为什么在范围循环中延迟以相反的顺序调用？

Question

I've run across an answer on a question about "reverse range" and I was going to down-vote it as it seemed ridiculously wrong but checked and it actually works(!):

https://play.golang.org/p/4K2fDlSoCm

package main

import (
    "fmt"
)

func main() {
    s := []int{1, 2, 3, 4, 5}
    for i, _ := range s {
        defer fmt.Println(s[i])
    }
}

And the output is:

5
4
3
2
1
Program exited.

Any ideas why it works this way? Am I right it's not guaranteed to be executed exactly in the reverse order? Also I don't think it's a good way to write programs but curious to find why are we getting this result.

Answer 1

defer is a LIFO, or a stack - it is guaranteed to executed in reverse order. It gets the first defer and puts it on some internal stack (probably, I don't know the gory details), and then puts the next defer on top of that one, and then when it hits the end of the function, it unwinds, starting at the top. It seems contrived in a for-loop (I know this is Go's example, not yours), but in other cases where one function depends upon the clean-up of some other function, it makes more sense why it SHOULD be, and hence IS, guaranteed to be reverse order of execution.

Here's a different example, all pseudo-code, but hopefully the point is clear.

open stream1
defer close stream1
defer write stream1 "stream2 better be closed, or we are in trouble..."
open stream2
defer close stream2
defer stream2 "this is the last you'll ever hear from stream2"

connect stream2 to stream1

write stream2 "hey, we are in stream2, this feeds into stream1"

Should print something like:

"hey, we are in stream2, this feeds into stream1"
"this is the last you'll ever hear from stream2"
"stream2 better be closed, or we are in trouble..."

If you didn't have guarantees about reverse ordering, you couldn't be sure stream1 was still open during your defer stream2 write.