The trick here is that in one pass you have to have a choice that selects a
branch based on the type string in the header.
Here's the idea, many details omitted, but I think you should get the gist:
<!--
Each record contains a vector of fields, the number of fields is the
number of entries in the field descriptor array.
The name and the type are from the field descriptor array entry
corresponding to the field position.
-->
<element name="record">
<complexType>
<sequence>
<element name="field" minOccurs="1" maxOccurs="200"
dfdl:occursCountKind="expression" dfdl:occursCount="{
fn:count(../../fieldDescriptorArray/Field) }">
<complexType>
<sequence dfdl:separator="%NL;">
<element name="name" type="xs:string"
dfdl:inputValueCalc='{ pathToFieldArray[dfdl:occursIndex()]/name }'/>
<!--
Here we use a choice where we choose a branch based on
the name of the type
in the field descriptor
-->
<element name="value">
<complexType>
<choice dfdl:choiceDispatchKey='{ path to field array
[dfdl:occursIndex()]/type }' >
<element name="s" type="xs:string"
dfdl:choiceBranchKey="string"
dfdl:lengthKind="explicit" dfdl:length='{
pathToFieldArray[dfdl:occursIndex]/length }'/>
<element name="b" type="xs:boolean"
dfdl:choiceBranchKey="boolean"
dfdl:lengthKind="explicit" dfdl:length="1" />
....
</choice>
</complexType>
</element>
</sequence>
</complexType>
</element>
</sequence>
</compllexType>
</element>
So the data you will get out looks like:
<record>
<field>
<name>station-name</name>
<value><s>Van Dorn Street</s></value>
</field>
<field>
<name>line</name>
<value><s>blue</s></value>
</field>
<field>
<name>isActive</name>
<value><b>true</b></value>
</field>
</record>
When parsing, the above schema will enforce that if the header says
"boolean" type, that the data must parse as
a boolean, i.e., the types are required to match between header and data.
There is no way in DFDL to generate element names dynamically. That
requires the 2-pass approach.
The above is the best you can do in one pass.
Note that an XSLT (which chatGPT can write for you) can translate this XML
into:
<record>
<station-name>Van Dorn Street</station-name>
<line>blue</line>
<isActive>true</isActive>
</record>
or even:
<record stationName="Van Dorn Street" line="blue" isActive="true"/>
I hope that helps.
On Tue, Oct 22, 2024 at 9:10 AM Mark Kozak <mark.ko...@adeptus-cs.com>
wrote:
> Thank you Mike,
>
>
>
> I have had some success implementing the two pass approach, but that will
> require some major restructuring of the parent project. So a single pass
> approach is still needed in at least the short term. Any help getting that
> to work will be very helpful.
>
>
>
> Thanks again,
>
> Mark Kozak
>
>
>
> *From:* Mike Beckerle <mbecke...@apache.org>
> *Sent:* Friday, October 18, 2024 7:15 PM
> *To:* users@daffodil.apache.org
> *Subject:* Re: revisit of parsing dynamic types?
>
>
>
> Sorry this got buried. I'll be back tuesday and will try to respond then
> if nobody else already has.
>
>
>
> On Tue, Oct 8, 2024 at 2:42 PM Mark Kozak <mark.ko...@adeptus-cs.com>
> wrote:
>
> Hello daffodil community.
>
>
>
> I am working on project that could be implemented by a schema similar to
> what was described in
> lists.apache.org/thread/1y9rl31q4hcnmtylz1dd0gdgrxty4onf. But I can't
> figure out how to iterate over the attribute types while looking back in
> the header.
>
>
>
> In the referenced example:
>
>
>
> <record>
> <field><s>blue</s></field>
> <field><b>true</b></field>
> </record>
>
>
> How do you get s and the b? I know we can look back it ../.. But how do I
> go back down multiple paths to first get the s and then the b?
>
>
>
> Thank you for any enlightenment.
>
>
>
>
>
> Mark Kozak
>
> Director of Engineering
>
> Adeptus Cyber Solutions
>
> Adeptus-CS.com
>
>
>
>
