Describe the feature

At least with Iceberg and Hive tables, we have observed that describe <table> is consistently slower than the equivalent query against information_schema.columns to retrieve column information:

select column_name, data_type
from information_schema.columns
where
    table_catalog = '{{ relation.database | lower }}'
    and table_schema = '{{ relation.schema | lower }}'
    and table_name = '{{ relation.identifier }}'

Given how often describe statements are run, we would like to switch to using the faster query against information_schema.columns.

Describe alternatives you've considered

We considered enabling metastore caching in trino. This is disabled in Iceberg and cannot be enabled so we don't consider it a valid option.

Why the describe statement is consistently slower in trino is an unknown. We feel updating the dbt-trino adapter is a quicker solution than tracking the down the issue in trino.

Who will benefit?

This will speed up dbt pipelines without requiring any trino configuration changes to enable any caching.

Are you willing to submit PR?

• Yes I am willing to submit a PR!