Ethan Steinberg

@fnrizzi This looks great, thanks for the addition! One important note is that we technically already had this feature, but it was a poorly (IMHO) designed thing involving proxies. See...

@rwgk I think you are right here. Seems best to avoid churn by keeping the proxies, but adding those helper asserts to guide people towards the proxies if they need...

I can confirm that I can reproduce this bug, investigating now.

The cause of the problem is that pybind11 is using PyGILState_Ensure and PyGILState_Release, which explicitly doesn't support sub interpreters. We would need to migrate pybind11 over to PyEval_SaveThread and PyEval_RestoreThread...

@rwgk It doesn't matter if his code is buggy or not since we are using an API which itself declares in the Python docs that it shouldn't work with sub-interpreters....

Yeah, this is pretty much a feature request and feels low priority given that sub-interpreters are a pretty niche feature. > It would be much better to tell people pybind11_fail("sorry,...

I'm not sure what the bug would be? All the timestamps used within FEMR should be "naive" timestamps, with no timezone attached (either UTC or otherwise)

Actually, you can optimize this even further, reducing the runtime by half at the cost of doubling the size of the HashMap. Instead of a HashMap that only contains entries...

I can't reproduce this bug locally using gcc 9.4 and Python 3.10.6 on Ubuntu 20 Can you tell us more about your environment where you have the bug? What compiler,...

My apologies for this. It's been tough finding time to contribute to femr after graduating. That meds_v3 branch is the version that is compatible with MEDS v3, and I think...