What the reader will get out of the article
In an imperative paradigm, you get things done by giving the computer a sequence of tasks and then it executes them. While executing them, it can change states. For example, let’s say you originally set A to 5, then later on you change the value of A. You have variables in the sense that the value inside the varable varies.
In a functional paradigm, you don’t tell the computer what to do but rather you tell it what stuff is. What the greatest common divisor of a number is, what the product from 1 to n is and so on.
Because of this, variables cannot vary. Once you set a variable, it stays that way forever (note, in purely functional languages they are nto called variables because of this). Because of this, functions have no side effects in the functional paradigm. An side effect is where the function changes something outside of it. Let’s look at an example of some typical Python code:
a = 3 def some_func(): global a a = 5 some_func() print(a)
The output for this code is 5. In the functional paradigm, changing variables is a big no-no and having functions effect things outside of their scope is also a big no-no. The only thing a function can do is calculate something and return it as a result.
Now you might be thinking: “no variables, no side effects? Why is this good?”. Good question, gnarly stranger reading this.
If a function is called twice with the same parameters, it’s guranteed to return the same result. If you’ve learnt about mathematical functions, you’ll know to appreciate this benefit. This is called referential transparency. Because functions have no side effects, if you are building a program which computes things, you can speed up the program. If the program knows that func(2) equates to 3, we can store this in a table. This prevents the program from repeatedly running the same function when we already know the answer.
Typically, in functional programming we do not use loops. We use recursion. Recursion is a mathematical concept, usually it means “feeding into itself”. With a recursive function, the function repeatedly calls itself as a sub-function. Here’s a nice example of a recursive function in Python:
def factorial_recursive(n): # Base case: 1! = 1 if n == 1: return 1 # Recursive case: n! = n * (n-1)! else: return n * factorial_recursive(n-1)
To understand map, let’s look at what iterables are. An iterable is anything you can iterate over. Typically these are lists or arrays, but Python has many different types of iterates. You can even create your own objects which are iterable by implementing magic methods. A magic method is like an API that help your objects become more pythonic. You need to implement 2 magic methods to make an object an iterable:
class Counter: def __int__(self, low, high): # set class attributes inside the magic method __init__ # for "inistalise" self.current = low self.high = high def __iter__(self): # first magic method to make this object iterable return self def __next__(self): # second magic method if self.current > self.high: raise StopIteration else: self.current += 1 return self.current - 1
The first magic method, ```iter`` or dunder iter (double underscore iter) returns the iterative object, this is often used at the start of a loop. Dunder next returns what the next object is.
Let’s go into a quick termianl session and check this out:
for c in Counter(3, 8): print(c)
This will print
3 4 5 6 7 8
So now we know what an iterable object is, let’s go back to the map function. The map function let’s us apply a function to every item in an iterable. Typically we want to apply a function to every item in a list, but know that it’s possible for most iterables. Map takes 2 inputs, the function to apply and the iterable object.
Let’s say we have a list of numbers like so:
[1, 2, 3, 4, 5]
And we want to square every number, we can write code like this:
x = [1, 2, 3, 4, 5] def square(num): return num*num print(list(map(square, x)))
Functional functions in Python are lazy. If we didn’t include the
list() the function would store the definition of that, and only compute it when it is needed to compute.
It’s a bit weird to go from non-lazy evaluation to lazy evaluation all of a sudden in Python.
Now it’s nice to write a normal function like
square(num) but it doesn’t look right. We have to define a whole function just to use it once in a map? Well, we can define a function in map using a lambda (anonymous) function.
A lambda expression is a one line function. Take, for instance, this lambda expression which squares a number given to it:
square = lambda x: x * x
Now let’s run this:
>>> square(3) 9
I hear you. “Brandon, where are the arguments? what the heck is this? that doesn’t look anything like a function???”
Well, it’s kind of confusing but can be explained. So we’re assignign something to the variable “square”. this part:
Tells Python that this is a lambda function, and the input is called x. Anything after the colon is what you do with the input, and it automatically returns whatever the resultant of that is.
To simplfy our square program into one line we can do:
x = [1, 2, 3, 4, 5] print(list(map(lambda num: num * num, x)))
So in a lambda expression, all the arguments go on the left and the stuff you want to do with them go on the right.
You could actually do something insane like this:
x = [1, 2, 3, 4, 5] print(list(map(lambda x: x * map(lambda a: a + a, x), x)))
Which times each number in the list by the sum of all of the numbers in the list. It gets a little messy, no one can deny that. The truth is that there’s a certain pleasure in writing code that only other functional programmers can read. Also, it’s super cool to take a function and turn it into a one liner.
Reduce is a function that turns an interable into one thing. Typically you perform a computation on a list to reduce it down to one number. Reduce looks like this:
We can (and often will) use lambda expressions as the function.
The product of a list is every single number multiplied together. In order to do this normally you would program:
product = 1 x = [1, 2, 3, 4] for num in x: product = product * num
But with reduce you can just write:
product = reduce((lambda x, y: x * y),[1, 2, 3, 4])
To get the same product. The code is shorter, and with knowledge of functional programming it is neater.
The filter function takes an iterable and filters out all the things you don’t want in that iterable.
Normally filter takes a function and a list. It applies the function to each item in the list and if that function returns True, it does nothing. If it returns False, it removes that from the list.
