Here are the updates to the restrictions that we discussed on the call. Action 165 raised.

Constraints on element lengthKind 'endOfParent'...

- element maxOccurs = 1
- no terminator on element
- no trailingSkip on element
- if element is in a sequence
- sequence must be the content of a complex type
- element must be the last object in the sequence
- separatorPosition of sequence must not be 'postFix'
- sequenceKind of sequence must be 'ordered'
- no terminator on sequence
- no trailingSkip on sequence
- no floating elements in the sequence
- if element is in a choice
- if choiceLengthKind is 'implicit'
- choice must be the content of a complex type
- no terminator on choice
- no trailingSkip on choice
- parent element lengthKind must not be 'implicit' or 'delimited'
- not sensitive to any in-scope markup

Notes:
- complex element can have 'endOfParent', & its last child element can be any lengthKind including 'endOfParent'
- element must be the last thing in its box
- a box is defined as a portion of the data stream that has an established length prior to the parsing of its children
- element with lengthKind 'explicit', 'prefixed', 'pattern' & no sequence/choice right framing
- choice with choiceLengthKind 'explicit'

Regards

Steve Hanson
Architect, Data Format Description Language (DFDL)
Co-Chair, OGF DFDL Working Group
IBM SWG, Hursley, UK
smh@uk.ibm.com
tel:+44-1962-815848
----- Forwarded by Steve Hanson/UK/IBM on 31/01/2012 16:54 -----

From: Steve Hanson/UK/IBM
To: Mike Beckerle <mbeckerle.dfdl@gmail.com>
Cc: dfdl-wg@ogf.org
Date: 23/01/2012 17:53
Subject: Re: [DFDL-WG] Fw: * DFDL Errata* Clarification: Limitations on use of endOfParent

I have typically used 'implicit' for this because the children of the root element should match the data leaving no bytes unconsumed. Alternatively using 'delimited' should be equivalent, you are never actually scanning. However when the last child is endOfParent according to the rules below I can't use 'implicit' or 'delimited'....so in that scenario I am stuck. Need a rethink.

Regards

Steve Hanson
Architect, Data Format Description Language (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: Steve Hanson/UK/IBM@IBMGB
Cc: dfdl-wg@ogf.org
Date: 23/01/2012 17:00
Subject: Re: [DFDL-WG] Fw: * DFDL Errata* Clarification: Limitations on use of endOfParent

Only question I have is on "endOfParent" not being allowed on root element.

If you have a DFDL implementation processing message buffers, and the root element's content ends at the end of the buffer/end-of-data, how do we express that?

I expected that to be end-of-parent, the notion being that there's an implicit parent for all content, which has an end which is the true end-of-data.

...mikeb

On Mon, Jan 23, 2012 at 6:19 AM, Steve Hanson <smh@uk.ibm.com> wrote:
----- Forwarded by Steve Hanson/UK/IBM on 23/01/2012 11:16 -----

From: Steve Hanson/UK/IBM
To: Mike Beckerle <mbeckerle.dfdl@gmail.com>, Tim Kimber/UK/IBM
Date: 18/01/2012 16:26
Subject: * DFDL Errata* Clarification: Limitations on use of endOfParent

As agreed on WG extra call on 18th Jan.

Will be raised as a separate issue on next DFDL WG call.

Constraints on element lengthKind 'endOfParent'...

- element maxOccurs = 1
- no terminator on element
- if element is in a sequence
- separatorPolicy of sequence must not be 'postFix'
- sequenceKind of sequence must be 'ordered'
- no floating elements in the sequence
- must be the 'last' in the sequence statically **
- if element is in a choice it is always 'last' statically **
- parent element lengthKind must not be 'implicit' or 'delimited'
- if element is complex then all possible 'last' elements ** must also be 'endOfParent'
- not sensitive to any in-scope markup
- not allowed on root element

** Need a concise description of walking the content of a complex element and building the list of 'last elements'. Involves factoring out local sequences and coping with choices.

