While using custom I tried splitting the tuples into separate DataFrame columns after fetching the results with pipeline_output() but since this is a view of the actual DataFrame I get a SettingWithCopyWarning.
To avoid the annoying warning I created a new DataFrame instance from the one returned by pipeline_output() and I operate on this, I can see the tuples in values property are the same objects of the initial DataFrame but the external structure is obviously not.
Is there a proper way of splitting tuples in pipeline results? I think this is sub-optimal since new DataFrames are built and the previous left to the gc every loop iteration, but there's no other way I could think of.
Should this rather be a new feature request of Factor class so that custom factors using multiple output can explicitly declare names of their outputs?