Structural pattern matching in Python 3.10


The biggest incentive to upgrade my ython version


February 20, 2022


Structural Pattern Matching is a new feature introduced in python 3.10, and perhaps the most intriguing one. It takes as input an object to inspect (following match), and multiple patterns to match against (following one or more case). If there is a match, a pattern-specific code block runs. The basic syntax is

error_code = 403
match error_code:
    case 400:
        print("Bad request")
    case 403:
    case 500:
        print("Internal server error")
    case _:
        print("Something's wrong with the Internet")

Here, we check case by case if error_code is equal to a value, if it’s not one of 400, 403, 500, _ will be a catch-all wildcard. Also, pattern matching will break when it hits a match, so each case should be independent.

To run examples in this article, be sure you are using python higher than 3.10. To check, use

import sys
'3.10.8 (main, Oct 13 2022, 09:48:40) [Clang 14.0.0 (clang-1400.0.29.102)]'

Switch statements

The most apparent usage of pattern matching is implementing swich ... case statements in many other programming languages, which is a replacement for multiple if else statements. Imagine in some game app we could will have a move class method which controls figure movement

def move(self, direction):
    match direction:
        case "up":
        case "down":
        case "left":
        case "right":

Object destructuring

The real power of pattern matching resides in destructuring data structures like lists and dictionaries.

match {"first_name": "Jane", "last_name": "Doe", "middle_name": ""}:
    case {"first_name": first_name , "last_name": last_name}:
        print(first_name, last_name)
Jane Doe

In the example above, we supplied a pattern that says the object should contains keys “first_name” and “last_name”, if the pattern matches, their values will be captured into the variables first_name and last_name. Note that for an dictionary, structural pattern matching matches the structure instead of the exact content, the pattern is valid as long as “first_name” and “last_name” exists in the object’s keys, and we don’t have to worry about the unmentioned “middle_name”.

For a more advanced example, suppose we are working with the SpaceX API to retrieve launch data, an individual record of launch looks like

        "flight_number": 1,
        "mission_name": "FalconSat",
        "upcoming": false,
        "launch_year": "2006",
        "launch_date_unix": 1143239400,
        "rocket": {
            "rocket_id": "falcon1",
            "rocket_name": "Falcon 1",
        "details": "Engine failure at 33 seconds and loss of vehicle",
        //... other fields

We only want to get the fields mission_name, rocket_name and details. Note that

import requests
import json
from pprint import pp
res = requests.get("")
data = res.json()
out = []
for launch in data:
    match launch:
        case {"mission_name": mname, "rocket": {"rocket_name": rname}, "details": details}:
            out.append({"mission_name": mname, "rocket_name": rname, "details": details})

[{'mission_name': 'FalconSat',
  'rocket_name': 'Falcon 1',
  'details': 'Engine failure at 33 seconds and loss of vehicle'},
 {'mission_name': 'DemoSat',
  'rocket_name': 'Falcon 1',
  'details': 'Successful first stage burn and transition to second stage, '
             'maximum altitude 289 km, Premature engine shutdown at T+7 min 30 '
             's, Failed to reach orbit, Failed to recover first stage'},
 {'mission_name': 'Trailblazer',
  'rocket_name': 'Falcon 1',
  'details': 'Residual stage 1 thrust led to collision between stage 1 and '
             'stage 2'}]

Here, not only are we matching top level key mission_name and details, we also use a nested pattern to extract rocket name.

List is another common data structure for this sort of extraction.

match [1, 2, 3]:
    case [1, 2]:
        print("match 1")
    case [2, *rest]:
        print("match 2")
    case[*other, last]:
        print("match 3")
match 3
[1, 2]

I mentioned that pattern matching for dict does not require use to explicitly declare all fields. For lists this is different, we have to declare all elements, this is why match 1 fails. Match 2 attempts to use the * syntax to capture all values after 2 into a list called rest. This is helpful in case the list is really large and it’s impossible to exhaust all elements, or we simply don’t care about some elements. However, the second pattern does not specify the correct order, as 2 is not the first element in the matching object. Match 3 captures all elements but the last into other, and assign the last element to last.

It’s also possible to do or logic with | in case statement. For example we may add a case 401 | 403 | 404: "Not allowed" statement to our first example. Here is another example from pep:

match command.split():
    ... # Other cases
    case ["north"] | ["go", "north"]:
        current_room = current_room.neighbor("north")
    case ["get", obj] | ["pick", "up", obj] | ["pick", obj, "up"]:
        ... # Code for picking up the given object

Literal patterns can also be captured using the as keyword

match command.split():
    case ["go", ("north" | "south" | "east" | "west") as direction]:
        current_room = current_room.neighbor(direction)

Adding conditions

Pattern matching allows if conditions the same way as list comprehension, we can append a if ... else clause like so

all_directions = ["up", "down"]

match commad.split():
    case ["go", direction] if direction in all_directions:
    case ["go", _]:
        print("direction not supported")

For conditions that checks if an object is of a data type, we can directly use the class creator function:

def sum_list(numbers):
    match numbers:
        case []:
            return 0
        case [int(first) | float(first), *rest]:
            return first + sum_list(rest)
        case _:
            raise ValueError(f"Can only sum lists of numbers")

sum_list([1, 2, 3])
sum_list([1, "2", 3])
ValueError: Can only sum lists of numbers

Here int(first) | float(first) checks if the first element is an integer or float. The next example uses a dataclass based pattern as well as its attributes

from dataclasses import dataclass

class Point:
    x: float
    y: float

p = Point(x = 1, y = 1)
# data class for validation
match p:
    case Point(x = x):

In general, we can use any python class for such validation. This is also called a class pattern.

Matching constants

Care should be taken when matching constants, since it’s common to store the constant in an variable and then use it in case. The following code will not work:

direction = "up"

match "down":
    case direction:
    case _:
        print("no match")

We are trying to match the value “up”, but python interprets it as a capture pattern that says store whatever the match object is into the variable direction, so we are effectively overriding direction (recall the list matching example) and the first case always matches. Python will not allow this with an error message SyntaxError: name capture 'direction' makes remaining patterns unreachable.

As an alternative to match literal values, we can use a enum type for constant matching instead

import enum
class Direction(str, enum.Enum):
    up = "up"
    down = "down"

match "down":
    case Direction.up:
        print("match 1")
    case Direction.down:
        print("match 2")
    case _:
        print("match 3")
match 2