Difference between numpy ediff1d and diff

Question

I have recently discovered numpy.ediff1d Is there an advantage of using numpy.ediff1d instead of numpy.diff or what is it's usecase?

Answer 1

Prior to numpy 1.16.0, one could use ediff1d(ar, to_end, to_begin) to pad the result of ar[1:] - ar[-1:]. But from 1.16.0 diff() supports padding and covers all of ediff1d() functionality and much more. Also, diff() has same performance in most cases and outperforms ediff1d() for boolean arrays.