[MAPREDUCE-2343] repeated replace() function calls damage the performance

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.21.0
• Component/s: tools/rumen
• Assignee: Unassigned
• Reporter: Xiaoming Shi
• Created: 23/Feb/11 20:14
• Updated: 13/May/16 23:11

Description

In the file

.hadoop-0.21.0/mapred/src/tools/org/apache/hadoop/tools/rumen/LoggedTaskAttempt.java                line:362   

hadoop-0.21.0/mapred/src/tools/org/apache/hadoop/tools/rumen/LoggedTask.java                               line:249

consecutive replace() is called to remove the special characters. It's 5+ times slower than using a for loop replace them all.

e.g.
 - str.replace('a', '|');
 - str.replace('b', '|');

 + StringBuilder sb = new StringBuilder( str.length() );
 + for (int i=0; i < str.length(); i++)
 +  {
 +           char c = str.charAt(i);
 +         if ( c == 'a'  || c =='b' )
 +               sb.append('|');
 +       else
 +                sb.append('c');
 +  }
 +  str  = sb.toString();

This bug has the same problem as the MySQL bug : http://bugs.mysql.com/bug.php?id=45699

Issue Links

• duplicates: MAPREDUCE-2380 Multiple replace function call can be replaced with a single for loop to improve performance