[Bug] Incremental models on partitioned string column fails about dbt-bigquery HOT 4 CLOSED

@ivan-fishbrain Thanks for the issue. As it is, BigQuery doesn't support string-type partition columns, so I do think it makes sense to raise some kind of error here. The question is whether BigQuery should be the one to raise that error (as it's doing here), or if the dbt-bigquery adapter should be doing it.

The BigQuery error would be clearer if this were your first time running the model, to the effect of:

PARTITION BY expression must be DATE(<timestamp_column>), DATE(<datetime_column>), DATETIME_TRUNC(<datetime_column>, DAY/HOUR/MONTH/YEAR), a DATE column, TIMESTAMP_TRUNC(<timestamp_column>, DAY/HOUR/MONTH/YEAR), DATE_TRUNC(<date_column>, MONTH/YEAR), or RANGE_BUCKET(<int64_column>, GENERATE_ARRAY(<int64_value>, <int64_value>[, <int64_value>]))

Alternatively, dbt-bigquery could be more aggressive about validating partition column types in the PartitionConfig, and raise an error if it differs from the supported values. I think that's a reasonable approach, since we already hard-code a lot of type-logic in the render method; it just means that we'll need a PR to the dbt-bigquery adapter if ever / whenever BigQuery adds support for partitioning by a new datatype.

I'm going to transfer the issue over to the dbt-bigquery repo, since that's where the change will need to happen. Here's some pseudo-code to get you started:

class PartitionConfig(dbtClassMixin):
    ...
    
    def validate(cls, data):
         super().validate(data)
         dtype = data.get('data_type')
         supported = ("date", "timestamp", "datetime", "int")
         if dtype not in supported:
             raise ValidationError(f"Got partition data type {dtype}, expected one of {supported}")

from dbt-bigquery.

@jtcohen6 Thank you for explanation! I wasn't aware about impossibility of creation string-partitioned tables.
Then, I think dbt should at least create warn about this impossibility explicitly.

from dbt-bigquery.

This issue has been marked as Stale because it has been open for 180 days with no activity. If you would like the issue to remain open, please remove the stale label or comment on the issue, or it will be closed in 7 days.

from dbt-bigquery.

I think you can partition on anything now. And also under no circumstances should dbt-core be calling a function that's not even supported in bigquery, surely!

from dbt-bigquery.