James
From your answers, you need IBM DFDL to implement the errata that sets
zero length string in the infoset. That gives you the round tripping you
require. I'd like to add this to the next release of IBM DFDL but can't
guarantee that it will get done.
Here's the specific rules from the errata, the relevant stuff in blue:
Empty representation when parsing
If empty representation is established when parsing, the possibility of
applying a default value
arises. Essentially, if a required occurrence of an element has empty
representation, then a
default value will be applied if present, though there are a couple of
variations on this rule.
Remember that in order to have established empty representation, the
occurrence must be
compliant with the emptyValueDelimiterPolicy for the element, and for a
complex element the
parser must have descended into the type and returned with no unsuppressed
processing error.
There are three main cases to consider. In what follows the term ‘string’
encompasses both
xs:string and xs:hexBinary as these are the two data types for which a
zero length (empty) string
is valid for the type. This behaviour is independent of occursCountKind.
Simple element (non-string)
Required occurrence: If a XSD ‘default’ or ‘fixed’ property is specified
then an item is added to the
Infoset using the value of the property, otherwise nothing is added to the
Infoset. (This may cause
a subsequent processing error depending on occursCountKind).
Optional occurrence: Nothing is added to the Infoset.
Simple element (string)
Required occurrence: If a XSD ‘default’ or ‘fixed‘ property is specified
then an item is added to the
infoset using the value of the property, otherwise an item is added to the
Infoset using empty
string as the value.
Optional occurrence: If emptyValueDelimiterPolicy is not ‘none’ then an
item is added to the
Infoset using empty string as the value, otherwise nothing is added to the
Infoset.
(To prevent unwanted empty strings from being added to the Infoset, use
minLength > ‘0’ and a
dfdl:assert that uses the dfdl:checkConstraints() function, to raise a
processing error.)
Complex element
Required occurrence: An item is added to the Infoset.
Optional occurrence: If emptyValueDelimiterPolicy is not ‘none’ then an
item is added to the
Infoset, otherwise nothing is added to the Infoset.
For both required and optional occurrences, the Infoset item may also have
a child item.
A) If the first child element of the complex type is a required simple
element, then an empty
string or default value will also be added to the Infoset.
B) If the first child element of the complex type is a required complex
element, then an item
is added to the Infoset (which may itself have a child via A)
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: "Garriss Jr., James P."
The current behaviour in IBM DFDL when it finds an empty xs:string is to give an error for a required (minOccurs '1') element
That is the behavior I am experiencing now, even with the emptyValueDelimiterPolicy set to none. Mar 5, 2013 9:25:17 AM error: CTDP3059E: Element 'PlainEmail' has minOccurs='1' and no default value but the input document contained only '0' occurrences. That’s not the behavior I expect or desire, as <> is completely valid input.
there is an errata in this area which has just been concluded, which changes this behaviour for xs:string so that a zero-length xs:string is added to the infoset under certain circumstances
That would be more appropriate for my situation.
what do you want ideally to appear in the infoset for the <> case? A zero length string, or nothing at all, or the special value 'nil', or a default value?
Hmmm, I think a zero length string gives the correct sense. Nil implies the value is unknown or not applicable, which is not true. Nothing indicates there was no input, which is not true, as <> is a legal value. Default value doesn’t make sense.
And what would you want to appear when the infoset was serialized: <> or is nothing also acceptable?
<>, because we need to reconstruct the input.
So I guess I know how to model this, I just need to wait for MTBK to be
updated to reflect this errata.
From: Steve Hanson [mailto:smh@uk.ibm.com]
Sent: Monday, March 04, 2013 4:40 AM
To: Garriss Jr., James P.
Cc: dfdl-wg@ogf.org; dfdl-wg-bounces@ogf.org
Subject: Re: [DFDL-WG] Empty element with initiator and terminator
James, if you look in the trace viewer and work back to the element in
question, I suspect that you will see an error like:
CTDP3138E: An unexpected initiator was found for an empty element
DFDL lets you specify what the syntax is for an empty element, using a
property called dfdl:emptyValueDelimiterPolicy here. It is probably set to
'none'. If you set it to 'both' then the parser will expect to find an
initiator and terminator when the content is empty. Note that eliminates
the need for the choice - you just have a single element with lengthKind
'delimited' and emptyValueDelimiterPolicy set appropriately.
The current behaviour in IBM DFDL when it finds an empty xs:string is to
give an error for a required (minOccurs '1') element, or to not add
anything to the infoset for an optional (minOccurs '0') element. However,
there is an errata in this area which has just been concluded, which
changes this behaviour for xs:string so that a zero-length xs:string is
added to the infoset under certain circumstances. So in order to guide you
down the right path, I need to know a bit more info. Specifically, what do
you want ideally to appear in the infoset for the <> case? A zero length
string, or nothing at all, or the special value 'nil', or a default value?
And what would you want to appear when the infoset was serialized: <> or
is nothing also acceptable?
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: "Garriss Jr., James P."