Jupyter PySpark Notebook

Introduction

• Following this
• Target local Linux
• Future: hopefully easier to run Jupyter or similar notebook with pyspark in AWS/GC
• Non-target: Mac laptop

Requirements

• Recent Linux server with four or more cores.
• Podman, aliased to docker
• A directory to save state into

Installation

• mkdir $WORKBOOK
• ed conf.sh # add $WORKBOOK
• mkdir $WORKBOOK
• sudo ./install-podman.sh

Running

• ./doit.sh
• Visit the 127.0.0.1 link it prints out
• Add these cells

# Setup
from pyspark.sql import SparkSession

# local = this host only, *=use all cores
spark = SparkSession.builder.master("local[*]").getOrCreate()

# add your imports here

from pyspark.sql.functions import *

# Read your data
df = spark.read.option("header", True).csv("10000.csv")

# Analyze your data
df.count

# df.limit(10).toPandas()

References

• Installation: https://jupyter-docker-stacks.readthedocs.io/en/latest/using/common.html
• PySpark test: https://medium.com/@suci/running-pyspark-on-jupyter-notebook-with-docker-602b18ac4494 (The first cell is wrong; use the above Setup instead.)
• PySpark CSV: https://sparkbyexamples.com/pyspark/pyspark-read-csv-file-into-dataframe/
• PySpark getting started: https://spark.apache.org/docs/3.1.1/api/python/getting_started/index.html