Currently, the pandas bindings make two queries even when the only data being requested is from the primary corpus. If there's no division or other call for the superset, this makes queries take longer than they should. (Not twice as long, because of caching, but it's still something). A pretty easy fix for a moderate gain. The only reason this isn't completely trivial is because we rely on the automatic renaming of queries in the merge for some of the final calculations.
Currently, the pandas bindings make two queries even when the only data being requested is from the primary corpus. If there's no division or other call for the superset, this makes queries take longer than they should. (Not twice as long, because of caching, but it's still something). A pretty easy fix for a moderate gain. The only reason this isn't completely trivial is because we rely on the automatic renaming of queries in the merge for some of the final calculations.