https://blog.golang.org/context

Introduction

In Go servers, each incoming request is handled in its own goroutine. Request handlers often start additional goroutines to access backends such as databases and RPC services. The set of goroutines working on a request typically needs access to request-specific values such as the identity of the end user, authorization tokens, and the request's deadline. When a request is canceled or times out, all the goroutines working on that request should exit quickly so the system can reclaim any resources they are using.

At Google, we developed a context package that makes it easy to pass request-scoped values, cancelation signals, and deadlines across API boundaries to all the goroutines involved in handling a request. The package is publicly available as context. This article describes how to use the package and provides a complete working example.

Context

 The core of the context package is the Context type:

// A Context carries a deadline, cancelation signal, and request-scoped values
// across API boundaries. Its methods are safe for simultaneous use by multiple
// goroutines.
type Context interface {
    // Done returns a channel that is closed when this Context is canceled
    // or times out.
    Done() <-chan struct{}

    // Err indicates why this context was canceled, after the Done channel
    // is closed.
    Err() error

    // Deadline returns the time when this Context will be canceled, if any.
    Deadline() (deadline time.Time, ok bool)

    // Value returns the value associated with key or nil if none.
    Value(key interface{}) interface{}
}

(This description is condensed; the godoc is authoritative.)

The Done method returns a channel that acts as a cancelation signal to functions running on behalf of the Context: when the channel is closed, the functions should abandon their work and return. The Err method returns an error indicating why the Context was canceled. The Pipelines and Cancelation article discusses the Done channel idiom in more detail.

A Context does not have a Cancel method for the same reason the Done channel is receive-only: the function receiving a cancelation signal is usually not the one that sends the signal. In particular, when a parent operation starts goroutines for sub-operations, those sub-operations should not be able to cancel the parent. Instead, the WithCancel function (described below) provides a way to cancel a new Context value.

A Context is safe for simultaneous use by multiple goroutines. Code can pass a single Context to any number of goroutines and cancel that Context to signal all of them.

The Deadline method allows functions to determine whether they should start work at all; if too little time is left, it may not be worthwhile. Code may also use a deadline to set timeouts for I/O operations.

Value allows a Context to carry request-scoped data. That data must be safe for simultaneous use by multiple goroutines.

Derived contexts

The context package provides functions to derive new Context values from existing ones. These values form a tree: when a Context is canceled, all Contexts derived from it are also canceled.

Background is the root of any Context tree; it is never canceled:

// Background returns an empty Context. It is never canceled, has no deadline,
// and has no values. Background is typically used in main, init, and tests,
// and as the top-level Context for incoming requests.
func Background() Context

WithCancel and WithTimeout return derived Context values that can be canceled sooner than the parent Context. The Context associated with an incoming request is typically canceled when the request handler returns. WithCancel is also useful for canceling redundant requests when using multiple replicas. WithTimeout is useful for setting a deadline on requests to backend servers:

// WithCancel returns a copy of parent whose Done channel is closed as soon as
// parent.Done is closed or cancel is called.
func WithCancel(parent Context) (ctx Context, cancel CancelFunc)

// A CancelFunc cancels a Context.
type CancelFunc func()

// WithTimeout returns a copy of parent whose Done channel is closed as soon as
// parent.Done is closed, cancel is called, or timeout elapses. The new
// Context's Deadline is the sooner of now+timeout and the parent's deadline, if
// any. If the timer is still running, the cancel function releases its
// resources.
func WithTimeout(parent Context, timeout time.Duration) (Context, CancelFunc)

WithValue provides a way to associate request-scoped values with a Context:

// WithValue returns a copy of parent whose Value method returns val for key.
func WithValue(parent Context, key interface{}, val interface{}) Context

The best way to see how to use the context package is through a worked example.

Example: Google Web Search

Our example is an HTTP server that handles URLs like /search?q=golang&timeout=1s by forwarding the query "golang" to the Google Web Search API and rendering the results. The timeout parameter tells the server to cancel the request after that duration elapses.

