在 pyspark UDF 中使用类方法 [英] Using a class method inside of a pyspark UDF

问题描述

数据工程师您好！

我正在尝试使用名为 星界

I am trying to write a pyspark udf using a method from a class called Astral

这里是udf:

def time_from_solar_noon(d, y):
    noon = astral.Astral().solar_noon_utc
    time = noon(d, y)
    return time 

solarNoon = F.udf(lambda d, y: time_from_solar_noon(d,y), TimestampType())

按照我的理解，该类将针对数据框中的每一行进行实例化，从而导致工作非常缓慢.

Now the way I understand it, the class will be instantiated for every single line in my dataframe, resulting in a very slow job.

如果我从我的函数中取出类实例化:

If I take the class instantiation out of my function :

noon = astral.Astral().solar_noon_utc

def time_from_solar_noon(d, y):
    time = noon(d, y)
    return time 

我收到以下错误消息:

  [Previous line repeated 326 more times]
    RecursionError: maximum recursion depth exceeded while calling a Python object

所以这是我的问题，我认为应该可以通过执行程序/线程至少有一个类实例化，而不是在我的数据框中逐行实例化，我该怎么做?

So here is my question, I think it should be possible to have at least one class instantiation by executor/thread, instead of one by line in my dataframe, how would I do that ?

感谢您的帮助

推荐答案

就像数据库连接一样，您可以使用 mapPartitions 来实例化有限数量的这些类实例:

Just like with database connections, you can instantiate only a limited number of these class instances, by using mapPartitions:

In [1]: from datetime import date
   ...: from astral import Astral
   ...: 
   ...: df = spark.createDataFrame(
   ...:     ((date(2019, 10, 4), 0),
   ...:      (date(2019, 10, 4), 19)),
   ...:     schema=("date", "longitude"))
   ...: 
   ...: 
   ...: def solar_noon(rows):
   ...:     a = Astral()  # initialize the class once per partition
   ...:     return ((a.solar_noon_utc(date=r.date, longitude=r.longitude), *r)
   ...:             for r in rows)  # reuses the same Astral instance for all rows in this partition
   ...: 
   ...: 
   ...: (df.rdd
   ...:  .mapPartitions(solar_noon)
   ...:  .toDF(schema=("solar_noon_utc", *df.columns))
   ...:  .show()
   ...:  )
   ...: 
   ...:  
+-------------------+----------+---------+                                      
|     solar_noon_utc|      date|longitude|
+-------------------+----------+---------+
|2019-10-04 13:48:58|2019-10-04|        0|
|2019-10-04 12:32:58|2019-10-04|       19|
+-------------------+----------+---------+

这是相当有效的，因为函数 (solar_noon) 被赋予每个工人，并且每个分区只初始化一次类，它可以容纳多行.

This is fairly efficient, as the function (solar_noon) is given to each worker and the class is only initialized once per partition, which can hold many rows.

这篇关于在 pyspark UDF 中使用类方法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！