Reinout van Rees: Amsterdam Python meetup, november 2018

My summary of the 28 november python meetup at the Byte office. I myself also gave a talk (about cookiecutter) but I obviously haven't made a summary of that. I'll try to summarize that one later :-)

Project Auger - Chris Laffra

One of Chris' pet projects is auger, automated unittest generation. He wrote it when lying in bed with a broken ankle and thought about what he hated most: writing tests.

Auger? Automated Unittest GEneRator. It works by running a tracer

The project's idea is:

• Write code as always
• Don't worry about tests
• Run the auger tracer to record function parameter values and function results.
• After recording, you can generate mocks and assertions.

"But this breaks test driven development"!!! Actually, not quite. It can be useful if you have to start working on an existing code base without any tests: you can generate a basic set of tests to start from.

So: it records what you did once and uses that as a starting point for your tests. It makes sure that what ones worked keeps on working.

It works with a "context manager". A context manager normally has __enter__() and __exit__(). But you can add more interesting things. If in the __enter__()` you call sys.settrace(self.trace), you can add a def trace(self, frame, event, args), which is then fired upon everything that happens within the context manager. You can use it for coverage tracking or logging or visualization of what happens in your code. He used the last for algorithm visualizations on http://pyalgoviz.appspot.com/

So... this sys.settrace() magic is used to figure out which functions get called with which parameters.

Functions and classes in the modules you want to check are tested, classes from other modules are partially mocked.

Python LED animation system BiblioPixel - Tom Ritchford

Bibliopixel (https://github.com/ManiacalLabs/BiblioPixel) is his pet project. It is a python3 program that runs on basically everything (raspberry pi, linux, osx, windows. What it does? It controls large numbers of lights in real-time without programming.

There are lots of output drivers form led strips and philips hue to an opengl in-browser renderer. There are also lots of different ways to steer it. Here is the documentation.

He actually started on a lot of programs having to do with audio and lights and so. Starting with a PDP-11 (which only produced beeps). Amiga, macintosch (something that actually worked and was used for real), java, javascript, python + C++. And now python.

The long-term goal is to programmatically control lights and other hardware in real time. And... he wants to define the project in text files. The actual light "program" should not be in code. Ideally, bits of projects ought to be reusable. And any input ought to be connectable to any output.

Bibliopixel started with the AllPixel LED controller which had a succesful kickstarter campaign (he got involved two years later).

An "animation" talks to a "layout" and the layout talks to one or more drivers (one could be a debug visualization on your laptop and the other the real physical installation). Animations can be nested.

Above it all is the "Project". A YAML (or json) file that defines the project and configures everything.

Bibliopixel is quite forgiving about inputs. It accepts all sorts of colors (red, #ff0000, etc). Capitalization, missing spaces, extraneous spaces: all fine. Likewise about "controls": a control receives a "raw" message and then tries to convert it into something it understands.

Bibliopixel is very reliable. Lots of automated tests. Hardware test boards to test the code with the eight most common types of hardware. Solid error handling and readable error messages help a lot.

There are some weak points. The biggest is lack of developers. Another problem is that it only supports three colors (RGB). So you can't handle RGBW (RGB plus white) and other such newer combinations. He hopes to move the code over completely to numpy, that would help a lot. Numpy is already supported, but the existing legacy implementation currently also still needs to be work.

He showed some nice demos at the end.

from Planet Python
via read more