[Bug] dbt build with --empty flag compiles unit tests with empty mock fixtures

[zendesk-cisenbe]

Is this a new bug in dbt-core?

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

Current Behavior

When you perform dbt build -s <some_model> --empty, and there are unit tests associated with <some_model>, the fixtures for the mocks in the unit test will return 0 records, but the expectations continue to use the full fixture. This always leads to a failure of the unit test.

The compiled code for the fixture looks like this:

select * from __dbt__cte__upstream_ref where false limit 0

Expected Behavior

Unit tests should ignore the --empty flag. Fixtures for both mocks and expectations should be respected regardless of tags.

Steps To Reproduce

1. In a dbt project containing a model with unit tests with fixtures (i. e., ref() or source())
2. run dbt build -s <some_model> --empty
3. The unit test will run first and fail because the fixture will return 0 rows, which will never compare to a populated expectation

Relevant log output

No response

Environment

- OS: Ubuntu 22.04.4
- Python: 3.11.9
- dbt: 1.8.3

Which database adapter are you using with dbt?

snowflake

Additional Context

No response

[dbeatty10]

Thanks for reporting this @zendesk-cisenbe.

I was able to reproduce this with the following files and commands:

Reprex

models/my_upstream_model.sql

select 1 as id

models/my_downstream_model.sql

select *
from {{ ref("my_upstream_model") }}

models/_unit_tests.yml

unit_tests:
  - name: test_10516

    model: my_downstream_model

    given:
      - input: ref("my_upstream_model")
        rows:
        - {id: 2}

    expect:
      rows:
        - {id: 2}

Commands:

dbt build -s my_downstream_model --empty

Output:

$ dbt build -s my_downstream_model --empty

23:59:31  Running with dbt=1.8.3
23:59:32  Found 2 models, 444 macros, 1 unit test
23:59:32  
23:59:34  Concurrency: 10 threads (target='snowflake')
23:59:34  
23:59:34  1 of 2 START unit_test my_downstream_model::test_10516 ......................... [RUN]
23:59:37  1 of 2 FAIL 1 my_downstream_model::test_10516 .................................. [FAIL 1 in 2.63s]
23:59:37  2 of 2 SKIP relation dbt_dbeatty.my_downstream_model ........................... [SKIP]
23:59:37  
23:59:37  Finished running 1 unit test, 1 view model in 0 hours 0 minutes and 4.89 seconds (4.89s).
23:59:37  
23:59:37  Completed with 1 error and 0 warnings:
23:59:37  
23:59:37  Failure in unit_test test_10516 (models/_unit_tests.yml)
23:59:37    

actual differs from expected:

@@ ,ID
---,2


23:59:37  
23:59:37    compiled code at target/compiled/my_project/models/_unit_tests.yml/models/test_10516.sql
23:59:37  
23:59:37  Done. PASS=0 WARN=0 ERROR=1 SKIP=1 TOTAL=2

Take a look at the unit code logic:

target/run/my_project/models/_unit_tests.yml/models/test_10516.sql

-- Build actual result given inputs
with dbt_internal_unit_test_actual as (
  select
    id, 'actual' as "actual_or_expected"
  from (
    with __dbt__cte__my_upstream_model as (

-- Fixture for my_upstream_model
select 
    
        try_cast('2' as NUMBER(1,0))
     as id
) select *
from (select * from __dbt__cte__my_upstream_model where false limit 0)
  ) _dbt_internal_unit_test_actual
),
-- Build expected result
dbt_internal_unit_test_expected as (
  select
    id, 'expected' as "actual_or_expected"
  from (
    select 
    
        try_cast('2' as NUMBER(1,0))
     as id
  ) _dbt_internal_unit_test_expected
)
-- Union actual and expected results
select * from dbt_internal_unit_test_actual
union all
select * from dbt_internal_unit_test_expected

As mentioned by @zendesk-cisenbe, note in the unit test fixture within the reprex where it is including (select * from __dbt__cte__my_upstream_model where false limit 0) instead of just __dbt__cte__my_upstream_model.

[jonathanneo]

This is affecting my company from running dbt unit tests in CI, as we run our CI pipelines with the --empty flag. This is impacting ~350 dbt-core users.