Plurrrr

Problem:

The startup time of CPython including loading big libraries like PyTorch or TensorFlow is too slow. In case of slow file systems, I have seen startup times including such import of 10-20 seconds.

Very simple idea:

Keep the state of CPython right after we imported the big libraries and make it available instantly when needed. When loading the state, we can continue to run any random Python script (we can use runpy).

Hynek Schlawack recently described graduality as Python’s super power: the ability to prototype in the REPL, and gradually add linting, type checking, and other practices to refine your code into maintainable, production-ready software. You can also apply graduality within tools, activating checks one at a time and fixing the resulting errors as you go.

One place you can apply graduality with Mypy is in the type hints for third party packages. The default course of action is to add type hints for a whole package at once. You do this either through the package itself if it has a py.typed file, or with a *-stubs package (both specified in PEP 561). But when you add full hints for a package used widely in your code base, you might find yourself facing an insurmountable mountain of errors.

Instead, you can gradually add type hints for the package, piece by piece, fixing small batches of errors as you go. This iterative approach is more psychologically pleasing, and it reduces the chance for mistakes.

Lánczos interpolation is one of the most popular methods to resize images, together with linear and cubic interpolation. I’ve spent a lot of time staring at images resampled with Lánczos, and a few years ago, I wondered where it came from. While many sources evaluate interpolation filters visually, I couldn’t find a good explanation of how Lánczos interpolation is derived. So here it is!