JavaPackage object is not callable error: Pyspark

问题:

Operations like dataframe.show() , sQLContext.read.json works fine , but most functions gives "JavaPackage object is not callable error" . eg : when i do

dataFrame.withColumn(field_name, monotonically_increasing_id())

I get an error

File "/tmp/spark-cd423f35-9572-45ee-b159-1b2732afa2a6/userFiles-3a6e1729-95f4-468b-914c-c706369bf2a6/Transformations.py", line 64, in add_id_column
    self.dataFrame = self.dataFrame.withColumn(field_name, monotonically_increasing_id())
  File "/home/himaprasoon/apps/spark-1.6.0-bin-hadoop2.6/python/pyspark/sql/functions.py", line 347, in monotonically_increasing_id
    return Column(sc._jvm.functions.monotonically_increasing_id())
TypeError: 'JavaPackage' object is not callable

I am using apache-zeppelin interpreter and have added py4j to python path.

When I do

import py4j
print(dir(py4j))

the import succeeds

['__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', 'compat', 'finalizer', 'java_collections', 'java_gateway', 'protocol', 'version']

When I tried

print(sc._jvm.functions)

in pyspark shell it prints

<py4j.java_gateway.JavaClass object at 0x7fdaf9727ba8>

But when I try this in my interpreter it prints

<py4j.java_gateway.JavaPackage object at 0x7f07cc3f77f0> 

回答1:

In zeppelin interpreter code

java_import(gateway.jvm, "org.apache.spark.sql.*")

was not getting executed. Adding this to the import fixed the issue