numpy unique rows

urows.py

def unique_rows(a, **kwargs):

    rowtype = np.dtype((np.void, a.dtype.itemsize * a.shape[1]))
    b = np.ascontiguousarray(a).view(rowtype)
    return_index = kwargs.pop('return_index', False)
    out = np.unique(b, return_index=True, **kwargs)
    idx = out[1]
    uvals = a[idx]
    if (not return_index) and (len(out) == 2):
        return uvals
    elif return_index:
        return (uvals,) + out[1:]
    else:
        return (uvals,) + out[2:]