Skip to content

sleibrock/functional-decorators

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Functional Decorators

Writing code is overrated - decorators is the new code.

What is Functional Decorators?

Functional Decorators is a project based on the concept of decorators in Python. A wrapper is a function that accepts a function and returns a new function that performs a task around the given function. Python has syntax to easily craft new functions from old ones via the syntax of

@decorator_function
def my_function(x):
    # code goes here
    ...

You've probably used it to route paths in web frameworks or some other kind of function manipulation. But what if instead of writing code, we wrote code that generated a function for us?

@_foldl(add)
@_fmap(cube)
@_range
def fold_cubes(x): return x

Looks like a real dumb function definition. But guess what - it does exactly what the decorators say! It applies a foldl of the built-in add function, over a list of numbers with the function cube applied to each number.

How does one run this? Simply call it afterwards!

>>> fold_cubes(100)
24502500.0

Neat, right? We really didn't do anything at all to create this function.

The Downside

You have to write it in an order backwards of what you want to do. Naturally the function we wrote, we would have verbally said "create a list of numbers, apply a function to them, and sum them up". However the decorator defintion is completely the opposite. It's "Add up a list of numbers that have an operation mapped over them".

So the order that which we write these functions may be considered confusing at first, but it's not as bad as you might think initially. The formal definition of this code block would have looked like:

def fold_cubes(x):
    return sum([cube(x) for z in range(x)])

Obviously easier to write, but the modularity of it is glued in place. We can't take functions apart and put them in other places like we can with these decorators. In the Functional Decorators, we layer decorators slowly on top of eachother to compose new functions without having to do much work at all. This is akin to writing point-free expressions in functional languages like Haskell where the variables aren't important, but the functions we compose are.

Note that the function definition at the bottom of the decorator chain isn't very important. So long as it returns x, it will do as the decorators say. But you can modify the bottom function to execute one last expression in the chain, so you can do whatever the heck you want with it! But mostly it's just there so you can name the final product function.

Should I ever use this?

Bottom-line answer: NO. But if you want to have some fun, yes!

Give me something crazy

Okay, here's a function that will return to you a sum of the list of absolute values of a function sin(x^2) where x is a radian-value converted from degrees.

@_foldl(add)
@_fmap(abs)
@_fmap(sin)
@_fmap(square)
@_fmap(radians)
@_range
def something(x): return x

Let's test it just to make sure it works.

>>> something(360)
215.36402018499368

Let's prove it works just to be sure.

>>> x = range(360)
>>> rads = map(radians, x)
>>> sqs = map(square, rads)
>>> sins = map(sin, sqs)
>>> abss = map(abs, sins)
>>> sum(abss)
215.36402018499368

Seems to work! The only downside is that we did several maps of individual functions where we can possibly just compose them all together into one big function. Unfortunately the list comprehension just looks a little bit worse for wear...

def something(x):
    return sum([abs(sin(square(radians(z)))) for z in range(x)])

That one actually isn't even remotely equal to what the decorators are doing. A real equal would be...

def something(n):
	return sum([abs(x) for x in
		[sin(y) for y in
			[square(z) for z in
				[radians(j) for j in range(n)]]]])

Disgusting, yes. But that's what the expanded form looks like in reality.

Can we log the state changes in the decorator chains?

There's a decorator called puts that allows us to print out the output of the current state of the program. It can be wedged inbetween decorators so you can see what's being passed on inbetween decorators.

>>> @_puts
... @_add1
... @_puts
... @_add1
... @_puts
... @_add1
... def f(x): return x
... 
>>> f(3)
Current state is 4
Current state is 5
Current state is 6
6

Closing Remarks

This was a project inspired by functional programming, and since Python is trying it's hardest not to be a functional language by any means, I thought I would adhere to Python rules and try something unconventional. It's funny to write an entire function without actually writing any meaningful code inside the function definition. But the performance surely takes a hit when not using list comps due to CPython's list comp internals.

About

Define your logic without a definition

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published