Table of Contents generated with DocToc

• Astro :rocket:
• Overview
• Philosophy
• :mega: Disclaimer :mega:
• Setup
• Using Astro as a SQL Engineer
  • Schemas
  • Setting up SQL files
• Using Astro as a Python Engineer
  • Setting Input and Output Tables
  • Loading Data
  • Transform
  • Putting it All Together
• Other SQL functions
  • Appending data
  • Merging data
  • Truncate table
• Dataframe functionality
  • dataframe
  • ML Operations
  • train
  • predict

Astro :rocket:

Your new Airflow DAG writing experience. Maintained with ❤️ by Astronomer.

Overview

The astro library is a suite of tools for writing ETL and ELT workflows in Airflow. It lets SQL engineers focus on writing SQL, Python engineers focus on writing Python, and all engineers focus on data engineering instead of configuration. By design, astro modules automatically pass database contexts to your tasks, meaning that you can focus on writing code and leave metadata definitions for load time.

• Without astro: Database connections and context are defined in operators. This metadata does not pass automatically from task to task, meaning that you have to redefine it for each new task/ operator. SQL queries must be static.
• With astro: Database connections and context are defined at load time. Metadata is automatically passed into functions, meaning that you can write generic queries in your task and run these queries without knowing the context ahead of time. SQL queries can be templated with JINJA.

Philosophy

With the astro library, we want to redefine the DAG writing experience from the bottom up. Our goal is to empower data engineers and data scientists to write DAGs based around data instead of task dependencies. With this in mind, we built a library focused on data movement and simplifying data transformations between different environments. Our first two integrations are SQL and pandas, but we are planning many more in the coming months.

With our SQL and dataframe modules, you should have the ability to treat SQL tables as if they're python objects. You can manipulate them, join them, templatize them, and ultimately turn them into dataframes if you want to run python functions against them. We hope that this library creates a cleaner Airflow ELT experience, as well as an easier onboarding for those who want to focus on data transformations instead of DAGs.

Please feel free to raise issues and propose improvements. Community contributions are highly welcome!

Thank you,

:sparkles: The Astro Team :sparkles:

:mega: Disclaimer :mega:

This project is still very early and the API will probably change as it progresses. We are actively seeking alpha users and brave souls to test it and offer feedback, but please know that this is not yet ready for production.

Setup

To start using astro:

1. Install astro by running the following command:

   pip install astro-projects
   

   Alternatively, you can add astro-projects to your requirements.txt file.

2. Set the following environment variable so that astro can pass table objects between tasks:

   AIRFLOW__CORE__ENABLE_XCOM_PICKLING=True
   

Using Astro as a SQL Engineer

Schemas

By default, running astro for a given database creates a schema called tmp_astro in the database. This default behavior assumes that you have permissions to create schemas on the fly, and that there is only one user adding/removing from this schema.

For production usage, we recommend that you work with your database administrator to create a shared schema for temporary tables. This schema can be shared across multiple users, but should be created with security in mind (e.g. don't place high security data in a shared schema).

Once you create this schema, an Airflow admin can replace tmp_astro by setting AIRFLOW__ASTRO__SQL_SCHEMA="<new-temp-schema>", or by setting the following in airflow.cfg:

[astro]
sql_schema=<new-temp-schema>

Setting up SQL files

When writing out a SQL DAG using astro, you can think of each SQL file as its own Airflow task. For example, if you wanted to aggregate orders, aggregate customers, and then join customers and orders, you could have the following directory of files:

|
ingest_models/
|
 -- customers_table.sql
 -- orders_table.sql
 -- join_customers_and_orders.sql

In each of these SQL files, standard SELECT statements automatically creates a table that can be referenced in downstream SQL files via a data dependency. astro handles creating all of the temporary tables required for this process.

# join_customers_and_orders.sql
SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM orders c LEFT OUTER JOIN customers p ON c.customer_id = p.customer_id

Defining metadata

Once your SQL is working as expected, you might want to define the query's database and schema during its runtime. To configure this for Airflow while keeping your SQL easy to run in your favorite SQL notebook, you can create a frontmatter:

# join_customers_and_orders.sql
---
database: foo
schema: bar
---
SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM orders c LEFT OUTER JOIN customers p ON c.customer_id = p.customer_id

One benefit of putting all metadata into a frontmatter block is that all of your SQL is still valid outside of the context of Airflow. If you want to develop your SQL locally, comment out the frontmatter block.

