There seems to be thread-safety issues while invoking python within C++ UDF's especially when the python code calls numpy or pandas. It turns out numpy and pandas aren't inherently thread-safe. As a consequence, scalar UDF's invoking the python code fail, and also it appears only one transform UDF call can be made in a session.
Any pointers on how to ensure thread safety while invoking python within C++ UDF? I see a few links like http://www.codeproject.com/Articles/118 ... -C-Part-II
Would this help? Any recommendations on how to get around this problem?
Thread Safety for numpy, pandas for embedded python in UDF
Moderator: NorbertKrupa
Re: Thread Safety for numpy, pandas for embedded python in UDF
You may want to ask this question in the HP Developer Community, where there’s more of an HP presence.
https://community.dev.hpe.com/t5/Vertic ... ms_vertica
https://community.dev.hpe.com/t5/Vertic ... ms_vertica
Sharon Cutter
Vertica Consultant, Zazz Technologies LLC
Vertica Consultant, Zazz Technologies LLC