What is concurrency?

Concurrent applications have multiple computations executing during overlapping periods of time. Respectively sequential programs in which no computations can be executed in overlapping periods of time.

Getting started with Context package

The Context package is responsible for signal cancelation and operation deadlines for processes and server requests.

The package has an context.Context interface:

type Context interface {
	Deadline() (deadline time.Time, ok bool)
	Done() <-chan struct{}
	Err() error
	Value(key interface{}) interface{}
}

The interface provides four functions to observe the context state:

  • Deadline returns the time when work done on behalf of this context should be canceled. It returns false when no deadline is set.
  • Done returns a channel that’s closed when work done on behalf of this context should be canceled.
  • Err returns a non-nil error value after Done is closed. It returns Canceled if the context was canceled or DeadlineExceeded if the context’s deadline passed.
  • Value returns the value associated with this context for key, or nil

There are two types of contexts:

  • context.TODO should be used context.TODO when it’s unclear which Context to use.
  • context.Background is typically used by the main function, initialization, and tests, and as the top-level Context for incoming requests.

Both are never canceled, have no values, and has no deadline.

In order to setup a deadline you should use one of the following constructors:

  • context.WithDeadline returns a copy of the parent context with the deadline adjusted to be no later than a specified time.Time. The returned context’s Done channel is closed when the deadline expires, when the returned cancel function is called, or when the parent context’s Done channel is closed, whichever happens first.
  • context.WithTimeout just calls context.WithDeadline for particular time.Duration

If you want to have a context that could be canceled only, you should use context.WithCancel function. Canceling this context releases resources associated with it, so code should call cancel as soon as the operations running in this Context complete.

Context package in practice

Lets have an application that process a payment transactions like that:

type Payment struct {
	Payee  string
	Amount float64
}

The program is asking the user to [C]onfirm or [A]bort his payment transaction within a 1 minute. If he does not anything, it will be terminated automacitally.

The ProcessPayment function is started as go routine that is waiting for user input.

go ProcessPayment(ctx, &Payment{
	Payee:  "John Doe",
	Amount: 128.54})

The function is observing the context state to terminate, cancel or proceed the payment:

func ProcessPayment(ctx context.Context, payment *Payment) {
	confirmed := ctx.Value("confirmed").(chan struct{})

	for {
		select {
		case <-confirmed:
			fmt.Printf("Your payment of %f GBP has been completed succefully.\n", payment.Amount)
			return
		case <-ctx.Done():
			if ctx.Err() == context.Canceled {
				fmt.Printf("Your payment transaction is canceled. The amount of %f GBP has been refunded.\n", payment.Amount)
				return
			} else if ctx.Err() == context.DeadlineExceeded {
				fmt.Println("Your payment transaction expired. You can complete it later.")
				os.Exit(0)
			}
		default:
			time.Sleep(1 * time.Second)
		}
	}
}

The confirmation channel is used to notify the function that the payment should be processed. If the Done channel returns a value before that the operation is aborted due to canceleation or exceeded deadline.

The ctx variable is a background context that has a deadline of 1 minute:

var (
	ctx    context.Context
	cancel context.CancelFunc
)

confirmed := make(chan struct{})
ctx = context.WithValue(context.Background(), "confirmed", confirmed)
ctx, cancel = context.WithTimeout(ctx, 1*time.Minute)

The full implementation of the example could be downloaded from here.

Usually the incomming request should create a context.Context object that underlying algorithm comply with.

  • Do not store Contexts inside a struct type; instead, pass a Context explicitly to each function that needs it.
  • Do not pass a nil Context, even if a function permits it. Pass context.TODO if you are unsure about which Context to use.
  • Do not use context.WithValue for passing optional parameters to functions. Use it for request data only.

Working with HTTP Request

On top of Context API there is a ctxhttp package that provides helper functions for performing context-aware HTTP requests. All of them are calling internally the Do function that is performing an http request that could be canceled or expired via the provided context.

func Do(ctx context.Context, client *http.Client, req *http.Request)
func Get(ctx context.Context, client *http.Client, url string)
func Head(ctx context.Context, client *http.Client, url string)
func PostForm(ctx context.Context, client *http.Client, url string, data url.Values)