You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Reynold Xin (JIRA)" <ji...@apache.org> on 2015/05/31 08:35:17 UTC
[jira] [Comment Edited] (SPARK-7197) Join with DataFrame Python API
not working properly with more than 1 column
[ https://issues.apache.org/jira/browse/SPARK-7197?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14566389#comment-14566389 ]
Reynold Xin edited comment on SPARK-7197 at 5/31/15 6:34 AM:
-------------------------------------------------------------
Actually I don't think this is a bug. If you do this, then it works:
{code}
a.join(b, (a.year==b.year) & (a.month==b.month), 'inner').explain(True)
{code}
This is a little bit unfortunate because it's not possible to override && and "and" in Python, so we took the same way Pandas handle ands -- by overriding "&".
was (Author: rxin):
Actually I think this is not a bug. If you do this, then it works:
{code}
a.join(b, (a.year==b.year) & (a.month==b.month), 'inner').explain(True)
{code}
This is a little bit unfortunate because it's not possible to override && and "and" in Python, so we took the same way Pandas handle ands -- by overriding "&".
> Join with DataFrame Python API not working properly with more than 1 column
> ---------------------------------------------------------------------------
>
> Key: SPARK-7197
> URL: https://issues.apache.org/jira/browse/SPARK-7197
> Project: Spark
> Issue Type: Bug
> Components: PySpark, SQL
> Affects Versions: 1.3.1
> Reporter: Ali Bajwa
> Priority: Critical
>
> It looks like join with DataFrames API in python does not return correct results if using more 2 or more columns. The example in the documentation
> only shows a single column.
> Here is an example to show the problem:
> ****Example code****
> {code}
> import pandas as pd
> from pyspark.sql import SQLContext
> hc = SQLContext(sc)
> A = pd.DataFrame({'year': ['1993', '2005', '1994'], 'month': ['5',
> '12', '12'], 'value': [100, 200, 300]})
> a = hc.createDataFrame(A)
> B = pd.DataFrame({'year': ['1993', '1993'], 'month': ['12', '12'],
> 'value': [101, 102]})
> b = hc.createDataFrame(B)
> print "Pandas" # try with Pandas
> print A
> print B
> print pd.merge(A, B, on=['year', 'month'], how='inner')
> print "Spark"
> print a.toPandas()
> print b.toPandas()
> print a.join(b, a.year==b.year and a.month==b.month, 'inner').toPandas()
> {code}
> *****Output****
> {code}
> Pandas
> month value year
> 0 5 100 1993
> 1 12 200 2005
> 2 12 300 1994
> month value year
> 0 12 101 1993
> 1 12 102 1993
> Empty DataFrame
> Columns: [month, value_x, year, value_y]
> Index: []
> Spark
> month value year
> 0 5 100 1993
> 1 12 200 2005
> 2 12 300 1994
> month value year
> 0 12 101 1993
> 1 12 102 1993
> month value year month value year
> 0 12 200 2005 12 102 1993
> 1 12 200 2005 12 101 1993
> 2 12 300 1994 12 102 1993
> 3 12 300 1994 12 101 1993
> {code}
> It looks like Spark returns some results where an inner join should
> return nothing.
> Confirmed on user mailing list as an issue with Ayan Guha.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org