Reliably manage your python dev environment.
xe stands for eXecutable Environment.
There are a ton of “best practices” for python projects. We can all agree that you should use a virtualenv and a test runner such as pytest or nose. Pip is another good tool to use. Sphinx is great for documentation. The list goes on.
The problem is that while we agree on the tools, we seldom agree on how we should use the tools. Some people use virtualenvwrapper and prefer to hide all their virtualenv’s in a hidden directory. Others prefer to create a directory within the project root. Others like to do a virtualenv . to create a virtualenv for a project so activation does the “Right Thing” with the prompt and uses the project name. Not to mention tools like tox that help create virtualenvs for different versions of Python and make sure your dev environment doesn’t mix with your test environment.
While I’m positive that I can’t bring the Python world together in harmony, I can write a tool to make this sort of environment management and automate my own standard practices. I’m calling this tool xe.
xe is slightly opinionated in decisions it makes regarding a project’s dev environment. These opinions are never meant to be controversial! In fact the goal is to be as benign as possible in hopes that xe will easily support any developer’s workflow.
With that in mind, there are some ideals that xe tries to maintain in order to be general and easy to work with.
- all commands should be runnable without a required “activation” step
- a project should have its own environment
If you’ve ever deployed a project only to realize that you failed to update a dependency, then there is a good chance you were bitten by a fragile environment. While it is handy to be able to “activate” your environment, it makes it really easy to miss things when you are updating dependencies.
Xe explicitly avoids shell level activation and instead finds your virtualenv on every command.
Another area xe helps is when you use an IDE type tool. Most editors and IDEs have the idea of a project. In the project settings you can configure builds that typically map to running tests or project tasks. If you are using a non-virtualenv aware tool, you usually have to configure environment variables in order to make sure the correct virtualenv is used. Even if your tool does understand virtualenvs, you still need to supply some configuration to the correct environment.
Using xe you can easily configure the default build as xe test $fn. There is not a list of environment variables you have to setup and configure. You don’t have to specify a virtualenv directory or testing tool. You can let xe takes care of it.
A project should have its own environment because you should be testing and using that project in isolation. That doesn’t mean you shouldn’t have subrepos or install other packages as editable. It simply means that if you are working in a project directory, you should be using that project’s environment.
Along similar lines, a project environment should be easy to delete and rebuild from scratch. Using xe, the default behavior is to create a directory local virtualenv that can be removed and rebuilt from scratch when necessary.
Here are the standard project files and directories that xe utilizes.
- venv/ (virtualenv directory)
Most of these can be configured to point to other non-root directories, but if you use these files, xe will try to work out of the box without extra configuration.
It would be nice to eventually support different build tools, but I imagine that will be implemented via plugins. The idea being that an organization could implement their own build entry points and use xe to run them correctly.
Typically you’d read this first, but as this is our code we’re talking about we needed to get the prereq’s out of the way and make sure that your project isn’t going to get borked by xe. Assuming things look reaonable, you can get started by doing:
$ xe bootstrap
xe will create a virtualenv if one hasn’t been created yet. It will then look for a dev_requirements.txt and run that in the newly create environment. From there you can use xe to run tasks. A good default is simply running python:
$ xe python
That will start up the python in the xe virtualenv. If you are using a django project you can use the following shortcut to have access to your manage.py commands:
$ xe manage runserver
If you use pytest, you can run your tests too:
$ xe test -x
Say you build your docs with make, you can use xe to run make and be confident your environment will be in place.
$ xe make html
To use rdo for all commands, add to your .xerc:
If you only want to use rdo for specific commands, specify them via the RDO_COMMANDS field:
RDO_COMMANDS: - make - python