Coroutines have always seemed to be an interesting concept to me, although I’ve never dived deep in that concept. When I first heard the name while trying to learn Lua (which I didn’t pursue), I thought this was just another fancy name of subroutines.
Before going further, we have to notice the similarities and differences between the concepts of subroutines and coroutines.
Subroutines are the normal functions that we have, they may or may not return a value. Coroutines are a lot like subroutines, but instead of returning values they yield values.
A subroutine works in context to a parent function. So the parent code can be made up of multiple subroutines, with each one executing, doing its work, and getting destroyed once the work is complete. If the parent calls again, a new instance of subroutine is created and executed which is independent of the previous call.
In contrast, a coroutine can start executing, do its work, pause, and return to the parent. When the parent calls again, the coroutine starts from where it paused during the previous call.
The best real world application I could find was using Coroutines for implementing finite state machines, since these too store their current state and move to another state based on inputs, and resume execution once the next state has finished its job.
Before implementing finite state machines using coroutines, let’s start with the syntax.
Syntax of Coroutines
Coroutine derive its syntax from generator because both of these work on the concept of pausing and yielding values.
As we’ve seen in Generator functions, we can either return a value which
completes the execution of a function, or we can yield a value which
temporarily suspends the execution until the function is called again.
Coroutines work the same way, though a bit in reverse than generators –
generators provide values, coroutines consume values and do something
with them. While a piece of code can be written in a way so that it
produces and consumes values at the same time,
it’s not a good idea
to do so Generator functions return value using a
Coroutines use the same yield statement to consume values, like this:
When, inside the function body, a yield statement is hit, the execution
pauses and the function waits for some value from the caller coroutine,
coroutine.send() to send the value.
send is called, the execution starts again, the caller does its
work and pauses again when the next
yield statement is hit. This goes
again until the caller calls the
close method which raises
GeneratorExit exception inside the callee, which we have to catch and
Using the above concepts, the following code provides a concrete example of coroutines:
A coroutine, once instanciated will need to be primed by calling its
*next*() method, this executes the code till the first
statement. It can then start functioning as required:
Coroutines as pipelines of data
Since coroutines can call other coroutines, these can be used as pipelines which pass around the data and do something with it.
Conceptually, each coroutine will require data on which to act upon. This data is sent by either another coroutine or by some source function.
As a trivial example, let’s read a file and count the number of vowels in lines which contain the word ‘gene’ in the text.
Coroutines as finite state machines
Author Tushar Tyagi
LastMod Jan 15, 2015