Regards

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

From: Steve Hanson/UK/IBM
To: Mike Beckerle <mbeckerle.dfdl@gmail.com>
Cc: Tim Kimber/UK/IBM@IBMGB
Date: 08/11/2011 18:22
Subject: Re: Coping with Character code U+0000 - and how to end-of-parent in an array

Hi Mike

For the record, I knew we said something about U+0000 - it's in section 5...

String – In DFDL a string can contain any character codes. None are reserved. (Including the character with character code U+0000, which is not permitted in XML documents.)

After discussion with Sandy Gao, this is what we wrote in the DFDL to XDM mapping document:
Note: SimpleElement [datavalue] values may contain characters that are illegal in XML, for example, DFDL strings can contain the character code 0 (zero) within them, but XML does not allow this character code in any XML content even if it is represented as a character entity. Nevertheless, a DFDL described string is mapped to an XDM string data value.

and later for the actual mapping to XDM:

SimpleElement: If the value of [datavalue] is special value “nil”, then the empty string, otherwise the value of [datavalue] converted to its canonical lexical representation.

On to your examples (and assuming separatorPosition is 'infix')...

You are right about the first (endOfParent) example being odd. This example would work fine if the lengthKind was 'delimited'. Remember that the 'explicit' length of the parent element creates a box which scopes the delimited behaviour.

The second (delimited) works fine.

endOfParent and delimited behave almost identically most of the time. When the element is an array, this looks to be one of the differences. I am thinking that endOfParent should not be allowed when maxOccurs > 1.

Regards

Steve Hanson
Architect, Data Format Description Language (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: Steve Hanson/UK/IBM@IBMGB, Tim Kimber/UK/IBM@IBMGB
Date: 08/11/2011 17:02
Subject: Coping with Character code U+0000 - and how to end-of-parent in an array

We have this incompatiability with XML infoset around U+0000 aka NUL.

However, one can model data containing character code U+0000 in the content as an array of strings with NUL termination. That is, we split the string on the NUL characters so as to avoid putting them in the infoset.

So I tried to do this and ran into issues. E.g., if data contains a string of length 80, but inside it the character code 0 can appear, then this could be modeled as:

<element name="stringsWithNul" dfdl:lengthKind="fixed" dfdl:length="80">
<sequence dfdl:separator="%x0000;">
<element name="substring" type="string" maxOccurs="80" dfdl:lengthKind="endOfParent" /> 
</sequence>
</element>

Problem: is that use of endOfParent length kind right? It's the last thing in the group, but the same element decl also describes the prior elements. If there are no NULs in the string, then endOfParent is exactly what you want. There will be only one substring, it will have all 80 characters. But if there are NULs in the middle, then you want the earlier array elements to be delimited by the sequence's separators, and only the last element to be delimited by endOfParent.

This semantics where the parent is providing the constraints on length, but sometimes its separator, just for the last thing it's endOfParent, is not something we can express I believe.

I was actually even unclear on this one: If the data 'string' has a terminator of ! then perhaps:

<element name="stringsWithNul" dfdl:lengthKind="delimited" dfdl:terminator="!">
<sequence dfdl:separator="%x0000;">
<element name="substring" type="string" maxOccurs="unbounded" dfdl:lengthKind="delimited"/> 
</sequence>
</element>

Is the array element delimited? Is that the right length kind for this situation?

Thanks for comments

...mikeb

...mikeb

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

--
Mike Beckerle | OGF DFDL WG Co-Chair
Tel: 781-330-0412

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

From:	Mike Beckerle <mbeckerle.dfdl@gmail.com>
To:	Steve Hanson/UK/IBM@IBMGB, Tim Kimber/UK/IBM@IBMGB
Date:	08/11/2011 17:02
Subject:	Coping with Character code U+0000 - and how to end-of-parent in an array