The code is split across three packages:

• server provides the main function and the handler for /search.
• userip provides functions for extracting a user IP address from a request and associating it with a Context.
• google provides the Search function for sending a query to Google.

The server program

The server program handles requests like /search?q=golang by serving the first few Google search results for golang. It registers handleSearch to handle the /search endpoint. The handler creates an initial Context called ctx and arranges for it to be canceled when the handler returns. If the request includes the timeout URL parameter, the Context is canceled automatically when the timeout elapses:

func handleSearch(w http.ResponseWriter, req *http.Request) {
    // ctx is the Context for this handler. Calling cancel closes the
    // ctx.Done channel, which is the cancellation signal for requests
    // started by this handler.
    var (
        ctx    context.Context
        cancel context.CancelFunc
    )
    timeout, err := time.ParseDuration(req.FormValue("timeout"))
    if err == nil {
        // The request has a timeout, so create a context that is
        // canceled automatically when the timeout expires.
        ctx, cancel = context.WithTimeout(context.Background(), timeout)
    } else {
        ctx, cancel = context.WithCancel(context.Background())
    }
    defer cancel() // Cancel ctx as soon as handleSearch returns.
}

    // Check the search query.
    query := req.FormValue("q")
    if query == "" {
        http.Error(w, "no query", http.StatusBadRequest)
        return
    }

    // Store the user IP in ctx for use by code in other packages.
    userIP, err := userip.FromRequest(req)
    if err != nil {
        http.Error(w, err.Error(), http.StatusBadRequest)
        return
    }
    ctx = userip.NewContext(ctx, userIP)

The handler calls google.Search with ctx and the query:

    // Run the Google search and print the results.
    start := time.Now()
    results, err := google.Search(ctx, query)
    elapsed := time.Since(start)

If the search succeeds, the handler renders the results:

    if err := resultsTemplate.Execute(w, struct {
        Results          google.Results
        Timeout, Elapsed time.Duration
    }{
        Results: results,
        Timeout: timeout,
        Elapsed: elapsed,
    }); err != nil {
        log.Print(err)
        return
    }

Package userip

The userip package provides functions for extracting a user IP address from a request and associating it with a Context. A Context provides a key-value mapping, where the keys and values are both of type interface{}. Key types must support equality, and values must be safe for simultaneous use by multiple goroutines. Packages like userip hide the details of this mapping and provide strongly-typed access to a specific Context value.

To avoid key collisions, userip defines an unexported type key and uses a value of this type as the context key:

// The key type is unexported to prevent collisions with context keys defined in
// other packages.
type key int

// userIPkey is the context key for the user IP address.  Its value of zero is
// arbitrary.  If this package defined other context keys, they would have
// different integer values.
const userIPKey key = 0

FromRequest extracts a userIP value from an http.Request:

func FromRequest(req *http.Request) (net.IP, error) {
    ip, _, err := net.SplitHostPort(req.RemoteAddr)
    if err != nil {
        return nil, fmt.Errorf("userip: %q is not IP:port", req.RemoteAddr)
    }
}

NewContext returns a new Context that carries a provided userIP value:

func NewContext(ctx context.Context, userIP net.IP) context.Context {
    return context.WithValue(ctx, userIPKey, userIP)
}

FromContext extracts a userIP from a Context:

func FromContext(ctx context.Context) (net.IP, bool) {
    // ctx.Value returns nil if ctx has no value for the key;
    // the net.IP type assertion returns ok=false for nil.
    userIP, ok := ctx.Value(userIPKey).(net.IP)
    return userIP, ok
}

Package google

The google.Search function makes an HTTP request to the Google Web Search API and parses the JSON-encoded result. It accepts a Context parameter ctx and returns immediately if ctx.Done is closed while the request is in flight.

