It is not possible for DFDL v1.0 to lift this restriction as an erratum.
We can explore removing it for DFDL v2.0 in the future, and create a
Daffodil experimental feature to implement this.
The process works that way, implement/prototype the change, then with
experience from an implementation, propose it for the standard.
In general, in DFDL, you do not get to design "the XML you want", and then
use DFDL to populate it. It simply doesn't work that way. Often tiers of
elements are required where one is not desired strictly speaking in the
model.
The way I have seen this sort of thing modeled is like so:
<myElement>
<nil/>
<myElement>
<myElement>
<subElement1>....</subElement1>
<subElement2>....</subElement2>
....
<myElement>
So that you always have myElement at the top to identify the thing. Then
you either have child element content, or a reserved element to indicate
the nil/nulled status.
Arguably, <myElement xsi:nil="true"/>, should be possible vs.
<myElement><nil/></myElement>.
Another alternative is to use the lack of an element to indicate that the
element is nil.
<choice>
<sequence dfdl:initiator="-"/>
<element name="myElement" ....>
... complex type here...
</element>
</choice>
Then, if the "-" is found, there will be no element at all in the infoset.
When unparsing, this choice will default to the first branch and output the
"-".
On Thu, Dec 2, 2021 at 8:00 AM Roger L Costello <coste...@mitre.org> wrote:
> Hi Mike,
>
>
>
> 1. Would you get the DFDL committee to remove this restriction, please?
>
>
>
> 2. I have 350 data formats, each having dozens of fields with nillable
> complexTypes. So there is a strong use case for removing the restriction.
>
>
>
> 3. The choice idea that you presented, although clever and interesting, is
> not suitable. In the example I presented, the output XML should be one of
> these:
>
>
>
> <MyField>-</MyField>
>
>
>
> <MyField>
>
> <DateTimeIso>…</DateTimeIso>
>
> </MyField>
>
>
>
> <MyField>
>
> <MonthName>…</MonthName>
>
> </MyField>
>
>
>
> With the technique you showed, the output would be one of these:
>
>
>
> <noValue>-</noValue>
>
>
>
> <MyField>
>
> <DateTimeIso>…</DateTimeIso>
>
> </MyField>
>
>
>
> <MyField>
>
> <MonthName>…</MonthName>
>
> </MyField>
>
>
>
> That first one -- <noValue> -- is not unacceptable to my community.
>
>
>
> /Roger
>
>
>
> *From:* Mike Beckerle <mbecke...@apache.org>
> *Sent:* Wednesday, December 1, 2021 4:06 PM
> *To:* users@daffodil.apache.org
> *Subject:* [EXT] Re: Huh? Nillable complex type elements can only have
> '%ES;' as their dfdl:nilValue property
>
>
>
> This is definitely a DFDL language restriction, not a Daffodil-specific
> thing.
>
>
>
> I have made inquiries as to what the rationale is for this restriction,
> and I'll report back.
>
>
>
> In the mean time, maybe this sort of technique.
>
>
>
> <choice>
>
> <element name="noValue" type="xs:string" dfdl:lengthKind="explicit"
> dfdl:length="0" dfdl:initiator="-"/>
>
> <element name="myComplexType">
>
> .... non nillable complex type definition ....
>
> </element>
>
> </choice>
>
>
>
> I have to admit that this workaround sort of suggests that there isn't a
> coherent rationale for the restriction, as this is pretty obviously
> equivalent to the complex type being nillable with nilValue="-".
>
>
>
>
>
> On Wed, Dec 1, 2021 at 12:43 PM Roger L Costello <coste...@mitre.org>
> wrote:
>
> Hi Folks,
>
> I am creating a DFDL schema for a data format that has a field whose value
> is a hyphen when no data is available and when data is available, then the
> field contains either an ISO datetime or a 3-letter month name. Here's how
> I specified the field:
>
> <xs:element name="MyField" nillable="true" dfdl:nilValue="-">
> <xs:complexType>
> <xs:choice dfdl:choiceLengthKind="implicit">
> <xs:element name="DateTimeIso">
> <xs:complexType>
> <xs:sequence>
> ...
> </xs:sequence>
> </xs:complexType>
> </xs:element>
> <xs:element name="MonthName" type="xs:string"
> dfdl:length="3" dfdl:lengthKind="explicit" />
> </xs:choice>
> </xs:complexType>
> </xs:element>
>
> When I run it, Daffodil gives this error message:
>
> [error] Schema Definition Error: Nillable complex type elements can only
> have '%ES;' as their dfdl:nilValue property.
>
> Is that an error in Daffodil? Or, is it an actual DFDL restriction? If the
> latter, why is there such a restriction? How do I work around this error?
>
> /Roger
>
>
