Long table names can be truncated on Postgres (at least) silently

[drewbanin]

Describe the bug

Long table names appear to be truncated by Postgres, but this behavior could also be present on other databases. This was just reproduced with a temp table suffix in a Snapshot query on Postgres, but I imagine this could just happen with exceptionally long table names:

The truncation will cause lots of problems across dbt, but the proximate issue here is that a subsequent information schema query will return zero rows (it's looking for the wrong identifier) which bricks a snapshot run. The quick user-space fix is to shrink the "base" name of the table, but we should also take action here to:

1. Reduce the length of the table suffix appended by dbt
2. Show a warning (or probably an error!) if a table is created with a name long enough for PG to truncate

More info on this PG constraint here.

Example informations schema query

      select
          column_name,
          data_type,
          character_maximum_length,
          numeric_precision,
          numeric_scale

      from information_schema.columns
      where table_name = 'github_issues_linked_to_open_pull_requests__dbt_tmp20200311221237629328'
      order by ordinal_position

> Returns 0 rows

System information

Which database are you using dbt with?

• postgres
• [?] redshift
• [?] bigquery
• [?] snowflake

The output of dbt --version:

Reproduced on 0.14.3

[jtcohen6]

Add a runtime error to create table as that checks for relation identifier length