Faster Python has stopped becoming a pipe dream, and is now a major topic for its development. Sometimes that comes by way of new syntax (lazy imports), sometimes by JIT compilation, and sometimes by generating C code from Python. Sometimes, it’s also by way of a whole new programming language (Mojo) that’s intended to be a powerful Python companion.
Top picks for Python readers on InfoWorld
Speed boost your Python programs with new lazy imports
With lazy imports in Python 3.15, the evaluation of imports can be delayed until they’re actually used, instead of when your Python program declares them. Best of all, you don’t need to rewrite everything to use this feature.
CPython vs. PyPy: Which Python runtime has the better JIT?
Conventional wisdom tells us that PyPy’s built-from-scratch and time-tested JIT should beat CPython’s own new native JIT. Conventional wisdom isn’t always right.
First look: Mojo 1.0 mixes Python and Rust
Is Mojo likely to outmuscle Python anytime soon? Probably not, but so far it’s shaping up to be as speedy as Rust without so much syntactical overhead.
PythoC: A new way to generate C code from Python
The traditional way to use Python to generate C code is Cython. But PythoC offers a far more streamlined experience for those who want to hitch C’s speed to Python’s convenience.
More good reads and Python updates elsewhere
PEP 836 – JIT go brrr: The path to a supported JIT compiler for CPython
The Python Software Foundation has delivered a roadmap for moving Python’s experimental JIT compiler towards a full-blown, supported, enabled-by-default part of Python’s future. But the road ahead could be bumpy.
Are you really expected to run five type-checkers now?
Well, are you? (Spoiler: not really!) Which of the big five type checkers for Python should you use? Even if you’re already committed to one type checker, this article is well worth reading. The context for why so many exist is useful.
Making Python faster with free threading and Mypyc
Mypyc is an underrated way to convert Python to C, and it’s now compatible with Python’s free-threaded build. Combining the two can unleash truly hair-raising speedups.
MicroGPT: A GPT in 200 lines of pure Python
You won’t get blazing GPU-powered performance, but you’ll get a hands-on under-the-hood example of how, exactly, one can create a GPT-2-esque neural network architecture. Try it out on your favorite public domain text!