Re:The Day Python Embarassed Imperative Programming
by hythloday on Monday February 27, 2012 @01:28PM
(#3639985)
I'm not disputing that Haskell makes it easier, but from a practical perspective I'd challenge the idea that nullability can't be turned off in Java with @NotNull annotations (and nullability inference tools like JastAdd[0]), or with decorators in Python. Do you see an essential difference in kind, or just a spectrum of ease of reasoning?
Re:The Day Python Embarassed Imperative Programming
by yew on Monday February 27, 2012 @05:55PM
(#3641032)
Would you mind expounding on that use of decorators? I'm not making the connection and it sounds like an interesting one.
Re:The Day Python Embarassed Imperative Programming
by hythloday on Monday February 27, 2012 @07:51PM
(#3641322)
Sure, it's just a specific kind of type-checking decorator in the same vein as the ones you can see at e.g. [0]. Obviously it can be made a lot more sophisticated to only pick up certain arguments. You can do a similar thing with metaclasses to prevent object attributes being set as None, too.
def isnotnone(fn):
def decorator(*args, **kwargs):
for arg in args:
if arg is None:
raise TypeError("{0} does not accept None".format(fn.func_name))
for arg in kwargs.values():
if arg is None:
raise TypeError("{0} does not accept None".format(fn.func_name))
return fn(*args,**kwargs)
return decorator
@isnotnone
def f(a, b=dict):
return a(), b()
>>> f(None)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "<stdin>", line 5, in decorator
TypeError: f does not accept None
>>> f(b=None)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "<stdin>", line 8, in decorator
TypeError: f does not accept None
>>> f()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "<stdin>", line 9, in decorator
TypeError: f() takes at least 1 argument (0 given)
>>> f(str)
('', {})
>>>
[0]http://code.activestate.com/recipes/454322-type-checking-dec...
But the thing that makes Haskell unique is not that you have explicit monads. It's that you can (1) use monads other than the IO monad (like Maybe, [], STM, ST), and (2) that you can choose to use no monads at all in a given function.
To put it another way, Haskell is unique because it allows you write functions without using monads, but in other languages you can't turn them off. Like nullable types, which can't be turned off in Java (ever debug NullPointerExceptions?) or Python, but which have to be enabled on a case-by-case basis in Haskell.