Why does Go panic on writing to a closed channel?

Question

Why does Go panic on writing to a closed channel?

While one can use the value, ok := <-channel idiom for reading from channels, and thus the ok result can be tested for hitting a closed channel:

// reading from closed channel

package main

import "fmt"

func main() {
    ch := make(chan int, 1)
    ch <- 2
    close(ch)

    read(ch)
    read(ch)
    read(ch)
}

func read(ch <-chan int) {
    i,ok := <- ch   
    if !ok {
        fmt.Printf("channel is closed\n")
        return
    }
    fmt.Printf("read %d from channel\n", i)
}

Output:

read 2 from channel
channel is closed
channel is closed

Run "reading from closed channel" on Playground

Writing to a possibly closed channel is more convoluted, because Go will panic if you simply try to write when the channel is closed:

//writing to closed channel

package main

import (
    "fmt"
)

func main() {
    output := make(chan int, 1) // create channel
    write(output, 2)
    close(output) // close channel
    write(output, 3)
    write(output, 4)
}

// how to write on possibly closed channel
func write(out chan int, i int) (err error) {

    defer func() {
        // recover from panic caused by writing to a closed channel
        if r := recover(); r != nil {
            err = fmt.Errorf("%v", r)
            fmt.Printf("write: error writing %d on channel: %v\n", i, err)
            return
        }

        fmt.Printf("write: wrote %d on channel\n", i)
    }()

    out <- i // write on possibly closed channel

    return err
}

Output:

write: wrote 2 on channel
write: error writing 3 on channel: send on closed channel
write: error writing 4 on channel: send on closed channel

Run "writing to closed channel" on Playground

As far as I know, there is not a simpler idiom for writing into a possibly closed channel without panicking. Why not? What is the reasoning behind such an asymmetric behavior between read and write?

Answer 1

From the Go Language Spec:

For a channel c, the built-in function close(c) records that no more values will be sent on the channel. It is an error if c is a receive-only channel. Sending to or closing a closed channel causes a run-time panic. Closing the nil channel also causes a run-time panic. After calling close, and after any previously sent values have been received, receive operations will return the zero value for the channel's type without blocking. The multi-valued receive operation returns a received value along with an indication of whether the channel is closed.

If you write to a closed channel, your program will panic. You could potentially catch this error with recover if you really want to do that, but being in a situation where you don't know whether the channel you are writing to is open is usually a sign of a bug in the program.

Some quotes:

Here is a motivation:

A channel "close" is really just a send of a special value on a channel. It is a special value that promises that no more values will be sent. Attempting to send a value on a channel after it has been closed will panic, since actually sending the value would violate the guarantee provided by close. Since a close is just a special kind of send, it is also not permitted after the channel has been closed.

Here is another:

The only use of channel close is to signal to the reader that there are no more values to come. That only makes sense when there is a single source of values, or when multiple sources coordinate. There is no reasonable program in which multiple goroutines close a channel without communicating. That would imply that multiple goroutines would know that there are no more values to send--how could they determine that if they don't communicate?

(Ian Lance Taylor)

--

Here is another:

Closing a channel releases it as a resource. It makes no more sense to close a channel multiple times than it makes to close a file descriptor multiple times, or free a block of allocated memory multiple times. Such actions imply the code is broken, which is why closing a closed channel triggers a panic.

(Rob Pike)

--

Source: Go design detail rationale question - channel close

Answer 2

Go channels are designed for 1 writer, and multiple readers. Thus the writer has to create, share, and close the channel once is done writing. In a way, the writer owns the channel. Although you could have multiple writers on a single channel (this is not advised/recommended).