session = cache.get_session_data(session_id) re-download data even *.nwb already exists

[AlbertYuChen]

Describe the bug
After I upgrade pandas from 1.0.1 to 1.1.1, cache.get_session_data(session_id) begins to behave very weird.
It does not recognize the existing .nwb files and re-download the data. The worst part is that it always re-try after it finishes the downloading (I have already change the timeout time to super long).

I'm uisng allensdk 1.3.0.

[achristensen56]

+1
I have the same issue.

OSX Catalina
Python 3.7
allensdk 1.6.0
pandas 0.25.3
tables 3.5.1

I manually downloaded the files through download links because even after increasing the timeout allowance to be 2+ hours, downloads were failing. I arranged the files in my directory, as as instructed (session_id# --> session_id#.nwb). When I call session = cach.get_session_data(session_id#) for a file I have, it tries and fails to redownload the file.

[AlbertYuChen]

@achristensen56 It seems allensdk 1.x.x downloading is no longer supported. It will just hang there for an hour and restart. I have switched to 2.x.x, it solved my download issue. But the nwb downloaded before using 1.xx CANNOT be read by 2.x.x. Not backward compatible. I have to redownload everything.

[achristensen56]

Ok, after reinstalling the newest AllenSDK in a new environemnt with Python3.7, this works as expected. Notably I had to regenerate the download links and redownload the files with the new AllenSDK. This is a bit of a PITA because I need the older AllenSDK for compatibility for a different project, so it would be nce if downloading the data could work in the older AllenSDK at some point.

As a side point, I'm still unable to get downloading to work directly through the SDK, even setting timeout much longer than an hour. It seems like the connection repeatedly gets interrupted. I'm not sure if that's an issue on my end or on y'alls side, but since I can download them easily through the download links, 🤷‍♀️.