Dim Element Validation Error

[lesserwhirls]

From the netCDF-C test DMRs:

$ xmllint --noout --schema xml.xsd test_struct_array.8.nc.dmr
test_struct_array.8.nc.dmr:11: Schemas validity error : Element '{http://xml.opendap.org/ns/DAP/4.0#}Dim': This element is not expected.
test_struct_array.8.nc.dmr fails to validate

test_struct_array.8.nc.dmr

[Mikejmnez]

Thanks @lesserwhirls !

Interesting. I think the schema expects in a Structure that Dim elements be declared before BaseTypes like Ints, Floats, etc. So it expects:

    <Structure name="s">
        <Dim size="2"/>
        <Dim size="2"/>
        <Int32 name="x"/>
        <Int32 name="y"/>
    </Structure>

instead of what I see in the example dmr above.

[ndp-opendap]

I think the schema expects in a Structure that Dim elements be declared before BaseTypes like Ints, Floats, etc

But I think the narrative in the specification document contradicts the schema in this matter.

I don't like that the ordering of the components in the Container types is different than the ordering in the Group type (which honestly is pretty much a Container type by a different name!)

[Mikejmnez]

How do we proceed? When I created the schema, I assumed that Structures and Sequences behaved like "Containers" such as a Group type. I can certainly modify the schema to match the dap4 specification ( I have not looked extensively into examples). I also agree that Structures are pretty much a Container type, and would not expect it to behave different from a Group type.

[ndp-opendap]

@Mikejmnez asked:

How do we proceed?

Well I think like this:

1. Find out what Hyrax does here. When we have arrays of Structures how does libdap4 produce the associated DMR?
2. If the TDS and libdap4 are producing the same ordering then great! Fix the schema
3. If they are different then we need to fix one, and I claim the one that isn't like the specification needs to be altered. Of course we should ask @jgallagher59701

One thing to remember here is:
Does it matter? Could we just decide not be so hung up on the ordering of the things? Or does that make the parsers impossible?

[Mikejmnez]

Luckily, there is a test file with a structure with dimensions on both the TDS and Hyrax. This file was provided by Sean for dap4 inter comparison. The file's DMR is:

<Dataset xml:base="http://test.opendap.org/opendap/dap4/d4ts/test_struct_array.nc.h5" dapVersion="4.0" dmrVersion="1.0" name="test_struct_array.nc.h5">
<Dimension name="dx" size="4"/>
<Dimension name="dy" size="3"/>
<Structure name="s">
<Int32 name="x"/>
<Int32 name="y"/>
<Dim name="/dx"/>
<Dim name="/dy"/>
</Structure>
</Dataset>

And so that motivates to fix the schema (see PR #12 )

[Mikejmnez]

One thing to remember here is:
Does it matter? Could we just decide not be so hung up on the ordering of the things? Or does that make the parsers impossible?

As someone who is embedded in parsers, it is nice to have these rules spelled out. Less logic needed on the parser is better (IMHO).