cookiecutter-pyspark-cloud

Run PySpark code in the 'cloud' with Amazon Web Services (AWS) Elastic MapReduce (EMR) service in a few simple steps with this cookiecutter project template!

Quickstart

pip install -U "cookiecutter>=1.7"
cookiecutter --no-input https://github.com/daniel-cortez-stevenson/cookiecutter-pyspark-cloud.git
cd pyspark-cloud
make install
pyspark_cloud

Your console will look something like:

Features

• AWS ☁️ Cloudformation Template for EMR: Simple Spark cluster deployment with infrastructure as code

  • JupyterHub is installed to the EMR Master node for development, and is backed by AWS S3 for persistent storage
  • JupyterLab endpoint available at https://master-dns:9443/lab
  • Jupyter Notebook 📔 endpoint available at https://master-dns:9443/tree with sparkmagic kernel
  • Includes examples of bootstrapping your cluster with bash scripts and EMR Steps 👀

• A Command-Line Interface for Running PySpark 'Jobs': For production 🚀 runs via EMR Step API

  • Uses the concept of 'jobs', which run PySpark scripts as a Python function via a common entrypoint - this an important point
  • Checkout the Medium article, which inspired a lot of this

• Log Like a Pro: Save time debugging in style 💃

• Wrap Scala with Python 🐍: Use libraries that haven't been included in the PySpark API!

  • An example of wrapping Scala Spark API code with PySpark API code is provided with SnowballStemmer
  • Could be extended to other Scala MLlib classes (and other Scala classes that implement the UDF interface)

• Simplify Workflows with Make ✅: A Makefile with commands for installation, development, and deployment.

  • use with make [COMMAND]
  • For example, distribute an executable .egg 🥚 distribution of your PySpark code to AWS S3 with make s3dist

• Organize Your Code: Package code shared between 'jobs' in a Python module of your package called common

• Extend the PySpark API: An example of extending the PySpark SQL DataFrame class, which allows chaining custom transformations with dot . notation

  • checkout this awesome PySpark utilities & extensions repo Quinn

• Development Framework: All the tools you need

  • Use bump2version to version your project
  • Use CodeCov to track the completeness of unit tests - see codecov.yml
  • Use [Flake8] to write Python code with common style & formatting conventions

Infrastructure Overview

As defined in the Cloudformation template

Usage

1. Clone this repo:

git clone https://github.com/daniel-cortez-stevenson/cookiecutter-pyspark-cloud.git
cd cookiecutter-pyspark-cloud

2. Create a Python environment with dependencies installed:

conda create -n cookiecutter -y "python=3.7"
pip install -r requirements.txt

conda activate cookiecutter

3. Make any changes to the template, as you wish.

4. Create your project from the template:

cd ..
cookiecutter ./cookiecutter-pyspark-cloud

5. Initialize git:

cd *your-repo_name*
git init
git add .
git commit -m "Initial Commit"

6. Create a new Conda environment for your new project & install project development dependenices:

conda deactivate
conda create -n *your-repo_name* -y "python=3.6"

make install-dev

Contribute

Contributions are welcome! Thanks!

Submit an Bug or Feature Request

Submit a Pull Request

Acknowledgements

Most of the ideas expressed in this repo are not new, but rather expressed in a new way. Thanks, folks! 🙌

• @MrPowers for the DataFrame extension snippet
• @ekampf for the original concept for the pyspark_entrypoint