openzim
Update to Python 3.14
Python 3.14 has few if any breaking changes and potentially offers up to 30% performance increase.
We should upgrade when we get a chance. It should just be a matter of bumping the versions in the Docker images and double checking that tests pass, and smoke testing the website.
The "up to 30% performance increase" is only opt-in for now, decided at interpreter compilation time. I doubt the Python interpreter we will use for WP1 will be compiled with the corresponding flag.
Upgrading to 3.14 is still necessary and useful for maintenance.
Well we control which Python image we build from: https://github.com/openzim/wp1/blob/main/docker/web/Dockerfile
So maybe we can use one that has the flag flipped on?
@benoit74 @rgaudin Shall we proceed with this upgrade? Do you have any hints as to what flags I need to enable on Python to get the performance increase? Is there a prebuilt image that I can base on?
I would recommend to wait for Python 3.14 to reach final release (in four days) but other than that, it is mandatory to upgrade to 3.14 indeed.
I would not mind at all about finding the right flags / image to get the performance increase. It seems risky both in terms of risk of bugs impossible to understand and in terms of finding the right prebuilt image which would be correctly maintained. Performance increase seems to have a geometric mean of only 3-5% in the end. It is not like we are eating resources and we would save lots of OPEX/CAPEX by saving 5%.
Yes to upgrading to 3.14 when it's released. Depending on your other python dependencies, that might not be possible immediately and you might have to upgrade only to 3.13 for a while.
As for using free-threaded python, I'm not in favor. I doubt it would have visible beneficial impact but the potential for issues is huge and given there's an effort to upgrade a lot of things, that sound like a bad timing.
