How to link raw data to prefix maps?

[matentzn]

SSSOM allows embedding prefix maps into data to reduce the risk of uncoupled data from prefix maps. For example, we can embed prefix maps into our tsvs files like this:

# curie_map:
#   FOODON: http://purl.obolibrary.org/obo/FOODON_
#   KF_FOOD: https://kewl-foodie.inc/food/
#   skos: http://www.w3.org/2004/02/skos/core#
#   sssom: https://w3id.org/sssom/
# license: https://creativecommons.org/licenses/by/4.0/
subject_id  subject_label   predicate_id    object_id   object_label    mapping_justification   author_id   object_source_version   mapping_date    confidence  comment
KF_FOOD:F001    apple   skos:exactMatch FOODON:00002473 apple (whole)   semapv:ManualMappingCuration    orcid:0000-0002-7356-1779   http://purl.obolibrary.org/obo/foodon/releases/2022-02-01/foodon.owl    2022-05-02  0.95    "We could map to FOODON:03310788 instead to cover sliced apples, but only ""whole"" apple types exist."
KF_FOOD:F002    gala    skos:exactMatch FOODON:00003348 Gala apple (whole)  semapv:ManualMappingCuration    orcid:0000-0002-7356-1779   http://purl.obolibrary.org/obo/foodon/releases/2022-02-01/foodon.owl    2022-05-02  1.0 
KF_FOOD:F003    pink    skos:exactMatch FOODON:00004186 Pink apple (whole)  semapv:ManualMappingCuration    orcid:0000-0002-7356-1779   http://purl.obolibrary.org/obo/foodon/releases/2022-02-01/foodon.owl    2022-05-02  0.9 "We could map to FOODON:00004187 instead which more specifically refers to ""raw"" Pink apples. Decided against to be consistent with other mapping choices."
KF_FOOD:F004    braeburn    skos:exactMatch sssom:NoMapping     semapv:ManualMappingCuration    orcid:0000-0002-7356-1779   http://purl.obolibrary.org/obo/foodon/releases/2022-02-01/foodon.owl    2022-05-02  1.0 
KF_FOOD:F004    braeburn    skos:broadMatch FOODON:00002473 apple (whole)   semapv:ManualMappingCuration    orcid:0000-0002-7356-1779   http://purl.obolibrary.org/obo/foodon/releases/2022-02-01/foodon.owl    2022-05-02  1.0 

In hindsight, choosing the term curie_map was probably not ideal, prefix_map would have been better.

In any case, maybe bioregistry should develop a recommendations for data providers to document their use of prefixes. It could be, that say, a link to a context like https://bioregistry.io/context/obo is sufficient, but it would go a long way to standardise this.

[cthoyt]

@matentzn maybe this is a good time to deprecate curie_map and create a new field extended_prefix_map that can either be 1. a URI to a JSON file or 2. a list of records

[matentzn]

deprecate curie_map

I cant say I am not tempted but I fear that ship has sailed. This is such a fundamental part of the model that it will require a storm of people to convince me to really get rid of it.

That said, introducing a new field, "external_prefix_map" would be one way to deal with this issue.

To get the debate furthered - what exactly is the reason for allowing EPMs in this context? In my current view, this just complicates processing IMO. What is required here is a bimap, not an EPM.

[cthoyt]

my alternate idea to extend what's allowed in curie_map is here: https://github.com/mapping-commons/sssom/issues/339

[matentzn]

The suggestion is not unreasonable, but what is the use case of allowing EPMs at all in this context?