datapunt-processing 0.0.1a6

Datapunt generic ETL command line scripts and functions for shell scripting in Docker.

Data-processing

At the City of Amsterdam we deal with many different types of structured and unstructered data. Much of the data is not of high quality and are missing proper semantics to do proper analytics with.

This repository combines generic command line functions to create extract, transform and load steps we can then use for creating a reproducable data for analytics and usage in dashboards and maps.

For more information about the how we use these functions in our workflow, read the data-pipeline guide.

How to use

amsterdam.github.io/data-processing

To use a function in python you can use:

from datapunt_processing.extract import download_from_catalog

or

from datapunt_processing.helpers.connections import objectstore_connection

To use the functions directly from the command line in your virtual environment or docker shell you can use it like this:

download_from_data_amsterdam -h

To see the list of command line functions see the modules below or directly in setup.py

Getting Started

To get the functions up and running:

pip install datapunt-processing

To develop the functions locally use these steps:

1. Clone the repository:

git clone https://github.com/Amsterdam/data-processing.git
cd data-processing

2. Create Virtual environment in Windows

# Create and activate a virtual environment, for example with:
python -m venv --copies --prompt data-processing .venv
.venv\Scripts\activate

2. Create Virtual environment in OSX

virtualenv --python=$(which python3) venv
source venv/bin/activate

3. Install the data-processing modules in editable mode

pip install -e .

4. A database is required for the transform and load functions. You can setup your postgres database credentials in the config.ini file to apply to the functions.

If want to use Docker, you can start a database server for your project in a new terminal. The name, port and login of the database can be changed in the docker-compose.yml. Also change them in the config.ini file which will be used by the functions to connect to that database.

docker-compose up -d database

Notebooks

Some of the examples are in the form of runnable Jupyter notebooks. Copies of these with all the images and output included are hosted at Anaconda Cloud. To run these notebooks on your own system, start up a Jupyter notebook server:

To install jupyter:

pip install -e .\[dev\]

jupyter notebook --NotebookApp.iopub_data_rate_limit=100000000

How to Contribute

If you want to contribute please follow the contribute guidelines

0. Prequisites

Fork this repository to your local github account to add add and test new functions.

git clone https://github.com/Amsterdam/data-processing.git

Install the docs,test,dev packages using this command:

pip install -e .[docs,test,dev]
or when using zsh
pip install -e .\[docs,test,dev\]

This package is build by using setuptools to be able to release stable versions on PyPi. It follows some of these guidelines of setting up a python package.

1. Add function

We try to use command line functions as much as possible to ensure we create functions to work easily with different environments and to force yourself creating more generic functions with input variables.

If possible, convert your function into a python-package command line script using the boilerplate_function.py

Add your function to the appropriate folder:

• extract

• transform

• load

• helpers

side note: not all functions are suitable for CL. Machine learning preprocessing steps or general API calls for instance, (that often require parameters in the form of dicts or lists) as input are not suitable and can be used as stand-alone scripts.

2. Add tests

Add test to the test folder and run:

python setup.py test

to test if no other functions are breaking. Correct those issues as well if needed.

3. Add documentation

Create a awesome_module.rst file with Sphinx Argparse extension fields to generate the description and argument fields by reusing an existing rst file. The helpers docs will generate automatically, so you can skip this step if it is placed in the helper function.

Add the rst filename to the list in modules.rst to be found on the main page.

Regenerate the documentation to test the docs output using this command line function:

sphinx/make docs
and test if the readme is not broken:
open docs/index.html

4. Add a Pull Request

Make a PR to add the add your awesome function to our processing code to be reused by many other developpers and data analists.