spark UDF Java 错误：方法 col([class java.util.ArrayList]) 不存在

Question

我有一个 python 字典：

fileClass = {'a1' : ['a','b','c','d'], 'b1':['a','e','d'], 'c1': ['a','c','d','f','g']}

元组列表为：

C = [('a','b'), ('c','d'),('e')]

我想最终创建一个火花数据框：

Name (a,b) (c,d) (e)
a1     2     2    0
b1     1     1    1
c1     1     2    0

它只包含出现在dict A中每个项目中的每个元组中元素的计数 为此，我创建了一个 dict 来将每个元素映射到 col 索引

classLoc = {'a':0,'b':0,'c':1,'d':1,'e':2}

那我用udf来定义

import numpy as np
def convertDictToDF(v, classLoc, length) :

    R = np.zeros((1,length))
    for c in v:
        try:
            loc = classLoc[c]
            R[loc] += 1
        except:
            pass 
    return(R)
udfConvertDictToDF = udf(convertDictToDF, ArrayType(IntegerType())) 

df = sc.parallelize([
    [k] + list(udfConvertDictToDF(v, classLoc, len(C)))
    for k, v in fileClass.items()]).toDF(['Name']+ C)

然后我收到错误消息

---------------------------------------------------------------------------
Py4JError                                 Traceback (most recent call last)
<ipython-input-40-ab668a12838a> in <module>()
      1 df = sc.parallelize([
      2     [k] + list(udfConvertDictToDF(v,classLoc, len(C)))
----> 3     for k, v in fileClass.items()]).toDF(['Name'] + C)
      4 
      5 df.show()

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/pyspark/sql/functions.pyc in __call__(self, *cols)
   1582     def __call__(self, *cols):
   1583         sc = SparkContext._active_spark_context
-> 1584         jc = self._judf.apply(_to_seq(sc, cols, _to_java_column))
   1585         return Column(jc)
   1586 

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/pyspark/sql/column.pyc in _to_seq(sc, cols, converter)
     58     """
     59     if converter:
---> 60         cols = [converter(c) for c in cols]
     61     return sc._jvm.PythonUtils.toSeq(cols)
     62 

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/pyspark/sql/column.pyc in _to_java_column(col)
     46         jcol = col._jc
     47     else:
---> 48         jcol = _create_column_from_name(col)
     49     return jcol
     50 

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/pyspark/sql/column.pyc in _create_column_from_name(name)
     39 def _create_column_from_name(name):
     40     sc = SparkContext._active_spark_context
---> 41     return sc._jvm.functions.col(name)
     42 
     43 

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/lib/py4j-0.9-src.zip/py4j/java_gateway.py in __call__(self, *args)
    811         answer = self.gateway_client.send_command(command)
    812         return_value = get_return_value(
--> 813             answer, self.gateway_client, self.target_id, self.name)
    814 
    815         for temp_arg in temp_args:

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/pyspark/sql/utils.pyc in deco(*a, **kw)
     43     def deco(*a, **kw):
     44         try:
---> 45             return f(*a, **kw)
     46         except py4j.protocol.Py4JJavaError as e:
     47             s = e.java_exception.toString()

/home/yizhng/spark-1.6.0-bin-hadoop2.6/python/lib/py4j-0.9-src.zip/py4j/protocol.py in get_return_value(answer, gateway_client, target_id, name)
    310                 raise Py4JError(
    311                     "An error occurred while calling {0}{1}{2}. Trace:\n{3}\n".
--> 312                     format(target_id, ".", name, value))
    313         else:
    314             raise Py4JError(

Py4JError: An error occurred while calling z:org.apache.spark.sql.functions.col. Trace:
py4j.Py4JException: Method col([class java.util.ArrayList]) does not exist
    at py4j.reflection.ReflectionEngine.getMethod(ReflectionEngine.java:335)
    at py4j.reflection.ReflectionEngine.getMethod(ReflectionEngine.java:360)
    at py4j.Gateway.invoke(Gateway.java:254)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:133)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:209)
    at java.lang.Thread.run(Thread.java:745)

我不明白我的 UDF 有什么问题会导致该错误消息。请帮忙

Answer 1

我认为这与你使用这条线的方式有关

[k] + list(udfConvertDictToDF(v, classLoc, len(C)))

在底部。

当我做一个简单的 python 版本时，我也会得到一个错误。

import numpy as np

C = [('a','b'), ('c','d'),('e')]

classLoc = {'a':0,'b':0,'c':1,'d':1,'e':2}

import numpy as np
def convertDictToDF(v, classLoc, length) :

    # I also got rid of (1,length) for (length)
    # b/c pandas .from_dict() method handles this for me
    R = np.zeros(length)  
    for c in v:
        try:
            loc = classLoc[c]
            R[loc] += 1
        except:
            pass 
    return(R)


[[k] + convertDictToDF(v, classLoc, len(C))
    for k, v in fileClass.items()]

产生这些错误

TypeError: ufunc 'add' did not contain a loop with signature matching types dtype('S32') dtype('S32') dtype('S32')

如果您要将列表理解更改为 dict 理解，您可以让它工作。

dict = {k:convertDictToDF(v, classLoc, len(C))
    for k, v in fileClass.items()}

它的输出看起来像这样

> {'a1': array([ 2.,  2.,  0.]), 'c1': array([ 1.,  2.,  0.]), 'b1': array([ 1.,  1.,  1.])}

在不知道您的最终用例是什么的情况下，我将让您获得您请求的输出，但使用稍微不同的方式，这可能无法按您的意愿扩展，所以我确定有更好的方法。

以下代码将为您提供数据框的其余部分，

import pandas as pd
df = pd.DataFrame.from_dict(data=dict,orient='index').sort_index() 
df.columns=C

产生你想要的输出

    (a, b)  (c, d)    e
a1     2.0     2.0  0.0
b1     1.0     1.0  1.0
c1     1.0     2.0  0.0

这将为您提供一个 Spark 数据帧

from pyspark.sql import SQLContext
sqlContext = SQLContext(sc)
df_s = sqlContext.createDataFrame(df)
df_s.show()

+----------+----------+---+
|('a', 'b')|('c', 'd')|  e|
+----------+----------+---+
|       2.0|       2.0|0.0|
|       1.0|       1.0|1.0|
|       1.0|       2.0|0.0|
+----------+----------+---+