Ehsan Totoni
Ehsan Totoni
I just hit this issue too. The compiler takes quite a while to throw the error which seems strange to me.
Sounds like a good plan. Looking at the CPython code, it doesn't seem to be too bad.
Thank you @stuartarchibald . This all makes sense (I suspected and tried to explore these things to some extent). I'd add that `set` object doesn't handle reference counted objects yet....
by using a dictionary instead of set
I'm currently exploring this idea and would appreciate feedback. Also, can these changes be merged in Numba (considering maintenance, packaging etc. requirements)?
Thanks @sklam. Will try to build CPython from source with JIT enabled and try it. But this will take at least over a year to be available and even then...
A bit of context about our use cases. Bodo compiler (which we are hoping to open source soon) may compile large code or generate long Python code from SQL and...
Here is the trace file: [trace.json](https://github.com/user-attachments/files/16763608/trace.json) Regarding extra lowering by partial-type-inference: we have separated typing and lowering for many of the functions to reduce compilation time already. Also, the recursive...