Pampy: Pattern Matching for Python
Pampy is pretty small (150 lines), reasonably fast, and often makes your code more readable
and hence easier to reason about. There is also a JavaScript version, called Pampy.js.
You can write many patterns
Patterns are evaluated in the order they appear.
You can write Fibonacci
The operator _ means "any other case I didn't think of".
from pampy import match, _
def fibonacci(n):
return match(n,
1, 1,
2, 1,
_, lambda x: fibonacci(x-1) + fibonacci(x-2)
)
You can write a Lisp calculator in 5 lines
from pampy import match, REST, _
def lisp(exp):
return match(exp,
int, lambda x: x,
callable, lambda x: x,
(callable, REST), lambda f, rest: f(*map(lisp, rest)),
tuple, lambda t: list(map(lisp, t)),
)
plus = lambda a, b: a + b
minus = lambda a, b: a - b
from functools import reduce
lisp((plus, 1, 2)) # => 3
lisp((plus, 1, (minus, 4, 2))) # => 3
lisp((reduce, plus, (range, 10))) # => 45
You can match so many things!
match(x,
3, "this matches the number 3",
int, "matches any integer",
(str, int), lambda a, b: "a tuple (a, b) you can use in a function",
[1, 2, _], "any list of 3 elements that begins with [1, 2]",
{'x': _}, "any dict with a key 'x' and any value associated",
_, "anything else"
)
You can match [HEAD, TAIL]
from pampy import match, HEAD, TAIL, _
x = [1, 2, 3]
match(x, [1, TAIL], lambda t: t) # => [2, 3]
match(x, [HEAD, TAIL], lambda h, t: (h, t)) # => (1, [2, 3])
TAIL
and REST
actually mean the same thing.
You can nest lists and tuples
from pampy import match, _
x = [1, [2, 3], 4]
match(x, [1, [_, 3], _], lambda a, b: [1, [a, 3], b]) # => [1, [2, 3], 4]
You can nest dicts. And you can use _ as key!
pet = { 'type': 'dog', 'details': { 'age': 3 } }
match(pet, { 'details': { 'age': _ } }, lambda age: age) # => 3
match(pet, { _ : { 'age': _ } }, lambda a, b: (a, b)) # => ('details', 3)
It feels like putting multiple _ inside dicts shouldn't work. Isn't ordering in dicts not guaranteed ?
But it does because
in Python 3.7, dict maintains insertion key order by default
You can match class hierarchies
class Pet: pass
class Dog(Pet): pass
class Cat(Pet): pass
class Hamster(Pet): pass
def what_is(x):
return match(x,
Dog, 'dog',
Cat, 'cat',
Pet, 'any other pet',
_, 'this is not a pet at all',
)
what_is(Cat()) # => 'cat'
what_is(Dog()) # => 'dog'
what_is(Hamster()) # => 'any other pet'
what_is(Pet()) # => 'any other pet'
what_is(42) # => 'this is not a pet at all'
Using Dataclasses
Pampy supports Python 3.7 dataclasses. You can pass the operator _
as arguments and it will match those fields.
@dataclass
class Pet:
name: str
age: int
pet = Pet('rover', 7)
match(pet, Pet('rover', _), lambda age: age) # => 7
match(pet, Pet(_, 7), lambda name: name) # => 'rover'
match(pet, Pet(_, _), lambda name, age: (name, age)) # => ('rover', 7)
Using typing
Pampy supports typing annotations.
class Pet: pass
class Dog(Pet): pass
class Cat(Pet): pass
class Hamster(Pet): pass
timestamp = NewType("year", Union[int, float])
def annotated(a: Tuple[int, float], b: str, c: E) -> timestamp:
pass
match((1, 2), Tuple[int, int], lambda a, b: (a, b)) # => (1, 2)
match(1, Union[str, int], lambda x: x) # => 1
match('a', Union[str, int], lambda x: x) # => 'a'
match('a', Optional[str], lambda x: x) # => 'a'
match(None, Optional[str], lambda x: x) # => None
match(Pet, Type[Pet], lambda x: x) # => Pet
match(Cat, Type[Pet], lambda x: x) # => Cat
match(Dog, Any, lambda x: x) # => Dog
match(Dog, Type[Any], lambda x: x) # => Dog
match(15, timestamp, lambda x: x) # => 15
match(10.0, timestamp, lambda x: x) # => 10.0
match([1, 2, 3], List[int], lambda x: x) # => [1, 2, 3]
match({'a': 1, 'b': 2}, Dict[str, int], lambda x: x) # => {'a': 1, 'b': 2}
match(annotated,
Callable[[Tuple[int, float], str, Pet], timestamp], lambda x: x
) # => annotated
For iterable generics actual type of value is guessed based on the first element.
match([1, 2, 3], List[int], lambda x: x) # => [1, 2, 3]
match([1, "b", "a"], List[int], lambda x: x) # => [1, "b", "a"]
match(["a", "b", "c"], List[int], lambda x: x) # raises MatchError
match(["a", "b", "c"], List[Union[str, int]], lambda x: x) # ["a", "b", "c"]
match({"a": 1, "b": 2}, Dict[str, int], lambda x: x) # {"a": 1, "b": 2}
match({"a": 1, "b": "dog"}, Dict[str, int], lambda x: x) # {"a": 1, "b": "dog"}
match({"a": 1, 1: 2}, Dict[str, int], lambda x: x) # {"a": 1, 1: 2}
match({2: 1, 1: 2}, Dict[str, int], lambda x: x) # raises MatchError
match({2: 1, 1: 2}, Dict[Union[str, int], int], lambda x: x) # {2: 1, 1: 2}
Iterable generics also match with any of their subtypes.
