pyspark基于window实现列数据偏移

假设有如下场景：

df = spark.createDataFrame(
 [("anhui", 1, '2019-06-15 13:20'),
 ("anhui",2, '2019-06-17 13:42'),("anhui",3, '2019-06-15 13:42'),
 ("anhui",4, '2019-06-6 13:40'),
 ("anhui",5, '2019-06-14 14:40'),
 ("anhui",6, "'2019-06-15 13:42'"),
 ("beijing",1, '2019-06-9 13:42'),
 ("beijing",2, '2019-06-14 13:42'),
 ("beijing",3, '2019-06-12 13:42')], 
 ["province","nums", "time"]
)
df.show()
+--------+----+------------------+
|province|nums|              time|
+--------+----+------------------+
|   anhui|   1|  2019-06-15 13:20|
|   anhui|   2|  2019-06-17 13:42|
|   anhui|   3|  2019-06-15 13:42|
|   anhui|   4|   2019-06-6 13:40|
|   anhui|   5|  2019-06-14 14:40|
|   anhui|   6|'2019-06-15 13:42'|
| beijing|   1|   2019-06-9 13:42|
| beijing|   2|  2019-06-14 13:42|
| beijing|   3|  2019-06-12 13:42|
+--------+----+------------------+

如果需要以“province”分组，“nums”递增顺序，对“time”进行列数据向下偏移，在pyspark中可以用基于window函数的方式完成，代码如下：

from pyspark.sql import Window
from pyspark.sql.functions import lag
w=Window.partitionBy("province").orderBy("nums")
for i in range(1,6):
    df=df.withColumn("time_offset_"+str(i),lag(col="time",count=i).over(w))

结果如下：

+--------+----+------------------+----------------+----------------+----------------+----------------+----------------+
|province|nums|              time|   time_offset_1|   time_offset_2|   time_offset_3|   time_offset_4|   time_offset_5|
+--------+----+------------------+----------------+----------------+----------------+----------------+----------------+
| beijing|   1|   2019-06-9 13:42|            null|            null|            null|            null|            null|
| beijing|   2|  2019-06-14 13:42| 2019-06-9 13:42|            null|            null|            null|            null|
| beijing|   3|  2019-06-12 13:42|2019-06-14 13:42| 2019-06-9 13:42|            null|            null|            null|
|   anhui|   1|  2019-06-15 13:20|            null|            null|            null|            null|            null|
|   anhui|   2|  2019-06-17 13:42|2019-06-15 13:20|            null|            null|            null|            null|
|   anhui|   3|  2019-06-15 13:42|2019-06-17 13:42|2019-06-15 13:20|            null|            null|            null|
|   anhui|   4|   2019-06-6 13:40|2019-06-15 13:42|2019-06-17 13:42|2019-06-15 13:20|            null|            null|
|   anhui|   5|  2019-06-14 14:40| 2019-06-6 13:40|2019-06-15 13:42|2019-06-17 13:42|2019-06-15 13:20|            null|
|   anhui|   6|'2019-06-15 13:42'|2019-06-14 14:40| 2019-06-6 13:40|2019-06-15 13:42|2019-06-17 13:42|2019-06-15 13:20|
+--------+----+------------------+----------------+----------------+----------------+----------------+----------------+