adapter.aasx: resolution of file:// references

[jkhsjdhjs]

When loading an AASX file, the aasx checks all submodels for File elements and tries to resolve the referenced files relative to the aasx container root. In Version 2.0, a document named TestFile.pdf in the AASX packages root would be referenced by a File element via the path /TestFile.pdf.
Since version 3.0, the schemata prohibit values like /TestFile.pdf and require the file:// protocol to be used, e.g. file:///TestFile.pdf. The problem is that File also supports remote references to files, for example via https://. This isn't allowed by the new schemata.

We should discuss whether the restrictions imposed by the schemata are actually correct, and if so, change the aasx adapter and examples accordingly.

See also:

basyx-python-sdk/basyx/aas/examples/data/example_aas.py

Line 453 in 8304c2a

value='/TestFile.pdf',

basyx-python-sdk/basyx/aas/examples/data/example_aas.py

Lines 470 to 471 in 8304c2a

	value='https://www.plattform-i40.de/PI40/Redaktion/DE/Downloads/Publikation/Details-of-the-Asset-'
	'Administration-Shell-Part1.pdf?__blob=publicationFile&v=5',

basyx-python-sdk/basyx/aas/adapter/aasx.py

Lines 233 to 259 in 8304c2a

	def _collect_supplementary_files(self, part_name: str, submodel: model.Submodel,
	file_store: "AbstractSupplementaryFileContainer") -> None:
	"""
	Helper function to search File objects within a single parsed Submodel, extract the referenced supplementary
	files and update the File object's values with the absolute path.
	:param part_name: The OPC part name of the part the submodel has been parsed from. This is used to resolve
	relative file paths.
	:param submodel: The Submodel to process
	:param file_store: The SupplementaryFileContainer to add the extracted supplementary files to
	"""
	for element in traversal.walk_submodel(submodel):
	if isinstance(element, model.File):
	if element.value is None:
	continue
	# Only absolute-path references and relative-path URI references (see RFC 3986, sec. 4.2) are considered
	# to refer to files within the AASX package. Thus, we must skip all other types of URIs (esp. absolute
	# URIs and network-path references)
	if element.value.startswith('//') or ':' in element.value.split('/')[0]:
	logger.info("Skipping supplementary file %s, since it seems to be an absolute URI or network-path "
	"URI reference", element.value)
	continue
	absolute_name = pyecma376_2.package_model.part_realpath(element.value, part_name)
	logger.debug("Reading supplementary file {} from AASX package ...".format(absolute_name))
	with self.reader.open_part(absolute_name) as p:
	final_name = file_store.add_file(absolute_name, p, self.reader.get_content_type(absolute_name))
	element.value = final_name

[s-heppner]

Actually, I believe this may be a problem with the specification. In Part 5 Package File Format, it is specified that Submodels should reference supplementary AASX-files via a "relative URI" in the File SubmodelElement.

[s-heppner]

Some further research proves constraint that the file:// prefix is mandatory in the PathType is wrong: PathType Definition. Therefore, it would appear that the schemata is wrong here. I'll look into it.

[s-heppner]

I opened an Issue at aas-core.

[s-heppner]

RFC 8089 may not allow for relative paths, have you tried to start with a . though?

[jkhsjdhjs]

Thanks for the investigation!

RFC 8089 may not allow for relative paths, have you tried to start with a . though?

I haven't tried that yet, but /TestFile.pdf looks more like an absolute path to me anyways, where / is the root of the AASX package.

[s-heppner]

As an ugly temporary fix, so that AASX-packages can used, we decided to remove the regex check from the schemata.
We need to document this though and revert these changes with the upcoming spec fix.