redshift_admin_table_stats.is_encoded is always true

[tpilewicz]

Describe the bug

From my understanding, field is_encoded from model redshift_admin_table_stats should indicate whether the table we're looking at has at least one encoded column. But currently the field is computed by looking in pg_attribute if any table attribute is encoded. The result is that if at least one attribute is encoded, the table appears as having at least one column encoded. Worse: in our redshift, the tableoid attribute is almost always encoded. Hence all tables have their is_encoded field set to true.

Steps to reproduce

Run the following in your redshift:

select attencodingtype, count(1)
from pg_attribute
where attname = 'tableoid'
group by 1;

And the following:

select is_encoded count(1)
from redshift_admin_table_stats
group by 1;

Expected results

We have some tables for which no column is encoded. They should appear when running the above query.

Actual results

In redshift_admin_table_stats, the tables I just mentioned are flagged as is_encoded, which should not be the case.

System information

The contents of your packages.yml file:

packages:
  - package: fishtown-analytics/dbt_utils
    version: 0.6.3
  - package: fishtown-analytics/redshift
    version: 0.4.1
  - package: fivetran/jira_source
    version: [">=0.1.0","<0.2.0"]
  - package: fivetran/jira
    version: 0.1.5
  - package: fivetran/intercom_source
    version: 0.1.2
  - package: fivetran/intercom
    version: 0.1.2

Which database are you using dbt with?

• postgres
• redshift
• bigquery
• snowflake
• other (specify: ____________)

The output of dbt --version:

installed version: 0.18.0
   latest version: 0.19.1

Your version of dbt is out of date! You can find instructions for upgrading here:
https://docs.getdbt.com/docs/installation

Plugins:
  - bigquery: 0.18.0
  - snowflake: 0.18.0
  - redshift: 0.18.0
  - postgres: 0.18.0

The operating system you're using:
OSX 11.2.3

The output of python --version:
Python 3.6.8

Additional context

Willing to open a PR if you confirm this should be the desired behavior. From what I understand we just need to restrict the max(attencodingtype) to attributes that have a strictly positive attnum?