Rank and the leading axis

Rank forces the operand function to work on a suffix of the argument's axes. For functions that work on a single axis, the leading one is a more flexible choice:

Rank's advantage over older models is that it also works on multi-axis functions.