ForkingPickler can't pickle

[biswajitSM]

Hi,
I was trying out the pycro-manager with democamera config in microanager-gamma (14May nightly build) in a jupyter notebook with the following codes as documented.

from pycromanager import Acquisition, multi_d_acquisition_events

with Acquisition(directory=r'C:\Users\biswajitpradha\Reseach\analysis\PycroManager\data', name='acquisition_name') as acq:
    events = multi_d_acquisition_events(
                                    num_time_points=4, time_interval_s=0,
                                    channel_group='channel', channels=['DAPI', 'FITC'],
                                    z_start=0, z_end=6, z_step=0.4,
                                    order='tcz')

But it showed an error:

AttributeError: Can't pickle local object 'Acquisition.__init__.<locals>.event_sending_fn'

full error pic:

Do you know what's going wrong here?

Thanks, Biswajit

[henrypinkard]

Appears to be something specific to python launching new processes on Windows. I just pushed a new version that may fix this. Upgrade to version 0.0.3 with pip install pycromanager --upgrade and give it a try

[dpshepherd]

Hi @henrypinkard ,

Thanks for the hard work on this! Same issue here. I just tried to install version 0.0.3 and it isn't found. pip install reports that only 0.0.1 and 0.0.2 are available.

Looking forward to trying the package out on a couple of our rigs,
Doug

[henrypinkard]

Whoops, look like I somehow forgot to push. Its up now. Let me know how it goes

[dpshepherd]

This code,

from pycromanager import Acquisition, multi_d_acquisition_events

with Acquisition(directory=r'C://testing//data', name='acquisition_name') as acq:
    events = multi_d_acquisition_events(
                                    num_time_points=4, time_interval_s=0,
                                    channel_group='channel', channels=['DAPI', 'FITC'],
                                    z_start=0, z_end=6, z_step=0.4,
                                    order='tcz')

used with version 0.0.3 now gives the following error:

TypeError: no default __reduce__ due to non-trivial __cinit__

full error picture:

[henrypinkard]

looks like this is a quirk with how windows handles multiple processes. For the time being, you can fix this by putting your whole script in a block with:

if __name__ == '__main__':
     with Acquisition...

I'm going to keep playing around to see if I can find a more durable solution

[henrypinkard]

I guess the above is the best solution possible due to how multiprocessing is implemented on windows. This should be fixed by #18, and I just pushed version 0.2.0 to pip that has these changes.

You'll need to get the next MM nightly build after this PR merges (micro-manager/micro-manager#805).

Closing for now since I think this is fixed. Open a new one if you run into something else