if the namespace of an object is changed it should be reflected in the object

[JMante1]

obj = sbol3.Component('hello', [sbol3.SBO_DNA]) obj.name = 'hello' obj.namespace = 'parts.igem.org' doc.add(obj)

This code leads to an object with http://sbols.org/unspecified_namespace used everywhere, except for the has namespace property. if the namespace property is updated it should change the namespace used throughout the property

[jakebeal]

pySBOL3 intentionally uses a "permissive" model of property changes, in which it does not check for potential inconsistencies until the user requests validation, since the user may be performing a given operation in the middle of some sort of multi-stage operation.

If you run a document validation after running the code that you provided, however, you will find that it does identify an error due to the invalid state that has been created:

>>> print(doc.validate())
http://sbols.org/unspecified_namespace/hello sbol3-10301: The namespace parts.igem.org is not a prefix of the identity http://sbols.org/unspecified_namespace/hello

Changing the namespace of an object is a potentially dangerous operation, since it can break links with other objects that point to this object, as well with with its child objects. As such, we recommend changing namespace via the "safe" function provided for that purpose on the Document class:

def change_object_namespace(top_levels: Iterable[TopLevel], new_namespace: str) -> Any:

This function provides a guaranteed safe change for a group of objects, moving all of them as a unit into the new namespace and preserving links both within the group and with external references.

[JMante1]

The change_object_namespace almost does what I want, unfortunately it doesn't update other references to the object (e.g. if a component is used in a combinatorial derivation the component is updated but the reference to it in the combinatorial derivation is not). Is there a function to do that too, i.e. to preserve the external references?

[tcmitchell]

@JMante1 this could be a bug. Document.change_object_namespace is supposed to maintain referential integrity among the top_levels passed to the method. If that is not what is happening for you, could you please share a reproducible test case so that the bug can be fixed?

[JMante1]

For this code

import sbol3
import os

direct = os.path.split(__file__)[0]
file_path_out = os.path.join(direct, 'out.nt')

doc = sbol3.Document()
comp = sbol3.Component('http://sbolstandard.org/testfiles/hello', sbol3.SBO_DNA)
seq = sbol3.Sequence('http://sbolstandard.org/testfiles/hello_sequence', elements='ATGC')
doc.add(comp)
doc.add(seq)

comp.sequences = [seq]

doc.change_object_namespace([seq], 'http://parts.igem.org')

doc.write(file_path_out)

I get out

http://parts.igem.org/hello_sequence http://sbols.org/v3#hasNamespace http://parts.igem.org .
http://parts.igem.org/hello_sequence http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://sbols.org/v3#Sequence .
http://sbolstandard.org/testfiles/hello http://sbols.org/v3#hasSequence http://sbolstandard.org/testfiles/hello_sequence . <===
http://sbolstandard.org/testfiles/hello http://sbols.org/v3#displayId "hello" .
http://parts.igem.org/hello_sequence http://sbols.org/v3#displayId "hello_sequence" .
http://sbolstandard.org/testfiles/hello http://sbols.org/v3#type https://identifiers.org/SBO:0000251 .
http://sbolstandard.org/testfiles/hello http://www.w3.org/1999/02/22-rdf-syntax-ns#type http://sbols.org/v3#Component .
http://sbolstandard.org/testfiles/hello http://sbols.org/v3#hasNamespace http://sbolstandard.org/testfiles .
http://parts.igem.org/hello_sequence http://sbols.org/v3#elements "ATGC" .

[JMante1]

I wrote this as a work around:

def update_uri_refs(doc, update_dict, use_derived=True, derived_ls = ['_sequence']):
    """
    This updates a set of referenced uris (may be a namespace or identity update)

    Args:
        doc (SBOL3 Document): document to be updated
        update_dict (dict): dictionary of the form {old_uri:new_uri}
        use_derived (bool, optional): Whether or not to also update derived uris. Defaults to True.
        derived_ls (list, optional): List of derivations e.g. also version of the uri
                                     with _sequence added to the end. Defaults to ['_sequence'].

    Returns:
        doc (SBOL3 Document): updated document
    """
    # create all the additional uris that will need to be updated
    derived_keys = []
    for deriv in derived_ls:
        der_update = [f'{x}{deriv}' for x in  update_dict.keys()]
        derived_keys.extend(der_update)

    # pull the graph from the document
    g = doc.graph()
    for index, (subject, predicate, _object) in enumerate(g):
        # if the object is one of the items to be updated do so
        if str(_object) in update_dict:
            g.remove((subject, predicate, _object))
            new = rdflib.URIRef(update_dict[str(_object)])
            g.add((subject, predicate, new))
        # update any derived objects
        elif use_derived and str(_object) in derived_keys:
            suffix = str(_object).split('_')[-1]  # assumes suffix starts with '_'
            suffix = f'_{suffix}'
            g.remove((subject, predicate, _object))
            old = str(_object)
            new = f"{update_dict[old.replace(suffix, '')]}{suffix}"
            new = rdflib.URIRef(new)
            g.add((subject, predicate, new))
        # update any derived subjects
        if use_derived and str(subject) in derived_keys:
            suffix = str(subject).split('_')[-1]  # assumes suffix starts with '_'
            suffix = f'_{suffix}'
            g.remove((subject, predicate, _object))
            old = str(subject)
            new = f"{update_dict[old.replace(suffix, '')]}{suffix}"
            new = rdflib.URIRef(new)
            g.add((new, predicate, _object))
    doc._parse_graph(g)
    return doc

[tcmitchell]

@JMante1 Thank you for the short and simple test case to demonstrate the problem, and thank you for taking the time to develop a workaround. Both are appreciated.
I think I see where Document.change_object_namespace is coming up short. It only takes one list of TopLevel objects. Those objects all have their namespace changed and their references are made consistent after the change. Your use case is slightly different. You have two distinct sets of objects. One set should have their namespace changed and get their references updated. The other set should get their references updated without changing their namespace. We had not considered this case when we designed and implemented Document.change_object_namespace.
I am testing a fix that will add a third argument, optional, to Document.change_object_namespace. This argument is an iterable of TopLevel objects that should have their references updated. It can be list or a Document (which is itself an Iterable[TopLevel].

[tcmitchell]

@JMante1 I have merged an enhancement to provide the functionality I think you need. With this enhancement you can add a third argument to Document.change_object_namespace. In your sample code above to demonstrate the error, you would change the invocation to Document.change_object_namespace to include the document:

doc.change_object_namespace([seq], 'http://parts.igem.org', doc)

Are you able to test from the "main" branch to see if this fix works for your original case? If possible I'd like to hear back from you before I cut a new release of pySBOL3 that includes this fix.
Thanks again for your assistance in creating the test case. It was tremendously helpful in understanding the issue and providing a fix.

[jakebeal]

Reopening because the autoclose happened before @JMante1 could test the fix.