airsql 0.6.4

A decorator-based SQL execution framework for Airflow

AirSQL

A decorator-based SQL execution framework for Airflow that provides clean, Python-like syntax for data operations.

Features

• 🎯 Decorator-based syntax - Clean, intuitive Python decorators for SQL operations
• 🔗 Native Airflow integration - Uses Airflow connections and follows Airflow patterns
• 🗃️ Multi-database support - Works with Postgres, BigQuery, and more
• 📄 SQL file support - Keep complex queries in separate .sql files with Jinja templating
• ⚡ Flexible outputs - Write to tables, return DataFrames, or save to files
• 🔄 Smart operations - Built-in support for replace, merge/upsert operations
• 🌐 Cross-database queries - Query across different databases (planned with DataFusion)
• 🔍 Data quality checks - Built-in SQL check operators compatible with dbt tests
• 📊 Transfer operators - Move data between BigQuery, Postgres, and GCS
• 👁️ Smart sensors - SQL sensors with retry logic for BigQuery and Postgres

Installation

pip install airsql

Or if you're using uv:

uv add airsql

Quick Start

Basic Usage

1. Simple DataFrame Query

from airsql import sql, Table, File

@sql.dataframe(source_conn="postgres_conn")
def get_active_users():
    return "SELECT * FROM users WHERE active = true"

# Use in DAG
df_task = get_active_users()

2. Query with Table References

@sql.dataframe
def user_activity_analysis(users_table, events_table):
    return """
    SELECT u.id, u.name, COUNT(e.id) as event_count
    FROM {{ users_table }} u
    LEFT JOIN {{ events_table }} e ON u.id = e.user_id
    GROUP BY u.id, u.name
    """

# Use in DAG
analysis_task = user_activity_analysis(
    users_table=Table("postgres_conn", "users.active_users"),
    events_table=Table("bigquery_conn", "analytics.user_events")
)

3. Replace Table Content

@sql.replace(output_table=Table("postgres_conn", "reports.daily_summary"))
def create_daily_report(transactions_table):
    return """
    SELECT DATE(created_at) as date, SUM(amount) as total
    FROM {{ transactions_table }}
    GROUP BY DATE(created_at)
    """

# Use in DAG
report_task = create_daily_report(
    transactions_table=Table("postgres_conn", "transactions.orders")
)

4. Data Quality Checks

@sql.check(conn_id="bigquery_conn")
def test_no_nulls(table):
    return "SELECT COUNT(*) FROM {{ table }} WHERE id IS NULL"

@sql.check(conn_id="postgres_conn")
def test_row_count(table):
    return "SELECT CASE WHEN COUNT(*) > 0 THEN 1 ELSE 0 END FROM {{ table }}"

# Use in DAG
null_check = test_no_nulls(table=Table("bigquery_conn", "analytics.users"))
count_check = test_row_count(table=Table("postgres_conn", "staging.orders"))

5. Transfer Operations

from airsql import BigQueryToPostgresOperator, PostgresToBigQueryOperator

# Transfer from BigQuery to Postgres
bq_to_pg = BigQueryToPostgresOperator(
    task_id="transfer_users",
    source_project_dataset_table="my-project.analytics.users",
    postgres_conn_id="postgres_default",
    destination_table="staging.users",
    gcs_bucket="temp-bucket",
    gcp_conn_id="google_cloud_default"
)

# Transfer from Postgres to BigQuery
pg_to_bq = PostgresToBigQueryOperator(
    task_id="transfer_orders",
    postgres_conn_id="postgres_default",
    sql="SELECT * FROM orders WHERE date >= '2024-01-01'",
    destination_project_dataset_table="my-project.staging.orders",
    gcs_bucket="temp-bucket",
    gcp_conn_id="google_cloud_default"
)

For more examples and detailed documentation, see the full documentation.

Migration from retize.sql

This package is the evolution of retize.sql. The main changes:

• Package renamed from retize.sql to airsql
• Table class schema field renamed to dataset (avoids Pydantic warnings)
• Asset URIs changed from rtz:// to airsql://
• Improved organization with sensors and transfers in submodules

License

This project is licensed under the MIT License.