pickle.loads给出 "模块 "对象在Pyspark Pandas Udf中没有属性'<ClassName>'。

Question

我试图在PySpark Pandas udf中对一个类实例进行pickle和unpickle。在udf之外，pickling工作得很好。

class ExampleModel:
    pass

clf = ExampleModel(args)
pickled_val = base64.b64encode(pickle.dumps(clf))
clf2 = pickle.loads(base64.b64decode(pickled_val))
print(clf2)
# <__main__.ExampleModel instance at 0x7f04d7444780>

然而，在pandas udf里面，我可以访问ExampleModel类，但不能解开字符串列。

df = spark_session.createDataFrame(
    [
        (1, pickled_val, '') 
    ],
    ['id', 'txt', 'error'] 
)

@pandas_udf(df.schema, PandasUDFType.GROUPED_MAP)
def example_unpickle(pdf):
    try:
        clf_obj = ExampleModel()
    except Exception as e:
        pdf.loc[:,'error'] = "1:" + str(e)
        return pdf

    try:
        clf3 = pickle.loads(base64.b64decode(pdf.iloc[0,1]))
    except Exception as e:
        pdf.loc[:,'error'] = "2: " + str(e)
        return pdf


df_clf = df\
            .groupby('id')\
            .apply(example_unpickle)

df_clf.show(truncate = False)

给出错误。

AttributeError: 'module' object has no attribute 'ExampleModel'

+---+------------------------------------------------+--------------------------------------------------+
|id |txt                                             |error                                             |
+---+------------------------------------------------+--------------------------------------------------+
|1  |KGlfX21haW5fXwpFeGFtcGxlTW9kZWwKcDAKKGRwMQpiLg==|2: 'module' object has no attribute 'ExampleModel'|
+---+------------------------------------------------+--------------------------------------------------+

Answer 1

解决的办法是把这个类做成一个单独的文件，并创建一个名为 "pickled_val "的文件。__init__.py 在同一目录下。

然后将该类导入为。

from ExampleFileName import ExampleModel