Concurrency‎ > ‎


Generators are functions that return the next value in a sequence each time the function is called:

generateInteger() => 0
generateInteger() => 1
generateInteger() => 2

This pattern can be used to implement iterators and introduce parallelism into loops.  Generators in Go are implemented with goroutines, though in other languages coroutines are often used.


If the generator's task is computationally expensive, the generator pattern can allow a consumer to run in parallel with a generator as it produces the next value to consume.  For example, the goroutine behind the "produce" generator below can execute in parallel with "consume".

for {

In some cases, the generator itself can be parallelized.  When a generator's task is computationally expensive and can be generated in any order (constrast with iterators), then the generator can be parallelized internally:

func generateRandomNumbers (n int) {
    ch := make (chan float)
    sem := make (semaphore, n)

    for i := 0; i < n; i++ {

        go func () {
            ch <- rand.Float()
        } ()


    // launch extra goroutine to eventually close ch
    go func () {
    return ch


The following for-loop will print 100 random numbers.  The random numbers are generated in parallel and arrive in a random order.  Since the order doesn't matter, this prevents us from blocking unnecessarily.

for x := range generateRandomNumbers(100) {

Note that generating random numbers in this way isn't very practical, since the parallelizing random number generator probably isn't worth the overhead of spawning so many goroutines.