Redis1️⃣4️⃣课 HyperLogLog

1)数据结构

基于HyperLogLog算法，极小空间完成独立数量统计

本质还是string。

127.0.0.1 6379: type hyperLogLog_key 
string

2)三个命令

pfadd key element [element ...] 向hyperloglog添加元素
pfcount key [key ...] 计算hyperloglog的独立总数
pfmerge destkey sourcekey [sourcekey ...] 合并多个hyperloglog
pfselftsest
pfdebug arg arg ...options...

127.0.0.1:6379> pfadd userIds "uuid-1" "uuid-2" "uuid-3" "uuid-4"
(integer) 1
127.0.0.1:6379> pfcount userIds
(integer) 4
127.0.0.1:6379> pfadd userIds "uuid-1" "uuid-2" "uuid-3" "uuid-90"
(integer) 1
127.0.0.1:6379> pfcount userIds
(integer) 5
127.0.0.1:6379> pfadd userIds2 "uuid-4" "uuid-5" "uuid-6" "uuid-7"
(integer) 1
127.0.0.1:6379> pfcount userIds2
(integer) 4
127.0.0.1:6379> pfcount userIds1 userIds2
(integer) 4
127.0.0.1:6379> pfmerge userIdMerge userIds1 userIds2
OK
127.0.0.1:6379> pfcount userIdMerge
(integer) 4
127.0.0.1:6379> pfmerge userIdMerge userIds userIds2
OK
127.0.0.1:6379> pfcount userIdMerge
(integer) 8
127.0.0.1:6379> PFSELFTEST
OK

3)内存消耗：百万独立用户

elements=""
key="2019_05_01:unique:ids"
for i in `seq 1 1000000`
do
  elements="${elements} uuid-"${i}
  if[[ $(i % 1000) == 0]] # ???
  then
    redis-cli pfadd ${key}  ${elements}
    elements=""
  fi
done

memory use

相对于bigmap 和set节省内存，小的惊人。

4)使用经验

1. 是否能容忍错误？（错误率：0.81%）

pfcount 2019_05_01:unique:ids (integer)
1009886

2. 是否需要单条数据？（无法获取到）