The Google Web Search API request includes the search query and the user IP as query parameters:

func Search(ctx context.Context, query string) (Results, error) {
    // Prepare the Google Search API request.
    req, err := http.NewRequest("GET", "https://ajax.googleapis.com/ajax/services/search/web?v=1.0", nil)
    if err != nil {
        return nil, err
    }
    q := req.URL.Query()
    q.Set("q", query)

    // If ctx is carrying the user IP address, forward it to the server.
    // Google APIs use the user IP to distinguish server-initiated requests
    // from end-user requests.
    if userIP, ok := userip.FromContext(ctx); ok {
        q.Set("userip", userIP.String())
    }
    req.URL.RawQuery = q.Encode()
}

Search uses a helper function, httpDo, to issue the HTTP request and cancel it if ctx.Done is closed while the request or response is being processed. Search passes a closure to httpDo handle the HTTP response:

    var results Results
    err = httpDo(ctx, req, func(resp *http.Response, err error) error {
        if err != nil {
            return err
        }
        defer resp.Body.Close()

        // Parse the JSON search result.
        // https://developers.google.com/web-search/docs/#fonje
        var data struct {
            ResponseData struct {
                Results []struct {
                    TitleNoFormatting string
                    URL               string
                }
            }
        }
        if err := json.NewDecoder(resp.Body).Decode(&data); err != nil {
            return err
        }
        for _, res := range data.ResponseData.Results {
            results = append(results, Result{Title: res.TitleNoFormatting, URL: res.URL})
        }
        return nil
    })
    // httpDo waits for the closure we provided to return, so it's safe to
    // read results here.
    return results, err

The httpDo function runs the HTTP request and processes its response in a new goroutine. It cancels the request if ctx.Done is closed before the goroutine exits:

func httpDo(ctx context.Context, req *http.Request, f func(*http.Response, error) error) error {
    // Run the HTTP request in a goroutine and pass the response to f.
    tr := &http.Transport{}
    client := &http.Client{Transport: tr}
    c := make(chan error, 1)
    go func() { c <- f(client.Do(req)) }()
    select {
    case <-ctx.Done():
        tr.CancelRequest(req)
        <-c // Wait for f to return.
        return ctx.Err()
    case err := <-c:
        return err
    }
}

Adapting code for Contexts

Many server frameworks provide packages and types for carrying request-scoped values. We can define new implementations of the Context interface to bridge between code using existing frameworks and code that expects a Context parameter.

For example, Gorilla's github.com/gorilla/context package allows handlers to associate data with incoming requests by providing a mapping from HTTP requests to key-value pairs. In gorilla.go, we provide a Context implementation whose Value method returns the values associated with a specific HTTP request in the Gorilla package.

Other packages have provided cancelation support similar to Context. For example, Tomb provides a Kill method that signals cancelation by closing a Dying channel. Tomb also provides methods to wait for those goroutines to exit, similar to sync.WaitGroup. In tomb.go, we provide a Context implementation that is canceled when either its parent Context is canceled or a provided Tomb is killed.

Conclusion

At Google, we require that Go programmers pass a Context parameter as the first argument to every function on the call path between incoming and outgoing requests. This allows Go code developed by many different teams to interoperate well. It provides simple control over timeouts and cancelation and ensures that critical values like security credentials transit Go programs properly.

Server frameworks that want to build on Context should provide implementations of Context to bridge between their packages and those that expect a Context parameter. Their client libraries would then accept a Context from the calling code. By establishing a common interface for request-scoped data and cancelation, Context makes it easier for package developers to share code for creating scalable services.

By Sameer Ajmani

https://rakyll.org/leakingctx/

The context package makes it possible to manage a chain of calls within the same call path by signaling context’s Done channel.

In this article, we will examine how to use the context package to avoid leaking goroutines.

Assume, you have a function that starts a goroutine internally. Once this function is called, the caller may not be able to terminate the goroutine started by the function.

// gen is a broken generator that will leak a goroutine.
func gen() <-chan int {
    ch := make(chan int)
    go func() {
        var n int
        for {
            ch <- n
            n++
        }
    }()
    return ch
}

The generator above starts a goroutine with an infinite loop, but the caller consumes the values until n is equal to 5.

