Skip to content

butla/mountepy

Repository files navigation

mountepy

https://coveralls.io/repos/butla/mountepy/badge.svg?branch=master&service=github

Utilities for creating HTTP (micro)service tests. Based on Mountebank.

Mountepy works by spawning and cleaning after given HTTP service processes and Mountebank. Thanks to that you no longer need that "start X before running the tests" for your application. No. Your tests start "X", it's put up or down only when it needs to and as many times as you need.

  • Test-framework-agnostic (use unittest, nose, py.test or whatever... but I like py.test).
  • Enables fast and reliable end-to-end testing of microservices. They won't be aware that they are in some testing mode.
  • Tested on Python 3.4, Ubuntu 14 x64.
  • Planned features in the road map below. If you have suggestions, just post them as Github issues. Pull requests are also welcome :)

I recommend Pytest for elastic composition of service process test fixtures. Your process may start once per test suite, once per test, etc.

As you can probably see, there's no build status icon. That's becaused Snap CI was decommissioned and I didn't have the time to make a switch yet. But don't worry, the project is stable and perfectly usable, it just doesn't require any code changes at the time. I use it at my company and it's infecting more and more projects :)

Installation

$ pip install mountepy

A standalone distribution of Mountebank (including NodeJS) will be downloaded on first run.

If you don't want Mountepy to download Mountebank:

  1. Install NodeJS and NPM. On Ubuntu it's
$ sudo apt-get install -y nodejs-legacy npm
  1. Install Mountebank yourself
$ npm install -g mountebank --production

Examples

Mountebank acts as a mock for external HTTP services. Here's how you spawn a Mountebank process, configure it with a stub of some HTTP service, assert that it's actually responding. Mountebank process is killed after the with block.

# requests is installed alongside Mountepy
import mountepy, requests

with mountepy.Mountebank() as mb:
    imposter = mb.add_imposter_simple(path='/something', response='mock response')
    stub_url = 'http://localhost:{}/something'.format(imposter.port)
    assert requests.get(stub_url).text == 'mock response'

It's a good idea to test your service as a whole process. Let's say that you have an one-file WSGI (e.g. Flask or Bottle) app that responds to a GET on its root path ('\') with a string it sees in RET_STR environment variable. Also, the app needs to know on what port to run, so we also pass it as an environment variable. {port} is a special value for Mountepy. It will be filled with the application's port, whether it's passed during object construction or automatically selected from free ports.

# port_for is installed alongside Mountepy
import mountepy, requests, port_for, os, sys

service_port = port_for.select_random()
service = mountepy.HttpService(
    [sys.executable, 'sample_app.py'],
    port=service_port,
    env={
        'PORT': '{port}',
        'RET_STR': 'Just some text.'
    })
with service:
    assert requests.get(service.url).text == 'Just some text.'

Starting a more complex service running on Gunicorn can look like this:

import os, sys

gunicorn_path = os.path.join(os.path.dirname(sys.executable), 'gunicorn')
service_command = [
    gunicorn_path,
    'your_package.app:get_app()',
    '--bind', ':{port}',
    '--enable-stdio-inheritance',
    '--pythonpath', ','.join(sys.path)]

service = HttpService(service_command)

# You can use start/stop methods instead of using the "with" statement.
# It's the same for Mountebank objects.
service.start()

# now you test stuff...
service.stop()

"Real world" use of mountepy can be found in PyDAS.

Measuring test coverage

Mountepy starts your code in a separate process, so it's normally hard to get information about the code covered by the tests. Fortunately, this problem is solved by Coverage. See this documentation page.

In short, you need to:

  • run coverage.process_startup() in each new Python process (this can be enforced by installing coverage_pth, but some caution is required)
  • set COVERAGE_PROCESS_START environment variable to location of your .coveragerc
  • run the tests themselves: coverage run (...), coverage combine and then coverage report -m

Again, see PyDAS's tox.ini for demonstration.

Running tests

Clone the repo with submodules, then install and run tox.

$ git clone --recursive git@github.com:butla/mountepy.git
$ sudo pip install tox
$ cd mountepy
$ tox

Motivation (on 2015-12-30)

  • Why Mountebank? It can be deployed as standalone application, is actively developed and supports TCP mocks which can be used to simulate broken HTTP messages.
  • Why not Pretenders? Doesn't support TCP and the development doesn't seem to be really active.
  • Why not WireMock? Doesn't support TCP and I don't want to be forced to install Java to run tests and it doesn't seem to have more features than Mountebank.
  • Why create a new project? There already is a Python Mountebank wrapper, but it doesn't offer much.

License

Mountepy is licensed under BSD Zero Clause license.

Why I didn't use one of the more popular licenses like MIT, 2 or 3-Clause BSD or Apache2? Well, this one is practically equal to 2-Clause BSD (and I don't see any functional differences between it and MIT license) with the exception of the rule about retaining the original license text in derivative work. So if you'd happen to redistribute my library along with your software you don't have to attach a copy of my license. So you won't break any copyright laws by being lazy (which I like to be, for instance). You're welcome.

About

Utilities for creating HTTP (micro)service tests. Based on Mountebank.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages