Using Hive functions in Spark Job via hiveContext

Question

I am using Hive 1.2 and Spark 1.4.1. The Following query runs perfectly fine via Hive CLI:

hive> select row_number() over (partition by one.id order by two.id) as sk,
two.id, two.name, one.name, current_date() 
from avant_source.one one 
inner join avant_source.two two 
on one.id = two.one_id;

but when I try to use it via HiveContext in a pyspark job it gives me an error:

py4j.protocol.Py4JJavaError: An error occurred while calling o26.sql.
: java.lang.RuntimeException: Couldn't find function current_date

Code snippet:

from pyspark import HiveContext

conf = SparkConf().setAppName('DFtest')
sc = SparkContext(conf=conf)
sqlContext = HiveContext(sc)

df = sqlContext.sql("select row_number() over (partition by one.id order by two.id) as sk, two.id, two.name, one.name, current_date() from avant_source.one one inner join avant_source.two two on one.id = two.one_id")

df.show()

sc.stop()

Is there a way to get the current date or timestamp in pyspark? I tried importing date, datetime, but it always throws an error saying function not found.

I tried to use current_date in Data Frames in pyspark 1.5 Sandbox, but then also I get a different error.

df = sqlContext.createDataFrame([(current_date,)],[‘d’])
df.select(date_sub(df.d,1).alias('d')).collect()

Error:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/mapr/spark/spark-1.5.2/python/pyspark/sql/dataframe.py", line 769, in select
    jdf = self._jdf.select(self._jcols(*cols))
  File "/opt/mapr/spark/spark-1.5.2/python/lib/py4j-0.8.2.1-src.zip/py4j/java_gateway.py", line 538, in __call__
  File "/opt/mapr/spark/spark-1.5.2/python/pyspark/sql/utils.py", line 40, in deco
    raise AnalysisException(s.split(': ', 1)[1])
pyspark.sql.utils.AnalysisException: cannot resolve 'datesub(d,1)' due to data type mismatch: argument 1 requires date type, however, 'd' is of struct<> type.;

Please advise.

Answer 1

For my scenario, I used the following

import datetime 
now =  datetime.datetime.now()
df = df.withColumn('eff_start', lit(now.strftime("%Y-%m-%d")))

For the error of not being able to use HiveContext for HiveQL correctly for Hive functions, it was a cluster issue, where one of the nodes on which HiveServer2 was running had too many alarms due to memory issues. That was causing the problem. It was tested successfully on a MapR Sandbox running Spark 1.5 and Hive 1.2