hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Tom White (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HADOOP-7350) Use ServiceLoader to discover compression codec classes
Date Thu, 02 Jun 2011 21:19:47 GMT

     [ https://issues.apache.org/jira/browse/HADOOP-7350?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Tom White updated HADOOP-7350:
------------------------------

    Attachment: HADOOP-7350.patch

{quote}
- We should probably remove the codecs from core-default.xml now that they're loaded via ServiceLoader
{quote}

Done - see new patch.

{quote}
- Is there a way to inject a new codec programatically through the ServiceLoader interface?
If so, we could entirely deprecate io.compression.codecs. If not, maybe we should rename it
to something like io.compression.extra.codecs and specify that it's only necessary if you
have a codec that doesn't expose itself through ServiceLoader?
{quote}

I don't think we need to deprecate or rename io.compression.codecs - it's just used to specify
_additional_ codecs to the ones that are loaded through a ServiceLoader. Note that duplicates
are ignored, so there's no problem with users older configs having codecs that could be loaded
through ServiceLoader.

{quote}
- hdfs-default.xml has an item dfs.image.compression.codec that needs to be updated
{quote}

This doesn't need to be updated, although with HADOOP-7323 (and a corresponding HDFS change)
it could be changed to "default".

> Use ServiceLoader to discover compression codec classes
> -------------------------------------------------------
>
>                 Key: HADOOP-7350
>                 URL: https://issues.apache.org/jira/browse/HADOOP-7350
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: conf, io
>            Reporter: Tom White
>            Assignee: Tom White
>         Attachments: HADOOP-7350.patch, HADOOP-7350.patch
>
>
> By using a ServiceLoader users wouldn't have to add codec classes to io.compression.codecs
for codecs that aren't shipped with Hadoop (e.g. LZO), since they would be automatically picked
up from the classpath.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message