Support Python models

[davidwitk]

Context: Currently, adding a Python models blocks the entire dry run.

Questions:

• What's needed to support Python models?
• Alternative: Can we imagine a way to skip Python models in model validation?

Exception is thrown here:

dbt-dry-run/dbt_dry_run/execution.py

Lines 70 to 76 in f69d6fb

	if node.language is not None and node.language != "sql":
	failing_nodes.append(
	(
	"NODE_NOT_SQL : Only SQL language models are supported",
	node.unique_id,
	)
	)

[ccharlesgb]

Hey @davidwitk, this could be possible if there was a way to predict the schema of a python model without running it but it would require quite a lot of refactoring to get the dry runner to accept non-sql models. You could possibly do something tactical by making it so that the python model must have run and just taking the existing schema of the model but that would limit the dry runners usefulness as a CICD tool.

If you mostly have SQL models and the Python models don't frequently change you might get some value out of just skipping them but it can mess up downstream models

[davidwitk]

I think the tactical approach makes a lot of sense. We have a use case of 1 Python model among hundreds of SQL models and I guess that's the typical setup for projects with Python models. So definitely some value already in having the Python models be skipped, and keeping the dry run active for pure SQL