The syntax looks like:
Let’s see a small example, without reduce we’ll write:
x = range(-5, 5) new_list =  for num in x: if num < 0: new_list.append(num)
With reduce, this becomes:
x = range(-5, 5) all_less_than_zero = list(filter(lambda num: num < 0, x)) print(all_less_than_zero)
Higher order functions can take functions as parameters and return functions. A very simple example would look like:
def returnSix(): return 6 def addition(num, num2): return num + num2 print(addition(4, returnSix())) # Output is 10
Or an even simpler example of the second definition, “return functions” is:
def rtnBrandon(): return "brandon" def rtnJohn(): return "john" def rtnPerson(): age = int(input("What's your age?")) if age == 21: return rtnBrandon() else: return rtnJohn()
You know earlier how I said that pure functional programming languages didn’t have variables? Well, higher order functions are what makes this easier. You don’t need to store a variable anywhere if all you’re doing is passing data through a long tunnel of functions, until it eventually prints out onto the screen.
All functions in Python are first class objects. A firstclass object is defined as having one or more of these features:
So all functions in Python are first class and can be used as a higher order function.
Partial application is a bit wacky, but super cool. You can call a function without supplying all of the arguments it requires. Let’s see this in an example. We want to create a function which takes 2 arguments, a base and an exponent, and returns base to the power of the exponent, like so:
def power(base, exponent): return base ** exponent
Now we want to have a dedicated square function, in order to work out the square of a number using the power function:
def square(base): return power(base, 2)
This works, but what if we want a cube function? or a function to the power of 4? Can we just keep on writing them forever? Well, not really. We can use partial applications here:
from functools import partial square = partial(power, exponent=2) print(square(2)) # output is 4
Isn’t that cool! We can call functions which require 2 arguments, using only 1 argument by telling Python what the second argument is.
We can also use a loop, to generate a power function that works from cubed all the way up to powers of 1000.
from functools import partial powers =  for x in range(2, 1001): powers.append(partial(power, exponent = x)) print(powers(2)) # output is 4
Real world application of partials https://github.com/tkaemming/django-subdomains/blob/master/subdomains/utils.py#L65-L72
You might of noticed, but a lot of the things we want to do in functional programming revolve around lists. Other than the reduce function, all of the funtions you have seen just generate lists. Guido (the inventor of Python) disliks functional stuff in Python becasue Python already has its own way to generate lists.
If you write “import this” into a Python IDLE session, you’ll get:
>>> import this The Zen of Python, by Tim Peters Beautiful is better than ugly. Explicit is better than implicit. Simple is better than complex. Complex is better than complicated. Flat is better than nested. Sparse is better than dense. Readability counts. Special cases aren't special enough to break the rules. Although practicality beats purity. Errors should never pass silently. Unless explicitly silenced. In the face of ambiguity, refuse the temptation to guess. There should be one-- and preferably only one --obvious way to do it. Although that way may not be obvious at first unless you're Dutch. Now is better than never. Although never is often better than *right* now. If the implementation is hard to explain, it's a bad idea. If the implementation is easy to explain, it may be a good idea. Namespaces are one honking great idea -- let's do more of those!
The important part is:
There should be one– and preferably only one –obvious way to do it.
In Python, map & filter can do the same things as a list comprehension (discussed next) can do. This breaks one of the rules of the Zen of Python.
Another talking point is Lambda. In Python, a lambda function is literally just a function. lambda is syntactic sugar. Both of these are equivalent:
foo = lamda a: 2 def foo(a): return 2
A regular function can do everything a lambda function can, but it doesn’t work the other way around. A lambda function cannot do everything that a regular function can do.
And that was a short argument about why Guido and some pythonista’s hate the idea of having the functional paradigm in Python.
Earlier, I mentioned that anything you could do with map or filter, you could do with a list comprehension. This is the part where we’ll learn about them.
A list comprehension is a way to generate lists in Python. The syntax is:
[function for item in iterable]
So let’s square every number in a list, as an example:
print([x * x for x in [1, 2, 3, 4]])
Okay, so we can see how we can apply a function to every item in a list. We didn’t even need to import any libraries, as list comprehensions are built into python. How do we go around applying a filter? Well, look at this code from earlier:
x = range(-5, 5) all_less_than_zero = list(filter(lambda num: num < 0, x)) print(all_less_than_zero)
We can convert this into a list comprehension like so:
x = range(-5, 5) all_less_than_zero = [num for num in x if num < 0]
List comprehensions support if statements like this. You no longer need to apply a million functions to something in order to get what you want. In fact, if you’re trying to make some kind of list chances are that it’ll look cleaner and easier using a list comprehension.
What if we want to square every number below 0 in a list? Well, with a lambda & map expression you’ll write:
x = range(-5, 5) all_less_than_zero = list(map(lamda num: if num < 0 then num * num, x)
So that’s seems really long and slightly complicated. With a list comprehension it’s just:
x = range(-5, 5) all_less_than_zero = [num * num for num in x if num < 0]
A list comprehension is only good for, well, lists. Map and filter work on any iterable, so what’s up with that? Well, you can use a generator expression to generate any iterable. Let’s take a look at that now.
A generator expression is like a list comprehension, but for any iterable. You can generate a tuple like so:
x = range(-5, 5) all_less_than_zero = (num * num for num in x if num < 0)
All you need to change is from the list’s square brackets [ ] to a tuple’s parathesis’ ( ).
How you can generate any iterable in Python (dictionaries, hashmaps, objects)
My key values of my blog: