You need to store your data into "column-based" format, checking out Hive RCFile, and its InputFormat option.
Date: Mon, 23 Dec 2013 21:37:23 +0800
Subject: Any method to get input splits by column?
From: [EMAIL PROTECTED]
To: [EMAIL PROTECTED]
By default, MR inputformat classes break input file into splits by rows. However, we have a specilal requirement on MR app: get input splits by column.
Is there any good method?