CODEWITHSUNDEEP
javahadoopcascading
yosi
yosi

Reputation: 639

Cascading - merge 2 aggregations

I have the following problem whicj I am trying to solve with cascading: I have csv file of records with the structure: o,a,f,i,c

I need to to aggregate the records by o,a,f and to sum the i's and c's per group.

For example:

100,200,300,5,1

100,200,300,6,2

101,201,301,20,5

101,201,301,21,6

should yield:

100,200,300,11,3

101,201,301,41,11

I could not understand how to merge the 2 Every instances that I have (can I aggregate both fields in the same time?).

Do you have any idea?

Yosi

public class CascMain {

public static void main(String[] args){

    Scheme sourceScheme = new TextLine(new Fields("line"));
    Tap source = new Lfs(sourceScheme, "/tmp/casc/group.csv");

    Scheme sinkScheme = new TextDelimited(new Fields("o", "a", "f", "ti", "tc"), ",");
    Tap sink = new Lfs(sinkScheme, "/tmp/casc/output/", SinkMode.REPLACE);

    Pipe assembly = new Pipe("agg-pipe");

    Function function = new RegexSplitter(new Fields("o", "a", "f", "i", "c"), ",");
    assembly = new Each(assembly, new Fields("line"), function);

    Pipe groupAssembly = new GroupBy("group", assembly, new Fields("o", "a", "f"));

    Sum impSum = new Sum(new Fields("ti"));
    Pipe i = new Every(groupAssembly, new Fields("i"), impSum);

    Sum clickSum = new Sum(new Fields("tc"));
    Pipe c = new Every(groupAssembly, new Fields("c"), clickSum);

    // WHAT SHOULD I DO HERE

    Properties properties = new Properties();
    FlowConnector.setApplicationJarClass(properties, CascMain.class);

    FlowConnector flowConnector = new FlowConnector(properties);
    Flow flow = flowConnector.connect("agg", source, sink, assembly);
    flow.complete();

}

}

Upvotes: 3

Views: 1576

Answers (1)

J-B
J-B

Reputation: 400

Use AggregateBy to aggregate multiple fields at the same time:

SumBy impSum = new SumBy(new Fields("i"), new Fields("ti"), long.class);
SumBy clickSum = new SumBy(new Fields("c"), new Fields("tc"), long.class);
assembly = new AggregateBy("totals", Pipe.pipes(assembly), new Fields("o", "a", "f"), 2, impSum, clickSum);

Upvotes: 7

Related Questions