1、WindowFunnel
关于官网的解释:
Returned value:Integer. The maximum number of consecutive triggered conditions from the chain within the sliding time window. All the chains in the selection are analyzed.
返回值:int类型。返回满足在指定滑动窗口内的连续触发条件的最大值。所有被选择的条件链都会被分析(这句翻译的不准确,主要看前面一句翻译即可)
具体实例分析:
建如下一张表,
CREATE TABLE funnel.funnel_test ( uid String, eventid String, eventTime UInt64) ENGINE = MergeTree PARTITION BY (uid, eventTime) ORDER BY (uid, eventTime) SETTINGS index_granularity = 8192
有三个字段:
uid:用户id
eventid:事件id
eventTime:事件发生时间(秒)
插入如下数据作为测试数据:
uid1 event1 1551398404
uid1 event2 1551398406
uid1 event3 1551398408
uid2 event2 1551398412
uid2 event3 1551398415
uid3 event3 1551398410
uid3 event4 1551398413
1.
select uid,windowFunnel(4)(toDateTime(eventTime),eventid = 'event2',eventid = 'event3') as funnel from funnel_test group by uid;
当我们设置的滑动窗口为4秒,条件链为event2->event3时,上述查询得到的结果为:
uid funnel
uid1 2
uid2 2
uid3 0
下面我们看看他是怎么得到这个结果的,首先将所有的数据根据uid聚合和排序(排序是windowFunnel里自己实现的),得到:
uid1: (event1,1551398404) -> (event2,1551398406) -> (event3,1551398408)
uid2: (event2,1551398412) -> (event3,1551398415)
uid3: (event3 ,1551398410) -> (event4,1551398413)
由上述聚合和排序之后的条件链中,只有uid1和uid2有event2->event3的条件链,且时间差分别为2(1551398408-1551398406) 和 3(1551398415-1551398412),小于滑动窗口4,所以满足条件,故uid1和uid2的结果都为2(event2,event3),而uid3为0(没有满足条件的条件链)
2、 如果滑动窗口改为2
select uid,windowFunnel(2)(toDateTime(eventTime),eventid = 'event2',eventid = 'event3') as funnel from funnel_test group by uid;
则由上述得到的条件链知道,结果为
uid funnel
uid1 2
uid2 1
uid3 0
为什么uid2变成了1,因为uid2的条件链中的event3和event2的时间差是3,大于了滑动窗口时间2,所以只有第一个条件event2满足查询,故结果为1
3、 如果滑动窗口为4,条件链改为event3,event4,
select uid,windowFunnel(4)(toDateTime(eventTime),eventid = 'event3',eventid = 'event4') as funnel from funnel_test group by uid;
则查询结果为
uid funnel
uid1 1
uid2 1
uid3 2
因为uid1和uid2只有事件event3,没有事件event4.
而uid3既有event3,也有event4,且两个事件的时间差小于滑动窗口4,故uid3的结果为2