A PyCon Inspired Post - Artisan Testing

Being at PyCon, I was inspired to do something. At work, I am typically writing CFEngine3 policy and not writing a lot of code. I should be filling my free time with more projects. Being around all the other great Python developers and getting to talk to them about their work (I cannot rave about BreakfastSerial enough), I was inspired to pick up where I left off with a project of mine. Part of it probably comes from Kenneth Reitz's talk, Python for Humans. He said that if something in python is confusing, you should work to make it more clear, more predictable, and easier to get started with. The project I left off on was 'artisan' - a library for generating complete instances for testing, rather than stubbing, mocking, or any of that other stuff.

Why create a library when there are copious amounts of them already in PyPI? Well, I didn't want to use stubs mocks. I don't like stubs or mocks. When testing, I like to use the logic already in existence for my instances. Also, if you look at the work that it takes to do mocking, it seems like a waste of time, it isn't straight forward, nor is it clear. You're duplicating a lot of the energy you have already spent writing your models.

Part of my views might be skewed by the fact that I have done a lot of ruby (I'm a recovering rubyist), but in ruby, testing is a core part of the culture. Testing is important. Ruby has the ability to write beautiful tests. Python is beautiful, so it should have the same ability.

Artisan Testing

Installing Artisan

Installing is nice and simple. It is available in PyPI.

pip install artisan

If you are not using a virtualenv, you probably will want to add the '--user' flag. However, you will also probably want to start using virtualenvs. Seriously, they're awesome.

Using Artisan

Let's assume there is a simple app you need to test with the Foo class. The Foo class looks a lot like:

class Foo(object):
    ''' This is a simple object '''

    def __init__(self):
        self.name = None

Before you can use artisan, you need to setup blueprints. Blueprints are simply dictionaries that contain lambdas used to generate the random data. In my example, I use the great python-faker library (it is available on PyPI, but I find the git repo is far more current).

In 'blueprints.py':

import faker

Foo = {
  'name': lambda: faker.name.name(),
}

In your tests, you then do:

import blueprints
import artisan

artisan.prepare(blueprints)

artisan.craft(Foo)                # Creates an instance of the Foo class with a randomly generated name
artisan.craft(Foo, name='BarBaz') # Creates an instance of the Foo class with the name BarBaz

A Complete Example

Python 3.3.0 (default, Mar 13 2013, 18:04:53) 
[GCC 4.2.1 20070719 ] on openbsd5
Type "help", "copyright", "credits" or "license" for more information.
>>> import blueprints
>>> import artisan
>>> 
>>> class Foo(object):
...     def __init__(self):
...         self.name = None
... 
>>> 
>>> artisan.prepare(blueprints)
>>> artisan.craft(Foo).name
'Mrs. Savion Kuhn'
>>> artisan.craft(Foo).name
'Ms. Pearlie Collier'
>>> type(artisan.craft(Foo))    
<class '__main__.Foo'>

Miscellaneous

I've been trying to keep artisan Python3 and Python2.X compatible. Also, Artisan in no way replaces your testing library, but helps to craft better data to test with. There are a number of improvements I'd like to integrate as well:

• I want to make the use of blueprints an optional component. Given a class to create, it would be cool to do introspection and determine the attributes to generate information for.
• I want abstract calls to python-faker behind a fabricate submodule to make it easier for developers to use, namely not having to write out lambdas. Although lambdas are a fantastic tool, and incredibly powerful, not everyone knows them and we're supposed to target the '90% Use Case'.

Disclaimer

Artisan is very much a Work In Progress. It will likely change. I am more than happy to accept Pull Requests, complaints, or general criticisms. Feel free to submit them on the GitHub repo.

References

• Artisan Github Repo
• Artisan on PyPI
• A review of Mocking