JS采用双精度浮点数(Double-precision floating-point format, Binary64)表示数值(Number),关于计算机中浮点数的表示,可参考这里。
1 数值常量的含义
- Number.MAX_VALUE
The largest number thar can be represented in JavaScript. Equal to approximately 1.79E+308.
二进制表示:0111 1111 1110 52个1
计算公式:
(
2
−
2
−
52
)
×
2
1023
≈
1.8
×
1
0
308
(2 - 2^{-52}) \times 2^{1023} \approx 1.8\times10^{308}
(2−2−52)×21023≈1.8×10308
// true, 约等于1.79E+308
Number.MAX_VALUE === (2 - Math.pow(2, -52)) * Math.pow(2, 1023)
- Number.MIN_VALUE
The closest number to zero that can be represented in JavaScript. Equal to approximately to 5.00E-304.
二进制表示:0000 0000 0000 48个0 0001
计算公式(指数位全0)是:
2
−
1074
≈
5
−
324
2^{-1074} \approx 5^{-324}
2−1074≈5−324
// true, 约等于5E-304
Number.MIN_VALUE === Math.pow(2, -52) * Math.pow(2, -1022)
- Number.MAX_SAFE_INTEGER Number.MIN_SAFE_INTEGER
The value of the largest(smallest) integer n such that n and n+1 are both exactly representable as a Number value. Equal to 9007199254740991(-9007199254740991).
二进制表示:0100 0011 0011 52个1
计算公式:
2
−
53
−
1
=
9007199254740991
2^{-53} - 1 = 9007199254740991
2−53−1=9007199254740991
// Number.MIN_SAFE_INTEGER is the opposite of Number.MAX_SAFE_INTEGER
// true, 9007199254740991
Number.MAX_SAFE_INTEGER === Math.pow(2, 53) - 1
- Number.EPSILON
The difference of 1 and the smallest value of number greater than 1 that is representable as Number value. Equal to approximately 2.22E-16.
二进制表示:0100 0011 0011 48个0 0001
计算公式:
2
−
52
≈
2.22
×
1
0
−
16
2^{-52} \approx 2.22 \times 10^{-16}
2−52≈2.22×10−16
// true, 约等于2.22E-16
Number.EPSILON === Math.pow(2, -52)
- Number.NaN
A value that is not a number. In equality comparations, NaN dose not equal to any value, including itself. To test whether a value is equalvalant to NaN, use isNaN function.
二进制表示:0111 1111 1111 52个不全为0
// isNaN先将参数toNumber, Number.isNaN要求参数必须是Number
isNaN === Number.isNaN() // false
isNaN('a') // true
isNaN('NaN') // true
Number.isNaN('a') // false
Number.isNaN('NaN') // false
- Number.POSITIVE_INFINITY Number.NEGETIVE_INFINITY
A value greater than the largest number that can be represented in JavaScript. JavaScript displays it as infinity.
二进制表示:0111 1111 1111 52个不全为0
// isFinite先将参数toNumber, Number.isFinite要求参数必须是Number
isFinite === Number.isFinite // false
isFinite('2') // true
Number.isFinite('2') // false
2 数值运算
- 0.1和0.2的二进制表示
// 0.0001100110011001100110011001100110011001100110011001101
// JS采用Binary64,已有精度损失,是最接近0.1的number
0.1.toString(2)
// 十进制:7205759403792794, 实际的53有效数字
let a = 0b11001100110011001100110011001100110011001100110011010
// 十进制:7205759403792793
let b = 0b11001100110011001100110011001100110011001100110011001
// JS:0.1, 计算器:0.10000000000000000555111512312578
let x = a * Math.pow(2, -56)
// JS:0.09999999999999999, 计算器:0.09999999999999999167332731531133
let y = b * Math.pow(2, -56)
// x比y更接近于0.1
// 0.001100110011001100110011001100110011001100110011001101
// 已有精度损失,是最接近0.2的number, 0.2 = 0.1 * 2
0.2.toString(2)
// 52位有效数字和0.1相同, Binary64中的52位有效数字也和0.1相同
// JS:0.2, 计算器:0.20000000000000001110223024625157
let x = a * Math.pow(2, -55)
// JS:0.19999999999999998, 计算器:0.19999999999999998334665463062265
let y = b * Math.pow(2, -55)
// x比y更接近于0.2
- 0.1 + 0.2 = ?
// 0.30000000000000004
0.1 + 0.2
// JS: 0.30000000000000004
// 计算器:0.30000000000000001665334536937735
0.10000000000000000555111512312578 + 0.20000000000000001110223024625157
// 0.0100110011001100110011001100110011001100110011001101
0.30000000000000004.toString(2)
// 十进制:5404319552844596, a = b + 1
let a = 10011001100110011001100110011001100110011001100110100
// 十进制:5404319552844595
let b = 10011001100110011001100110011001100110011001100110011
// JS:0.2, 计算器:0.30000000000000004440892098500626
let x = a * Math.pow(2, -55)
// JS:0.19999999999999998, 计算器:0.29999999999999998889776975374843
let y = b * Math.pow(2, -55)
let z = 0.30000000000000001665334536937735
x === z // true
y === z // false
// x比y更接近于z
x - z = 0.00000000000000002775557561562891 = 2.775557561562891E-17
z - y = 0.00000000000000002775557561562892 = 2.775557561562892E-17
3 数值的Binary64表示
- 整数的Binary64表示,MAX_SAFE_INTEGER的由来
0 ~ MAX_SAFE_INTEGER
间的整数,正数的符号位是0
关键是确定指数位E
和有效数字S
,指数位有1023
的偏移,011 1111 1111
例子1:S = 0b1.001,E = 0b1000 = 2 3 2^3 23,value = 9
例子2:S = 1.11…还有50个1
,E = 100…还有50个0
=
2
52
2^{52}
252
v
a
l
u
e
=
(
2
−
2
−
52
)
×
2
52
=
2
53
−
1
value = (2 - 2^{-52}) \times 2^{52} = 2^{53} - 1
value=(2−2−52)×252=253−1
小数点后的有效数字最多有52位,而E的范围足够大,能够使小数变成整数。
最后1位有效数字的权重最小,为 2 E − 52 2^{E - 52} 2E−52,E从小变大,权重也在增大,能表达的数值间隔也在增大。
- 0.1的Binary64表示
// 0.0001100110011001100110011001100110011001100110011001101
0.1.toString(2)
小数点后55
位,有效数字52
位,有效数字前面去掉1
,后面补上0
Binary64中52位有效数字是:1001100110011001100110011001100110011001100110011010
符号位和指数位: 0011 1111 1011 52位有效数字