Hey team 👋 I'm using the

dagster-dbt

integration and specifically

load_assets_from_dbt_manifest

. This function defines two parameters: • `select`: a dbt selection string for the models in a project that you want to include. • `selected_unique_ids`: the set of dbt unique_ids that you want to load as assets. They appear quite similar so I guessed that

select

limits what gets executed by

dbt run

while

selected_unique_ids

selects what gets exposed as an asset in Dagster. Is it correct? I wanted to test that assumption so I created this simple DAG in dbt

a --> b

and loaded the asset this way:

assets = load_assets_from_dbt_manifest(manifest, select="+b", selected_unique_ids={"model.dbt_project.b"})

but I got the following error:

dagster._core.errors.DagsterInvariantViolationError: Core compute for op "run_dbt_dbt_project_aeba1" returned an output "a" that does not exist. The available outputs are ['b']

After digging into the integration code, I saw that the op does not account for the

selected_unique_ids

and yield materialization for every dbt node. Bug? An other related question is: why is the

selected_unique_ids

parameter not present in the definition of

load_assets_from_dbt_project

?

hi @Flavien! the

selected_unique_ids

parameter is generally is not useful to set yourself (and the docs should be updated to reflect this). basically, it represents the "resolved" unique ids from a dbt select statement, and it should either be

None

, or exactly match the set of unique ids specified by the

select

parameter. It mostly exists for legacy reasons -- originally

load_assets_from_dbt_manifest

did not support the

select

parameter (for various reasons), so this was an alternate way for specifying that selection. Nowadays, the select parameter is generally a much more convenient way of specifying that information.

Thank you @owen for the response and also congratulations for the 1.0 release. It would be quite useful to also have an

exclude

parameter. I saw you created an issue last month to do just that. Are you open to contributions?

hi @Flavien! That's definitely on our radar, and we're absolutely open to contributions