@dimitri-yatsenko / @eywalker / @guzman-raphael -
Would be good to get some feedback here before un-WIP, my familiarity with pandas is a bit lacking to feel comfortable saying 'this is the fix'. See also example test script issue666.txt in #666 comments.

Thanks for the feedback @gucky92
Just to make sure I've got it - these both refer to passing the 'drop' argument to reset_index, correct? e.g. you're proposing either:

• rows.reset_index(drop=not isinstance(rows.index, pandas.MultiIndex)).to_records(index=False)
  or:
• rows.reset_index(drop=set(self.primary_key) != set(rows.index.names)).to_records(index=False)

correct?

@ixcat Yes, that is what i meant.

Simpler and more in alignment with what we actually want to check is whether there is any "named" index/indices. Given this, a good check would be

rows.reset_index(drop=len(rows.index.names)==1 and not rows.index.names[0]).to_records(index=False)

The default case of an unnamed index will have rows.index.names with single entry of None and can be safely dropped.

back to 'WIP' until unit test

Ah sorry I removed WIP as this was set as Draft and figured that suffice to indicate the WIP status.