为什么不能将Go slice用作Go映射中的键，就像将数组用作键一样？

Question

Why can't Go slice (which is an implementation of Go arrays) be used as keys in Go maps pretty much the same way arrays can be used as keys?

Answer 1

Here's Nigel Tao's answer from https://groups.google.com/forum/#!topic/golang-nuts/zYlx6sR4F8Y:

One reason is that arrays are value types. If a0 is an [N]int (an array) then doing

a1 := a0
a1[0] = 0

will not affect a0[0] at all.

In comparison, slices refer to an underlying array. Copying a slice value is O(1) instead of O(length). If s0 is an []int (a slice) then doing

s1 := s0
s1[0] = 0

will affect what s0[0] is.

http://play.golang.org/p/TVkntIsLo8

Map keys need some notion of equality. For arrays, this is simply element-wise equality. For slices, there is more than one way to define equality: one is element-wise equality, another is referring to the same array backing store. Furthermore, does map insertion need to make an (expensive) copy of the entire backing array? Copying would probably be less surprising, but it is inconsistent with what assignment does.

What should this code snippet print?

m := make(map[[]int]bool)
s0 := []int{6, 7, 8}
s1 := []int{6, 7, 8}
s2 := s0
m[s0] = true
s2[0] = 9
println(m[s0])
println(m[s1])
println(m[s2])

Different programmers can have different expectations. To avoid confusion, we have simply decided not to allow slices as map keys for now.