# join_customers_and_orders.sql
-- ---
-- database: foo
-- schema: bar
-- ---
SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM orders c LEFT OUTER JOIN customers p ON c.customer_id = p.customer_id

Defining dependencies

When running SQL queries in Airflow DAGs, you need to define dependencies that break up your SQL into multiple, reproducible steps. We offer two ways to define dependencies within an astro SQL file:

You can define your dependency via the template_vars frontmatter argument. In the following example, we set template_vars so that our join_customers_and_orders query cannot run until both our customers_table and agg_orders queries finish. This data dependency is equivalent to a task dependency in our DAG. The only difference is that we're defining it directly in our SQL instead of using Airflow's dependency operators.

# join_customers_and_orders.sql
---
database: foo
schema: bar
template_vars:
    customers: customers_table
    orders: agg_orders
---
SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM orders c LEFT OUTER JOIN customers p ON c.customer_id = p.customer_id

Defining outputs

With certain SQL models, you will want to specify an output based on a table name, schema, and/or database.

Any table created without an output_table will be placed in the temporary schema with a generated table name.

---
database: foo
schema: bar
template_vars:
    customers: customers_table
    orders: agg_orders
output_table:
    table_name: my_pg_table
    database: foo
    schema: my_prod_schema
---
SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM orders c LEFT OUTER JOIN customers p ON c.customer_id = p.customer_id

Supported arguments

Here is a list of supported frontmatter arguments:

Incorporating SQL directory into DAG

Now that you have developed your SQL, we can attach your directory to any DAG using the aql.render function.

The following example DAG pulls a directory of CSV files from s3 into a postgres table, then passes that table into a directory of SQL models that will process the table in an ELT fashion.

import os
from datetime import datetime, timedelta

from airflow.models import DAG

from astro import sql as aql
from astro.sql.table import Table

default_args = {
    "retries": 1,
    "retry_delay": 0,
}

dag = DAG(
    dag_id="sql_file_dag",
    start_date=datetime(2019, 1, 1),
    max_active_runs=3,
    schedule_interval=timedelta(minutes=30),
    default_args=default_args,
)


dir_path = os.path.dirname(os.path.realpath(__file__))
with dag:
    raw_orders = aql.load_file(
        path="s3://my/path/{{ execution_date }}/",
        task_id="pull_from_s3",
        file_conn_id="my_s3_conn",
        output_table=Table(table_name="foo", conn_id="my_postgres_conn"),
    )
    ingest_models = aql.render(dir_path + "/ingest_models", orders_table=raw_orders)

Once the render function completes, it returns a dictionary of all of the models based on the sql file name. If you have a sql file named join_orders_and_customers.sql, then the result would be stored in model["join_orders_and_customers].

Passing on tables to subsequent tasks

The following DAG is an example of passing a single model to a subsequent rendering (perhaps you want to separate your ingest and transforms):

dir_path = os.path.dirname(os.path.realpath(__file__))
with dag:
    raw_orders = aql.load_file(
        path="s3://my/path/{{ execution_date }}/",
        file_conn_id="my_s3_conn",
        output_table=Table(table_name="foo", conn_id="my_postgres_conn"),
    )
    ingest_models = aql.render(dir_path + "/ingest_models", orders_table=raw_orders)
    aql.render(
        dir_path + "/transform_models",
        orders_and_customers=ingest_models["join_orders_and_customers"],
    )

You can also pass the entire dictionary of models to the subsequent task by dereferencing the dictionary using **.

In this example, we pass all tables to the next round of SQL files.

dir_path = os.path.dirname(os.path.realpath(__file__))
with dag:
    raw_orders = aql.load_file(
        path="s3://my/path/{{ execution_date }}/",
        file_conn_id="my_s3_conn",
        output_table=Table(table_name="foo", conn_id="my_postgres_conn"),
    )
    ingest_models = aql.render(dir_path + "/ingest_models", orders_table=raw_orders)
    aql.render(
        dir_path + "/transform_models",
        orders_and_customers=ingest_models["join_orders_and_customers"],
    )

You can even pass the resulting tables into a python function that uses the astro.dataframe to automatically convert your table into a dataframe. We'll discuss how to do this more in "Using Astro as a Python Engineer".

