[CT-1598] [Regression] Granting select does not work with dbt-postgres==1.3.1

[keurcien]

Is this a regression in a recent version of dbt-core?

• I believe this is a regression in dbt-core functionality
• I have searched the existing issues, and I could not find an existing issue for this regression

Current Behavior

Granting select to a certain user by adding this config to a model used to work with dbt 1.1.x, but after an upgrade to dbt 1.3.0, this raises an error when running dbt run --models xxx

# xxx.sql
{{ 
    config(
        grants = {'select': ['user_a']}
    )
}}

Expected/Previous Behavior

The dbt command should not fail.

Steps To Reproduce

1. Use a postgres profile
2. Create a dbt model one.sql

select 1

3. Add the config

{{ 
    config(
        grants = {'select': ['user_a']}
    )
}}

4.Run the command dbt run --models one

Relevant log output

No response

Environment

- OS: macOS 13.0.1
- Python: python 3.10
- dbt (working version): 1.1.0
- dbt (regression version): 1.3.1

Which database adapter are you using with dbt?

postgres

Additional Context

No response

[jtcohen6]

@keurcien What's the actual error you're seeing?

Granting select to a certain user by adding this config to a model used to work with dbt 1.1.x, but after an upgrade to dbt 1.3.0, this raises an error when running dbt run --models xxx

This is surprising — we only added built-in support for grants in dbt-core v1.2.x, and we didn't make any significant changes from v1.2 → v1.3.

Is it possible that you hand-rolled your own implementation of a grants config? In that case, upgrading from v1.1 → v1.3 would mean that dbt's built-in grants implementation would start taking effect, and it could conflict with your own custom one.

[keurcien]

Ok actually my user was not 'user_a' but 'metabase-proxy' and it might be due to the dash -. We did not have a custom implementation and I'm pretty sure we were under 1.1 and had it running, so this is strange.

{{ 
    config(
        grants = {'select': ['metabase-proxy']}
    )
}}

> dbt run --models one.sql

14:48:30  Running with dbt=1.3.1
14:48:30  Found 106 models, 1 test, 0 snapshots, 0 analyses, 733 macros, 0 operations, 3 seed files, 43 sources, 3 exposures, 0 metrics
14:48:30  
14:48:34  Concurrency: 4 threads (target='dev')
14:48:34  
14:48:34  1 of 1 START sql table model core.one .......................................... [RUN]
14:48:34  1 of 1 ERROR creating sql table model core.one ................................. [ERROR in 0.28s]
14:48:35  
14:48:35  Finished running 1 table model in 0 hours 0 minutes and 4.22 seconds (4.22s).
14:48:35  
14:48:35  Completed with 1 error and 0 warnings:
14:48:35  
14:48:35  Database Error in model one (models/marts/core/one.sql)
14:48:35    syntax error at or near "-"
14:48:35    LINE 4: ...    grant select on "dwh"."core"."one" to metabase-proxy

[jtcohen6]

Ah! So either we need to add quotes to the grantees here, similar to what we do on dbt-bigquery:

dbt-core/core/dbt/include/global_project/macros/adapters/apply_grants.sql

Lines 72 to 74 in c25260e

	{%- macro default__get_grant_sql(relation, privilege, grantees) -%}
	grant {{ privilege }} on {{ relation }} to {{ grantees | join(', ') }}
	{%- endmacro -%}

Or you need to add (double) quotes in your own config:

{{ 
    config(
        grants = {'select': ['"metabase-proxy"']}
    )
}}

There isn't usually a downside to quoting on Postgres/Redshift, but worth considering edge cases.

I still don't see how this is a regression in v1.3, though.

[keurcien]

Oh ok! Yeah my bad because I thought it was broken, gonna close the issue. Thanks for your help!