Python environments

We are currently transitioning from Python 2.7 to Python 3.5. Most lab projects (e.g. soundsig, zeebeez, songephys) are not yet Python 3 compatible, but we are working on it. For now, use Python 2.7 but try to write Python 3 compatible code.

The two preferred ways to manage Python packages are with anaconda (miniconda), or with virtualenv.

Setup

Python 2 and Python 3 come pre-installed on new versions of Ubuntu and Mac, and should already be on your workstations. If you choose to use anaconda or anaconda3 (Python 3 version), it will modify your bashrc so that you use an installation of Python that comes with anaconda (rather than your default system version).

Creating an environment

anaconda

conda create ENV_NAME

virtualenv

• In the root directory of each project you work on, type virtualenv env --python [python|python3] to create a new environment in env/

• Activate the environment by running source env/bin/activate. This modifies your PATH to point to the virtualenv packages. Do this every time you need to install any packages or run any code, including starting a jupyter notebook.

Installing packages

anaconda

conda install PKG

virtualenv

pip install PKG

Packages installed with pip with a virtualenv activated are installed in env/ and cannot be seen outside the env. Similarly, packages installed with pip outside the virtualenv cannot be seen inside the virtualenv.

To start with a blank slate, just delete the env/ directory and create a new environment.

Defining project dependencies in requirements.txt

Dependencies can be defined in a "requirements" file usually called requirements.txt. Typically, it lists pypi packages with version numbers in the form PKG==VERSION. It can also include references to git repositories and commit hashes to install from a repository.

Install from a requirements file:

pip install -r requirements.txt

Python 2 to 3 migration

Here are some of the common issues you may hit when trying to move from Python 2 to 3

print

The print statement print "hello" has been replaced with the print function print("hello"). Maintain 2 and 3 compatibility by adding from __future__ import print_function and using the new syntax.

division

The division operator / will do floating point division by default. What used to be 7 / 2 == 3 will now be 7 / 2 == 3.5. In Python 3, integer (floor) division can be explicitly done with the // operator, i.e. 7 // 2 == 3. Maintain 2 and 3 compatibility by replacing any instance of integer division with // and adding from __future__ import division.

dictionary items

The old dict.items() would return a list of tuples of (key, value) pairs. In python 3, it will return a dict_items iterable. Iteration as for key, value in d.items() will still work, but other things may break...?

xrange

xrange used to be a iterator version of range. Now, range is an iterator by default and xrange no longer exists. Just replace all usages of xrange with range.

Style guide