时间过滤 Pyspark dataframe 中的结构列

Question

我有一个 dataframe 的列，其中的结构具有日期和值，因此架构看起来像

root
 |-- col1: struct (nullable = true)
 |    |-- dates: array (nullable = true)
 |    |    |-- element: timestamp (containsNull = true)
 |    |-- values: array (nullable = true)
 |    |    |-- element: double (containsNull = true)
 |-- col2: struct (nullable = true)
 |    |-- dates: array (nullable = true)
 |    |    |-- element: timestamp (containsNull = true)
 |    |-- values: array (nullable = true)
 |    |    |-- element: double (containsNull = true)
 |-- id: string (nullable = true)

给定一些时间索引：

time_index = datetime.datetime(2015, 12, 12, 4, 45)

以及前后天数：

min_diff = -1 and max_diff = 2

我想要新的列col1_filt和col2_filt ，它们具有相同的结构，返回落入由time_index和min_diff和max_diff定义的 window 以及相应值的日期。 如果没有日期或值属于该 window 我希望它返回None 。

下面是一个可以使用的示例 DataFrame。

示例 DataFrame：

example_input = [
    Row(
        id = "A", 
        col1 = Row(
            dates = [datetime.datetime(2015, 12, 11, 5, 28), datetime.datetime(2015, 12, 12, 4, 45), datetime.datetime(2015, 12, 13, 5, 9)], 
            values = [17.7, 19.1, 19.1]
        ),
        col2 = Row(
            dates = [datetime.datetime(2015, 12, 13, 4, 48), datetime.datetime(2015, 12, 15, 5, 8)], 
            values = [19.1, 19.1]
        )
    ),
    Row(
        id = "B", 
        col1 = Row(
            dates = [datetime.datetime(2017, 1, 13, 5, 9)], 
            values = [19.1]
        ),
        col2 = Row(
            dates = [datetime.datetime(2017, 1, 12, 2, 48), datetime.datetime(2017, 1, 15, 5, 8)], 
            values = [19.5, 29.1]
        )
    ),
]

df = spark.createDataFrame(example_input)

显示df：

+-------------------------------------------------------------------------------------+----------------------------------------------------------+---+
|col1                                                                                 |col2                                                      |id |
+-------------------------------------------------------------------------------------+----------------------------------------------------------+---+
|[[2015-12-11 05:28:00, 2015-12-12 04:45:00, 2015-12-13 05:09:00], [17.7, 19.1, 19.1]]|[[2015-12-13 04:48:00, 2015-12-15 05:08:00], [19.1, 19.1]]|A  |
|[[2017-01-13 05:09:00], [19.1]]                                                      |[[2017-01-12 02:48:00, 2017-01-15 05:08:00], [19.5, 29.1]]|B  |
+-------------------------------------------------------------------------------------+----------------------------------------------------------+---+

我有一些代码将采用 Pyspark 行 object 并返回过滤后的 Pyspark 行 ZA8CFDE6331BD59EB62AC96F891 怎么做

Answer 1

下面是使用 UDF 进行过滤的示例：

import datetime
import pyspark.sql.functions as F
from pyspark.sql import Row

time_index = datetime.datetime(2015, 12, 12, 4, 45)
min_diff = -1
max_diff = 2

def time_filter(r):
    ret = list(zip(*[
        x for x in list(zip(r['dates'], r['values'])) 
        if x[0] > time_index + datetime.timedelta(days=min_diff) 
        and x[0] < time_index + datetime.timedelta(days=max_diff)
    ]))
    return Row(dates=ret[0], values=ret[1]) if len(ret) != 0 else None

time_filter_udf = F.udf(time_filter, 'struct<dates:array<timestamp>,values:array<double>>')

df2 = df.withColumn('col1_filt', time_filter_udf('col1')).withColumn('col2_filt', time_filter_udf('col2'))

df2.show(truncate=False)
+-------------------------------------------------------------------------------------+----------------------------------------------------------+---+-------------------------------------------------------------------------------------+-------------------------------+
|col1                                                                                 |col2                                                      |id |col1_filt                                                                            |col2_filt                      |
+-------------------------------------------------------------------------------------+----------------------------------------------------------+---+-------------------------------------------------------------------------------------+-------------------------------+
|[[2015-12-11 05:28:00, 2015-12-12 04:45:00, 2015-12-13 05:09:00], [17.7, 19.1, 19.1]]|[[2015-12-13 04:48:00, 2015-12-15 05:08:00], [19.1, 19.1]]|A  |[[2015-12-11 05:28:00, 2015-12-12 04:45:00, 2015-12-13 05:09:00], [17.7, 19.1, 19.1]]|[[2015-12-13 04:48:00], [19.1]]|
|[[2017-01-13 05:09:00], [19.1]]                                                      |[[2017-01-12 02:48:00, 2017-01-15 05:08:00], [19.5, 29.1]]|B  |null                                                                                 |null                           |
+-------------------------------------------------------------------------------------+----------------------------------------------------------+---+-------------------------------------------------------------------------------------+-------------------------------+