You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Ed Lee (JIRA)" <ji...@apache.org> on 2017/05/07 11:51:04 UTC
[jira] [Updated] (SPARK-20617) pyspark.sql filtering fails when
using ~isin when there are nulls in column
[ https://issues.apache.org/jira/browse/SPARK-20617?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ed Lee updated SPARK-20617:
---------------------------
Summary: pyspark.sql filtering fails when using ~isin when there are nulls in column (was: pyspark.sql, filtering with ~isin missing rows)
> pyspark.sql filtering fails when using ~isin when there are nulls in column
> ---------------------------------------------------------------------------
>
> Key: SPARK-20617
> URL: https://issues.apache.org/jira/browse/SPARK-20617
> Project: Spark
> Issue Type: Bug
> Components: PySpark, SQL
> Affects Versions: 2.2.0
> Environment: Ubuntu Xenial 16.04, Python 3.5
> Reporter: Ed Lee
> Fix For: 2.2.0
>
>
> Hello encountered a filtering bug using 'isin' in pyspark sql on version 2.2.0, Ubuntu 16.04.
> Enclosed below an example to replicate:
> from pyspark.sql import SparkSession
> from pyspark.sql import functions as sf
> import pandas as pd
> spark = SparkSession.builder.master("local").appName("Word Count").getOrCreate()
> test_df = pd.DataFrame({"col1": [None, None, "a", "b", "c"],
> "col2": range(5)
> })
> test_sdf = spark.createDataFrame(test_df)
> test_sdf.show()
> |col1|col2|
> |null| 0|
> |null| 1|
> | a| 2|
> | b| 3|
> | c| 4|
> # Below shows null entries in col1 are considered 'isin' the list ["a"] (it is not in the list so it should show):
> test_sdf.filter(sf.col("col1").isin(["a"]) == False).show()
> Or:
> test_sdf.filter(~sf.col("col1").isin(["a"])).show()
> *Expecting*:
> |col1|col2|
> |null| 0|
> |null| 1|
> | b| 3|
> | c| 4|
> *Got*:
> |col1|col2|
> | b| 3|
> | c| 4|
> My workarounds:
> 1. null is considered 'in', so add OR isNull conditon:
> test_sdf.filter((sf.col("col1").isin(["a"])== False) | (
> sf.col("col1").isNull())).show()
> To get:
> |col1|col2|isin|
> |null| 0|null|
> |null| 1|null|
> | c| 4|null|
> | b| 3|null|
> 2. Use left join and filter
> join_df = pd.DataFrame({"col1": ["a"],
> "isin": 1
> })
> join_sdf = spark.createDataFrame(join_df)
> test_sdf.join(join_sdf, on="col1", how="left") \
> .filter(sf.col("isin").isNull()) \
> .show()
> To get:
> |col1|col2|isin|
> |null| 0|null|
> |null| 1|null|
> | c| 4|null|
> | b| 3|null|
> Thank you
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org