This question was posed in Slack in 2019:
When running schema tests, such as
not_null
, the SQL is executed over the entire table. As the table grows in size the cost of running such tests increases… Is there a way to limit the schema tests to the partition of data that was inserted on the currentrun
?
One proposed solution was “write a custom test that only filters on the latest [data].”
What’s the best practice?