art049 / art049/my-python-event-loop
Code Restructure
- Dominant language
- Python
- Stars
- 15
- Forks
- 1
- PR merge metrics
- No merged PRs in 30d
Description
Enjoyed reading your code. It was quite comprehensive for your EuroPython 2024 talk. However, there were a few ideas that might make it more concise. This is a worthwhile project that you should pick back up. These are a few, but not all, of my quick observations. Implementing these may allow plugins for SELinux and osquery. Since coreutils is your home, these modifications could enable PyO3 stub files for an overlay over coreutils.
1. Binary search of cron schedules isnt very scalable.
- I'd use a heap or BST for these times. That way you can have horizontally distributed cron reads
- If you were to use your FastAPI lambda endpoint as a front-end for a SIEM, which isnt a bad idea, you would want to load balance this
- Might even be worth decoupling the call_at fxn
2. Passable locks associated with your handle data structure: parking-lot based queues
- Sometimes theres multiple handlers: for instance, high end TDM systems
- If you were to pass the per-thread loop context to PyO3/Rust, you may need to consider handling CFFI concurrency interrupts
3. Due to 2, _run_once should be rewritten to preempt... OnceLock and RwLocks would be seen here
- If you have a high end TDM system, you could have different loop contexts lock expensive resources and/or operations
4. Some of your code is redundant... abstract yourself of simple getters/setters
- Create a metaclass to help your trivial methods
5. Make the selector a TypedDict (inline dictionary in 3.14) or a dataclass
- _add_reader() and _add_writer() have been made inflexible
- Cut the implementation specific dependency so that events can be plugins
- This permits contextual EVENT_WRITEs...
Contributor guide
No contributing guide indexed for this repository
Research direction
The issue names the event-loop entry points call_at, _run_once, _add_reader(), _add_writer(), the selector, and handle data. Start by mapping those entry points and their current relationships; no files or tests are named. The requested work is not bounded to one change, and the issue does not define a clear completion criterion.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- fastapi, python, rust
- Domain
- backend, operating-systems
- Issue type
- Refactor
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 15/100