examples/data_quality/pandera/run_spark.py

"""
This script runs a Hamilton DAG whose intent is to create some features for input to a model.
The way the code is set up, is that you can do one thing:

1. It uses Pandas on Spark to parallelize & scale computation. Use {"execution": "spark"} for this mode.

The model fitting steps are not represented here, just the feature ingestion and transformation logic.

Using Spark only makes sense if you hit "big data" scale. Otherwise, it's a lot of overhead for little value.

To run:
> python run_spark.py
"""

import logging

import data_loaders

# we need to tell hamilton where to load function definitions from
import feature_logic_spark
import pyspark.pandas as ps
from pyspark.sql import SparkSession

from hamilton import base, driver, log_setup
from hamilton.experimental import h_spark

if __name__ == "__main__":
    log_setup.setup_logging(log_level=log_setup.LOG_LEVELS["INFO"])
    spark = SparkSession.builder.getOrCreate()
    # spark.sparkContext.setLogLevel('info')
    ps.set_option(
        "compute.ops_on_diff_frames", True
    )  # we should play around here on how to correctly initialize data.
    ps.set_option("compute.default_index_type", "distributed")  # this one doesn't seem to work?
    logger = logging.getLogger(__name__)

    # passing in execution to help set up the right nodes for the DAG
    config = {"location": "Absenteeism_at_work.csv", "execution": "spark"}
    skga = h_spark.SparkKoalasGraphAdapter(
        spark_session=spark,
        result_builder=base.PandasDataFrameResult(),
        # result_builder=h_spark.KoalasDataFrameResult(),
        spine_column="index_col",
    )
    dr = driver.Driver(
        config, data_loaders, feature_logic_spark, adapter=skga
    )  # can pass in multiple modules
    # we need to specify what we want in the final dataframe.
    output_columns = [
        "age",
        "age_zero_mean_unit_variance",
        "has_children",
        "is_summer",
        "has_pet",
        "day_of_the_week_2",
        "day_of_the_week_3",
        "day_of_the_week_4",
        "day_of_the_week_5",
        "day_of_the_week_6",
        "seasons_1",
        "seasons_2",
        "seasons_3",
        "seasons_4",
        "absenteeism_time_in_hours",
        "index_col",
    ]
    # To visualize do `pip install "sf-hamilton[visualization]"` if you want these to work
    # dr.visualize_execution(output_columns, './my_dag.dot', {})
    # dr.display_all_functions('./my_full_dag.dot')

    # let's create the dataframe!
    df = dr.execute(output_columns)
    print(df.head().to_string())
    spark.stop()