Switch query to ebrain-kg-core

[mih]

So far we have used hard-crafted queries. With ebrains-kg-core available, it makes sense to remove this complication and switch to standard queries provided by this API wrapper. Here are examples:

Establish query setup:

from kg_core.kg import kg
k=kg()
k.with_token()
kg_client= k.build()

Get info on a dataset (by ID):

>>> pprint(list(kg_client.instances.get_by_id('e472a8c7-d9f9-4e75-9d0b-b137cecbc6a2').data.keys()))
['https://openminds.ebrains.eu/vocab/repository',
 '@type',
 'https://openminds.ebrains.eu/vocab/keyword',
 'https://openminds.ebrains.eu/vocab/description',
 'https://openminds.ebrains.eu/vocab/preparationDesign',
 'https://openminds.ebrains.eu/vocab/studyTarget',
 'https://core.kg.ebrains.eu/vocab/meta/revision',
 'https://openminds.ebrains.eu/vocab/studiedSpecimen',
 'http://schema.org/identifier',
 'https://openminds.ebrains.eu/vocab/behavioralProtocol',
 'https://openminds.ebrains.eu/vocab/versionIdentifier',
 'https://openminds.ebrains.eu/vocab/fullName',
 'https://openminds.ebrains.eu/vocab/shortName',
 'https://core.kg.ebrains.eu/vocab/meta/space',
 'https://openminds.ebrains.eu/vocab/releaseDate',
 '@id',
 'https://openminds.ebrains.eu/vocab/relatedPublication',
 'https://openminds.ebrains.eu/vocab/ethicsAssessment',
 'https://openminds.ebrains.eu/vocab/type',
 'https://openminds.ebrains.eu/vocab/funding',
 'https://openminds.ebrains.eu/vocab/author',
 'https://openminds.ebrains.eu/vocab/homepage',
 'https://openminds.ebrains.eu/vocab/license',
 'https://openminds.ebrains.eu/vocab/custodian',
 'https://openminds.ebrains.eu/vocab/accessibility',
 'https://openminds.ebrains.eu/vocab/isAlternativeVersionOf',
 'https://openminds.ebrains.eu/vocab/digitalIdentifier',
 'https://openminds.ebrains.eu/vocab/versionInnovation',
 'https://openminds.ebrains.eu/vocab/experimentalApproach',
 'https://openminds.ebrains.eu/vocab/dataType',
 'https://openminds.ebrains.eu/vocab/howToCite',
 'https://openminds.ebrains.eu/vocab/fullDocumentation',
 'https://openminds.ebrains.eu/vocab/technique',
 'https://core.kg.ebrains.eu/vocab/meta/lastReleasedAt',
 'https://core.kg.ebrains.eu/vocab/meta/firstReleasedAt']

Get info on a file repository (by ID from dataset version record)

>>> pprint(list(kg_client.instances.get_by_id('00932cbe-f90f-4968-a91f-da717c554320').data.keys()))
['https://core.kg.ebrains.eu/vocab/lastSync',
 '@type',
 'https://core.kg.ebrains.eu/vocab/bytes',
 'https://core.kg.ebrains.eu/vocab/meta/revision',
 'https://openminds.ebrains.eu/vocab/hostedBy',
 'https://openminds.ebrains.eu/vocab/IRI',
 'http://schema.org/identifier',
 'https://core.kg.ebrains.eu/vocab/lastModified',
 'https://core.kg.ebrains.eu/vocab/objectsCount',
 'https://core.kg.ebrains.eu/vocab/lastSyncIRI',
 'https://openminds.ebrains.eu/vocab/repositoryType',
 'https://core.kg.ebrains.eu/vocab/lastSeenCount',
 'https://core.kg.ebrains.eu/vocab/meta/space',
 'https://core.kg.ebrains.eu/vocab/public',
 'https://openminds.ebrains.eu/vocab/name',
 'https://openminds.ebrains.eu/vocab/contentTypePattern',
 '@id',
 'https://core.kg.ebrains.eu/vocab/meta/lastReleasedAt',
 'https://core.kg.ebrains.eu/vocab/meta/firstReleasedAt']

At present it is unclear to me how to get the actual file repository content listing. The only way I see it to visit the https://core.kg.ebrains.eu/vocab/lastSyncIRI-type URL, which yields an XML-formatted response that has the container content list. There is likely a better way that does not require a different query/parsing paradigm.

