Feathr Python Project Developer Guide


  • Navigate to feathr_project folder
  • Install the project by python3 -m pip install -e . This will install a feathr CLI for you. Type feathr in the terminal to see the instructions.

If you get an error similar to “fatal error: ‘librdkafka/rdkafka.h’ file not found”, see the installing librdkafka section of this document.

CLI Usage

  • Run feathr in your terminal to see the instructions.
  • Run feathr init to create a new workspace
  • Navigate to the new workspace
  • Feathr requires a local engine packaged in the jar to test features locally. you can download the jar by feathr start.
  • Run feathr test, then type in a feature, like feature_a
  • After features are fully tested, you can create your training dataset by feathr join. You can also materialize your features to online storage by feathr deploy.
  • You can register your features to the metadata registry by feathr register

Python Coding Style Guide

We use Google Python Style Guide.

Integration Test

Run pytest in this folder to kick off the integration test. The integration test will test the creation of feature dataset, the materialization to online storage, and retrieve from online storage, as well as the CLI. It usually takes 5 ~ 10 minutes. It needs certain keys for cloud resources.

Using Docker

We provide a dockerfile with feathr installed, and with embedded vscode server, and jupyter lab. To build the docker, navigate to Feathr root folder, and use following command to build a docker file: docker build -t <Your Image Name> docker

Run the docker image with port mapping. dockerportmapping

8080 : VsCode (default password is ‘feathr’)

9090: Jupyter Lab (default password is ‘feathr’)

With the docker image running, open your browser and access localhost:<your mapped port> to access apps. codeserver jupyterlab

Using Virtual Environment

It’s recommended to use virtual environment for Python project development.

Using Python VENV

  • Install virtualenv: python3 -m pip install --user virtualenv
  • Make sure you are not using any other virtualenv(either Python or Conda) with: deactivate or conda deactivate
  • Create virtualenv in my_env folder: python3 -m venv my_env. Use a unique name(here my_env), so it doesn’t confuse with other virtual environments.
  • Activate my_env virtualenv: source my_env/bin/activate.
  • After activated, you should see your terminal started with (my_env)
  • To confirm your virtual environment is working, you can type which python and it should show python path is in my_env folder
  • Then follow Installation and Usage.
  • To deactivate virtualenv: deactivate Ref: Installing packages using pip and virtual environments

Using Conda VENV

  • To create an environment: conda create --name myenv
  • To create an environment with a specific version of Python: conda create -n myenv python=3.6
  • To activate yourenvname: conda activate yourenvname
  • Then follow Installation and Usage.
  • To deactivate: conda deactivate Ref: Managing environments

Installing librdkafka

feathr assumes that you have the Apache Kafka C/C++ client library installed. To install it, follow the install instructions on the librdkafka home page.

For the Mac, if you get this error message when installing the project:

 fatal error: 'librdkafka/rdkafka.h' file not found
    #include <librdkafka/rdkafka.h>
    1 error generated.
    error: command '/usr/bin/clang' failed with exit code 1

If this happens,

  1. Run brew info librdkafka, and take note of the library install path (for example, “/opt/homebrew/Cellar/librdkafka/1.8.2/include”),
  2. run export C_INCLUDE_PATH=$LIBRDKAFKA_INCLUDE_PATH, where $LIBRDKAFKA_INCLUDE_PATH is the include path found in step 2.
  3. Rerun the python setup install.