dagster-sqlmesh

WARNING: THIS IS A WORK IN PROGRESS

SQLMesh library for dagster integration.

Current features

• A @sqlmesh_assets decorator akin to dagster-dbt's @dbt_assets decorator.
• A SQLMeshResource that allows you to call sqlmesh from inside an asset (likely one defined by the @sqlmesh_assets decorator)
• A SQLMeshDagsterTranslator that allows customizing the translation of sqlmesh models into dagster assets.

Basic Usage

This dagster sqlmesh adapter is intended to work in a similar pattern to that of dagster-dbt in the most basic case by using the @sqlmesh_assets

Assuming that your sqlmesh project is located in a directory /home/foo/sqlmesh_project, this is how you'd setup your dagster assets:

from dagster import (
    AssetExecutionContext,
    Definitions,
)
from dagster_sqlmesh import sqlmesh_assets, SQLMeshContextConfig, SQLMeshResource

sqlmesh_config = SQLMeshContextConfig(path="/home/foo/sqlmesh_project", gateway="name-of-your-gateway")

@sqlmesh_assets(environment="dev", config=sqlmesh_config)
def sqlmesh_project(context: AssetExecutionContext, sqlmesh: SQLMeshResource):
    yield from sqlmesh.run(context)

defs = Definitions(
    assets=[sqlmesh_project],
    resources={
        "sqlmesh": SQLMeshResource(config=sqlmesh_config),
    },
)

Contributing

We are very open to contributions!

In order to build the project you'll need the following:

• python 3.11 or 3.12
• node 18+
• pnpm 8+

Note: this is a python project but some of our dependent tools are in typescript. As such all this is needed

Installing

The project uses Make commands to simplify the development setup process. To get started:

make init

This will:

• Set up a Python virtual environment with Python 3.12
• Install all Python dependencies
• Install Node.js dependencies via pnpm

Note: All Make commands automatically use the correct virtual environment - you don't need to activate it manually.

To upgrade dependencies:

make upgrade-python-deps  # Upgrade Python dependencies
make upgrade-node-deps   # Upgrade Node.js dependencies

Running tests

We have tests that should work entirely locally. You may see a db.db file appear in the root of the repository when these tests are run. It can be safely ignored or deleted.

To run tests:

make test

Running the "sample" dagster project

In the sample/dagster_project directory, is a minimal dagster project with the accompanying sqlmesh project from sample/sqlmesh_project configured as an asset. To run the sample dagster project deployment with a UI:

make dagster-dev 

or

make dev

If you'd like to materialize the dagster assets quickly on the CLI:

make dagster-materialize

Note: The sqlmesh project that is in the sample folder has a dependency on a table that doesn't exist by default within the defined duckdb database. You'll notice there's a test_source asset in the dagster project. This asset will automatically populate that table in duckdb so that the sqlmesh project can be run properly. Before you run any materializations against the sqlmesh related assets in dagster, ensure that you've run the test_source at least once.

Future Plans

• Create a new "loader" for sqlmesh and dagster definitions to allow for automatic creation of administrative jobs for sqlmesh (e.g. migrations). Additionally, we may want to have this generate assets outside of the multi_asset paradigm within dagster such that assets can have independent partitions. There is an existing issue for this in dagster itself.