdl:occursIndex() property must only be used within the context of an array (as defined by the glossary), and returns the index of the current occurrence of that array. It is a schema definition error if the function is used otherwise.

Regards

Steve Hanson
Architect, IBM DFDL
Co-Chair, OGF DFDL Working Group
IBM SWG, Hursley, UK
smh@uk.ibm.com
tel:+44-1962-815848

From: Mike Beckerle <mbeckerle.dfdl@gmail.com>
To: Tim Kimber/UK/IBM@IBMGB,
Cc: Steve Hanson/UK/IBM@IBMGB, "dfdl-wg@ogf.org" <dfdl-wg@ogf.org>
Date: 04/09/2014 18:28
Subject: Re: [DFDL-WG] Action 269: clarification on dfdl:occursIndex() function

I understand the principle you are suggesting, but this is a dfdl-specific function whose name shares concept with dfdl properties.

It feels very inconsistent to say that scalar elements disregard dfdl:occursCountKind and dfdl:occursCount, but at runtime they respond to dfdl:occursIndex(...) with a value.

My view: where we are using XPath functions directly, keeping their semantics makes some sense. Where we are introducing new dfdl functions I believe the semantics should be as strict and narrow and schema-aware as possible.

My bias comes from discovering the hard way that XPath is full of compromises that only are justifiable if you have to consider execution of the XPath against documents where there is no schema available.

In DFDL we're never in that situation, so we can simply do much better at helping users create correct expressions that do what is intended.

My favorite whipping post is fn:nilled(...). In XPath fn:nilled(...) always returns a value or empty node set, even if used in entirely non-sensical ways. Consider fn:nilled($x) where x is a DFDL variable. Note that DFDL variables can only be simple type values. They can't be nilled meaningfully ever. But fn:nilled says if called on a non-element value returns empty node list, so technically fn:nilled($x) is a valid expression.

I can accept this because fn:nilled is already documented as having this behavior. But it is a miserable semantics when you have schema information available that can inform you where calling fn:nilled is meaningful versus when it is simply always going to produce empty-node-list.

I'm reworking the daffodil expression language implementation to discard Saxon and do our own. I'm planning some flags for expression-strictness checking so that users can either have robust schema-aware semantics, or can insist on XPath's gooey semantics.

Mike Beckerle | OGF DFDL Workgroup Co-Chair | Tresys Technology | www.tresys.com
Please note: Contributions to the DFDL Workgroup's email discussions are subject to the OGF Intellectual Property Policy

On Wed, Sep 3, 2014 at 4:28 PM, Tim Kimber <KIMBERT@uk.ibm.com> wrote:
I think it would be better to allow dfdl:occursIndex() to be used on non-array elements. XML Schema and XPath do not distinguish between 'array' and 'non-array'. DFDL does, but only in the representation properties. I think the parts of the specification that build purely on XSDL and XPath should conform to the norms of those languages.

regards,

Tim Kimber,
Technical Lead for IBM Integration Bus Healthcare Pack
Hursley, UK
Internet: kimbert@uk.ibm.com
Tel. 01962-816742
Internal tel. 37246742

From: Mike Beckerle <mbeckerle.dfdl@gmail.com>
To: Steve Hanson/UK/IBM@IBMGB
Cc: "dfdl-wg@ogf.org" <dfdl-wg@ogf.org>, "dfdl-wg-bounces@ogf.org" <dfdl-wg-bounces@ogf.org>
Date: 03/09/2014 18:21
Subject: Re: [DFDL-WG] Action 269: clarification on dfdl:occursIndex() function
Sent by: dfdl-wg-bounces@ogf.org

I need some further clarification on this.

Current description says "This function may be used on non-Array elements." In the old no-arg version, that meant that you didn't have to hang this directly on the array, but could be on some contained child element.

I believe what we mean now is this:

dfdl:occursIndex(0) = the current element itself must be an array (SDE otherwise), return its occurs index.
dfdl:occursIndex(1) = the parent element must be an array (SDE otherwise), return its occurs index.
...
dfdl:occursIndex(N) = the nth-parent element must be an array (SDE otherwise) return its occurs index.

Specifically note that 0 is a valid argument (I think we just neglected this case in prior discussion).

The alternative to SDE if the path is incorrect here is to mask the error by returning 1. But I don't think this is helpful really. At least I can't think of a use case for why this particular polymorphism would be useful. An SDE is the more conservative choice.

