Error when converting from spark dataframe with dates to pandas dataframe

Question

I have a spark dataframe with this schema:

root
 |-- product_id: integer (nullable = true)
 |-- stock: integer (nullable = true)
 |-- start_date: date (nullable = true)
 |-- end_date: date (nullable = true)

When trying to pass it to a pandas_udf or convert to a pandas dataframe with:

pandas_df = spark_df.toPandas()

It returns this error:

AttributeError        Traceback (most recent call last)
<ipython-input-86-4bccc6e8422d> in <module>()
     10 # spark_df.printSchema()
     11 
---> 12 pandas_df = spark_df.toPandas()

/home/.../lib/python2.7/site-packages/pyspark/sql/dataframe.pyc in toPandas(self)
   2123                         table = pyarrow.Table.from_batches(batches)
   2124                         pdf = table.to_pandas()
-> 2125                         pdf = _check_dataframe_convert_date(pdf, self.schema)
   2126                         return _check_dataframe_localize_timestamps(pdf, timezone)
   2127                     else:

/home.../lib/python2.7/site-packages/pyspark/sql/types.pyc in _check_dataframe_convert_date(pdf, schema)
   1705     """
   1706     for field in schema:
-> 1707         pdf[field.name] = _check_series_convert_date(pdf[field.name], field.dataType)
   1708     return pdf
   1709 

/home/.../lib/python2.7/site-packages/pyspark/sql/types.pyc in _check_series_convert_date(series, data_type)
   1690     """
   1691     if type(data_type) == DateType:
-> 1692         return series.dt.date
   1693     else:
   1694         return series

/home/.../lib/python2.7/site-packages/pandas/core/generic.pyc in __getattr__(self, name)
   5061         if (name in self._internal_names_set or name in self._metadata or
   5062                 name in self._accessors):
-> 5063             return object.__getattribute__(self, name)
   5064         else:
   5065             if self._info_axis._can_hold_identifiers_and_holds_name(name):

/home/.../lib/python2.7/site-packages/pandas/core/accessor.pyc in __get__(self, obj, cls)
    169             # we're accessing the attribute of the class, i.e., Dataset.geo
    170             return self._accessor
--> 171         accessor_obj = self._accessor(obj)
    172         # Replace the property with the accessor object. Inspired by:
    173         # http://www.pydanny.com/cached-property.html

/home/.../lib/python2.7/site-packages/pandas/core/indexes/accessors.pyc in __new__(cls, data)
    322             pass  # we raise an attribute error anyway
    323 
--> 324         raise AttributeError("Can only use .dt accessor with datetimelike "
    325                              "values")

AttributeError: Can only use .dt accessor with datetimelike values

If the date fields are dropped from the spark dataframe the conversion works without problems.

I checked that the data doesn't contains any nulls but it would be nice to know how to deal with those too.

I'm using python2.7 with:

• pyspark==2.4.0
• pyarrow==0.12.1
• pandas==0.24.1

Answer 1

Looks like a bug. Have the same issue with pyarrow==0.12.1 and pyarrow==0.12.0. Casting spark dataframe column to TIMESTAMP works for me.

spark.sql('SELECT CAST(date_column as TIMESTAMP) FROM foo')

Also rolling back to pyarrow==0.11.0 solves the issue. (my python is 3.7.1 and pandas 0.24.2)