Pyspark tutorial

Welcome to the Pyspark tutorial section.

The courses comprises of 4 folders containing notebooks. Folders and notebooks are sorted in order of difficulty given their name, so you should follow the numerotation. For example, you should finish all notebooks in 1-beginner before starting 2-novice. Likewise, when doing 2-novice finish the 1-... notebook before doing 2-....

Inside each notebook, we have documented a number of questions and unimplemented code cells answering the question, followed by a code cell which acts as test cases for the function called Graded cell. Your submission will be graded on how many test cases you pass given your implementation of the previous function.

PS: The instructor has hidden test cases on his side so don't try to circumvent the system by just returning the expected value in your method.

Prerequisites

• Anaconda 2019+
• Required Java 8/11.
  • You can set the JAVA_HOME environment variable to point to the Java 8/11 folder you want to use for the project, from Edit the system environment variables window or set JAVA_HOME=<path_to_java> in command-line before running jupyter notebook.
  • You may also install Java OpenJDK inside your Anaconda environment with conda install openjdk. The JAVA_HOME variable should be automatically updated for this environment only.

Run

We provide you with a requirements.txt which is used to download dependencies in a conda environment we will name pyspark-tutorial.

Using Anaconda Navigator

Go to Environments tab then tap Import button. Name it pyspark-tutorial. In the dropdown type of file select Pip requirement file .txt and browse to the requirements.txt file and press enter to create the environment. You should now be able to select the environment.

Go to Environments tab, select the pyspark-tutorial environment. When your mouse is over the environment, you should see a green arrow, click on it and select Open with Jupyter notebook. Then browse to the folder with all the notebooks.

• you may need to define the PYSPARK_PYTHON environment variable so Spark workers can point to the correct Python command.

Using Anaconda prompt

Create Environment:

conda create -n pyspark-tutorial python=3.8
conda activate pyspark-tutorial
pip install -r requirements.txt
# pip install bokeh jupyter numpy pandas psutil pyspark seaborn

Run Notebook:

set PYSPARK_PYTHON=python
jupyter lab

Run a JupyterLab session : jupyter lab from the root of your project, when in your pyspark-tutorial conda environment.

Notes:

• you may run into java.io.FileNotFoundException: HADOOP_HOME and hadoop.home.dir are unset. warnings on Windows. Do not worry about it, they are necessary for remote connections only.
• you may need to define the PYSPARK_PYTHON environment variable so Spark workers can point to the correct Python command.

When you are done with the environment, don't forget to deactivate your Anaconda environment : conda deactivate

Submit

At the end of the course, send your assignments by email to the instructor.

BEFORE SENDING:

• Check that your Jupyter notebook runs from start to finish when you run it from top to bottom, using Kernel > Restart & Run All. Clean up all unecessary cells. Your grade is based on how far the execution of the notebook goes from top to bottom without me going in.
• Don't rename the files, and preserve the folder structure when sending the zipped project, even if you did not complete the notebook. It should look like:

├── 1-beginner
│   └── 1-Initiation.ipynb
├── 2-novice
│   ├── FL_insurance.csv
│   ├── 1-Initiation-RDD.ipynb
│   └── 2-Pagerank-RDD.ipynb
└── 3-advanced
    ├── FL_insurance.csv
    ├── titanic.csv
    ├── 1-Initiation-SparkSQL.ipynb
    └── 2-Advanced-SQL-and-ML.ipynb

Contributing guide

This repo is generated from a private nbgrader project with the solutions to the problems, which I use to autograde the students solutions.

You may create a pull request with the change in the notebook so I copy-paste it in the private repo and overwrite the files. It is normal if I close and don't merge your PR while overwriting the file with your change. Hopefully I find a better way to regenerate the release folder.