match([1, 2, 3], Iterable[int], lambda x: x) # => [1, 2, 3]
match({1, 2, 3}, Iterable[int], lambda x: x) # => {1, 2, 3}
match(range(10), Iterable[int], lambda x: x) # => range(10)
match([1, 2, 3], List[int], lambda x: x) # => [1, 2, 3]
match({1, 2, 3}, List[int], lambda x: x) # => raises MatchError
match(range(10), List[int], lambda x: x) # => raises MatchError
match([1, 2, 3], Set[int], lambda x: x) # => raises MatchError
match({1, 2, 3}, Set[int], lambda x: x) # => {1, 2, 3}
match(range(10), Set[int], lambda x: x) # => raises MatchError
For Callable any arg without annotation treated as Any.
def annotated(a: int, b: int) -> float:
pass
def not_annotated(a, b):
pass
def partially_annotated(a, b: float):
pass
match(annotated, Callable[[int, int], float], lambda x: x) # => annotated
match(not_annotated, Callable[[int, int], float], lambda x: x) # => raises MatchError
match(not_annotated, Callable[[Any, Any], Any], lambda x: x) # => not_annotated
match(annotated, Callable[[Any, Any], Any], lambda x: x) # => raises MatchError
match(partially_annotated,
Callable[[Any, float], Any], lambda x: x
) # => partially_annotated
TypeVar is not supported.
All the things you can match
As Pattern you can use any Python type, any class, or any Python value.
The operator _
and built-in types like int
or str
, extract variables that are passed to functions.
Types and Classes are matched via instanceof(value, pattern)
.
Iterable
Patterns match recursively through all their elements. The same goes for dictionaries., Pattern Example, What it means, Matched Example, Arguments Passed to function, NOT Matched Example, ---------------, --------------, ---------------, -----------------------------, ------------------, "hello"
, only the string "hello"
matches, "hello"
, nothing, any other value, None
, only None
, None
, nothing, any other value, int
, Any integer, 42
, 42
, any other value, float
, Any float number, 2.35
, 2.35
, any other value, str
, Any string, "hello"
, "hello"
, any other value, tuple
, Any tuple, (1, 2)
, (1, 2)
, any other value, list
, Any list, [1, 2]
, [1, 2]
, any other value, MyClass
, Any instance of MyClass. And any object that extends MyClass., MyClass()
, that instance, any other object, _
, Any object (even None), that value, ANY
, The same as _
, that value, (int, int)
, A tuple made of any two integers, (1, 2)
, 1
and 2
, (True, False), [1, 2, _]
, A list that starts with 1, 2 and ends with any value, [1, 2, 3]
, 3
, [1, 2, 3, 4]
, [1, 2, TAIL]
, A list that start with 1, 2 and ends with any sequence, [1, 2, 3, 4]
, [3, 4]
, [1, 7, 7, 7]
, {'type':'dog', age: _ }
, Any dict with type: "dog"
and with an age, {"type":"dog", "age": 3}
, 3
, {"type":"cat", "age":2}
, {'type':'dog', age: int }
, Any dict with type: "dog"
and with an int
age, {"type":"dog", "age": 3}
, 3
, {"type":"dog", "age":2.3}
, re.compile('(\w+)-(\w+)-cat$')
, Any string that matches that regular expression expr, "my-fuffy-cat"
, "my"
and "puffy"
, "fuffy-dog"
, Pet(name=_, age=7)
, Any Pet dataclass with age == 7
, Pet('rover', 7)
, ['rover']
, Pet('rover', 8)
, Any
, The same as _
, that value, Union[int, float, None]
, Any integer or float number or None, 2.35
, 2.35
, any other value, Optional[int]
, The same as Union[int, None]
, 2
, 2
, any other value, Type[MyClass]
, Any subclass of MyClass. And any class that extends MyClass., MyClass
, that class, any other object, Callable[[int], float]
, Any callable with exactly that signature, def a(q:int) -> float: ...
, that function, def a(q) -> float: ...
, Tuple[MyClass, int, float]
, The same as (MyClass, int, float)
, Mapping[str, int]
Any subtype of Mapping
acceptable too, any mapping or subtype of mapping with string keys and integer values, {'a': 2, 'b': 3}
, that dict, {'a': 'b', 'b': 'c'}
, Iterable[int]
Any subtype of Iterable
acceptable too, any iterable or subtype of iterable with integer values, range(10)
and [1, 2, 3]
, that iterable, ['a', 'b', 'v']
, ## Using default
By default match()
is strict. If no pattern matches, it raises a MatchError
.
You can instead provide a fallback value using default
to be used when nothing matches.
>>> match([1, 2], [1, 2, 3], "whatever")
MatchError: '_' not provided. This case is not handled: [1, 2]
>>> match([1, 2], [1, 2, 3], "whatever", default=False)
False
Using Regular Expressions
Pampy supports Python's Regex. You can pass a compiled regex as pattern, and Pampy is going to run patter.search()
, and then pass to the action function the result of .groups()
.
def what_is(pet):
return match(pet,
re.compile('(\w+)-(\w+)-cat$'), lambda name, my: 'cat '+name,
re.compile('(\w+)-(\w+)-dog$'), lambda name, my: 'dog '+name,
_, "something else"
)
what_is('fuffy-my-dog') # => 'dog fuffy'
what_is('puffy-her-dog') # => 'dog puffy'
what_is('carla-your-cat') # => 'cat carla'
what_is('roger-my-hamster') # => 'something else'
Install for Python3
Pampy works in Python >= 3.6 Because dict matching can work only in the latest Pythons.
To install it:
$ pip install pampy
or
$ pip3 install pampy
If you really must use Python2
Pampy is Python3-first, but you can use most of its features in Python2 via this backport by Manuel Barkhau:
pip install backports.pampy
from backports.pampy import match, HEAD, TAIL, _