from astro.dataframe import dataframe as df


@df
def aggregate_data(agg_df: pd.DataFrame):
    customers_and_orders_dataframe = agg_df.pivot_table(
        index="DATE", values="NAME", columns=["TYPE"], aggfunc="count"
    ).reset_index()
    return customers_and_orders_dataframe


dir_path = os.path.dirname(os.path.realpath(__file__))
with dag:
    raw_orders = aql.load_file(
        path="s3://my/path/{{ execution_date }}/",
        file_conn_id="my_s3_conn",
        output_table=Table(table_name="foo", conn_id="my_postgres_conn"),
    )
    ingest_models = aql.render(dir_path + "/ingest_models", orders_table=raw_orders)
    aggregate_data(agg_df=ingest_models["agg_orders"])

Using Astro as a Python Engineer

For those who don't want to store their transformations in external SQL files or who want to create transformation functions that are extendable and importable, we offer a rich python API that simplifies the SQL experience for the python engineer!

Setting Input and Output Tables

Before we can complete any transformations, we need to define a way to get our tables in and out of Airflow. We can do this by defining either Table or TempTable objects in the input_table and output_table parameters of our table instantiations.

The Table class

To instantiate a table or bring in a table from a database into the astro ecosystem, you can pass a Table object into the class. This Table object will contain all of the metadata that's necessary for handling table creation between tasks. After you define a Table's metadata in the beginning of your pipeline, astro can automatically pass that metadata along to downstream tasks.

In the following example, we define our table in the DAG instantiation. In each subsequent task, we only pass in an input table argument because astro automatically passes in the additional context from our original input_table parameter.

from astro import sql as aql
from astro.sql.table import Table


@aql.transform
def my_first_sql_transformation(input_table: Table):
    return "SELECT * FROM {input_table}"


@aql.transform
def my_second_sql_transformation(input_table_2: Table):
    return "SELECT * FROM {input_table_2}"


with dag:
    my_table = my_first_sql_transformation(
        input_table=Table(table_name="foo", database="bar", conn_id="postgres_conn")
    )
    my_second_sql_transformation(my_table)

The TempTable Class

Following the traditional dev ops concept of pets vs. cattle, you can decide whether the result of a function is a "pet" (e.g. a named table that you would want to reference later), or a "cattle" that can be deleted at any time for garbage collection.

If you want to ensure that the output of your task is a cattle, you can declare it as a nameless TempTable. This places the output into your temp schema, which can be later bulk deleted. By default, all aql.transform functions will output to TempTables unless a Table object is used in the output_table argument.

In the following example DAG, we set an output_table to a nameless TempTable meaning that any output from this DAG will be deleted once the DAG completes. If we wanted to keep our output, we would simply update the parameter to instantiate a Table instead.

from astro import sql as aql
from astro.sql.table import Table, TempTable


@aql.transform
def my_first_sql_transformation(input_table: Table):
    return "SELECT * FROM {input_table}"


@aql.transform
def my_second_sql_transformation(input_table_2: Table):
    return "SELECT * FROM {input_table_2}"


with dag:
    my_table = my_first_sql_transformation(
        input_table=Table(table_name="foo", database="bar", conn_id="postgres_conn"),
        output_table=TempTable(database="bar", conn_id="postgres_conn"),
    )
    my_second_sql_transformation(my_table)

Loading Data

To create an ELT pipeline, users can first load CSV or parquet data from either local, S3, or GCS into a SQL database with the load_sql function. To interact with S3, you must set an S3 Airflow connection in the AIRFLOW__ASTRO__CONN_AWS_DEFAULT environment variable.

In the following example, we load data from S3 by specifying the path and connection ID for our S3 database in aql.load_file:

from astro import sql as aql
from astro.sql.table import Table

raw_orders = aql.load_file(
    path="s3://my/s3/path.csv",
    file_conn_id="my_s3_conn",
    output_table=Table(table_name="my_table", conn_id="postgres_conn"),
)

Transform

The transform function of the SQL decorator is the "T" of the ELT system. Each step of the transform pipeline creates a new table from the SELECT statement and enables tasks to pass those tables as if they were native Python objects. The following example DAG shows how we can quickly pass tables between tasks when completing a data transformation.

@aql.transform
def get_orders():
    ...


@aql.transform
def get_customers():
    ...


