What to use in replacement of an interface/protocol in python
Question:
I am making a chess game and wanted to make a standard piece interface/protocol. Python does not have those in the language, so what am I supposed to use? I read a bit about factories, but I’m not sure how they would help. Thanks in advance!
Answers:
In short, you probably don’t need to worry about it at all. Since Python uses duck typing – see also the Wikipedia article for a broader definition – if an object has the right methods, it will simply work, otherwise exceptions will be raised.
You could possibly have a Piece base class with some methods throwing NotImplementedError to indicate they need to be re-implemented:
class Piece(object):
def move(<args>):
raise NotImplementedError(optional_error_message)
class Queen(Piece):
def move(<args>):
# Specific implementation for the Queen's movements
# Calling Queen().move(<args>) will work as intended but
class Knight(Piece):
pass
# Knight().move() will raise a NotImplementedError
Alternatively, you could explicitly validate an object you receive to make sure it has all the right methods, or that it is a subclass of Piece by using isinstance or isubclass.
Note that checking the type may not be considered “Pythonic” by some and using the NotImplementedError approach or the abc module – as mentioned in this very good answer – could be preferable.
Your factory just has to produce instances of objects having the right methods on them.
I wrote a chess game in Python (with tkinter), and the way I did it was to have a Piece class, Queen/Knight/etc. classes that inherit from the Piece class, a Player class, a Square class, and a main program class for tkinter’s main loop. Each Piece has a color and location, and a method to help generate move sets for pieces that move in straight lines until blocked. Specific Piece subclasses each contain a method to determine their move set. A Square object contains a Piece and the square’s location on the board.
The main program class has an __init__ that sets up the board, places the pieces, loads piece icons and sound effects, and initializes options. A draw_board method redraws the board, resetting all pieces and rebinding the hotkeys. Then there are various other methods to load new icons, start a new game, set the volume, save, undo, castle, and so on.
I’m not done with version 10 yet, but you can get the source code and assets for version 9 here.
You can also check out the open source Shane’s Chess Information Database. I’ve never used it, but it looks pretty nice.
I don’t usually use interfaces in Python, but if you have to do it you can use zope.interface. You can then verify whether classes or objects implement certain interfaces. Also, it can also raise errors if classes don’t implement all methods or attributes. Twisted and other frameworks use this library.
Even though Python, being dynamic, can use duck-typing, it is still possible to implement what Java and C# call “interfaces”. This is done by declaring an Abstract Base Class. https://docs.python.org/2/library/abc.html or https://docs.python.org/3.4/library/abc.html
When you define the ABC, put all of the Interface-like methods in it and have in their body either pass or raise NotImplementedError. The child classes inherit from your ABC and override these methods just like any other child class overrides parent class methods. (Since Python has multiple inheritance, they can inherit from your ABC plus any other class you like.)
New in Python 3.8:
Some of the benefits of interfaces and protocols are type hinting during the development process using tools built into IDEs and static type analysis for detection of errors before runtime. This way, a static analysis tool can tell you when you check your code if you’re trying to access any members that are not defined on an object, instead of only finding out at runtime.
The typing.Protocol class was added to Python 3.8 as a mechanism for “structural subtyping.” The power behind this is that it can be used as an implicit base class. That is, any class that has members that match the Protocol‘s defined members is considered to be a subclass of it for purposes of static type analysis.
The basic example given in PEP 544 shows how this can be used.
from typing import Protocol
class SupportsClose(Protocol):
def close(self) -> None:
# ...
class Resource:
# ...
def close(self) -> None:
self.file.close()
self.lock.release()
def close_all(things: Iterable[SupportsClose]) -> None:
for thing in things:
thing.close()
file = open('foo.txt')
resource = Resource()
close_all([file, resource]) # OK!
close_all([1]) # Error: 'int' has no 'close' method
Note: The typing-extensions package backports typing.Protocol for Python 3.5+.
The beauty of Python is that interfaces aren’t necessary. Due to duck typing, you can just create several classes that all have the same method signature:
class Queen:
def move(self, x, y):
#do stuff
class Pawn:
def move(self, x, y):
# do stuff
Instances of these classes can be used interchangeably:
def make_move(piece, x, y):
piece.move(x, y)
q = Queen()
make_move(q, 0, 0)
p = Pawn()
make_move(p, 4, 5)
Note the above is by no means a good design for a fully fledged chess game. It is only for illustration purposes.
There is a good way to mimick Interfaces in Python.
Use metaclass=ABCMeta when generating an "Interface" class in Python
and the @abstractmethod decorator for all the methods which have to be implemented for this interface. Both are from abc class. (If not implementing any of such @abstractmethod-decorated methods, while inheriting, thus implementing the "Interface" class, a NotImplementedError will be raised, as soon as one generates an instance from such a concrete class.)
