SBOL 2 to 3 Converter: Empty ComponentDefinition conversion

[fxbuson]

I tried converting something very minimal with the python native converter:

import sbol2
from sbol_utilities import conversion

doc2 = sbol2.Document()
comp = sbol2.ComponentDefinition()
doc2.add(comp)

doc3 = conversion.convert2to3(doc2, use_native_converter = True)

(Note this converts with no problems in the JS converter)

The first error I get is related to namespaces. it seems like the current code does not accept None namespaces when it clearly should (there is even a default mode that should return None):

r._sbol3_namespace(self, obj2)
    351     return namespaces[0]
    352 # Check if the object starts with any of the provided namespaces
--> 353 for namespace in self.namespaces:
    354     if obj2.identity.startswith(namespace):
    355         return namespace

TypeError: 'NoneType' object is not iterable

I then tried giving it a namespace with:

doc3 = conversion.convert2to3(doc2, namespaces = [sbol2.DEFAULT_NS], use_native_converter = True)

I then got an error when extracting the identity to sbol3:

File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\sbol3\identified.py:60, in extract_display_id(identity)
     58 msg += ' A displayId MUST be composed of only alphanumeric'
     59 msg += ' or underscore characters and MUST NOT begin with a digit.'
---> 60 raise ValueError(msg)

We can see from the code in pysbol3.identified.py that it tries to extract whatever is after the last forward slash:

    [...]
    display_id = parsed.path.split('/')[-1]
    if is_valid_display_id(display_id):
        return display_id
    else:
        msg = f'"{display_id}" is not a valid displayId.'
        msg += ' A displayId MUST be composed of only alphanumeric'
        msg += ' or underscore characters and MUST NOT begin with a digit.'
        raise ValueError(msg)

But the identity in SBOL2 ends with the version, not the ID. In this case my component had identity = 'http://examples.org/ComponentDefinition/example/1'. If I trim the identity first:

comp.identity = "/".join(comp.identity.split('/')[:-1])

Then the conversion runs ok. Should I just implement the trimming or is this expected to break something else? And for the namespace, should the function be looking for a None namespace or is this expected behaviour?

[fxbuson]

The same applies to the conversion back to SBOL2. I first assumed the version should default to None (which it does right now), but according to the SBOL 3.1 specification:

When SBOL 3.x Identified object is converted to SBOL 2.x, if its identity is a URL, then the identity of
the SBOL 2.x object SHOULD be constructed as [SBOL3 identity]/[SBOL2 version]. If the object does
not have an SBOL2 version property, then its version SHOULD default to 1.

Re-checking the roundtrip conversion, going from 2 > 3 > 2 with component of version = "2.2", I first start with an identity of "http://examples.org/example/2.2" and end with "http://examples.org/example"

[fxbuson]

Adding to this, keeping the namespace as None and expecting pysbol3 to handle it does not work because it just uses everything before the last separator.

For http://examples.org/ComponentDefinition/example it assumes the namespace is http://examples.org/ComponentDefinition, when it should just be http://examples.org

Therefore to avoid changes to pysbol3 this will have to be handled inside the converter code