sqoop-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Devin G. Bost (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SQOOP-1312) One of mappers does not load data from mySql if double column is used as split key
Date Fri, 05 Oct 2018 02:05:00 GMT

    [ https://issues.apache.org/jira/browse/SQOOP-1312?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16639165#comment-16639165
] 

Devin G. Bost commented on SQOOP-1312:
--------------------------------------

I'm also getting this issue in Sqoop version: *1.4.7*.
{quote}/04 18:59:52 INFO db.DBInputFormat: Using read commited transaction isolation
18/10/04 18:59:52 INFO db.DataDrivenDBInputFormat: BoundingValsQuery: SELECT MIN(CUSTOMER_ID),
MAX(CUSTOMER_ID) FROM CUSTOMERS
18/10/04 18:59:54 WARN db.FloatSplitter: Generating splits for a floating-point index column.
Due to the
18/10/04 18:59:54 WARN db.FloatSplitter: imprecise representation of floating-point values
in Java, this
18/10/04 18:59:54 WARN db.FloatSplitter: may result in an incomplete import.
18/10/04 18:59:54 WARN db.FloatSplitter: You are strongly encouraged to choose an integral
split column.
18/10/04 18:59:54 DEBUG db.DataDrivenDBInputFormat: Creating input split with lower bound
'CUSTOMER_ID >= 98.0' and upper bound 'CUSTOMER_ID < 648440.0'
18/10/04 18:59:54 DEBUG db.DataDrivenDBInputFormat: Creating input split with lower bound
'CUSTOMER_ID >= 648440.0' and upper bound 'CUSTOMER_ID < 1296782.0'
18/10/04 18:59:54 DEBUG db.DataDrivenDBInputFormat: Creating input split with lower bound
'CUSTOMER_ID >= 1296782.0' and upper bound 'CUSTOMER_ID < 1945124.0'
18/10/04 18:59:54 DEBUG db.DataDrivenDBInputFormat: Creating input split with lower bound
'CUSTOMER_ID >= 2593466.0' and upper bound 'CUSTOMER_ID <= 2593466.0'
{quote}
+It's skipping the entire range from 1945124.0 to 2593466.0, which contains approximately
8338 rows.+ 

Also, it's doubtful that it's really a problem with the float conversion because all of the
floating point values are actually used as integers (so they always end in .0 and thus could
easily be converted into integers without any data loss). 

I confirmed that there are 0 rows with the CUSTOMER_ID value below 98.0, so I believe the
only issue is that the entire range from 1945124.0 to 2593466.0 is getting skipped. 

 