As a naming convention, begin all such classes with a capital I (for Interface).
from abc import ABCMeta, abstractmethod
class IPiece(metaclass=ABCMeta):
"The Piece Interface"
@abstractmethod
def move(<args>):
"NotImplementedError is superfluous, instead, one can use this space"
"To write some remarks, comments, annotations..."
class Queen(Piece):
def move(<args>):
# Specific implementation for the Queen's movements
typing.Protocol concrete mypy example (static type checking)
https://stackoverflow.com/a/50255847/895245 provided an example, but here’s a slightly more concrete one with sample mypy output. Related: How to implement virtual methods in Python?
protocol.py
from typing import Protocol
class CanFly(Protocol):
def fly(self) -> str:
pass
def fly_fast(self) -> str:
return 'CanFly.fly_fast'
class Bird(CanFly):
def fly(self):
return 'Bird.fly'
def fly_fast(self):
return 'Bird.fly_fast'
class FakeBird(CanFly):
pass
assert Bird().fly() == 'Bird.fly'
assert Bird().fly_fast() == 'Bird.fly_fast'
# mypy error
assert FakeBird().fly() is None
# mypy error
assert FakeBird().fly_fast() == 'CanFly.fly_fast'
If we run this file, the asserts pass, as we didn’t add any dynamic typechecking:
python protocol.py
but if we typecheck if mypy:
python -m pip install --user mypy
mypy protocol.py
we get an error as expected:
protocol.py:22: error: Cannot instantiate abstract class "FakeBird" with abstract attribute "fly"
protocol.py:24: error: Cannot instantiate abstract class "FakeBird" with abstract attribute "fly"
It is a bit unfortunate however that the error checking only picks up the error on instantiation, and not at class definition.
typing.Protocol counts methods as abstract when their body is "empty"
I’m not sure what they count as empty, but both all of the following count as empty:
pass... ellipsis objectraise NotImplementedError()
So the best possibility is likely:
protocol_empty.py
from typing import Protocol
class CanFly(Protocol):
def fly(self) -> None:
raise NotImplementedError()
class Bird(CanFly):
def fly(self):
return None
class FakeBird(CanFly):
pass
Bird().fly()
FakeBird().fly()
which fails as desired:
protocol_empty.py:15: error: Cannot instantiate abstract class "FakeBird" with abstract attribute "fly"
protocol_empty.py:15: note: The following method was marked implicitly abstract because it has an empty function body: "fly". If it is not meant to be abstract, explicitly return None.
but if e.g. we replace the:
raise NotImplementedError()
with some random "non-empty" statement such as:
x = 1
then mypy does not count them as virtual and gives no errors.
Tested on Python 3.10.7, mypy 0.982, Ubuntu 21.10.
I am making a chess game and wanted to make a standard piece interface/protocol. Python does not have those in the language, so what am I supposed to use? I read a bit about factories, but I’m not sure how they would help. Thanks in advance!
In short, you probably don’t need to worry about it at all. Since Python uses duck typing – see also the Wikipedia article for a broader definition – if an object has the right methods, it will simply work, otherwise exceptions will be raised.
You could possibly have a Piece base class with some methods throwing NotImplementedError to indicate they need to be re-implemented:
class Piece(object):
def move(<args>):
raise NotImplementedError(optional_error_message)
class Queen(Piece):
def move(<args>):
# Specific implementation for the Queen's movements
# Calling Queen().move(<args>) will work as intended but
class Knight(Piece):
pass
# Knight().move() will raise a NotImplementedError
Alternatively, you could explicitly validate an object you receive to make sure it has all the right methods, or that it is a subclass of Piece by using isinstance or isubclass.
Note that checking the type may not be considered “Pythonic” by some and using the NotImplementedError approach or the abc module – as mentioned in this very good answer – could be preferable.
Your factory just has to produce instances of objects having the right methods on them.
I wrote a chess game in Python (with tkinter), and the way I did it was to have a Piece class, Queen/Knight/etc. classes that inherit from the Piece class, a Player class, a Square class, and a main program class for tkinter’s main loop. Each Piece has a color and location, and a method to help generate move sets for pieces that move in straight lines until blocked. Specific Piece subclasses each contain a method to determine their move set. A Square object contains a Piece and the square’s location on the board.
The main program class has an __init__ that sets up the board, places the pieces, loads piece icons and sound effects, and initializes options. A draw_board method redraws the board, resetting all pieces and rebinding the hotkeys. Then there are various other methods to load new icons, start a new game, set the volume, save, undo, castle, and so on.
I’m not done with version 10 yet, but you can get the source code and assets for version 9 here.
You can also check out the open source Shane’s Chess Information Database. I’ve never used it, but it looks pretty nice.
I don’t usually use interfaces in Python, but if you have to do it you can use zope.interface. You can then verify whether classes or objects implement certain interfaces. Also, it can also raise errors if classes don’t implement all methods or attributes. Twisted and other frameworks use this library.
