You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Niels Basjes (JIRA)" <ji...@apache.org> on 2011/01/05 21:35:52 UTC
[jira] Updated: (HADOOP-7076) Splittable Gzip
[ https://issues.apache.org/jira/browse/HADOOP-7076?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Niels Basjes updated HADOOP-7076:
---------------------------------
Attachment: HADOOP-7076.patch
This patch now passes all unit tests.
Changes present in this file:
- Added additional method getBytesRead() to the org.apache.hadoop.io.compress.Decompressor interface to be able to query the position of the underlying file.
- Added the option to decrease the blocksize used by the DecompressorStream to read the disk file and feed the decompressor (Needed to get the required accuracy).
- Added SplittableGzipCodec that allows splitting Gzipped input files.
- Added TestSplittableCodecSeams that tests if all the splits are seamless: No duplicate records and no missing records.
- Fixes several bugs in TestCodec.java
- Reset of decompressor
- Writing an number in a binary form into a file that is later read and parsed as a text file (now all textual)
- Naming : no more "Splitable" in the touched unit test files.
> Splittable Gzip
> ---------------
>
> Key: HADOOP-7076
> URL: https://issues.apache.org/jira/browse/HADOOP-7076
> Project: Hadoop Common
> Issue Type: New Feature
> Components: io
> Reporter: Niels Basjes
> Attachments: HADOOP-7076.patch
>
>
> Files compressed with the gzip codec are not splittable due to the nature of the codec.
> This limits the options you have scaling out when reading large gzipped input files.
> Given the fact that gunzipping a 1GiB file usually takes only 2 minutes I figured that for some use cases wasting some resources may result in a shorter job time under certain conditions.
> So reading the entire input file from the start for each split (wasting resources!!) may lead to additional scalability.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.