条件复杂的Spark SQL窗口函数 [英] Spark SQL window function with complex condition

问题描述

这可能是最容易通过示例解释的.假设我有一个用户登录到网站的DataFrame，例如:

This is probably easiest to explain through example. Suppose I have a DataFrame of user logins to a website, for instance:

scala> df.show(5)
+----------------+----------+
|       user_name|login_date|
+----------------+----------+
|SirChillingtonIV|2012-01-04|
|Booooooo99900098|2012-01-04|
|Booooooo99900098|2012-01-06|
|  OprahWinfreyJr|2012-01-10|
|SirChillingtonIV|2012-01-11|
+----------------+----------+
only showing top 5 rows

我想在此添加一列，以指示他们何时成为站点上的活跃用户.但是有一个警告:在一段时间内将用户视为活动用户，并且在此时间段之后，如果他们再次登录，则会重置其became_active日期.假设此期限为 5天.然后，从上表派生的所需表将是这样的:

I would like to add to this a column indicating when they became an active user on the site. But there is one caveat: there is a time period during which a user is considered active, and after this period, if they log in again, their became_active date resets. Suppose this period is 5 days. Then the desired table derived from the above table would be something like this:

+----------------+----------+-------------+
|       user_name|login_date|became_active|
+----------------+----------+-------------+
|SirChillingtonIV|2012-01-04|   2012-01-04|
|Booooooo99900098|2012-01-04|   2012-01-04|
|Booooooo99900098|2012-01-06|   2012-01-04|
|  OprahWinfreyJr|2012-01-10|   2012-01-10|
|SirChillingtonIV|2012-01-11|   2012-01-11|
+----------------+----------+-------------+

因此，特别是SirChillingtonIV的became_active日期被重设，因为他们的第二次登录是在活动期到期之后进行的，但是Booooooo99900098的became_active日期未被第二次登录，因为他/她处于活动状态时期.

So, in particular, SirChillingtonIV's became_active date was reset because their second login came after the active period expired, but Booooooo99900098's became_active date was not reset the second time he/she logged in, because it fell within the active period.

我最初的想法是将窗口函数与lag一起使用，然后使用lag ged值来填充became_active列；例如，开始时大致如下:

My initial thought was to use window functions with lag, and then using the lagged values to fill the became_active column; for instance, something starting roughly like:

import org.apache.spark.sql.expressions.Window
import org.apache.spark.sql.functions._

val window = Window.partitionBy("user_name").orderBy("login_date")
val df2 = df.withColumn("tmp", lag("login_date", 1).over(window))

然后，如果tmp是null(即，如果是第一次登录)或login_date - tmp >= 5然后是became_active = login_date，则填写became_active日期的规则将是；否则，请转到tmp中的下一个最近值，并应用相同的规则.这暗示了一种递归方法，我在想像一种实现方法时遇到了麻烦.

Then, the rule to fill in the became_active date would be, if tmp is null (i.e., if it's the first ever login) or if login_date - tmp >= 5 then became_active = login_date; otherwise, go to the next most recent value in tmp and apply the same rule. This suggests a recursive approach, which I'm having trouble imagining a way to implement.

我的问题:这是一种可行的方法吗?如果是，我如何才能返回"并查看tmp的较早值，直到找到要停止的位置?据我所知，我无法遍历Spark SQL Column的值.还有另一种方法可以达到这个结果吗?

My questions: Is this a viable approach, and if so, how can I "go back" and look at earlier values of tmp until I find one where I stop? I can't, to my knowledge, iterate through values of a Spark SQL Column. Is there another way to achieve this result?

推荐答案

这就是诀窍.导入一堆函数:

Here is the trick. Import a bunch of functions:

import org.apache.spark.sql.expressions.Window
import org.apache.spark.sql.functions.{coalesce, datediff, lag, lit, min, sum}

定义窗口:

val userWindow = Window.partitionBy("user_name").orderBy("login_date")
val userSessionWindow = Window.partitionBy("user_name", "session")

找到新会话开始的时间点:

Find the points where new sessions starts:

val newSession =  (coalesce(
  datediff($"login_date", lag($"login_date", 1).over(userWindow)),
  lit(0)
) > 5).cast("bigint")

val sessionized = df.withColumn("session", sum(newSession).over(userWindow))

查找每个会话的最早日期:

Find the earliest date per session:

val result = sessionized
  .withColumn("became_active", min($"login_date").over(userSessionWindow))
  .drop("session")

数据集定义为:

val df = Seq(
  ("SirChillingtonIV", "2012-01-04"), ("Booooooo99900098", "2012-01-04"),
  ("Booooooo99900098", "2012-01-06"), ("OprahWinfreyJr", "2012-01-10"), 
  ("SirChillingtonIV", "2012-01-11"), ("SirChillingtonIV", "2012-01-14"),
  ("SirChillingtonIV", "2012-08-11")
).toDF("user_name", "login_date")

结果是:

+----------------+----------+-------------+
|       user_name|login_date|became_active|
+----------------+----------+-------------+
|  OprahWinfreyJr|2012-01-10|   2012-01-10|
|SirChillingtonIV|2012-01-04|   2012-01-04| <- The first session for user
|SirChillingtonIV|2012-01-11|   2012-01-11| <- The second session for user
|SirChillingtonIV|2012-01-14|   2012-01-11| 
|SirChillingtonIV|2012-08-11|   2012-08-11| <- The third session for user
|Booooooo99900098|2012-01-04|   2012-01-04|
|Booooooo99900098|2012-01-06|   2012-01-04|
+----------------+----------+-------------+

这篇关于条件复杂的Spark SQL窗口函数的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！