[
https://issues.apache.org/jira/browse/TINKERPOP3-703?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14559358#comment-14559358
]
Marko A. Rodriguez commented on TINKERPOP3-703:
-----------------------------------------------
Probably make an interface called {{InputRDD}} and {{OutputRDD}} so it has the
same look-and-feel as {{Input/OutputFormat}}. Then vendors/users can simply
implement these interfaces and bypass Hadoop interfaces.
> Provide a "INPUT_GRAPH_RDD_CLASS" for SparkGraphComputer vendors.
> -----------------------------------------------------------------
>
> Key: TINKERPOP3-703
> URL: https://issues.apache.org/jira/browse/TINKERPOP3-703
> Project: TinkerPop 3
> Issue Type: Improvement
> Components: hadoop
> Reporter: Marko A. Rodriguez
> Assignee: Marko A. Rodriguez
>
> If a vendor has a direct Spark connector to their database, then they should
> be able to bypass using an {{InputFormat}} to construct the {{graphRDD}}. It
> would be cool to have a configuration option that is
> {{INPUT_GRAPH_RDD_CLASS}}.
> {code}
> Class<BiFunction<SparkContext,Configuration,JavaPairRDD<Object,VertexWritable>>
> {code}
> This way, the SparkGraphComputer code would be like:
> {code}
> if(configuration.hasProperty(INPUT_GRAPH_RDD_CLASS)) {
>
> Class<BiFunction<SparkContext,Configuration,JavaPairRDD<Object,VertexWritable>>
> theClass =
> Class.forName(configuration.get(INPUT_GRAPH_RDD_CLASS)).newInstance();
> graphRDD = theClass.apply(sparkContext,configuration);
> } else {
> graphRDD = // use input format
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
