Implementation:MaterializeInc Materialize MaterializeAdapter Relations

Knowledge Sources	misc/dbt-materialize/dbt/adapters/materialize/impl.py, misc/dbt-materialize/dbt/adapters/materialize/relation.py
Domains	Declarative Data Modeling, Relation Type System, Catalog Introspection
Last Updated	2026-02-08

Overview

Concrete adapter and relation classes for Materialize provided by dbt-materialize, implementing relation type discrimination, catalog listing, constraint rendering, and column schema introspection.

Description

The MaterializeAdapter class extends both PostgresAdapter and SQLAdapter to provide Materialize-specific behavior for managing data models and relations. It binds together the MaterializeConnectionManager and MaterializeRelation classes, and overrides several PostgreSQL-specific methods.

MaterializeRelation is a frozen dataclass extending PostgresRelation that:

• Carries a type field of type MaterializeRelationType, an enum with values: Table, View, CTE, External, MaterializedView, Source, SourceTable, Sink, and the legacy MaterializedViewLegacy.
• Overrides the maximum identifier length to 255 characters (PostgreSQL limits to 63).
• Provides boolean properties: is_materialized_view, is_source, is_source_table, is_sink.

MaterializeAdapter provides:

• list_relations_without_caching(): Queries the Materialize catalog to list all relations in a schema, mapping each to its MaterializeRelationType.
• render_column_constraint(): Renders NOT NULL constraints as ASSERT NOT NULL (Materialize-specific syntax).
• render_raw_columns_constraints(): Processes raw column constraint definitions from model configurations.
• get_column_schema_from_query(): Creates a temporary view to introspect column names and types without executing the full query (avoids needing a valid cluster).
• CONSTRAINT_SUPPORT: Only not_null is enforced; check, unique, primary_key, and foreign_key are not supported.

Usage

Use this implementation when:

• Working with Materialize-specific relation types (sources, sinks, source tables) in dbt models.
• Understanding how dbt resolves and caches Materialize catalog metadata.
• Debugging constraint rendering for materialized views with ASSERT NOT NULL.
• Investigating how column type introspection works without requiring a valid cluster.

Code Reference

Source Location

File	Lines	Description
misc/dbt-materialize/dbt/adapters/materialize/impl.py	L110-358	MaterializeAdapter class
misc/dbt-materialize/dbt/adapters/materialize/relation.py	L27-43	MaterializeRelationType enum
misc/dbt-materialize/dbt/adapters/materialize/relation.py	L45-78	MaterializeRelation dataclass
misc/dbt-materialize/dbt/adapters/materialize/impl.py	L194-219	list_relations_without_caching()
misc/dbt-materialize/dbt/adapters/materialize/impl.py	L275-339	get_column_schema_from_query()

Signature

class MaterializeRelationType(StrEnum):
    Table = "table"
    View = "view"
    CTE = "cte"
    External = "external"
    MaterializedView = "materialized_view"
    Source = "source"
    SourceTable = "source_table"
    Sink = "sink"
    MaterializedViewLegacy = "materializedview"


@dataclass(frozen=True, eq=False, repr=False)
class MaterializeRelation(PostgresRelation):
    type: Optional[MaterializeRelationType] = None
    require_alias: bool = False

    def relation_max_name_length(self) -> int: ...

    @classproperty
    def get_relation_type(cls) -> Type[MaterializeRelationType]: ...

    @property
    def is_materialized_view(self) -> bool: ...

    @property
    def is_source(self) -> bool: ...

    @property
    def is_source_table(self) -> bool: ...

    @property
    def is_sink(self) -> bool: ...


class MaterializeAdapter(PostgresAdapter, SQLAdapter):
    ConnectionManager = MaterializeConnectionManager
    Relation = MaterializeRelation
    AdapterSpecificConfigs = MaterializeConfig

    def list_relations_without_caching(
        self, schema_relation: MaterializeRelation
    ) -> List[MaterializeRelation]: ...

    @classmethod
    def render_column_constraint(
        cls, constraint: ColumnLevelConstraint
    ) -> str: ...

    def get_column_schema_from_query(
        self, sql: str
    ) -> List[PostgresColumn]: ...

Import

from dbt.adapters.materialize import MaterializeAdapter
from dbt.adapters.materialize.relation import MaterializeRelation, MaterializeRelationType

I/O Contract

Inputs

Parameter	Type	Description
schema_relation (for list_relations_without_caching)	MaterializeRelation	A relation object identifying the target schema to list relations from.
constraint (for render_column_constraint)	ColumnLevelConstraint	A parsed column constraint object. Only ConstraintType.not_null produces output.
sql (for get_column_schema_from_query)	str	A SQL query whose result column names and types are to be introspected.

Outputs

Return	Type	Description
(from list_relations_without_caching)	List[MaterializeRelation]	All relations in the specified schema, each with its correct MaterializeRelationType.
(from render_column_constraint)	str	The string "assert not null" for NOT NULL constraints, or "" for unsupported types.
(from get_column_schema_from_query)	List[PostgresColumn]	Column objects with name and type information derived from a temporary view.

Usage Examples

# Checking if a relation is a materialized view
from dbt.adapters.materialize.relation import MaterializeRelation, MaterializeRelationType

rel = MaterializeRelation.create(
    database="materialize",
    schema="public",
    identifier="my_model",
    type=MaterializeRelationType.MaterializedView,
)

assert rel.is_materialized_view is True
assert rel.is_source is False
assert rel.relation_max_name_length() == 255

-- Example dbt model (materialized view) referencing a source
{{ config(materialized='materialized_view', cluster='quickstart') }}

SELECT
    id,
    name,
    updated_at
FROM {{ source('my_source', 'events') }}
WHERE updated_at > NOW() - INTERVAL '7 days'

Related Pages

Implements Principle

• Principle:MaterializeInc_Materialize_Source_and_Model_Definition

Requires Environment

• Environment:MaterializeInc_Materialize_Dbt_Materialize_Runtime