Incorrect data_type for BigQuery REPEATED fields.

[Thrasi]

Describe the bug

BigQuery supports arrays as REPEATED fields
Running generate_model_yaml for a model with a repeated field of a certain datatype will result in a
schema with data_type: datatype rather than data_type: array<datatype>
Running the model again with contract.enforced=true will show the error:

column_name	definition_type	contract_type	mismatch_reason
repeated_int	ARRAY<INT64>	INT64	data type mismatch

Repeated records should have data_type: array

Steps to reproduce

create a model_with_repeated_field.sql:

select 
    [1,2,3,4] as repeated_int

run it

$ dbt run -s model_with_repeated_field -f

Expected results

version: 2

models:
  - name: model_with_repeated_field
    description: ""
    columns:
      - name: repeated_int
        data_type: array<int64>
        description: ""

Actual results

version: 2

models:
  - name: model_with_repeated_field
    description: ""
    columns:
      - name: repeated_int
        data_type: int64
        description: ""

Screenshots and log output

Running the model with the following yml:

models:
  - name: model_with_repeated_field
    config:
      contract:
        enforced: true
    columns:
      - name: repeated_int
        data_type: int64
        description: ""

System information

packages:
  - package: dbt-labs/codegen
    version: 0.12.1
  - package: dbt-labs/dbt_utils
    version: 1.1.1

Which database are you using dbt with?

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

The output of dbt --version:

Core:
  - installed: 1.8.8
  - latest:    1.8.8 - Up to date!

Plugins:
  - bigquery: 1.8.3 - Up to date!

The operating system you're using:
MacOS Sequoia Version 15.1

The output of python --version:
Python 3.10.15

Additional context

generate_model_yaml gets the data_type using data_type_format_model

    {% if include_data_types %}
        {% do model_yaml.append('        data_type: ' ~ codegen.data_type_format_model(column)) %}
    {% endif %}

which in turn calls codegen.format_column

{% macro data_type_format_model(column) -%}
  {{ return(adapter.dispatch('data_type_format_model', 'codegen')(column)) }}
{%- endmacro %}

{# format a column data type for a model #}
{% macro default__data_type_format_model(column) %}
    {% set formatted = codegen.format_column(column) %}
    {{ return(formatted['data_type'] | lower) }}
{% endmacro %}

format_column is vendored in macros/vendored/dbt_core/format_column.sql
But it doesn't handle the specific case for repeated fields.

I am tempted to create a default__format_column and a bigquery__format_column to handle BigQuery specifically:

{% macro format_column(column) -%}
  {{ return(adapter.dispatch('format_column', 'codegen')(column)) }}
{%- endmacro %}

{% macro default__format_column(column) -%}
  {% set data_type = column.dtype %}
  {% set formatted = column.column.lower() ~ " " ~ data_type %}
  {{ return({'name': column.name, 'data_type': data_type, 'formatted': formatted}) }}
{%- endmacro -%}

{% macro bigquery__format_column(column) -%}
  {% if column.mode.lower() == "repeated" %}
    {% set data_type = "array" if column.dtype.lower() == "record" else "array<" ~ column.dtype ~ ">" %}
  {% else %}
    {% set data_type = column.dtype %}
  {% endif %}
  {% set formatted = column.column.lower() ~ " " ~ data_type %}
  {{ return({'name': column.name, 'data_type': data_type, 'formatted': formatted}) }}
{%- endmacro -%}

and moving it all into helpers.sql while removing the vendored format_column

Alternatively, bigquery__format_column could return a 'mode': column.mode field
and the adapter specific datatype conversion would be taken care of in
bigquery__data_type_format_model and bigquery__data_type_format_source

Are you interested in contributing the fix?

Yes. I would like some input on whether the proposed solution in Additional context is reasonable.