Made an initial pass though a Pandas conversion of AssociationTask. Ended up simplifying the selection of DiaObjects in update to an inefficient boolean mask on the data id as the indexing refused to work.

You can find outputs for the DiaObjects and DiaSources in /project/morriscb/src/ap_verify_hits2015/test_pandas/association.db

I made a small change to the AssociationTask usage of pandas to make use of the DataFrame object's built in indexing to allow for easy and fast look up of diaObjects and diaSources by their diaObjectId.

Looked into the data output from the current master to that of this ticket. Found a bug where if a column was not specified when inserting a new DiaObject dataframe in the Ppdb any columns not declared would default to a value of 0 instead of NaN/Null. Fixed in a new commit.

After running ap_verify_hits2015, all visits, ccds 24 and 25, I have confirmed that the output values are identical between the pandas and previous afw version.

Nice work!

Knew I forgot something. Jenkins run: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/29957/pipeline