a iteration can be saved for "func (r *Rand) Perm(n int) []int"

[yaojingguo]

In src/math/rand/rand.go, there is the following code:

func (r *Rand) Perm(n int) []int {
    m := make([]int, n)
    for i := 0; i < n; i++ {
        j := r.Intn(i + 1)
        m[i] = m[j]
        m[j] = i
    }
    return m
}

i can start from 1 instead of 0 since the iteration for i=0 always swap m[0] with m[0]. A fix is to change i := 0 to i := 1.

[randall77]

Indeed. Unfortunately, this is tricky. Perm also side-effects r, and making this change will affect the final state of r. I don't think we can do that for compatibility reasons.
You could start the loop at 1 and call r.Intn(1) (or Int63()) before the loop. Then add a big comment as to why that call is there and that we should remove it for Go 2.

Too bad Intn(1) wasn't originally coded to do the n==1 shortcut.

Care to make a patch?

[minux]

I think preserving the call to rand.Intn will negate any speed advantage of the saved iteration. Perhaps we should just add a note that we need to preserve the rand state here.

[randall77]

Yeah, it's not a big win.
Comment sounds fine. Bonus: that can go in during the freeze.

[yaojingguo]

@randall77, so the conclusion is for me to make a patch which adds a comment, right?

[randall77]

Yes, sounds good.

[minux]

Yes.

[yaojingguo]

A patch is created.

CL link: https://go-review.googlesource.com/16852

[gopherbot]

CL https://golang.org/cl/16803 mentions this issue.

[gopherbot]

CL https://golang.org/cl/16852 mentions this issue.