Filtering a spark partitioned table is not working in Pyspark

Question

I am using spark 2.3 and have written one dataframe to create hive partitioned table using dataframe writer class method in pyspark.

newdf.coalesce(1).write.format('orc').partitionBy('veh_country').mode("overwrite").saveAsTable('emp.partition_Load_table')

Here is my table structure and partitions information.

hive> desc emp.partition_Load_table;
OK
veh_code                varchar(17)
veh_flag                varchar(1)
veh_model               smallint
veh_country             varchar(3)

# Partition Information
# col_name              data_type               comment

veh_country              varchar(3)

hive> show partitions partition_Load_table;
OK
veh_country=CHN
veh_country=USA
veh_country=RUS

Now I am reading this table back in pyspark inside a dataframe.

    df2_data = spark.sql("""
    SELECT * 
    from udb.partition_Load_table
    """);

df2_data.show() --> is working

But I am not able to filter it using partition key column

from pyspark.sql.functions import col
newdf = df2_data.where(col("veh_country")=='CHN')

I am getting below error message:

: java.lang.RuntimeException: Caught Hive MetaException attempting to get partition metadata by filter from Hive. 
You can set the Spark configuration setting spark.sql.hive.manageFilesourcePartitions to false to work around this problem, 
however this will result in degraded performance. Please report a bug: https://issues.apache.org/jira/browse/SPARK
Caused by: MetaException(message:Filtering is supported only on partition keys of type string)

whereas when I am creating dataframe by specifying the hdfs absolute path of table. filter and where clause is working as expected.

newdataframe = spark.read.format("orc").option("header","false").load("hdfs/path/emp.db/partition_load_table")

below is working

newdataframe.where(col("veh_country")=='CHN').show()

my question is that why it was not able to filter the dataframe in first place. and also why it's throwing an error message " Filtering is supported only on partition keys of type string " even though my veh_country is defined as string or varchar datatypes.

Answer 1

I have stumbled on this issue also. What helped for me was to do this line:

spark.sql("SET spark.sql.hive.manageFilesourcePartitions=False")

and then use spark.sql(query) instead of using dataframe.

I do not know what happens under the hood, but this solved my problem.

Although it might be too late for you (since this question was asked 8 months ago), this might help for other people.