Do we alter the structure of input dataframes?

[bjschoenfeld]

From a comment on D3M Slack:

Hey, quick note to primitive developers: please don't destructively modify the inputs to your primitive. Make a copy if you need to alter the structure of the dataframe. We just chased down a bug in a promising primitive that was clobbering the learningData in the input dataset. This makes your primitive non-reentrant and more or less useless (at least in our TA2).

Do our primitives behave this way?