Include instances as attributes of the relevant class

[apdavison]

This is a proposal for how to include predefined metadata instances (e.g. controlled terms) in the Python library (cf #24). Usage example:

In [1]: from openminds.latest.controlled_terms import AgeCategory

In [2]: AgeCategory.adult
Out[2]: <openminds.latest.controlled_terms.age_category.AgeCategory at 0x104eb3970>

In [3]: AgeCategory.adult.id
Out[3]: 'https://openminds.ebrains.eu/instances/ageCategory/adult'

In [4]: AgeCategory.adult.definition
Out[4]: "'Adult' categorizes the life cycle stage of an animal or human that reached sexual maturity."

[ehennestad]

I don't quite understand from the jinja template whether all the instances will be embedded into the python class?

I had done something similar before in the openMINDS MATLAB package, but with suggestions from Lyuba ended up doing it differently.

This is the example of a schema definition for AgeCategory.
https://github.com/openMetadataInitiative/openMINDS_MATLAB/blob/main/code/schemas/latest/%2Bopenminds/%2Bcontrolledterms/AgeCategory.m#L1

As you can see, it has a hidden CONTROLLED_INSTANCES property, and this holds the name of all known instances of this schema. In the ControlledTerm superclass there is a case in the constructor to deserialise an json ld instance if the constructor is called with an input argument which matches one of the names in the CONTROLLED_INSTANCES property

Example:

adult = openminds.controlledterms.AgeCategory('adult')

adult = 

  AgeCategory (https://openminds.ebrains.eu/controlledTerms/AgeCategory) with properties:

                     definition: "''Adult'' categorizes the life cycle stage of an animal or human that reached sexual maturity."
                    description: ""
             interlexIdentifier: "http://uri.interlex.org/base/ilx_0729043"
             knowledgeSpaceLink: ""
                           name: "adult"
    preferredOntologyIdentifier: "http://purl.obolibrary.org/obo/UBERON_0000113"
                        synonym: ["adult stage"    "post-juvenile adult"    "post-juvenile adult stage"]

  Required Properties: name

>> adult.id

ans =

    'https://openminds.ebrains.eu/instances/ageCategory/adult'

[lzehl]

@ehennestad I let Andrew respond but it is not loading all instances into the class only the one of the stated type (as it is done for MATLAB as well, more or less)

[Peyman-N]

Python name generated for 4-(2-Hydroxyethyl)-1-piperazineEthanesulfonicAcid is four_(2_hydroxyethyl)_1_piperazine_ethanesulfonic_acid which is not a valid python variable name.
6,7-Dinitro-1,4-dihydroquinoxaline-2,3-dione is converted to six,7_dinitro_1,4_dihydroquinoxaline_2,3_dione.

[ehennestad]

Python name generated for 4-(2-Hydroxyethyl)-1-piperazineEthanesulfonicAcid is four_(2_hydroxyethyl)_1_piperazine_ethanesulfonic_acid which is not a valid python variable name.
6,7-Dinitro-1,4-dihydroquinoxaline-2,3-dione is converted to six,7_dinitro_1,4_dihydroquinoxaline_2,3_dione.

I also had this issue with a lot of instance names, as MATLAB is very strict about property names.

I ended up changing the constructor for controlled term classes to create the object from a controlled instance if the constructor input is a string which is recognised as a known controlled instance, for example:
openminds.controlledterms.MolecularEntity('4-(2-Hydroxyethyl)-1-piperazineEthanesulfonicAcid')

and it returns the instance:

ans = 

  MolecularEntity (https://openminds.ebrains.eu/controlledTerms/MolecularEntity) with properties:

                     definition: "HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) is a zwitterionic sulfonic acid buffering agent; one of the twenty Good's buffers. [adapted from W  ikipedia (https://en.wikipedia.org/wiki/HEPES)]"
                    description: ""
             interlexIdentifier: "http://uri.interlex.org/base/ilx_0484759"
             knowledgeSpaceLink: "https://knowledge-space.org/wiki/CHEBI:42334#2-4-2-hydroxyethyl-piperazin-1-yl-ethanesulfonic-acid"
                           name: "4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid"
    preferredOntologyIdentifier: "http://purl.obolibrary.org/obo/CHEBI_42334"
                        synonym: ["HEPES"    "2-[4-(2-hydroxyethyl)piperazin-1-yl]ethane-1-sulfonic acid"]

  Required Properties: name

I think this is a better solution because you can keep using all the original names and don't have to worry about remapping them

[apdavison]

@ehennestad and @Peyman-N I've fixed the Python name generation, but I've also added a by_name() class method to controlled-terms classes, similar to the solution suggested by @ehennestad. This method also uses synonyms for lookups, e.g.:

In [1]: from openminds.latest.controlled_terms import MolecularEntity

In [2]: MolecularEntity.by_name("HEPES") == MolecularEntity.by_name("4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid")
Out[2]: True