Today, look at the distribution of red envelopes, red envelopes micro-channel allocation algorithm reference circulating a few years ago, and today a version with Golang achieve, and verify the test results.
Random algorithm micro-letter envelopes is how to achieve? https://www.zhihu.com/question/22625187
Red core algorithm
分配:红包里的金额怎么算?为什么出现各个红包金额相差很大?
答:随机,额度在0.01和(剩余平均值*2)之间
Each split red, the amount ranges between 0.01 [2] * average residual, it is a wonderful design.
Such as hair 100, a total of 10 red hair, then the average value of $ 10, the first split out of the red line fluctuated between 0.01 yuan to 20 members, you can not ensure that a person picking up the whole red case, since the maximum residual is 2 times the average.
0.1 yuan such as hair, a total of 10 red hair, each 0.01 yuan, which would not have a random algorithm, direct it evenly.
No bb, show your code!
Design red envelope structure
//reward.go
//红包
type Reward struct {
Count int //个数
Money int //总金额(分)
RemainCount int //剩余个数
RemainMoney int //剩余金额(分)
BestMoney int //手气最佳金额
BestMoneyIndex int //手气最佳序号
MoneyList []int //拆分列表
}
- I am here to do with the amount calculated int integer, floating-point precision to avoid problems when displayed in addition to 100, the unit is the yuan.
The core algorithm randomly assigned red envelopes
//reward.go
// 抢红包
func GrabReward(reward *Reward) int {
if reward.RemainCount <= 0 {
panic("RemainCount <= 0")
}
//最后一个
if reward.RemainCount - 1 == 0 {
money := reward.RemainMoney
reward.RemainCount = 0
reward.RemainMoney = 0
return money
}`
//是否可以直接0.01
if (reward.RemainMoney / reward.RemainCount) == 1 {
money := 1
reward.RemainMoney -= money
reward.RemainCount--
return money
}
//红包算法参考 https://www.zhihu.com/question/22625187
//最大可领金额 = 剩余金额的平均值x2 = (剩余金额 / 剩余数量) * 2
//领取金额范围 = 0.01 ~ 最大可领金额
maxMoney := int(reward.RemainMoney / reward.RemainCount) * 2
rand.Seed(time.Now().UnixNano())
money := rand.Intn(maxMoney)
for money == 0 {
//防止零
money = rand.Intn(maxMoney)
}
reward.RemainMoney -= money
//防止剩余金额负数
if reward.RemainMoney < 0 {
money += reward.RemainMoney
reward.RemainMoney = 0
reward.RemainCount = 0
} else {
reward.RemainCount--
}
return money
}
After the allocation algorithm is complete, verify, verification unit testing approach
//reward_test.go
func TestGrabReward2(t *testing.T) {
chanReward := make(chan Reward)
rand.Seed(time.Now().UnixNano())
go func(){
//随机生成1000个红包
for i:=0; i < 1000; i++ {
//随机红包个数 1~50
count := rand.Intn(50) + 1
//随机红包总金额 1~100元
money := rand.Intn(10000) + 100
avg := money / count
for avg == 0 {
//保证金额足够分配
count = rand.Intn(50) + 1
money = rand.Intn(10000) + 100
avg = money / count
}
reward := Reward{Count: count, Money: money,
RemainCount: count, RemainMoney: money}
chanReward <- reward
}
close(chanReward)
}()
//打印拆包列表,带手气最佳
for reward := range chanReward {
for i := 0; reward.RemainCount > 0; i++ {
money := GrabReward(&reward)
if money > reward.BestMoney {
reward.BestMoneyIndex, reward.BestMoney = i, money
}
reward.MoneyList = append(reward.MoneyList, money)
}
t.Logf("总个数:%d, 总金额:%.2f", reward.Count, float32(reward.Money)/100)
for i := range reward.MoneyList {
money := reward.MoneyList[i]
isBest := ""
if reward.BestMoneyIndex == i {
isBest = " ** 手气最佳"
}
t.Logf("money_%d : (%.2f)%s\n", i+1, float32(money)/100, isBest)
}
t.Log("-------")
}
}
operation result
reward_test.go:106: 总个数:7, 总金额:86.59
reward_test.go:113: money_1 : (16.29)
reward_test.go:113: money_2 : (4.93)
reward_test.go:113: money_3 : (22.89) ** 手气最佳
reward_test.go:113: money_4 : (3.17)
reward_test.go:113: money_5 : (20.51)
reward_test.go:113: money_6 : (0.12)
reward_test.go:113: money_7 : (18.68)
reward_test.go:115: -------
reward_test.go:106: 总个数:10, 总金额:53.79
reward_test.go:113: money_1 : (3.56)
reward_test.go:113: money_2 : (6.39)
reward_test.go:113: money_3 : (0.36)
reward_test.go:113: money_4 : (2.60)
reward_test.go:113: money_5 : (10.11)
reward_test.go:113: money_6 : (5.76)
reward_test.go:113: money_7 : (2.84)
reward_test.go:113: money_8 : (14.04) ** 手气最佳
reward_test.go:113: money_9 : (1.95)
reward_test.go:113: money_10 : (6.18)
reward_test.go:115: -------
Performance Testing
//性能测试
func BenchmarkGrabReward(b *testing.B) {
chanReward := make(chan *Reward, b.N)
rand.Seed(time.Now().UnixNano())
go func(){
//随机生成红包
for i:=0; i < b.N; i++ {
//随机红包个数 1~50
count := rand.Intn(50) + 1
//随机红包总金额 1~100元
money := rand.Intn(10000) + 100
avg := money / count
for avg == 0 {
//保证金额足够分配
count = rand.Intn(50) + 1
money = rand.Intn(10000) + 100
avg = money / count
}
reward := Reward{Count: count, Money: money,
RemainCount: count, RemainMoney: money}
chanReward <- &reward
}
close(chanReward)
}()
//打印拆包列表,带手气最佳
for reward := range chanReward {
for i := 0; reward.RemainCount > 0; i++ {
money := GrabReward(reward)
if money > reward.BestMoney {
reward.BestMoneyIndex, reward.BestMoney = i, money
}
reward.MoneyList = append(reward.MoneyList, money)
}
_ = fmt.Sprintf("总个数:%d, 总金额:%.2f", reward.Count, float32(reward.Money)/100)
for i := range reward.MoneyList {
money := reward.MoneyList[i]
isBest := ""
if reward.BestMoneyIndex == i {
isBest = " ** 手气最佳"
}
_ = fmt.Sprintf("money_%d : (%.2f)%s\n", i+1, float32(money)/100, isBest)
}
}
}
Performance Test Results
BenchmarkGrabReward-8 4461 244842 ns/op
//4核8线的CPU运运行4461次,平均每次244842纳秒=0.244842毫秒
Performance can be said to be very good, because this test is a pure in-memory computing, there is no network IO, no storage disk write purely for verification algorithm, so performance is very high.
carry out!