问题
I'm trying to understand precisely why, when called from an external function, my time seeded random number generator returns sequences of identical numbers.
Minimal working example of issue:
package main
import (
"fmt"
"math/rand"
"time"
)
//Generates random int as function of range
func getRand(Range int) int {
s1 := rand.NewSource(time.Now().UnixNano())
r1 := rand.New(s1)
return r1.Intn(Range)
}
//Print 100 random ints between 0 and 100
func main() {
for i := 0; i < 100; i++ {
fmt.Print(getRand(100), ", ")
}
}
The output of this is
Out[1]: 40, 40, 40, 40, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 34,
34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 47,
47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47,
47,47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 99, 99, 99, 99, 99, 99, 99,
99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99, 99,
99, 99, 99, 99, 99, 99, 99,
I'd like to know why this is happening for my own education. I'm also open to suggestions for a solution.
Details: I need to call random numbers in lots of external functions of my code but, like this MWE, when seeded within a function other than main they repeatedly return the same numbers. Additionally, I need to dynamically update the range, so generating lists a priori is not an option. I would rather not have to generate the numbers in main() and pass them into each function-- ranges are calculated inside these and it would complicate things
回答1:
This is because time.Time has a granularity (which is 1 nanosecond) just like your system clock (which might even be multiple milliseconds–depends on many things), and if you call time.Now() multiple times within the greater of these granularities, chances are the returned time.Time
will be the same, meaning its Time.UnixNano() method will return you the same nanoseconds (the same number).
And if you use the same number as the seed, the random number generator is ought to return the same numbers.
You only need to seed the RNG once, on app startup, not before each use. You may use a package init()
function for that, or in the variable declaration:
var r = rand.New(rand.NewSource(time.Now().UnixNano()))
//Generates random int as function of range
func getRand(Range int) int {
return r.Intn(Range)
}
//Print 100 random ints between 0 and 100
func main() {
for i := 0; i < 100; i++ {
fmt.Print(getRand(100), ", ")
}
}
Example output (try it on the Go Playground):
0, 28, 27, 62, 63, 89, 24, 27, 88, 84, 82, 55, 49, 35, 2, 32, 84, 58, 78, 28, 26, 58, 30, 28, 74, 6, 39, 24, 40, 47, 49, 39, 61, 62, 67, 7, 94, 87, 37, 99, 90, 80, 93, 83, 27, 69, 25, 45, 99, 12, 44, 39, 34, 86, 18, 42, 76, 40, 44, 12, 70, 3, 70, 99, 57, 43, 90, 65, 97, 64, 68, 60, 65, 56, 3, 81, 54, 56, 43, 57, 92, 93, 54, 92, 9, 86, 16, 72, 29, 12, 97, 87, 55, 42, 87, 41, 94, 53, 23, 64,
One thing to note here: rand.NewSource() returns a source which is not safe for concurrent use. If you need to call getRand()
from multiple goroutines, you need to synchronize access to r
, or use a separate rand.Rand in each goroutine.
回答2:
I'm not a Go expert but I think the problem is a generic programming issue. It is related to the fact you set the seed to each call. The seed is based on time function. So what is happening is that in a very short time, you have a number of calls while the time is not changed (yet) and so you get the same value because you are setting the same seed again and again.
Try to set the seed, once only and outside the for
loop of calls.
来源:https://stackoverflow.com/questions/52276548/generating-identical-random-numbers-in-sequence-after-time-seed-running-on-my