panic and recover: only for the truly unexpected

What a panic is

A panic is Go's emergency stop. It happens when the program hits something it cannot continue from, like reading index 10 of a 3-item slice, following a nil pointer, or writing to a nil map. You can also call panic yourself. When a panic starts, the current function stops, deferred calls run (in reverse order) up through the calling functions, and then the program exits with a message and a stack trace (the list of function calls that led there).

func main() {
	defer fmt.Println("deferred: still runs")
	s := []int{1, 2, 3}
	i := 5
	fmt.Println(s[i])
	fmt.Println("never printed")
}
// deferred: still runs
// panic: runtime error: index out of range [5] with length 3
// goroutine 1 [running]: ...

Errors versus panics

The standard library follows this. regexp.MustCompile panics if the pattern is invalid, because the pattern is usually a fixed string in your code and a bad one is a bug you want to see immediately at start-up. regexp.Compile returns an error, for patterns that come from users.

var pinRe = regexp.MustCompile(`^[1-9][0-9]{5}$`) // fixed pattern: Must is fine

func userPattern(p string) (*regexp.Regexp, error) {
	return regexp.Compile(p) // user input: return the error
}

recover

recover() stops a panic, but only when called directly inside a deferred function in the goroutine that is panicking. It returns the value passed to panic (or nil if there is no panic). After recovering, the function that deferred it returns normally to its caller.

func safeDivide(a, b int) (q int, err error) {
	defer func() {
		if r := recover(); r != nil {
			err = fmt.Errorf("recovered: %v", r)
		}
	}()
	return a / b, nil
}

func main() {
	fmt.Println(safeDivide(10, 2)) // 5 <nil>
	fmt.Println(safeDivide(1, 0))  // 0 recovered: runtime error: integer divide by zero
}

Where recover is used in real code

• HTTP servers: net/http already recovers a panic in a handler, logs it and closes that connection, so one bad request does not kill the server. Many teams also add their own recovery middleware to return a clean 500 response and log details (module 8).
• Worker pools: a long-running worker recovers so one bad job is logged instead of stopping all processing.
• Libraries that use panic internally for deep recursion (like parsers) and convert it to an error at their public boundary.