@aql.transform
def join_orders_and_customers(orders_table: Table, customer_table: Table):
    """Join `orders_table` and `customers_table` to create a simple 'feature' dataset."""
    return """SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM {orders_table} c LEFT OUTER JOIN {customer_table} p ON c.customer_id = p.customer_id"""


with dag:
    orders = get_orders()
    customers = get_customers()
    join_orders_and_customers(orders, customers)

Note that the functions in this example use a custom templating system. Wrapping a value in single brackets (like {customer_table}) indicates the value needs to be rendered as a SQL table. The SQL decorator also treats values in double brackets as Airflow jinja templates.

Please note that this is NOT an f string. F-strings in SQL formatting risk security breaches via SQL injections.

For security, users MUST explicitly identify tables in the function parameters by typing a value as a Table. Only then will the SQL decorator treat the value as a table.

Transform File

Another option for larger SQL queries is to use the transform_file function to pass an external SQL file to the DAG. All of the same templating will work for this SQL query.

with self.dag:
    f = aql.transform_file(
        sql=str(cwd) + "/my_sql_function.sql",
        conn_id="postgres_conn",
        database="pagila",
        parameters={
            "actor": Table("actor"),
            "film_actor_join": Table("film_actor"),
            "unsafe_parameter": "G%%",
        },
        output_table=Table("my_table_from_file"),
    )

Raw SQL

Most ETL use cases can be addressed by cross-sharing task outputs, as shown above with @aql.transform. If you need to perform a SQL operation that doesn't return a table but might take a table as an argument, you can use @aql.run_raw_sql.

@aql.run_raw_sql
def drop_table(table_to_drop):
    return "DROP TABLE IF EXISTS {table_to_drop}"

Putting it All Together

The following is a full example DAG of a SQL + Python workflow using astro. We pull data from S3, run SQL transformations to merge our pulled data with existing data, and move the result of that merge into a dataframe so that we can complete complex work on it using Python / ML.

from datetime import datetime, timedelta

from airflow.models import DAG
from pandas import DataFrame

from astro import sql as aql
from astro import dataframe as df
from astro.sql.table import Table

default_args = {
    "owner": "airflow",
    "retries": 1,
    "retry_delay": 0,
}

dag = DAG(
    dag_id="astro_example_dag",
    start_date=datetime(2019, 1, 1),
    max_active_runs=3,
    schedule_interval=timedelta(minutes=30),
    default_args=default_args,
)


@aql.transform
def aggregate_orders(orders_table: Table):
    return """SELECT customer_id, count(*) AS purchase_count FROM {orders_table}
        WHERE purchase_date >= DATEADD(day, -7, '{{ execution_date }}')"""


@aql.transform(conn_id="postgres_conn", database="pagila")
def get_customers(customer_table: Table = Table("customer")):
    """Basic clean-up of an existing table."""
    return """SELECT customer_id, source, region, member_since
        FROM {customer_table} WHERE NOT is_deleted"""


@aql.transform
def join_orders_and_customers(orders_table: Table, customer_table: Table):
    """Now join those together to create a very simple 'feature' dataset."""
    return """SELECT c.customer_id, c.source, c.region, c.member_since,
        CASE WHEN purchase_count IS NULL THEN 0 ELSE 1 END AS recent_purchase
        FROM {orders_table} c LEFT OUTER JOIN {customer_table} p ON c.customer_id = p.customer_id"""


@df
def perform_dataframe_transformation(df: DataFrame):
    """Train model with Python. You can import any python library you like and treat this as you would a normal
    dataframe
    """
    recent_purchases_dataframe = df.loc[:, "recent_purchase"]
    return recent_purchases_dataframe


@df
def dataframe_action_to_sql(df: DataFrame):
    """
    This function gives us an example of a dataframe function that we intend to put back into SQL. The only thing
    we need to keep in mind for a SQL return function is that the result has to be a dataframe. Any non-dataframe
    return will result in an error as there's no way for us to know how to upload the object to SQL.
    """
    return df


SOURCE_TABLE = "source_finance_table"

s3_path = (
    f"s3://astronomer-galaxy-stage-dev/thanos/{SOURCE_TABLE}/"
    "{{ execution_date.year }}/"
    "{{ execution_date.month }}/"
    "{{ execution_date.day}}/"
    f"{SOURCE_TABLE}_"
    "{{ ts_nodash }}.csv"
)

with dag:
    """Structure DAG dependencies.
    So easy! It's like magic!
    """

    raw_orders = aql.load_file(
        path="s3://my/s3/path.csv",
        file_conn_id="my_s3_conn",
        output_table=Table(table_name="foo", conn_id="my_postgres_conn"),
    )
    agg_orders = aggregate_orders(raw_orders)
    customers = get_customers()
    features = join_orders_and_customers(customers, agg_orders)
    simple_df = perform_dataframe_transformation(df=features)
    # By defining the output_table in the invocation, we are telling astro where to put the result dataframe
    dataframe_action_to_sql(
        simple_df, output_table=Table(table_name="result", conn_id="my_postgres_conn")
    )

