cache=True & generators?

reproducing code:

import numba
import numpy as np

@numba.njit(cache=True)
def _thisfunc(vals):
    for v in vals:
        yield v

# calling code:
_thisfunc(np.random.rand(500))

results in:

C:\Anaconda3\lib\site-packages\numba\caching.py in _dump(self, obj)
    563 
    564     def _dump(self, obj):
--> 565         return pickle.dumps(obj, protocol=-1)
    566 
    567     @contextlib.contextmanager

PicklingError: Can't pickle <function _thisfunc at 0x0000014AF82B4D08>: it's not the same object as __main__._thisfunc

but with cache=False the function works. Is there any workaround for this?

[stuartarchibald]

@sklam fixed in #7957, support will be in Numba 0.56.

[esc]

This issue has been opened relatively long ago, and it looks like a genuine bug. At the same time, yield constructs are now fully supported by numba. Is there a timescale for fixing this?

@mtazzari thank you for asking about this, I just double checked on current master and the issue still persists. There is currently no timescale or ETA for this being fixed.