I'm pretty big time DRY extremist. Mostly because I have had so many issues in real life whenever I'm dealing with lists of things, I always wind up forgetting one ingredient in the recipe or step in a project... It's easier to DRY than to automatically verify stuff is im sync.
My new thing lately has been JSON schemas.
I use them to set defaults, to decide what properties to save and load, to eliminate unnecessary defaults when saving, to generate __slots__, and for verifying correctness at different steps in the program, like when accepting data from an API call.
I'm still using manual type annotations though, it doesn't seem like there are common tools for making a class from a schema in a way that MyPy understands.
They also serve as the source of truth documentation for the data. I'm not sure why I didn't start using them a long time ago!
I wonder what other similarly powerful DRY tools I'm not using but should be?
Author here: This specific example is a perfect case of the thing the post calls out! API boundaries are one of the most important places where we need to maintain invariants about our code base, one of the most important places that it is important not to have to worry about “keeping things in sync” as you say.
In the TypeScript world, there are a number of libraries that can handle creating types from deserialization logic: Zod [1], Superstruct [2], Typebox [3], etc. – and Typebox specifically does what you are looking to do with MyPy. I do not know the Python ecosystem at all anymore (it has been a decade!) but I would be surprised if nothing like that exists. If it does, there is a good chance it is associated with the functional programming enthusiast corner of the Python community, so you might look there!
That is so cool! Typescript has some of the best ecosystem support out there.
Python definitely does have tools like that, but many haven't been touched in a few years, and Python code sadly has a shelf life, although it seems to slowly be stabilizing.
My new thing lately has been JSON schemas.
I use them to set defaults, to decide what properties to save and load, to eliminate unnecessary defaults when saving, to generate __slots__, and for verifying correctness at different steps in the program, like when accepting data from an API call.
I'm still using manual type annotations though, it doesn't seem like there are common tools for making a class from a schema in a way that MyPy understands.
They also serve as the source of truth documentation for the data. I'm not sure why I didn't start using them a long time ago!
I wonder what other similarly powerful DRY tools I'm not using but should be?