What does |= (ior) do in Python?

|= performs an in-place⁺ operation between pairs of objects. In particular, between:

• sets: a union operation
• dicts: an update operation
• counters: a union (of multisets) operation
• numbers: a bitwise OR, binary operation

In most cases, it is related to the | operator. See examples below.

Sets

For example, the union of two sets assigned to s1 and s2 share the following equivalent expressions:

>>> s1 = s1 | s2                                           # 1
>>> s1 |= s2                                               # 2
>>> s1.__ior__(s2)                                         # 3

where the final value of s1 is equivalent either by:

1. an assigned OR operation
2. an in-place OR operation
3. an in-place OR operation via special method⁺⁺

Example

Here we apply OR (|) and the in-place OR (|=) to sets:

>>> s1 = {"a", "b", "c"}
>>> s2 = {"d", "e", "f"}

>>> # OR, | 
>>> s1 | s2
{'a', 'b', 'c', 'd', 'e', 'f'}
>>> s1                                                     # `s1` is unchanged
{'a', 'b', 'c'}

>>> # In-place OR, |=
>>> s1 |= s2
>>> s1                                                     # `s1` is reassigned
{'a', 'b', 'c', 'd', 'e', 'f'}

---

Dictionaries

In Python 3.9+, new merge (|) and update (|=) operators are proposed between dictionaries. Note: these are not the same as set operators mentioned above.

Given operations between two dicts assigned to d1 and d2:

>>> d1 = d1 | d2                                           # 1
>>> d1 |= d2                                               # 2

where d1 is equivalent via:

1. an assigned merge-right operation
2. an in-place merge-right (update) operation; equivalent to d1.update(d2)

Example

Here we apply merge (|) and update (|=) to dicts:

>>> d1 = {"a": 0, "b": 1, "c": 2}
>>> d2 = {"c": 20, "d": 30}

>>> # Merge, | 
>>> d1 | d2
{"a": 0, "b": 1, "c": 20, "d": 30}
>>> d1 
{"a": 0, "b": 1, "c": 2}

>>> # Update, |=
>>> d1 |= d2
>>> d1 
{"a": 0, "b": 1, "c": 20, "d": 30}

---

Counters

The collections.Counter is related to a mathematical datastructure called a multiset (mset). It is basically a dict of (object, multiplicity) key-value pairs.

Given operations between two counters assigned to c1 and c2:

>>> c1 = c1 | c2                                           # 1
>>> c1 |= c2                                               # 2

where c1 is equivalent via:

1. an assigned union operation
2. an in-place union operation

A union of multisets contains the maximum multiplicities per entry. Note, this does not behave the same way as between two sets or between two regular dicts.

Example

Here we apply union (|) and the in-place union (|=) to Counters:

import collections as ct


>>> c1 = ct.Counter({2: 2, 3: 3})
>>> c2 = ct.Counter({1: 1, 3: 5})

>>> # Union, |    
>>> c1 | c2
Counter({2: 2, 3: 5, 1: 1})
>>> c1
Counter({2: 2, 3: 3})

>>> # In-place Union, |=
>>> c1 |= c2
>>> c1
Counter({2: 2, 3: 5, 1: 1})

---

Numbers

Lastly, you can do binary math.

Given operations between two numbers assigned to n1 and n2:

>>> n1 = n1 | n2                                           # 1
>>> n1 |= n2                                               # 2

where n1 is equivalent via:

1. an assigned bitwise OR operation
2. an in-place bitwise OR operation

Example

Here we apply bitwise OR (|) and the in-place bitwise OR (|=) to numbers:

>>> n1 = 0
>>> n2 = 1

>>> # Bitwise OR, |
>>> n1 | n2
1
>>> n1
0

>>> # In-place Bitwise OR, |=
>>> n1 |= n2
>>> n1
1

---

Review

This section briefly reviews some bitwise math. In the simplest case, the bitwise OR operation compares two binary bits. It will always return 1 except when both bits are 0.

>>> assert 1 == (1 | 1) == (1 | 0) == (0 | 1)
>>> assert 0 == (0 | 0)

We now extend this idea beyond binary numbers. Given any two integral numbers (lacking fractional components), we apply the bitwise OR and get an integral result:

>>> a = 10 
>>> b = 16 
>>> a | b
26

How? In general, the bitwise operations follow some "rules":

1. internally compare binary equivalents
2. apply the operation
3. return the result as the given type

Let’s apply these rules to our regular integers above.

(1) Compare binary equivalents, seen here as strings (0b denotes binary):

>>> bin(a)
'0b1010'
>>> bin(b)
'0b10000'

(2) Apply a bitwise OR operation to each column (0 when both are 0, else 1):

01010
10000
-----
11010

(3) Return the result in the given type, e.g. base 10, decimal:

>>> int(0b11010)
26

The internal binary comparison means we can apply the latter to integers in any base, e.g. hex and octal:

>>> a = 10                                   # 10, dec
>>> b = 0b10000                              # 16, bin
>>> c = 0xa                                  # 10, hex
>>> d = 0o20                                 # 16, oct

>>> a | b
26
>>> c | d
26

See Also

• An example of overloading the __ior__() method to iterate iterables in a MutableSet abstract base class
• R. Hettinger’s OrderedSet recipe (see lines 3 and 10 respectively)
• A thread on Python-ideas on why to use |= to update a set
• A section B.8 of Dive in Python 3 on special methods of Python operators
• In-place binary operators fallback to regular methods, see cpython source code (eval.c and abstract.c). Thanks @asottile.
• A post on how python handles displaying prepended zeros in bitwise computations

⁺_{The in-place bitwise OR operator cannot be applied to literals; assign objects to names.}

⁺⁺_{Special methods return the same operations as their corresponding operators.}