Abdeali Kothari created SPARK-35741:
---------------------------------------
Summary: Variance of 1 record gives NULL in Spark 3.x and NaN in
Spark 2.x
Key: SPARK-35741
URL: https://issues.apache.org/jira/browse/SPARK-35741
Project: Spark
Issue Type: Bug
Components: SQL
Affects Versions: 3.1.1
Reporter: Abdeali Kothari
I have a few testcases in my suite which started failing after moving to Spark 3
And I noticed that the reason was that VARIANCE() function was earlier
returning NaN when it was run on 1 record. Now, it gives NULL.
{code:java}
export SPARK_HOME=/usr/local/hadoop/spark-2.4.6-bin-hadoop2.7/
python
>>> import pyspark
>>> spark = pyspark.sql.SparkSession.builder.getOrCreate()
>>> spark.sql('SELECT VARIANCE(1)').show()
+---------------------------+
|var_samp(CAST(1 AS DOUBLE))|
+---------------------------+
| NaN|
+---------------------------+{code}
With spark 3:
{code:java}
export SPARK_HOME=/usr/local/hadoop/spark-3.1.1-bin-hadoop2.7/
python
>>> import pyspark
>>> spark = pyspark.sql.SparkSession.builder.getOrCreate()
>>> spark.sql('SELECT VARIANCE(1)').show()
+---------------------------+
|variance(CAST(1 AS DOUBLE))|
+---------------------------+
| null|
+---------------------------+
{code}
Just thought I'd report it here as I didn't see it in any of the release notes
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
