[CT-254] dbt-core - Snapshot - macro dbt_macro__create_schema takes not more than 1 argument(s)

[rlshuhart]

dbt-core/core/dbt/include/global_project/macros/materializations/snapshots/snapshot.sql

Lines 9 to 11 in 3ad61d5

	{% if not adapter.check_schema_exists(model.database, model.schema) %}
	{% do create_schema(model.database, model.schema) %}
	{% endif %}

Hi I've raise this same issue in dbt-sqlserver, but the issue might actually lie here after further investigation. I'm running into the issue, "macro 'dbt_macro__create_schema' takes not more than 1 argument(s)", when attempting to do a simple snapshot. When I delete lines 9-11 in the above all goes well.

Please see additional details on dbt-sqlserver issue at dbt-msft/dbt-sqlserver#208

[jtcohen6]

@rlshuhart Thanks for opening, you're totally right! We need to make that create_schema call take a single Relation argument, rather than database + schema. This was a breaking change a long time ago... all the way back in v0.17 (#2411).

It's worth checking to see if we actually need this create_schema call at all. I have a sneaking suspicion that the snapshot may already be properly accounted for by this logic:

dbt-core/core/dbt/task/run.py

Lines 437 to 439 in 5e0a765

	def before_run(self, adapter, selected_uids: AbstractSet[str]):
	with adapter.connection_named("master"):
	self.create_schemas(adapter, selected_uids)

If that's indeed the case, your solution (deleting lines 9-11) is the right one.

[jtcohen6]

I've concluded these lines are totally unneeded. After a bit of digging through the git blame, I see that the origin of the create_schema in the snapshot (then archive) materialization was in: #625

At the time, Drew noted that:

There might be a smarter way to do this -- we could set the node schema to the value of target_schema, then this will happen automatically through the node_runner before_run flow.

This is in fact what we do:

dbt-core/core/dbt/parser/snapshots.py

Lines 38 to 40 in 204d535

	# the target schema must be set if we got here, so overwrite the node's
	# schema
	node.schema = node.config.target_schema

And have been doing since #1478 (here specifically). This is picked up accordingly by the before_run method in the snapshot task (inherited from the method in the run task shown above).

Worth noting that this piece still isn't perfect, if you're defining target_schema in a .yml file instead of snapshots/*.sql: #4000

[rlshuhart]

Thanks a lot for looking into this, @jtcohen6!