archive table making incorrect column lengths on PostgreSQL

[joevandyk]

(this might be related to a dbt seed issue regarding column types: #684)

I have a PostgreSQL 10 table with text columns.

However, the archive tool creates the archive table with columns of type character varying(255)-- causing errors like value too long for type character varying(255) when the archive table is populated.

[joevandyk]

I'm guessing this is the culprit: https://github.com/fishtown-analytics/dbt/blob/2d441f8ebdacb2d60ed0de1dd126a22a8cfad44d/dbt/schema.py#L38-L42

[drewbanin]

Thanks @joevandyk - does this happen on the first dbt archive? Or does it happen on subsequent archives?

Archive uses a create table as (....) to initially build the archive table, so i'd be surprised if this doesn't work on the first run. For subsequent runs, dbt will execute insert statements to add new data to the archive table. Sometimes the new data is "wider" than the existing columns, but dbt should be able to widen the column automatically.

Any other information you can share here would be really helpful!

[joevandyk]

I messaged you the log in the dbt slack channel

[drewbanin]

Thanks! Yeah - text has a max width of 256 on Redshift, but it can be an unlimited size on Postgres. We should 1) change our hardcoded 255 to 256 everywhere (don't know how that happened) and 2) fix the logic around text types on Postgres