How can I override comparisons between NumPy's ndarray and my type?

Question

In NumPy, it is possible to use the __array_priority__ attribute to take control of binary operators acting on an ndarray and a user-defined type. For instance:

class Foo(object):
  def __radd__(self, lhs): return 0
  __array_priority__ = 100

a = np.random.random((100,100))
b = Foo()
a + b # calls b.__radd__(a) -> 0

The same thing, however, doesn't appear to work for comparison operators. For instance, if I add the following line to Foo , then it is never called from the expression a < b :

def __rlt__(self, lhs): return 0

I realize that __rlt__ is not really a Python special name, but I thought it might work. I tried all of __lt__ , __le__ , __eq__ , __ne__ , __ge__ , __gt__ with and without a preceding r , plus __cmp__ , too, but I could never get NumPy to call any of them.

Can these comparisons be overridden?

UPDATE

To avoid confusion, here is a longer description NumPy's behavior. For starters, here's what the Guide to NumPy book says:

If the ufunc has 2 inputs and 1 output and the second input is an Object array
then a special-case check is performed so that NotImplemented is returned if the
second input is not an ndarray, has the array priority attribute, and has an
r<op> special method.

I think this is the rule that makes + work. Here's an example:

import numpy as np
a = np.random.random((2,2))


class Bar0(object):
  def __add__(self, rhs): return 0
  def __radd__(self, rhs): return 1

b = Bar0()
print a + b # Calls __radd__ four times, returns an array
# [[1 1]
#  [1 1]]



class Bar1(object):
  def __add__(self, rhs): return 0
  def __radd__(self, rhs): return 1
  __array_priority__ = 100

b = Bar1()
print a + b # Calls __radd__ once, returns 1
# 1

As you can see, without __array_priority__ , NumPy interprets the user-defined object as a scalar type, and applies the operation at every position in the array. That's not what I want. My type is array-like (but should not be derived from ndarray).

Here's a longer example showing how this fails when all of the comparison methods are defined:

class Foo(object):
  def __cmp__(self, rhs): return 0
  def __lt__(self, rhs): return 1
  def __le__(self, rhs): return 2
  def __eq__(self, rhs): return 3
  def __ne__(self, rhs): return 4
  def __gt__(self, rhs): return 5
  def __ge__(self, rhs): return 6
  __array_priority__ = 100

b = Foo()
print a < b # Calls __cmp__ four times, returns an array
# [[False False]
#  [False False]]

Answer 1

It looks like I can answer this myself. np.set_numeric_ops can be used as follows:

class Foo(object):
  def __lt__(self, rhs): return 0
  def __le__(self, rhs): return 1
  def __eq__(self, rhs): return 2
  def __ne__(self, rhs): return 3
  def __gt__(self, rhs): return 4
  def __ge__(self, rhs): return 5
  __array_priority__ = 100

def override(name):
  def ufunc(x,y):
    if isinstance(y,Foo): return NotImplemented
    return np.getattr(name)(x,y)
  return ufunc

np.set_numeric_ops(
    ** {
        ufunc : override(ufunc) for ufunc in (
            "less", "less_equal", "equal", "not_equal", "greater_equal"
          , "greater"
          )
    }
  )

a = np.random.random((2,2))
b = Foo()
print a < b
# 4

Answer 2

I cannot reproduce your problem. The correct approach is to use __cmp__ special-method. If I write

import numpy as np

class Foo(object):
    def __radd__(self, lhs): 
        return 0

    def __cmp__(self, this):
        return -1

    __array_prioriy__ = 100

a = np.random.random((100,100))
b = Foo()
print a<b

and set a break point in the debugger, execution stops at the return -1 .

Btw: __array_prioriy__ doesn't make any difference here: you have a typo in it!