File name should be represented by a byte array instead of a String

• Key: HADOOP-1759
• Type: Improvement
• Status: Closed
• Resolution: Fixed
• Priority: Major
• Assignee: Konstantin Shvachko
• Reporter: Konstantin Shvachko
• Affects Version/s: 0.14.0
• Fix Version/s: 0.15.0

Description

This patch changes INode.name type to byte[], as outlined in HADOOP-1687 (4).
I submit it as a separate patch since it does not touch a lot of code, just INode.java

Comments

[Owen O'Malley]

Can we please use a Text rather than a byte[]?

[Sameer Paranjpye]

Don't think we should. This is a memory conservation exercise. Using a Text would mean an additional 28 bytes/file (16 byte JVM header + 8 byte reference + 4 bytes of length), for 64-bit JVMs at any rate.

[Konstantin Shvachko]

This is about saving bytes.
Using Text as an object with a length field will add 32 bytes (28 as Sameer calculated + padding) for each name.
For regular (ascii) file names String and Text have almost the same memory overhead.

[Hadoop QA]

+1

http://issues.apache.org/jira/secure/attachment/12364342/byteName.patch applied and successfully tested against trunk revision r568776.

Test results: http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/599/testReport/
Console output: http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/599/console

[Doug Cutting]

I just committed this. Thanks, Konstantin!