The Python metaclass system is something every Python programmer uses all the time, perhaps without even knowing about it. Every class should be created through some (subclass of) type, which allows for endless customization and magic. Instead of statically defining your class, you can now have it created as you normally would and dynamically add, modify, or remove attributes from your class during definition; very magical but very useful. The magic component, however, is also the reason it should be used with a lot of caution. While metaclasses can be used to make your life much easier, they are also amongst the easiest ways of producing completely incomprehensible code.

Regardless, there are some great use-cases for metaclasses and many libraries such as SQLAlchemy and Django use metaclasses to make your code work much easier and arguably better. Actually comprehending the magic that is used inside is generally not needed for the usage of these libraries, which makes the cases defendable. The question becomes whether a much better experience for beginners is worth some dark magic internally, and looking at the success of these libraries, I would say yes in this case.

To conclude, when thinking about using metaclasses, keep in mind what Tim Peters once said: "Metaclasses are deeper magic than 99% of users should ever worry about. If you wonder whether you need them, you don't."

Now we will continue with a solution to remove some of the magic that metaclasses generate: documentation. The next chapter will show us how your code can be documented, how that documentation can be tested, and most importantly, how the documentation can be made smarter by annotating types in your documentation.