Snowflake Horizon Catalog

Write Iceberg tables to Snowflake through Snowflake Horizon Catalog.

Supported features

Feature	Status	Remarks
Read/write Snowflake-managed Iceberg tables	✅	Through Horizon's Iceberg REST catalog, without a Snowflake warehouse
Table descriptions in Snowflake	✅
Column descriptions in Snowflake	✅	Comments set in Snowflake take precedence
Read native Snowflake tables	❌️	Only Iceberg tables are available through Horizon, see Make Iceberg the default table format
Read/write views	❌️

Prerequisites

• A Snowflake database for Snowflake-managed Iceberg tables: with an external volume, or with Snowflake-managed storage.

Installation

pip install 'dbt-polars[snowflake]'

Profile configuration

Example

my_project:
  target: dev
  outputs:
    dev:
      type: polars
      catalog_type: snowflake
      schema: dev
      account: myorg-myaccount
      database_name: analytics
      role: transformer
      credential: "{{ env_var('SNOWFLAKE_PAT') }}"

To combine this catalog with others, put the same parameters in an entry of catalogs, with type: snowflake and a name, see Catalog overview.

Parameter Reference

The following parameters are supported, in addition to the common profile parameters:

• catalog_type - (Required) Must be snowflake.

• account - (Required) Account identifier in the form <organization>-<account>, as returned by SELECT CURRENT_ORGANIZATION_NAME() || '-' || CURRENT_ACCOUNT_NAME();.

• database_name - (Required) Snowflake database holding the Iceberg tables.

• schema - (Required) Default schema for models. dbt-polars creates missing schemas.

• role - (Optional) Role dbt-polars acts as.

• Authentication parameters - (Optional) Passed to pyiceberg's REST catalog, which exchanges them for an access token at Snowflake. Commonly used:

  • credential - Programmatic access token (PAT) of the user.
  • token - OAuth access token, e.g. from your identity provider.
  • scope - Overrides the session:role:<ROLE> scope derived from role.

Recommendations

Run models in parallel

For small models, most of the run time is spent waiting for Snowflake's REST API. We recommend threads: 4 with compute_threads: 1, so these calls run in parallel while one model at a time runs Polars. See Running models in parallel.

Make Iceberg the default table format

dbt-polars can only read Iceberg tables. We recommend making Iceberg the default table format in Snowflake, so a plain CREATE TABLE also creates an Iceberg table:

Database

ALTER DATABASE <database_name> SET DEFAULT_METADATA_WRITE_FORMAT = 'ICEBERG';

Schema

ALTER SCHEMA <database_name>.<schema> SET DEFAULT_METADATA_WRITE_FORMAT = 'ICEBERG';

Account

Applies to all databases in the account. Requires the ACCOUNTADMIN role.

ALTER ACCOUNT SET DEFAULT_METADATA_WRITE_FORMAT = 'ICEBERG';

Required permissions

The role dbt-polars uses needs these privileges:

Privilege	On	Needed for
USAGE	the database	Accessing the database at all
CREATE SCHEMA	the database	Creating schemas
USAGE	the external volume	Creating and writing tables
USAGE, CREATE ICEBERG TABLE	schemas dbt-polars didn't create	Creating tables in them
SELECT	tables dbt-polars doesn't own	Reading them, e.g. as a source
INSERT, UPDATE, DELETE, TRUNCATE	tables dbt-polars doesn't own	Writing to them