[apdavison]

Hi Michael,

You might find fairgraph helpful. This is a Python library which builds on top of ebrains-kg-core but also knows about openMINDS schemas. Documentation here: https://fairgraph.readthedocs.io/en/latest

Example for listing repository contents:

In [1]: from fairgraph import KGClient

In [2]: import fairgraph.openminds.core as omcore

In [3]: client = KGClient(host="core.kg.ebrains.eu")

In [4]: dv = omcore.DatasetVersion.from_id("e472a8c7-d9f9-4e75-9d0b-b137cecbc6a2", client)

In [5]: files = omcore.File.list(client, file_repository=dv.repository)

In [6]: for file in files:
   ...:     print(file.iri)
   ...:
https://object.cscs.ch/v1/AUTH_4791e0a3b3de43e2840fe46d9dc2b334/ext-d000017_DiFuMoAtlases_pub/64/DataDescriptor-DiFuMo(64-dimensions).pdf
https://object.cscs.ch/v1/AUTH_4791e0a3b3de43e2840fe46d9dc2b334/ext-d000017_DiFuMoAtlases_pub/64/labels_64_dictionary.csv
https://object.cscs.ch/v1/AUTH_4791e0a3b3de43e2840fe46d9dc2b334/ext-d000017_DiFuMoAtlases_pub/64/Licence-CC-BY.pdf
https://object.cscs.ch/v1/AUTH_4791e0a3b3de43e2840fe46d9dc2b334/ext-d000017_DiFuMoAtlases_pub/64/maps.nii.gz
https://object.cscs.ch/v1/AUTH_4791e0a3b3de43e2840fe46d9dc2b334/ext-d000017_DiFuMoAtlases_pub/64/3mm/resampled_maps.nii.gz

In [7]: files[0].download(".", client, accept_terms_of_use=True)
Out[7]: PosixPath('DataDescriptor-DiFuMo(64-dimensions).pdf')

[mih]

@apdavison that looks fantastic! I will take a closer look shortly! Thanks much!

[mih]

Hey @apdavison! I now had a chance to try this out, and it works just as advertized -- really cool! I particularly like the readily accessible content of properties, e.g. the type of a file hash

>>> f.hash
Hash(algorithm='MD5', digest='1dc869c088d4ebd615287fb79b5853b2')

I was hoping that you had also come up with a convention to derive local file paths from the IRIs of files, but I could not find something related to that. To be more specific:

I can know the repo a file is in:

>>> f.file_repository
KGProxy([<class 'fairgraph.openminds.core.data.file_repository.FileRepository'>], 'https://kg.ebrains.eu/api/instances/00932cbe-f90f-4968-a91f-da717c554320')

I can also know the IRI of the file, pointing into that repo

>>> f.iri
IRI(https://object.cscs.ch/v1/AUTH_4791e0a3b3de43e2840fe46d9dc2b334/ext-d000017_DiFuMoAtlases_pub/64/DataDescriptor-DiFuMo(64-dimensions).pdf)

But there seems to be no standard implementation to derive from this information that a suitable local path could be 64/DataDescriptor-DiFuMo(64-dimensions).pdf.

It seems to require something like "longest-common-prefix". However, this would quickly become messy when a dataset incorporates files from different file repositories (which I understand is not done (yet?), but possible.

Are you aware of an implementation for that?

Thanks in advance!

Unrelated, but worth noting: fairgraph also exhibits the really slow file repository queries (e.g. 20-30s for a 5-file repo, such as the one demo'ed above). I had suspected that I was somehow doing it suboptimally with my custom queries, but now it looks like a more general issue.

[apdavison]

We have some code that does this for CSCS Swift containers, which is the most commonly used file repository in EBRAINS, but nothing for the general case.