Data Science Template

This repository contains a project creation tool for Data Science models. The main usage is quite simple, but we have a lot of features working OUT OF THE BOX:

• Dockerfile template;
• Unit-tests architecture with coverage report;
• Makefile for project management and Docker abstraction;
• Python software distribution via setup.py;
• Multiple releases formats into /dist: .tar.gz, .egg and .whl;
• Documentation: Project Specification && Model Report;
• Notebook template for data analysis;
• Multiple linters to help catching bugs during development:
  • security: bandit;
  • style: flake8;
  • typing: mypy;
• CI/CD configuration with 3 stages: test, build and pages
  • tests: run unit-tests;
  • build: build docker image and populate /dist;
  • pages: build html documentation using Sphinx and publish using Read The Docs;

Usage

So let's create a new model! You just need clone this repository and use the bash script create-model as follow:

git clone git@github.com:NeowayLabs/data-science-template.git
cd data_science_template
./create-model ~/awesome-model

The basic placeholders like project name, python package name are automatically filled! If you have configured the git user.name and git user.email the system will assume to be equal as author and email and will replace these placeholders as well. The directory created will be initialized as a git repository, but you still need configure your remote url manually.

The create-model program will help you configure the remote printing some instructions to do that correctly, keep attention to blue exclamation marks [!] on output of create-model. But basically you need do two steps:

1. Create a new project in GitHub.
2. Get the ssh or https remote from it.

You can test the remote address executing a git clone command:

git clone <your-repository-address-here>

To enable properly CI/CD functionality for your new repository you can do the following:

1. Go to Travis CI.
2. Select your repository to use with Travis CI. These instructions may help.

After this, update your git remote origin and do the first commit:

cd ~/awesome-model
make check  # check if everything is ok and tests are passing
git add .
git commit -m "Initial commit" # Or your desired message
git remote add origin <your-repository-address-here>
git push -u origin master

Releasing Software

You can release a new version of your software with this command:

make release VERSION=X.Y.Z

That is all. That command will build the docker image, run the tests, linters and if everything is ok: a new git tag will be created and be pushed to GitHub.

Author

Neoway @ Data Analytics Team