Pyrefly vs. Pyright/Basedpyright comparison?

[IanQS]

Can anyone who has extensive experience with the two tools offer any insight into how and where they differ? For one I'm noticing that there is a huge spike in my CPU when I open up my editor using pyrefly when compared to basedpyright. Also, it seems like basedpyright offers feedback much faster (almost immediately when I open the file). I'm using Helix Editor and can share my configs if that might help

[rchen152]

Hi @IanQS - thanks for giving pyrefly a try! If you could share your configs, that would be really helpful for debugging these issues you're seeing. Could you also make sure you're on the latest pyrefly version (0.39.4)? We recently fixed a couple bugs that were causing long startup times and excessive memory consumption.

[IanQS]

Thanks for the quick response! In my helix languages.toml I only have the following related to my python files:

[[language]]
name = "python"
language-id = "python"
auto-format = true
language-servers = ["pyrefly", "ruff", "scls", "codebook"]

I don't think ruff would have any issues with pyrefly but I can share my config if that would help (there are many lines)

in my pyproject.toml I have:

[tool.pyrefly]
search-path = ["ph/**", "experiments/**"]
project-includes = ["ph/**", "experiments/**"]
project-excludes = [
    "**/node_modules",
    "**/__pycache__",
    "**/*venv/**/*",
]

I'm trying this across my Mac and Linux machines. Both are on 0.39.4. When I open my project file on either, all my CPUs are maxed out for a bit and it takes a few seconds to start displaying the first issue(s) and this seems persistent across both machines. Unfortunately, the repo I'm able to recreate this behavior on is private and I'm not at liberty to release it.

---

It can be seen to a lesser extent on: IanQS/langchain-postgres. When I open langchain_postgres/vectorstores.py it takes about 3-4 seconds before diagnostics from pyrefly kick in (line 877 has one of the issues). Do let me know how else I can help!

[rchen152]

Thanks, this is really helpful! I'll try to reproduce on langchain-postgres and let you know if I need more information.