Wow, having found msgspec i realized how shit pydantic is. How did pydantic even make it this far being so shit? Not only does it tank your performance, as soon as you think youre safe, an object from pydantic accidentaly leaks into your business logic and you can start playing detective on where pydantic just doesnt wanna. My opinion, stay away from this forbidden spaghetti

That's not all, for some reason it needs to be compiled and doesn't work on python 3.14. Pydantic was for me a short fetisj too.

@retoor doesn’t work on python π

@retoor thanks for reminding me to upgrade python. Whats the state of the freethreading build?

@retoor like it wasnt difficult to finish a side project before, with pydantic youre debugging the most worthless validation error shit. I do like fastapi, but not supporting msgspec as models and the functional shenanigans make me consider consider something else entirely as the feamework

@Chewbanacas it's only when self compile with build parameter given for experimental threading. This project ran harder because of it: https://molodetz.nl/project/isspam/....

Also, python crashed a few times my whole c just by spawning too much threads. The previous version didn't had a threadpoolexecutor and just spawned 900 threads. With old python, it's inefficient but OK. New python will wrack your system. That repository is a performance fight between me, Bordeddev, 12bitfloat and jestdotty. My python version was a joke to test new threading, I have made a C and C++ version. By now everyone has won once and lost afterwards. The race is to analize the 900 books of that repository so fast as possible. Originally, I wanted to use this data to train a spam filter. That's why project name is isspam.

@Chewbanacas I use aiohttp as framework for all my projects. It's very bare, even jinja and session handling has to be implemented. For database peewee or dataset. Dataset is amazing, it let's you use sqlite as a kinda doc-db. Perfect for fast development. I just made a base application for this that I use everytime. Also, I made the python buildin xmlrpc library completely asyncio. Ever tried xmlrpc? It's comfortable from python to python. Faster api development isn't possible. Cool thing about aiohttp is that it doesn't change much, it does what it does and it can have other apps as subapps under a prefix. There's no hierarchy in design of the apps like you would have with django.

@retoor tea i read something about data types staying in memory forever, if in this context immortalized means the same thing. Distributed systems fucked my head enough, my psyche wont handle going quantum after astronomy