编译器/转译器不一定会始终删除解构赋值,因为所有常绿浏览器都支持自 2020 年起原生解构。根据,有一些证据表明,至少在 2018 年,V8 中通过解构赋值生成的字节码要多得多比传统的函数参数更冗长:
功能参数:
function add(number1, number2){
return number1 + number2;
}
const result = add(1,5);
输出字节码:
[generating bytecode for function: add]
Parameter count 3
Frame size 0
74 E> 0x2a2a0affd2a2 @ 0 : 91 StackCheck
96 S> 0x2a2a0affd2a3 @ 1 : 1d 02 Ldar a1
111 E> 0x2a2a0affd2a5 @ 3 : 2b 03 00 Add a0, [0]
121 S> 0x2a2a0affd2a8 @ 6 : 95 Return
Constant pool (size = 0)
Handler Table (size = 16)
解构赋值:
function add({number1, number2}){
return number1 + number2;
}
const result = add({number1: 1, number2: 5});
输出字节码:
[generating bytecode for function: add]
Parameter count 2
Frame size 40
74 E> 0x2c1d63b7d312 @ 0 : 91 StackCheck
0x2c1d63b7d313 @ 1 : 1f 02 fb Mov a0, r0
0x2c1d63b7d316 @ 4 : 1d fb Ldar r0
0x2c1d63b7d318 @ 6 : 89 06 JumpIfUndefined [6] (0x2c1d63b7d31e @ 12)
0x2c1d63b7d31a @ 8 : 1d fb Ldar r0
0x2c1d63b7d31c @ 10 : 88 10 JumpIfNotNull [16] (0x2c1d63b7d32c @ 26)
0x2c1d63b7d31e @ 12 : 03 3f LdaSmi [63]
0x2c1d63b7d320 @ 14 : 1e f8 Star r3
0x2c1d63b7d322 @ 16 : 09 00 LdaConstant [0]
0x2c1d63b7d324 @ 18 : 1e f7 Star r4
0x2c1d63b7d326 @ 20 : 53 e8 00 f8 02 CallRuntime [NewTypeError], r3-r4
76 E> 0x2c1d63b7d32b @ 25 : 93 Throw
76 S> 0x2c1d63b7d32c @ 26 : 20 fb 00 02 LdaNamedProperty r0, [0], [2]
0x2c1d63b7d330 @ 30 : 1e fa Star r1
85 S> 0x2c1d63b7d332 @ 32 : 20 fb 01 04 LdaNamedProperty r0, [1], [4]
0x2c1d63b7d336 @ 36 : 1e f9 Star r2
98 S> 0x2c1d63b7d338 @ 38 : 1d f9 Ldar r2
113 E> 0x2c1d63b7d33a @ 40 : 2b fa 06 Add r1, [6]
123 S> 0x2c1d63b7d33d @ 43 : 95 Return
Constant pool (size = 2)
Handler Table (size = 16)
字节码行数从函数参数的 4 行显着增加到解构赋值的 19 行。总之,截至 2018 年的 V8 中,解构赋值的计算效率低于传统函数参数。在内存空间利用率方面,答案有点复杂,可以参考这里。
这可能是一个过早的优化,但是在计算繁重的代码中,建议考虑不使用解构赋值。