Even though Python, being dynamic, can use duck-typing, it is still possible to implement what Java and C# call “interfaces”. This is done by declaring an Abstract Base Class. https://docs.python.org/2/library/abc.html or https://docs.python.org/3.4/library/abc.html
When you define the ABC, put all of the Interface-like methods in it and have in their body either pass or raise NotImplementedError. The child classes inherit from your ABC and override these methods just like any other child class overrides parent class methods. (Since Python has multiple inheritance, they can inherit from your ABC plus any other class you like.)
New in Python 3.8:
Some of the benefits of interfaces and protocols are type hinting during the development process using tools built into IDEs and static type analysis for detection of errors before runtime. This way, a static analysis tool can tell you when you check your code if you’re trying to access any members that are not defined on an object, instead of only finding out at runtime.
The typing.Protocol class was added to Python 3.8 as a mechanism for “structural subtyping.” The power behind this is that it can be used as an implicit base class. That is, any class that has members that match the Protocol‘s defined members is considered to be a subclass of it for purposes of static type analysis.
The basic example given in PEP 544 shows how this can be used.
from typing import Protocol
class SupportsClose(Protocol):
def close(self) -> None:
# ...
class Resource:
# ...
def close(self) -> None:
self.file.close()
self.lock.release()
def close_all(things: Iterable[SupportsClose]) -> None:
for thing in things:
thing.close()
file = open('foo.txt')
resource = Resource()
close_all([file, resource]) # OK!
close_all([1]) # Error: 'int' has no 'close' method
Note: The typing-extensions package backports typing.Protocol for Python 3.5+.
The beauty of Python is that interfaces aren’t necessary. Due to duck typing, you can just create several classes that all have the same method signature:
class Queen:
def move(self, x, y):
#do stuff
class Pawn:
def move(self, x, y):
# do stuff
Instances of these classes can be used interchangeably:
def make_move(piece, x, y):
piece.move(x, y)
q = Queen()
make_move(q, 0, 0)
p = Pawn()
make_move(p, 4, 5)
Note the above is by no means a good design for a fully fledged chess game. It is only for illustration purposes.
There is a good way to mimick Interfaces in Python.
Use metaclass=ABCMeta when generating an "Interface" class in Python
and the @abstractmethod decorator for all the methods which have to be implemented for this interface. Both are from abc class. (If not implementing any of such @abstractmethod-decorated methods, while inheriting, thus implementing the "Interface" class, a NotImplementedError will be raised, as soon as one generates an instance from such a concrete class.)
As a naming convention, begin all such classes with a capital I (for Interface).
from abc import ABCMeta, abstractmethod
class IPiece(metaclass=ABCMeta):
"The Piece Interface"
@abstractmethod
def move(<args>):
"NotImplementedError is superfluous, instead, one can use this space"
"To write some remarks, comments, annotations..."
class Queen(Piece):
def move(<args>):
# Specific implementation for the Queen's movements
typing.Protocol concrete mypy example (static type checking)
https://stackoverflow.com/a/50255847/895245 provided an example, but here’s a slightly more concrete one with sample mypy output. Related: How to implement virtual methods in Python?
protocol.py
from typing import Protocol
class CanFly(Protocol):
def fly(self) -> str:
pass
def fly_fast(self) -> str:
return 'CanFly.fly_fast'
class Bird(CanFly):
def fly(self):
return 'Bird.fly'
def fly_fast(self):
return 'Bird.fly_fast'
class FakeBird(CanFly):
pass
assert Bird().fly() == 'Bird.fly'
assert Bird().fly_fast() == 'Bird.fly_fast'
# mypy error
assert FakeBird().fly() is None
# mypy error
assert FakeBird().fly_fast() == 'CanFly.fly_fast'
If we run this file, the asserts pass, as we didn’t add any dynamic typechecking:
python protocol.py
but if we typecheck if mypy:
python -m pip install --user mypy
mypy protocol.py
we get an error as expected:
protocol.py:22: error: Cannot instantiate abstract class "FakeBird" with abstract attribute "fly"
protocol.py:24: error: Cannot instantiate abstract class "FakeBird" with abstract attribute "fly"
It is a bit unfortunate however that the error checking only picks up the error on instantiation, and not at class definition.
typing.Protocol counts methods as abstract when their body is "empty"
I’m not sure what they count as empty, but both all of the following count as empty:
pass...ellipsis objectraise NotImplementedError()
So the best possibility is likely:
protocol_empty.py
from typing import Protocol
class CanFly(Protocol):
def fly(self) -> None:
raise NotImplementedError()
class Bird(CanFly):
def fly(self):
return None
class FakeBird(CanFly):
pass
Bird().fly()
FakeBird().fly()
which fails as desired:
protocol_empty.py:15: error: Cannot instantiate abstract class "FakeBird" with abstract attribute "fly"
protocol_empty.py:15: note: The following method was marked implicitly abstract because it has an empty function body: "fly". If it is not meant to be abstract, explicitly return None.
but if e.g. we replace the:
raise NotImplementedError()
with some random "non-empty" statement such as:
x = 1
then mypy does not count them as virtual and gives no errors.
Tested on Python 3.10.7, mypy 0.982, Ubuntu 21.10.