Skip to content


Boolean Logic

dirty-equals types can be combined based on either & (and, all checks must be True for the combined check to be True) or | (or, any check can be True for the combined check to be True).

Types can also be inverted using the ~ operator, this is equivalent to using != instead of ==.


Boolean Combination of Types
from dirty_equals import HasLen, Contains

assert ['a', 'b', 'c'] == HasLen(3) & Contains('a') #(1)!
assert ['a', 'b', 'c'] == HasLen(3) | Contains('z') #(2)!

assert ['a', 'b', 'c'] != Contains('z')
assert ['a', 'b', 'c'] == ~Contains('z')

  1. The object on the left has to both have length 3 and contain "a"
  2. The object on the left has to either have length 3 or contain "z"

Initialised vs. Class comparison


This does not work with PyPy.

dirty-equals allows comparison with types regardless of whether they've been initialised.

This saves users adding () in lots of places.


Initialised vs. Uninitialised
from dirty_equals import IsInt

# these two cases are the same
assert 1 == IsInt
assert 1 == IsInt()


Types that require at least on argument when being initialised (like IsApprox) cannot be used like this, comparisons will just return False.

__repr__ and pytest compatibility

dirty-equals types have reasonable __repr__ methods, which describe types and generally are a close match of how they would be created:

from dirty_equals import IsInt, IsApprox

assert repr(IsInt) == 'IsInt'
assert repr(IsInt()) == 'IsInt()'
assert repr(IsApprox(42)) == 'IsApprox(approx=42)'

However the repr method of types changes when an equals (==) operation on them returns a True, in this case the __repr__ method will return repr(other).

repr() after comparison
from dirty_equals import IsInt

v = IsInt()
assert 42 == v
assert repr(v) == '42'

This black magic is designed to make the output of pytest when asserts on large objects fail as simple as possible to read.

Consider the following unit test:

pytest error example
from datetime import datetime
from dirty_equals import IsPositiveInt, IsNow

def test_partial_dict():
    api_response_data = {
        'id': 1, #(1)!
        'first_name': 'John',
        'last_name': 'Doe',
        'phone': '+44 123456789',

    assert api_response_data == {
        'id': IsPositiveInt(),
        'first_name': 'John',
        'last_name': 'Doe',
        'created_at': IsNow(iso_string=True),
        # phone number is missing, so the test will fail
  1. For simplicity we've hardcoded id here, but in a test it could be any positive int, hence why we need IsPositiveInt()

Here's an except from the output of pytest -vv show the error details:

pytest output
E         Common items:
E         {'created_at': '2022-02-25T15:41:38.493512',
E          'first_name': 'John',
E          'id': 1,
E          'last_name': 'Doe'}
E         Left contains 1 more item:
E         {'phone': '+44 123456789'}
E         Full diff:
E           {
E            'created_at': '2022-02-25T15:41:38.493512',
E            'first_name': 'John',
E            'id': 1,
E            'last_name': 'Doe',
E         +  'phone': '+44 123456789',
E           }

It's easy to see that the phone key is missing, id and created_at are represented by the exact values they were compared to, so don't show as different in the "Full diff" section.


This black magic only works when using initialised types, if IsPositiveInt was used instead IsPositiveInt() in the above example, the output would not be as clean.