Does Keith's input format apply the necessary Accumulo iterators to provide a sane view of the data to MapReduce?
And what you're proposing is an input format that works over RFiles where perhaps multiple versions of the same row/column don't exist in multiple files and where there are no delete markers, etc?
On Mar 9, 2012, at 12:10 PM, John Vines (Created) (JIRA) wrote:
> RFile Input Format
> Key: ACCUMULO-454
> URL: https://issues.apache.org/jira/browse/ACCUMULO-454
> Project: Accumulo
> Issue Type: New Feature
> Components: client
> Reporter: John Vines
> Assignee: Billie Rinaldi
> Fix For: 1.4.1
> We currently provide InputFormats for reading from Accumulo and output formats for both direct input as well as outputting RFiles. But we provide no mechanism for doing a mapreduce over existing RFiles, which may be useful for optimizing data flow. We already have input formats which use RFiles directly for input (The offline input format Keith just finished), but that still relies on the Accumulo structure. We should go ahead and also create an input format that just hits RFiles like the other standard file input formats.
> This message is automatically generated by JIRA.
> If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
> For more information on JIRA, see: http://www.atlassian.com/software/jira