Parallel Letter Frequency

Instructions

Count the frequency of letters in texts using parallel computation.

Parallelism is about doing things in parallel that can also be done sequentially. A common example is counting the frequency of letters. Create a function that returns the total frequency of each letter in a list of texts and that employs parallelism.

Concurrency vs Parallelism

Go supports concurrency via "goroutines" which are started with the go keyword. It is a simple, lightweight and elegant way to provide concurrency support and is one of the greatest strengths of the language.

You may notice that while this exercise is called Parallel letter frequency you don't see the term "Parallel" used very often in Go. Gophers prefer to use the term Concurrent to describe the management of multiple independent goroutines ("processes" or "threads" in other language contexts). Although these terms are often used interchangeably Gophers like to be technically correct and use "concurrent" when discussing the seemingly simultaneous executions of goroutines.

While we can plan for our programs to run in parallel, and at times they may appear to run in parallel, without strict knowledge of the execution context of our code all we can guarantee is that processes will run concurrently. In other words they may be executing sequentially faster than we can distinguish but not strictly simultaneously.

For more take a look at The Go Blog's post: Concurrency is not parallelism.

Concurrency Resources

If you are new to the concurrency features in Go here are some resources to get you started. We recommend looking over these before starting this exercise:

• Concurrency in the Golang Book
• A Tour of Go's concurrency section
• DigitalOcean Golang Tutorial: How To Run Multiple Functions Concurrently in Go
• Synchronizing Go Routines with Channels and WaitGroups
• Buffered Channels and Worker Pools

For a really deep dive you can try the book Concurrency in Go by @kat-co.

Testing

For this exercise, the unit tests cannot determine whether you wrote a good concurrent solution. Instead, it is best to solve this exercise locally and execute the benchmarks with go test -bench .. For a good solution, you should see that the concurrent version shows a lower number of nano seconds per operation (ns/op) than the sequential version.