> One of mappers does not load data from mySql if double column is used as split key
> ----------------------------------------------------------------------------------
>
>                 Key: SQOOP-1312
>                 URL: https://issues.apache.org/jira/browse/SQOOP-1312
>             Project: Sqoop
>          Issue Type: Bug
>    Affects Versions: 1.4.4
>            Reporter: Jong Ho Lee
>            Assignee: Jong Ho Lee
>            Priority: Major
>         Attachments: splitter.patch, splitter.patch
>
>
> When we used Sqoop to load data from mySQL using one double column as split-key in Samsung
SDS,
>   the last mapper did not load data from mySQL at all. 
>   The number of mappers was sometimes increased by 1.
>   I think they were caused by some bugs in FloatSplitter.java
>   For the last split, lowClausePrefix + Double.toString(curUpper), may be lowClausePrefix
+ Double.toString(curLower).
>   In while (curUpper < maxVal) loop, because of round-off error, 
>   minVal + splitSize * numSplits can be smaller than maxVal.
>   Therefore, using for-loop would be better.
>   Attached is a proposed new FloatSplitter.java
> {code}
> /**
>  * Licensed to the Apache Software Foundation (ASF) under one
>  * or more contributor license agreements.  See the NOTICE file
>  * distributed with this work for additional information
>  * regarding copyright ownership.  The ASF licenses this file
>  * to you under the Apache License, Version 2.0 (the
>  * "License"); you may not use this file except in compliance
>  * with the License.  You may obtain a copy of the License at
>  *
>  *     http://www.apache.org/licenses/LICENSE-2.0
>  *
>  * Unless required by applicable law or agreed to in writing, software
>  * distributed under the License is distributed on an "AS IS" BASIS,
>  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
>  * See the License for the specific language governing permissions and
>  * limitations under the License.
>  */
> // modified by Jongho Lee at Samsung SDS.
> package org.apache.sqoop.mapreduce.db;
> import java.sql.ResultSet;
> import java.sql.SQLException;
> import java.util.ArrayList;
> import java.util.List;
> import org.apache.commons.logging.Log;
> import org.apache.commons.logging.LogFactory;
> import org.apache.hadoop.conf.Configuration;
> import org.apache.hadoop.mapreduce.InputSplit;
> import com.cloudera.sqoop.config.ConfigurationHelper;
> import com.cloudera.sqoop.mapreduce.db.DBSplitter;
> import com.cloudera.sqoop.mapreduce.db.DataDrivenDBInputFormat;
> /**
>  * Implement DBSplitter over floating-point values.
>  */
> public class FloatSplitter implements DBSplitter  {
>   private static final Log LOG = LogFactory.getLog(FloatSplitter.class);
>   private static final double MIN_INCREMENT = 10000 * Double.MIN_VALUE;
>   public List<InputSplit> split(Configuration conf, ResultSet results,
>       String colName) throws SQLException {
>     LOG.warn("Generating splits for a floating-point index column. Due to the");
>     LOG.warn("imprecise representation of floating-point values in Java, this");
>     LOG.warn("may result in an incomplete import.");
>     LOG.warn("You are strongly encouraged to choose an integral split column.");
>     List<InputSplit> splits = new ArrayList<InputSplit>();
>     if (results.getString(1) == null && results.getString(2) == null) {
>       // Range is null to null. Return a null split accordingly.
>       splits.add(new DataDrivenDBInputFormat.DataDrivenDBInputSplit(
>           colName + " IS NULL", colName + " IS NULL"));
>       return splits;
>     }
>     double minVal = results.getDouble(1);
>     double maxVal = results.getDouble(2);
>     // Use this as a hint. May need an extra task if the size doesn't
>     // divide cleanly.
>     int numSplits = ConfigurationHelper.getConfNumMaps(conf);
>     double splitSize = (maxVal - minVal) / (double) numSplits;
>     if (splitSize < MIN_INCREMENT) {
>       splitSize = MIN_INCREMENT;
>     }
>     String lowClausePrefix = colName + " >= ";
>     String highClausePrefix = colName + " < ";
>     double curLower = minVal;
>     double curUpper = curLower + splitSize;
>     for (int i = 0; i < numSplits - 1; i++) {
>       // while (curUpper < maxVal) {  // changed to for loop
>       splits.add(new DataDrivenDBInputFormat.DataDrivenDBInputSplit(
>           lowClausePrefix + Double.toString(curLower),
>           highClausePrefix + Double.toString(curUpper)));
>       curLower = curUpper;
>       curUpper += splitSize;
>     }
>     // Catch any overage and create the closed interval for the last split.
>     if (curLower <= maxVal || splits.size() == 1) {
>       splits.add(new DataDrivenDBInputFormat.DataDrivenDBInputSplit(
>           lowClausePrefix + Double.toString(curLower),
>           colName + " <= " + Double.toString(maxVal)));
>       // curUpper -> curLower // changed
>     }
>     if (results.getString(1) == null || results.getString(2) == null) {
>       // At least one extrema is null; add a null split.
>       splits.add(new DataDrivenDBInputFormat.DataDrivenDBInputSplit(
>           colName + " IS NULL", colName + " IS NULL"));
>     }
>     return splits;
>   }
> }
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message