Saturday, October 17, 2009

Devilishly Clever

I was doing my usual research on Python when I ran across this recipe to make tail recursive calls not blow the stack in Python: Tail Call Optimization Decorator. I read through the code and thought, "Wow! This is devilishly clever!" As a thought exercise, I think it's awesome. But, it got me thinking about clever and production code. In my opinion, clever is never good or wanted in production code. It's great to learn and understand clever code though. It's a great mental workout to keep you sharp.

So, what's my point with all of this besides to say that clever is bad? The example in the above link is factorial (which as everyone knows is hated by me, but that's another story). But, the amazing thing about factorial examples are that they are dead simple, yet there's several ways to get the answer. Here's a few that I came up with:
`def reg_fact(x):    if x is 1:        return 1    return reg_fact(x - 1) * xdef tail_fact(x):    def func(acc, x):        if x is 1:            return acc        return func(acc * x, x - 1)    return func(1, x)def not_rec_fact(x):    result = 1    for each in range(2, x + 1):        result *= each    return resultdef not_rec_fact_fancy(x):    return reduce(lambda result, each: result * each, range(1, x + 1))import operatordef not_rec_fact_super_fancy(x):    return reduce(operator.mul, range(1, x + 1))`

Each of these compute factorial. Amazingly, there's even more ways than what I listed (the math wizards in the audience know what they are). Now, think about this: Computing factorial is dead simple. What happens when we get to harder problems? Being clever can actually get in the way of making the code easy to understand. It might even kill performance. Let's think back to the devilishly clever code. The performance of making Python tail recursive is awful. Sure, it's pure and tail recursive, but in production code that is deadly. What we want is simple and to the point. It's why I generally like solutions that need less code. There's less noise to get in the way of understanding and generally can mean better performance as well.

Real world problems are hardly ever as straightforward as factorial. The balancing act comes when you drop a solution because it's not working for whatever reason. Raising and catching exceptions so you can have a tail recursive factorial is overkill. It took more code than the non-recursive version. It begs for the programmer to know their tool set and to know how to solve problems in that tool set that are straightforward. Tail recursion is powerful in languages like Haskell and Erlang. But, there's always another way of doing things that can make more sense in the language you are using. In our case, the other ways were just as easy yet more scalable for our tool set. Food for thought the next time you go down the path of clever and end up writing more noise than solution.

Saturday, October 10, 2009

Freezing Objects in Python

Someone stop me. I like freezing simple objects or what Eric Evans calls "Value Objects" in his excellent "Domain Driven Design" book. Python doesn't have immutable objects (ala freeze in Ruby) explicitly, but we can easily create it. Python gives us the power to get under its covers. Here's my implementation:
`class ValueObject(object):    def __setattr__(self, name, value):        if name == 'value' and hasattr(self, 'value'):            raise AttributeError("Can not change value attribute")        else:            self.__dict__[name] = value`

Do you have to ask? Yes, I made a test. Here they are:
`import unittestclass Test(unittest.TestCase):    def testSimple(self):        class Cents(ValueObject):            def __init__(self, value):                self.value = value            def __str__(self):                return str(self.value) + " cents"        subject = Cents(5)        self.assertEquals(5, subject.value)        def set_value():            subject.value = 6        self.failUnlessRaises(AttributeError, set_value)`

Monday, October 05, 2009

method_missing in Python

OK, I know there is a few implementations of this knocking around on the net already, but I'm in learning mode. So, here's my implementation of Ruby's method_missing:
`class PossibleMissingAttribute(object):    def __init__(self, object, name):        self._object = object        self._name = name    def __call__(self, *args, **kwargs):        return self._object._method_missing(self._name, *args, **kwargs)        def __str__(self):        return self.__class__.__name__ + ": " + str(self._object) + "." + self._name        def __getattr__(self, name):        return None        def __nonzero__(self):        return False    class MethodMissingError(Exception):    def __init__(self, object, name, *args, **kwargs):        self.object = object        self.name = name        self.args = args        self.kwargs = kwargs    def __str__(self):        return repr(str(self.object) + "." + self.name)class Missing(object):    def __getattr__(self, name):        return PossibleMissingAttribute(self, name)    def _method_missing(self, name, *args, **kwargs):        raise MethodMissingError(self, name, *args, **kwargs)`

Three objects and that's it! One to be a placeholder when an argument is missing. One object to represent the MethodMissingError and the last to be an abstract class to inherit from. But, you could easily just put these methods anywhere. Here's my tests:
`import unittestclass TestMissing(unittest.TestCase):    def test_missing(self):       class TestClass(Missing):           def __str__(self):               return self.__class__.__name__           def existing(self):               return "i am"           def _method_missing(self, name, *args, **kwargs):               return args[0]       self.assertEquals("am not", TestClass().missing("am not"))       self.assertEquals("i am", TestClass().existing())       self.assertEquals(None, TestClass().missing.something_else_missing)       self.assertFalse(TestClass().missing)       self.assertEquals("PossibleMissingAttribute: TestClass.missing", str(TestClass().missing))           def test_exception(self):        class TestClass(Missing):           def __str__(self):               return self.__class__.__name__        self.failUnlessRaises(MethodMissingError, lambda: TestClass().missing())`