Other SQL functions

While simple SQL statements such as SELECT statements are very similar between different flavors of SQL, certain functions can vary widely between different SQL systems. This wide variation can lead to issues if a user decides to switch from postgres to snowflake. To simplify this process, we created some high level APIs that handle certain common SQL use-cases to ensure universal interoperability of your DAGs across SQL flavors.

Appending data

After transforming a table, you might want to append the results of your transformation to a reporting table. For example, you might want to aggregate daily data on a "main" table that analysts use for timeseries analysis.

The aql.append function merges tables assuming that there are no conflicts. You can choose to merge the data 'as-is' or cast it to a new value if needed. Note that this query will fail if there is a merge conflict.

foo = aql.append(
    conn_id="postgres_conn",
    database="postgres",
    append_table=APPEND_TABLE,
    columns=["Bedrooms", "Bathrooms"],
    casted_columns={"Age": "INTEGER"},
    main_table=MAIN_TABLE,
)

Merging data

To merge data into an existing table in situations where there might be conflicts, the aql.merge function adds data to a table with either an "update" or "ignore" strategy. The "ignore" strategy does not add values that conflict, while the "update" strategy overwrites the older values. This function only handles basic merge statements. Use the run_raw_sql function for complex statements.

Note that the merge_keys parameter is a list in Postgres, but a map in Snowflake. This syntax decision was unavoidable due to the differences in how Postgres and Snowflake handle conflict resolution. Also note that * inserts are disabled for the merge function.

Postgres:

a = aql.merge(
    target_table=MAIN_TABLE,
    merge_table=MERGE_TABLE,
    merge_keys=["list", "sell"],
    target_columns=["list", "sell", "taxes"],
    merge_columns=["list", "sell", "age"],
    conn_id="postgres_conn",
    conflict_strategy="update",
    database="pagila",
)

Snowflake:

a = aql.merge(
    target_table=MAIN_TABLE,
    merge_table=MERGE_TABLE,
    merge_keys={"list": "list", "sell": "sell"},
    target_columns=["list", "sell"],
    merge_columns=["list", "sell"],
    conn_id="snowflake_conn",
    database="DWH_LEGACY",
    conflict_strategy="ignore",
)

Truncate table

a = aql.truncate(
    table=TRUNCATE_TABLE,
    conn_id="snowflake_conn",
    database="DWH_LEGACY",
)

Dataframe functionality

Finally, your pipeline might call for procedures that would be too complex or impossible in SQL. This could be building a model from a feature set, or using a windowing function which Pandas is more adept for. The df functions can easily move your data into a Pandas dataframe and back to your database as needed.

At runtime, the operator loads any Table object into a Pandas DataFrame. If the Task returns a DataFame, downstream Taskflow API Tasks can interact with it to continue using Python.

If after running the function, you wish to return the value into your database, simply include a Table in the reserved output_table parameters (please note that since this parameter is reserved, you can not use it in your function definition).

dataframe

from astro import dataframe as df
from astro import sql as aql
from astro.sql.table import Table
import pandas as pd


@df
def get_dataframe():
    return pd.DataFrame({"numbers": [1, 2, 3], "colors": ["red", "white", "blue"]})


@aql.transform
def sample_pg(input_table: Table):
    return "SELECT * FROM {input_table}"


with self.dag:
    my_df = get_dataframe(
        output_table=Table(
            table_name="my_df_table", conn_id="postgres_conn", database="pagila"
        )
    )
    pg_df = sample_pg(my_df)

ML Operations

We currently offer two ML based functions: train and predict. Currently these functions do the exact same thing as dataframe, but eventually we hope to add valuable ML functionality (e.g. hyperparam for train and model serving options in predict).

For now please feel free to use these endpoints as convenience functions, knowing that there will long term be added functionality.

train

from astro.ml import train


@train
def my_df_func():
    return pd.DataFrame(data={"col1": [1, 2], "col2": [3, 4]})

predict

from astro.ml import predict


@predict
def my_df_func():
    return pd.DataFrame(data={"col1": [1, 2], "col2": [3, 4]})