Mike Beckerle | OGF DFDL Workgroup Co-Chair | Tresys Technology | www.tresys.com
Please note: Contributions to the DFDL Workgroup's email discussions are subject to the OGF Intellectual Property Policy

On Wed, Sep 3, 2014 at 11:27 AM, Steve Hanson <smh@uk.ibm.com> wrote:
Mark

Yes, that makes more sense. The existing wording says:

dfdl:occursIndex()

Returns the position of the current item within an array as an xs:long.

The first element is at position 1.

The function may be used on non-array elements.

So I blindly used xs:long for the argument type as well. On consideration the argument type should be xs:nonNegativeInteger and so should the return type.

Could be xs:positiveInteger, but as you say it's not used anywhere in the spec (no XPath constructor, etc).

Regards

Steve Hanson
Architect, IBM DFDL
Co-Chair, OGF DFDL Working Group
IBM SWG, Hursley, UK
smh@uk.ibm.com
tel:+44-1962-815848

From: Mark Frost/UK/IBM
To: Steve Hanson/UK/IBM@IBMGB,
Cc: Mike Beckerle <mbeckerle.dfdl@gmail.com>, "dfdl-wg@ogf.org" <dfdl-wg@ogf.org>, dfdl-wg-bounces@ogf.org
Date: 03/09/2014 13:35
Subject: Re: [DFDL-WG] Action 269: clarification on dfdl:occursIndex() function

Would a parameter type xs:positiveInteger (right value space) or xs:NonNegativeInteger (less wrong value space, and already used in DFDL) be more appropriate?

Regards,
Mark



Mark Frost		IBM United Kingdom
Software Engineer		Hursley Park
IBM DFDL, IBM Integration Bus		Winchester
		SO21 2JN
Phone:	+44 (0)1962 817009	England
e-mail:	frostmar@uk.ibm.com

From: Steve Hanson/UK/IBM@IBMGB
To: Mike Beckerle <mbeckerle.dfdl@gmail.com>
Cc: "dfdl-wg@ogf.org" <dfdl-wg@ogf.org>, dfdl-wg-bounces@ogf.org
Date: 03/09/2014 13:16
Subject: Re: [DFDL-WG] Action 269: clarification on dfdl:occursIndex() function
Sent by: dfdl-wg-bounces@ogf.org

Closed. After discussion, settled on an xs:long argument being a 1 based count up the parent axis. It is a processing error if the argument reaches beyond the root. It is a schema definition error if the argument is <= 0. Erratum 2.167 created. To be included in draft r23.

Regards

Steve Hanson
Architect, IBM DFDL
Co-Chair, OGF DFDL Working Group
IBM SWG, Hursley, UK
smh@uk.ibm.com
tel:+44-1962-815848

From: Mike Beckerle <mbeckerle.dfdl@gmail.com>
To: "dfdl-wg@ogf.org" <dfdl-wg@ogf.org>,
Date: 02/09/2014 15:21
Subject: Re: [DFDL-WG] Action 269: clarification on dfdl:occursIndex() function
Sent by: dfdl-wg-bounces@ogf.org

I am currently implementing functionality in the area where dfdl:occursIndex( path ) will be in the Daffodil implementation.

It doesn't look very hard to put in a specific argument that is restricted to be upward only and which identifies exactly which array.

This function is a special case in any DFDL runtime no matter what, and will require a fair bit of test attention for any implementation so making it simpler by allowing only the innermost array to be examined isn't saving a lot of work really.

Question: is a crazy path like ../foo/../../bar/.././.. (which just so happens to be equivalent to ../../..), a schema definition error?

Mike Beckerle | OGF DFDL Workgroup Co-Chair | Tresys Technology | www.tresys.com
Please note: Contributions to the DFDL Workgroup's email discussions are subject to the OGF Intellectual Property Policy

On Tue, Sep 2, 2014 at 9:28 AM, Mike Beckerle <mbeckerle@tresys.com> wrote:

________________________________
From: Steve Hanson [smh@uk.ibm.com]
Sent: Monday, September 01, 2014 11:08 AM
To: Mike Beckerle
Subject: Fw: [DFDL-WG] Action 269: clarification on dfdl:occursIndex() function

269
dfdl:occursIndex() function (Steve)
29/7: Clarify behaviour and decide whether an argument is needed to make it context independent. Noted that fn:position() also does not take an argument, but it returns position in current sequence and not position in array.
4/8: With Steve but consensus is that an argument is needed.
26/8: Mike to review. If an argument is needed, we should make sure this is in next published spec as the function is used in MIL-STD-2045 schemas.

