Metadata-Version: 2.1
Name: schemaflow
Version: 0.1.0
Summary: a package to write data pipelines for data science systematically
Home-page: https://github.com/jorgecarleitao/schemaflow
Author: Jorge C. Leitao
License: UNKNOWN
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Description-Content-Type: text/markdown

[![Build Status](https://travis-ci.org/jorgecarleitao/schemaflow.svg?branch=master)](https://travis-ci.org/jorgecarleitao/schemaflow)
[![Coverage Status](https://coveralls.io/repos/github/jorgecarleitao/schemaflow/badge.svg)](https://coveralls.io/github/jorgecarleitao/schemaflow)
[![Documentation Status](https://readthedocs.org/projects/schemaflow/badge/?version=latest)](https://schemaflow.readthedocs.io/en/latest/?badge=latest)

# SchemaFlow

This is a a package to write data pipelines for data science systematically in Python.
Thanks for checking it out.

## The problem that this package solves

A major challenge in creating a robust data pipeline is guaranteeing interoperability between
pipes: how do we guarantee that the pipe that someone wrote is compatible
with my pipeline *without* running the whole pipeline multiple times until I get it right?

## The solution this package adopts

This package declares an interface to define a stateful data transformation that gives 
the developer the opportunity to declare what comes in, what comes out, and what states are modified
on each pipe and therefore the whole pipeline.

## Install 

    # git clone the repository
    pip install .

## Run tests

    pip install -r requirements_tests.txt
    python -m unittest discover

## Build documentation

    pip install -r requirements_docs.txt
    cd docs && make html && cd ..
    open docs/build/html/index.html

## Use cases

You have a hadoop cluster with csv/etc., use PySpark to process them
and fit a model. There are multiple processing steps developed by many people.


