PursuedPyBear is, above all other usages, a tool for learning. We continually find ways to reduce the amount of previous knowledge is required to get to your first functioning video game. The greatest example of this is the evolution of ppb start up throughout time:
Originally, ppb was a strict MVC framework with required dependency injection and little concept of sensible defaults. You had to know what each part of the system was, instantiate it, and then pass it to the next component.
We kept the dependency injection but rebuilt the engine to have strong opinions and defaults at every level of the system. However, you had to know what a context manager was, and how to use one.
Today, we can make a functional game in 15 lines of code, and you never need to see the underlying context manager.
Progressive Revealing of Complexity¶
We want to encourage exploration and flexibility of the underlying tool, and one of the ways we achieve this is through only revealing the complexity of the tool at the point you must understand it to do something. Our “Hello World!” example requires only understanding how to invoke functions and how to write your own: the fundamental building blocks of Python programming. In the next hour of exploration, it’s possible to learn what objects are, how classes are defined and using them yourself. And from there, you can begin to learn more complex features of Python.
Whenever possible, we prefer to provide powerful and sensible defaults, but with as many options for advanced users as possible.
Every language and tool ends up with a number of quirks known as “wats”. In
ppb we tend to call them “warts”: they’re places where the knowledge you
have of how a system works is thrown a curve ball that requires reassessing what
feel for what we mean.
One way this bears out is that no matter what level your knowledge of
learning something new should only add to that knowledge, not require
We also try to reduce the number of times a user is forced to ask “why is it like this and not like that?” Things that are like messages should use the event queue. State should be contained by objects at the right level of abstraction. Things should fit the model.