Coroutines are similar to generators with a few differences. The main differences are:

• generators are data producers
• coroutines are data consumers

First of all let's review the generator creation process. We can make generators like this:

def fib():
    a, b = 0, 1
    while True:
        yield a
        a, b = b, a+b

We then commonly use it in a for loop like this:

for i in fib():
    print(i)

It is fast and does not put a lot of pressure on memory because it generates the values on the fly rather than storing them in a list. Now, if we use yield in the above example, more generally, we get a coroutine. Coroutines consume values which are sent to it. A very basic example would be a grep alternative in Python:

def grep(pattern):
    print("Searching for", pattern)
    while True:
        line = (yield)
        if pattern in line:
            print(line)

Wait! What does yield return? Well we have turned it into a coroutine. It does not contain any value initially, instead we supply it values externally. We supply values by using the .send() method. Here is an example:

search = grep('coroutine')
next(search)
# Output: Searching for coroutine
search.send("I love you")
search.send("Don't you love me?")
search.send("I love coroutines instead!")
# Output: I love coroutines instead!

The sent values are accessed by yield. Why did we run next()? It is required in order to start the coroutine. Just like generators, coroutines do not start the function immediately. Instead they run it in response to the __next__() and .send() methods. Therefore, you have to run next() so that the execution advances to the yield expression.

We can close a coroutine by calling the .close() method:

search = grep('coroutine')
# ...
search.close()

There is a lot more to coroutines. I suggest you check out this awesome presentation by David Beazley.