fix(bigquery): qualify column refs in test_unique / test_not_null to avoid row-struct resolution

[1fanwang]

resolves dbt-labs/dbt-core#11067
docs N/A

Problem

When a column shares its name with the model, BigQuery's unique and not_null generic tests can incorrectly pass even with duplicate or null column values. Reproducer per dbt-core#11067:

# models/orders.sql
select 1 as orders union all
select 1 as orders union all
select cast(null as int64) as orders

# models/_models.yml
models:
  - name: orders
    columns:
      - name: orders
        data_tests:
          - unique
          - not_null

Root cause: BigQuery's identifier-resolution rule. When a column's name matches its containing relation's name and the relation is referenced without an alias, an unqualified reference like orders resolves to the row STRUCT containing all of the row's columns, not to the column value itself.

The current bigquery__test_unique and the inherited default__test_not_null emit unqualified column references:

{# bigquery__test_unique today #}
with dbt_test__target as (
  select {{ column_name }} as unique_field
  from {{ model }}
  where {{ column_name }} is not null
)
...

so the WHERE / GROUP BY apply against the row struct rather than the column, masking the violations the tests are meant to catch.

Solution

Alias the source relation as dbt_test__source and qualify column references in bigquery__test_unique. Add a bigquery__test_not_null override that does the same.

{% macro bigquery__test_unique(model, column_name) %}
with dbt_test__target as (
  select dbt_test__source.{{ column_name }} as unique_field
  from {{ model }} dbt_test__source
  where dbt_test__source.{{ column_name }} is not null
)
...
{% endmacro %}

{% macro bigquery__test_not_null(model, column_name) %}
{% set column_list = '*' if should_store_failures() else 'dbt_test__source.' ~ column_name %}
select {{ column_list }}
from {{ model }} dbt_test__source
where dbt_test__source.{{ column_name }} is null
{% endmacro %}

Same shape suggested in the review of the original stale attempt at dbt-core#2061 / PR dbt-core#2075. Limited the change to BigQuery because Postgres / Snowflake / Redshift / Spark resolve unqualified column references column-first, so the bug does not manifest there and changing the default macro would have a wider blast radius.

Tradeoffs considered:

• Updating default__test_unique / default__test_not_null instead. Cleaner long-term, but riskier for adapters that override those defaults or user macros that shim them. BigQuery-only is safer; the default macros can still be changed if reviewers want the broader fix.
• Different alias name. dbt_test__source mirrors the existing dbt_test__target CTE. A column literally named dbt_test__source would still resolve correctly because dbt_test__source.{{ column_name }} is alias-dot-column syntax. This can be renamed if reviewers prefer something less collision-prone.

Tests

Added dbt-bigquery/tests/functional/adapter/generic_tests/test_column_table_name_collision.py. It builds the reproducer model from the issue, runs dbt test, and asserts that both the unique and not_null tests fail with at least one failure each.

I have not been able to run the BigQuery integration tests locally because this environment has no project credentials. CI will need to validate the fixture shape.

Checklist

• I have read the contributing guide and understand what's expected of me
• I have run this code in development and it appears to resolve the stated issue (BQ integration tests not run locally; relying on CI)
• This PR includes tests, or tests are not required/relevant for this PR
• This PR has no interface changes (e.g. macros, cli, logs, json artifacts, config files, adapter interface, etc) or this PR has already received feedback and approval from Product or DX

[cla-bot]

Thanks for your pull request, and welcome to our community! We require contributors to sign our Contributor License Agreement and we don't seem to have your signature on file. Check out this article for more information on why we have a CLA.

In order for us to review and merge your code, please submit the Individual Contributor License Agreement form attached above above. If you have questions about the CLA, or if you believe you've received this message in error, please reach out through a comment on this PR.

CLA has not been signed by users: @1fanwang

[Copilot]

Pull request overview

Fixes BigQuery generic unique/not_null tests when a column name collides with its model name by qualifying column references against an explicit source alias.

Changes:

• Update bigquery__test_unique to alias the source relation and qualify column_name references.
• Add a BigQuery-specific bigquery__test_not_null override with the same qualification approach.
• Add a functional regression test covering the column/table-name collision scenario and a changelog entry.

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 4 comments.

File	Description
dbt-bigquery/src/dbt/include/bigquery/macros/adapters.sql	Qualifies column references in BigQuery unique and adds a BigQuery override for not_null.
dbt-bigquery/tests/functional/adapter/generic_tests/test_column_table_name_collision.py	New regression test ensuring both unique and not_null fail correctly for the collision case.
dbt-bigquery/.changes/unreleased/Fixes-20260427-083810.yaml	Adds a Fixes changelog entry describing the behavior change.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[cla-bot]

Thanks for your pull request, and welcome to our community! We require contributors to sign our Contributor License Agreement and we don't seem to have your signature on file. Check out this article for more information on why we have a CLA.

In order for us to review and merge your code, please submit the Individual Contributor License Agreement form attached above above. If you have questions about the CLA, or if you believe you've received this message in error, please reach out through a comment on this PR.

CLA has not been signed by users: @1fanwang

[1fanwang]

Pushed 92a39c5 — wraps model in (select * from {{ model }}) before the dbt_test__source alias in both macros so where:-filtered tests don't end up double-aliased, switched the result lookup to prefix matching, and added a where-filtered regression case. Updated the changelog accordingly.