切片vs数组用于已知长度的数据

Question

When dealing with something like a sha256 hash which has a known length of 32 bytes– is it more idiomatic to work with []byte slices or [32]byte arrays?

I'm a bit new to go– but intuitively it feels like while working with a [32]byte array to store a hash is nice and explicit, it's actually quite cumbersome to manipulate and requires all sorts of strange things like copying the array into a new buffer slice to manipulate as []byte.

Answer 1

Looking at the Go standard library, it looks like arrays are often used for something like a hash of known length. e.g sha256.Sum256 returns a [sha256.Size]byte where sha256.Size is 32. Or looking at the digest struct that implements the hash.Hash interface in sha256 it contains [chunk]byte where chunk is 64, the sha256 blocksize.

I think it is typical to use arrays to avoid unnecessary memory allocation. To take advantage of algorithms that operate on slices of a variable length you can convert the array to a slice without copying with x[:].

package main

import (
    "crypto/sha256"
    "fmt"
)

type myHash struct {
    x [32]byte
}

func print(b []byte) {
    fmt.Print(b)
}

func main() {
    b := []byte("Hello world!")
    h := myHash{x: sha256.Sum256(b)}
    print(h.x[:])
}