// The call site of gen doesn't have a 
for n := range gen() {
    fmt.Println(n)
    if n == 5 {
        break
    }
}

Once the caller is done with the generator (when it breaks the loop), the goroutine will run forever executing the infinite loop. Our code will leak a goroutine.

We can avoid the problem by signaling the internal goroutine with a stop channel but there is a better solution: cancellable contexts. The generator can select on a context’s Done channel and once the context is done, the internal goroutine can be cancelled.

// gen is a generator that can be cancellable by cancelling the ctx.
func gen(ctx context.Context) <-chan int {
    ch := make(chan int)
    go func() {
        var n int
        for {
            select {
            case <-ctx.Done():
                return // avoid leaking of this goroutine when ctx is done.
            case ch <- n:
                n++
            }
        }
    }()
    return ch
}

Now, the caller can signal the generator when it is done consuming. Once cancel function is called, the internal goroutine will be returned.

ctx, cancel := context.WithCancel(context.Background())
defer cancel() // make sure all paths cancel the context to avoid context leak

for n := range gen(ctx) {
    fmt.Println(n)
    if n == 5 {
        cancel()
        break
    }
}

// ...

The full program is available as a gist.

Context and Cancellation of goroutines

http://dahernan.github.io/2015/02/04/context-and-cancellation-of-goroutines/

Yesterday I went to the event London Go Gathering, where all the talks had a great level, but particulary Peter Bourgon gave me idea to write about the excelent package context.

Context is used to pass request scoped variables, but in this case I’m only going to focus in cancelation signals.

Lets say that I have a program that execute a long running function, in this case work and we run it in a separate go routine.

package main

import (
    "fmt"
    "sync"
    "time"

)

var (
    wg sync.WaitGroup
)

func work() error {
    defer wg.Done()

    for i := 0; i < 1000; i++ {
        select {
        case <-time.After(2 * time.Second):
            fmt.Println("Doing some work ", i)
        }
    }
    return nil
}


func main() {
    fmt.Println("Hey, I'm going to do some work")

    wg.Add(1)
    go work()
    wg.Wait()

    fmt.Println("Finished. I'm going home")
}

$ go run work.go
Hey, I'm going to do some work
Doing some work  0
Doing some work  1
Doing some work  2
Doing some work  3
...
Doing some work  999
Finished. I'm going home

Now imagine that we have to call that work function from a user interaction or a http request, we probably don’t want to wait forever for that goroutine to finish, so a common pattern is to set a timeout, using a buffered channel, like this:

package main

import (
    "fmt"
    "log"
    "time"
)

func work() error {
    for i := 0; i < 1000; i++ {
        select {
        case <-time.After(2 * time.Second):
            fmt.Println("Doing some work ", i)
        }
    }
    return nil
}

func main() {
    fmt.Println("Hey, I'm going to do some work")

    ch := make(chan error, 1)
    go func() {
        ch <- work()
    }()

    select {
    case err := <-ch:
        if err != nil {
            log.Fatal("Something went wrong :(", err)
        }
    case <-time.After(4 * time.Second):
        fmt.Println("Life is to short to wait that long")
    }

    fmt.Println("Finished. I'm going home")
}

$ go run work.go
Hey, I'm going to do some work
Doing some work  0
Doing some work  1
Life is to short to wait that long
Finished. I'm going home

Now, is a little bit better because, the main execution doesn’t have to wait for work if it’s timing out.

But it has a problem, if my program is still running like for example a web server, even if I don’t wait for the function work to finish, the goroutine it would be running and consuming resources. So I need a way to cancel that goroutine.

