Dev Machine

A simple, highly opinionated automated setup for an OS X developer workstation.

Prerequisites

Install iTerm: TODO: Explain syncing the preferences.

Install the Xcode command line developer tools: xcode-select --install

In the dialog, just choose the command line tools. You won't need the full Xcode unless you're building Cocoa apps.

Installation

Set GITHUB_USER for emacs.d and dotfiles

The installation script will attempt to download two repositories from Github called emacs.d and dotfiles. Set a GITHUB_USER environment variable to the name of the account where these repos can be found: export GITHUB_USER=user_name

If you don't already have these repos, and want to have something you can customize, you can fork mine.

Set DEV_HOME for the root directory of where you hack

This script assumes you use a common directory root for your software development. If you want to use ~/dev, for example, then export DEV_HOME="~/dev".

Running

To install everything: make bootstrap

Have a look at the makefile to see what's included. Run individual targets if you don't need everything.

What opinions are included?

Python

To ensure repeatablity, make python starts with a clean slate by removing any Python libraries previously installed by OS X or homebrew.

When the script completes, add the following to your .bashrc if you don't already have it:

# python env: virtualenv, virtualenvwrapper, and autoenv
export WORKON_HOME=$HOME/dev/.virtualenvs
export PROJECT_HOME=$HOME/dev
export VIRTUALENVWRAPPER_PYTHON=/usr/local/bin/python
export VIRTUALENVWRAPPER_VIRTUALENV=/usr/local/bin/virtualenv
source /usr/local/bin/virtualenvwrapper.sh
source /usr/local/bin/activate.sh
# end python env

Open a new shell to start with a fresh env. Close your old shell.

You now have virtual environments with a few libraries installed.

common-requirements.txt contains concerns truly common to all python projects. Add sparingly to this list.

emacs-requirements.txt contains flake8 and jedi and assumes you're using elpy.

The libraries installed in this top-level brew python installation will be visible in virtualenv's.

virtualenv workflow

Here's an example of how to use virtual environments to create a stats workspace.

mkvirtualenv stats
pip install --upgrade pip setuptools

Add a requirements.txt that installs the dependencies in setup.py, if applicable, along with dev/test dependencies pegged to a specific version:

--index-url https://pypi.python.org/simple/

# equivalent to python setup.py develop
-e .

numpy==1.9.2
scipy==0.15.1
matplotlib==1.4.3
statsmodels==0.6.1
pandas==0.15.2
ipython[notebook]==3.0.0
pytest==2.6.4
networkx==1.9.1
scikit-learn==0.15.2
moto==0.4.1
pytest==2.6.4
httmock==1.2.3

Create a .env file for autoenv to enter the virtualenv whenever you cd into this project directory:

#!/usr/bin/env bash

name=stats

if [[ "`dirname "${BASH_SOURCE}"`" == "`pwd`" && -z "`echo $VIRTUAL_ENV | grep $name`" ]]; then workon $name; fi

Now cd into the project directory, or cd . if you're already there. Answer y to allow the .env file to be sourced; you'll only need to do this the first time.

pip install -r requirements.txt

pip list should show you only the dependencies for this project.

Read more about a virtualenv workflow in the Python Guide.

Updating dependencies in requirements.txt

Use the update_requirements script to generate an updated requirements.txt from a manifest file that lists dependencies without versions, then install the new requirements:

make install-latest-requirements

This allows the requirements files to be updated in discrete intervals.

...and other opinionated choices

Check the makefile for more information.