You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@impala.apache.org by "Henry Robinson (JIRA)" <ji...@apache.org> on 2017/07/24 15:59:00 UTC
[jira] [Resolved] (IMPALA-5532) Don't heap-allocate compressor
objects in RowBatch
[ https://issues.apache.org/jira/browse/IMPALA-5532?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Henry Robinson resolved IMPALA-5532.
------------------------------------
Resolution: Fixed
Fix Version/s: Impala 2.10.0
https://github.com/apache/incubator-impala/commit/f3d8ccdf0f19b0b4077df517cf604a863c55bb37
> Don't heap-allocate compressor objects in RowBatch
> --------------------------------------------------
>
> Key: IMPALA-5532
> URL: https://issues.apache.org/jira/browse/IMPALA-5532
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Affects Versions: Impala 2.10.0
> Reporter: Henry Robinson
> Assignee: Henry Robinson
> Fix For: Impala 2.10.0
>
>
> Every call to {{RowBatch::RowBatch(..., const TRowBatch&, ...)}} or {{RowBatch::Serialize()}} creates a (de)compressor object. That uses the {{Codec::CreateCompressor()}} interface which returns a pointer to a {{Codec}} object, so that the virtual compression interface can be used.
> However, we always use LZ4 compression, and so needlessly heap-allocate the compressor objects to get the advantage of implementation hiding which we don't actually need. We should just declare a stack-allocated LZ4 (de)compressor when needed.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)