For cancelation of the goroutine we can use the context package. We have to change the function to accept an argument of type context.Context, by convention it’s usuallly the first argument.

package main

import (
    "fmt"
    "sync"
    "time"

    "golang.org/x/net/context"
)

var (
    wg sync.WaitGroup
)

func work(ctx context.Context) error {
    defer wg.Done()

    for i := 0; i < 1000; i++ {
        select {
        case <-time.After(2 * time.Second):
            fmt.Println("Doing some work ", i)

        // we received the signal of cancelation in this channel    
        case <-ctx.Done():
            fmt.Println("Cancel the context ", i)
            return ctx.Err()
        }
    }
    return nil
}

func main() {   
    ctx, cancel := context.WithTimeout(context.Background(), 4*time.Second)
    defer cancel()

    fmt.Println("Hey, I'm going to do some work")

    wg.Add(1)
    go work(ctx)
    wg.Wait()

    fmt.Println("Finished. I'm going home")
}

$ go run work.go
Hey, I'm going to do some work
Doing some work  0
Cancel the context  1
Finished. I'm going home

This is pretty good!, apart that the code looks more simple to manage the timeout, now we are making sure that the function work doesn’t waste any resource.

These examples are good to learn the basics, but let’s try to make it more real. Now the work function is going to do an http request to a server and the server is going to be this other program:

package main

// Lazy and Very Random Server 
import (
    "fmt"
    "math/rand"
    "net/http"
    "time"
)

func main() {
    http.HandleFunc("/", LazyServer)
    http.ListenAndServe(":1111", nil)
}

// sometimes really fast server, sometimes really slow server
func LazyServer(w http.ResponseWriter, req *http.Request) {
    headOrTails := rand.Intn(2)

    if headOrTails == 0 {
        time.Sleep(6 * time.Second)
        fmt.Fprintf(w, "Go! slow %v", headOrTails)
        fmt.Printf("Go! slow %v", headOrTails)
        return
    }

    fmt.Fprintf(w, "Go! quick %v", headOrTails)
    fmt.Printf("Go! quick %v", headOrTails)
    return
}

Randomly is going to be very quick or very slow, we can check that with curl

$ curl http://localhost:1111/
Go! quick 1
$ curl http://localhost:1111/
Go! quick 1
$ curl http://localhost:1111/
*some seconds later*
Go! slow 0

So we are going to make an http request to this server, in a goroutine, but if the server is slow we are going to Cancel the request and return quickly, so we can manage the cancellation and free the connection.

package main

import (
    "fmt"
    "io/ioutil"
    "net/http"
    "sync"
    "time"

    "golang.org/x/net/context"
)

var (
    wg sync.WaitGroup
)

// main is not changed
func main() {
    ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
    defer cancel()

    fmt.Println("Hey, I'm going to do some work")

    wg.Add(1)
    go work(ctx)
    wg.Wait()

    fmt.Println("Finished. I'm going home")

}

func work(ctx context.Context) error {
    defer wg.Done()

    tr := &http.Transport{}
    client := &http.Client{Transport: tr}

    // anonymous struct to pack and unpack data in the channel
    c := make(chan struct {
        r   *http.Response
        err error
    }, 1)

    req, _ := http.NewRequest("GET", "http://localhost:1111", nil)
    go func() {
        resp, err := client.Do(req)
        fmt.Println("Doing http request is a hard job")
        pack := struct {
            r   *http.Response
            err error
        }{resp, err}
        c <- pack
    }()

    select {
    case <-ctx.Done():
        tr.CancelRequest(req)
        <-c // Wait for client.Do
        fmt.Println("Cancel the context")
        return ctx.Err()
    case ok := <-c:
        err := ok.err
        resp := ok.r
        if err != nil {
            fmt.Println("Error ", err)
            return err
        }

        defer resp.Body.Close()
        out, _ := ioutil.ReadAll(resp.Body)
        fmt.Printf("Server Response: %s\n", out)

    }
    return nil
}

$ go run work.go
Hey, I'm going to do some work
Doing http request is a hard job
Server Response: Go! quick 1
Finished. I'm going home

$ go run work.go
Hey, I'm going to do some work
Doing http request is a hard job
Cancel the context
Finished. I'm going home

As you can see in the output, we avoid the slow responses from the server.

In the client the tcp connection is canceled so is not going to be busy waiting for a slow response, so we don’t waste resources.

Happy coding gophers!.