Should LinkML metamodel have a way to specify which slots should be "indexed" in various storage backends, e.g. SQL DDL, mongo, neo4j, etc...

[sierra-moxon]

From a recent discussion with @kheal( see: microbiomedata/nmdc-schema#2155)

The gist of the question is: Should LinkML have a way to specify that a particular slot or set of slots in a class (perhaps using the unique_keys meta slot) should result in an index on an attribute/column/property or a compound index (if the storage mechanism supports this—e.g., SQL DDL)?

Right now, our SQL DDL generator uses the keys related metamodel slots to establish FK and PK relationships between tables in that serialization. We get the indexes "for free" in many RDBM implementations.

For MongoDB, we don't have constraints, but we can specify indexes: https://www.mongodb.com/docs/manual/indexes/. Still, we currently don't serialize a "schema" into MongoDB "collections" because there is no such thing. But it would be helpful to hint at the code that creates a collection and where to index it. We could do this through the keys LinkML mechanism as the SQL DDL generator does to create constraints.

Thoughts?

[ialarmedalien]

I think it'd be useful, yep!