Python models not referencing to unittest data fixtures

[bjones325]

Python Models use context.resolve_table(model_name) to find the physical table name to query.

When unittesting a python model with data fixtures mocking upstream tables, resolve_table does not resolve to the mocked fixture, IF the upstream table is not a SQLMesh model.

For example, with a model such as:

def execute(...):
      not_a_sqlmesh_model = "catalog.schema.upstream_table_not_a_model"
      table_ref = context.resolve_table(not_a_sqlmesh_model)
      fetchdf(f"SELECT * FROM {table_ref}")

and a unittest yml of:

my_test:
  model: my_model
  inputs:
    catalog.schema.upstream_table_not_a_model:
      rows:
           - col_a: value_a

the table_ref field will resolve to "catalog.schema.upstream_table_not_a_model", and NOT the table generated by the unittest yml.

This makes unittesting Python models dependent on upstream tables, which are NOT sqlmesh models, impossible.
Either, context.resolve_table should resolve to the table created by the data fixture, or fetchdf should parse and replace the table reference to point to the data fixture.

[izeigerman]

@themisvaltinos can you please have a look at this

[themisvaltinos]

sure I will look into this @izeigerman