[jira] [Resolved] (FLINK-39) [GitHub] Port HBase access to ozone

["Fabian Hueske (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/FLINK-39?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Fabian Hueske resolved FLINK-39.
--------------------------------
       Resolution: Fixed
    Fix Version/s:     (was: pre-apache)
                   0.8-incubating

Fixed with a1100af4247f77632f07ee9cea9fc7452104fac6

> [GitHub] Port HBase access to ozone
> -----------------------------------
>
>                 Key: FLINK-39
>                 URL: https://issues.apache.org/jira/browse/FLINK-39
>             Project: Flink
>          Issue Type: Improvement
>            Reporter: GitHub Import
>              Labels: github-import
>             Fix For: 0.8-incubating
>
>
> During the IMR Hackathon, an HBase access implementation to Stratosphere 0.2.1 was created. 
> We need to port this to ozone and the latest version of HBase.
> This is needed by several projects, including IMR and Deutsche Telekom.
> Marcus provided some issues that that current prototype has that could be resolved:
> 1. Currently, we only support fetching a single row version at a time. If we want to support multiple versions per record, that needs to be added and we need to specify how we map this to pact records.
> 2. All columns of a row are currently squeezed into a single pact record field (basically, we serialize the result object returned by HBase). This was fine for the Hackathon and didn't really matter for the Meteor integration (here, we convert this to a JSON structure, and leave to the mapping tho Arvid's schema code).
> But I guess it's not the PACT way of handling this, since you'd rather have columns mapped to record fields for fast access/use as key, right?
> So we need to write some code so people can specify which columns go where.
> This may actually be more tricky than it sounds, because you might not know the column qualifier in advance, only the column family.
> The question here is: how do we handle nesting? There may be different/unknown numbers of qualifier/values pairs for a certain column family. Do we just unnest them and emit multiple records for a single row, or do we want to provide a way to nest all these values into a single records.
> ---------------- Imported from GitHub ----------------
> Url: https://github.com/stratosphere/stratosphere/issues/39
> Created by: [ktzoumas|https://github.com/ktzoumas]
> Labels: enhancement, 
> Created at: Thu Jun 20 13:54:04 CEST 2013
> State: open



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)