You've seen my implementation. Here's another one that I found on the net: method_missing in Python
`class MethodMissing(object):    def method_missing(self, attr, *args, **kwargs):        “”" Stub: override this function “”"        raise AttributeError(”Missing method %s called.”%attr)    def __getattr__(self, attr):        def callable(*args, **kwargs):            return self.method_missing(attr, *args, **kwargs)        return callable`

Um, this version is much shorter than mine. It gets the same job done and is more obvious of what it is doing. Simply returning a function on __getattr__ is the best way to go. It's what my PossibleArgumentMissing was doing. But, where I tried to get all fancy using __call__ to mimic a function which caused my version to be longer. I was also trying to get the argument to come back to return false in if conditions and be almost like None (again, I'm learning and was looking at what was possible). Simplest is best.

This was a fun thought exercise. I love all the hooks Python provides to allow you to get underneath the hood if need be. I will be soon doing a post on descriptors and even decorators. I find Python tends to sway heavier on the functional side of programming versus object-oriented. I'm loving the succinct, yet readable code. Too many languages get into the being so terse that they sacrifice readability or debugability (is that a word? Is now!).

Tuesday, September 29, 2009

Prototypes in Python

I'm in love with prototype-based languages, but there only a few to play with ECMAScript (Javascript) and Self. Since I'm learning Python, I thought it would be a good experiment to implement prototypes in Python. For one thing, I would get to learn some of the meta capabilities and other advanced features. A pure thought experiment just for learning and seeing how far I could push another paradigm onto Python. Here is what I came up with:
`class Prototype(object):    def __init__(self, *args):        self._parents = []        for each in args:            self._parents.append(each)            def clone(self, *args):        return self.__class__(self, *args)        def __getattr__(self, name):        for each in self._parents:            try:                return getattr(each, name)            except AttributeError:                pass        raise AttributeError(name)`

The implementation is suprisingly simple: three methods!

The first is the initialize method for new instances. It takes a list of objects that we will use as the parents of our new object. Of course, by simply calling Prototype's constructor with no arguments, we get a clean object. I made the ability to have multiple inheritance simple because of Self.

The second method is the clone method and it simply calls the constructor for another object makes itself the first parent with the rest of the arguments becoming the other parents. Nothing to it!

The last method is the real meat. It only get called by the Python engine when an attribute fails to be looked up. All I do is call each parent and if one succeeds, it returns the answer. This is a depth first search of the parent hierarchy. Simple.

And that's all we need to implement prototypes in Python. Now, I just need to explain some practical uses of this in some later blog entries.

How do I know all of this works? Here's my tests:
`import unittestclass Test(unittest.TestCase):    def testSimple(self):        parent = Prototype()        child = parent.clone()                parent.value = 0        self.assertEquals(0, child.value)        child.value = 1                self.assertEquals(1, child.value)        self.assertEquals(0, parent.value)                self.failUnlessRaises(AttributeError, lambda:parent.unknown)        def testMultipleParents(self):        parent1 = Prototype()        parent2 = Prototype()        child = parent1.clone(parent2)                parent2.value = 2        self.failUnlessRaises(AttributeError, lambda:parent1.value)        self.assertEquals(2, child.value)`

If anyone sees anything wrong in my implementation or un-Pythonic let me know. I'm still learning, but I thought it would be cool to show my progress.

Sunday, September 13, 2009

Constraints in Python

With every language that I learn, I pay close attention to my work flow. I take note of anything that gets in the way of getting my task accomplished. In a perfect language, I should never have to worry about the language ever. A free flow of ideas should come out. Whenever I find myself totally immersed in a problem, I keep track of what made it so. Recently, the one feature that surprised me the most was Python's forced indention. At first I thought, how nice not to use curly brackets to denote scope, but what about that white space? But, I put my best foot forward and worked on some code. I must admit that I love it now! It makes code more readable by keeping the rules consistent (you have to be or your code will have bugs). Formatting my code is now one less worry I have. I am forced to indent as I go along. This constraint is liberating in the fact that it allows me concentrate on the problem at hand and is one less thing to worry about. No more worrying about if the curly braces go on the same line or not depending on local coding standards. It makes code easier to read across Python programmers. In fact, I have yet to meet a Python program that I could not understand easily from first glance which is a rare accomplishment in other languages.

I also love the constraint about one line lambdas. I thought it was going to drive me crazy, but I have embraced completely. It keeps your lambda functions short and sweet like they really should be. But, the best side effect is that if that lambda needs to grow, you simply move it out to a named function automatically making your code more readable. Excellent.

Both of the features that I just mentioned are merely constraints that take away thinking about the language and force you to write readable code. I always like and embrace the design decisions to constrain the coder to make more maintainable code. The less I have to think about the language, the more time I spend on the problem I am trying to solve and that's the way it should be. It just amazed me that sometimes I find them in the strangest of ways. I would have never thought that forced indention or one lined lambdas would bring me so much coding pleasure.