Regards

Steve Hanson
Architect, IBM DFDL<http://www.ibm.com/developerworks/library/se-dfdl/index.html>
Co-Chair, OGF DFDL Working Group<http://www.ogf.org/dfdl/>
IBM SWG, Hursley, UK
smh@uk.ibm.com<mailto:smh@uk.ibm.com>
tel:+44-1962-815848
----- Forwarded by Steve Hanson/UK/IBM on 01/09/2014 16:03 -----

From: Steve Hanson/UK/IBM
To: dfdl-wg@ogf.org,
Cc: dfdl-wg-bounces@ogf.org
Date: 31/07/2014 10:57
Subject: Re: [DFDL-WG] clarification on dfdl:occursIndex() function

________________________________

Note: The reason why fn:position() takes no argument is because it gives position in the current sequence. Which is also why we have a separate DFDL specific function dfdl:occursIndex() which is intended for giving position in enclosing array. I think we need an argument that targets the specific enclosing array.

Regards

Steve Hanson
Architect, IBM DFDL<http://www.ibm.com/developerworks/library/se-dfdl/index.html>
Co-Chair, OGF DFDL Working Group<http://www.ogf.org/dfdl/>
IBM SWG, Hursley, UK
smh@uk.ibm.com<mailto:smh@uk.ibm.com>
tel:+44-1962-815848

From: Tim Kimber/UK/IBM@IBMGB
To: dfdl-wg@ogf.org,
Date: 28/07/2014 23:51
Subject: Re: [DFDL-WG] clarification on dfdl:occursIndex() function
Sent by: dfdl-wg-bounces@ogf.org
________________________________

I agree that it is inconvenient for the 'nearest array parent' to be inaccessible. However, experience with discriminators makes me fearful of any rule that includes the phrase 'nearest enclosing' :-)
I think at least one other DFDL function allows the target of the function to be specified as an argument, but insists that the argument must be in the dynamic scope of the element ( i.e. its parent/grandparent etc ). I would be much happier with that solution for occursCountIndex().

I can think of a use case where it may be useful to get a consistent behaviour for occursCountIndex(). If a DFDL schema is generated from some other data format description then the model generation code may want to refer to the nth occurrence of something else in the model, where n is the occurs index of the current element - regardless of whether this particular element is an array.

regards,

Tim Kimber,

From: Mike Beckerle <mbeckerle.dfdl@gmail.com>
To: "dfdl-wg@ogf.org" <dfdl-wg@ogf.org>
Date: 28/07/2014 23:27
Subject: [DFDL-WG] clarification on dfdl:occursIndex() function
Sent by: dfdl-wg-bounces@ogf.org
________________________________

This function says it can be called on non-array elements. However, it does not say what the result is.

If called when "." is not itself an array element there are only two possible behaviors consistent with the fact that it is explicitly allowed on non-array elements.

The result has to be either

(a) 1
(b) the occursIndex of the nearest enclosing array parent, or 1 if there is no enclosing array parent.

I claim (a) is fairly pointless. You will just end up having to create newVariableInstances to carry the array current index downward into expressions.

I cannot think of a use case where one would want to call occursIndex() polymorphically, i.e., where you want a number in the case of an array, but 1 otherwise.

So (b) is the preferable behavior.

Mike Beckerle | OGF DFDL Workgroup Co-Chair | Tresys Technology | www.tresys.com<http://www.tresys.com/>
Please note: Contributions to the DFDL Workgroup's email discussions are subject to the OGF Intellectual Property Policy<http://www.ogf.org/About/abt_policies.php>
--
dfdl-wg mailing list
dfdl-wg@ogf.org
https://www.ogf.org/mailman/listinfo/dfdl-wg

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU--
dfdl-wg mailing list
dfdl-wg@ogf.org
https://www.ogf.org/mailman/listinfo/dfdl-wg

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU -- dfdl-wg mailing list
dfdl-wg@ogf.org
https://www.ogf.org/mailman/listinfo/dfdl-wg

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU-- dfdl-wg mailing list dfdl-wg@ogf.org https://www.ogf.org/mailman/listinfo/dfdl-wg

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU -- dfdl-wg mailing list dfdl-wg@ogf.org https://www.ogf.org/mailman/listinfo/dfdl-wg

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU