/**
* Assembles the Plan of the Pairwise Shortest Paths example Pact program. The program computes
* one iteration of the Pairwise Shortest Paths algorithm.
*
* <p>For the first iteration, two input formats can be chosen: 1) RDF triples with foaf:knows
* predicates 2) Text-serialized paths (see PathInFormat and PathOutFormat)
*
* <p>To choose 1) set the forth parameter to "true". If set to "false" 2) will be used.
*/
@Override
public Plan getPlan(String... args) {
// parse job parameters
int numSubTasks = (args.length > 0 ? Integer.parseInt(args[0]) : 1);
String paths = (args.length > 1 ? args[1] : "");
String output = (args.length > 2 ? args[2] : "");
boolean rdfInput = (args.length > 3 ? Boolean.parseBoolean(args[3]) : false);
FileDataSource pathsInput;
if (rdfInput) {
pathsInput = new FileDataSource(new RDFTripleInFormat(), paths, "RDF Triples");
} else {
pathsInput = new FileDataSource(new PathInFormat(), paths, "Paths");
}
pathsInput.setDegreeOfParallelism(numSubTasks);
JoinOperator concatPaths =
JoinOperator.builder(new ConcatPaths(), StringValue.class, 0, 1)
.name("Concat Paths")
.build();
concatPaths.setDegreeOfParallelism(numSubTasks);
CoGroupOperator findShortestPaths =
CoGroupOperator.builder(new FindShortestPath(), StringValue.class, 0, 0)
.keyField(StringValue.class, 1, 1)
.name("Find Shortest Paths")
.build();
findShortestPaths.setDegreeOfParallelism(numSubTasks);
FileDataSink result = new FileDataSink(new PathOutFormat(), output, "New Paths");
result.setDegreeOfParallelism(numSubTasks);
result.setInput(findShortestPaths);
findShortestPaths.setFirstInput(pathsInput);
findShortestPaths.setSecondInput(concatPaths);
concatPaths.setFirstInput(pathsInput);
concatPaths.setSecondInput(pathsInput);
return new Plan(result, "Pairwise Shortest Paths");
}
@SuppressWarnings("unchecked")
@Override
public Plan getPlan(String... args) {
// parse job parameters
final int numSubTasks = (args.length > 0 ? Integer.parseInt(args[0]) : 1);
final String verticesInput = (args.length > 1 ? args[1] : "");
final String edgeInput = (args.length > 2 ? args[2] : "");
final String output = (args.length > 3 ? args[3] : "");
final int maxIterations = (args.length > 4 ? Integer.parseInt(args[4]) : 1);
// data source for initial vertices
FileDataSource initialVertices =
new FileDataSource(new CsvInputFormat(' ', LongValue.class), verticesInput, "Vertices");
MapOperator verticesWithId =
MapOperator.builder(DuplicateLongMap.class)
.input(initialVertices)
.name("Assign Vertex Ids")
.build();
DeltaIteration iteration = new DeltaIteration(0, "Connected Components Iteration");
iteration.setInitialSolutionSet(verticesWithId);
iteration.setInitialWorkset(verticesWithId);
iteration.setMaximumNumberOfIterations(maxIterations);
// create DataSourceContract for the edges
FileDataSource edges =
new FileDataSource(
new CsvInputFormat(' ', LongValue.class, LongValue.class), edgeInput, "Edges");
// create CrossOperator for distance computation
JoinOperator joinWithNeighbors =
JoinOperator.builder(new NeighborWithComponentIDJoin(), LongValue.class, 0, 0)
.input1(iteration.getWorkset())
.input2(edges)
.name("Join Candidate Id With Neighbor")
.build();
CoGroupOperator minAndUpdate =
CoGroupOperator.builder(new MinIdAndUpdate(), LongValue.class, 0, 0)
.input1(joinWithNeighbors)
.input2(iteration.getSolutionSet())
.name("Min Id and Update")
.build();
iteration.setNextWorkset(minAndUpdate);
iteration.setSolutionSetDelta(minAndUpdate);
// create DataSinkContract for writing the new cluster positions
FileDataSink result = new FileDataSink(new CsvOutputFormat(), output, iteration, "Result");
CsvOutputFormat.configureRecordFormat(result)
.recordDelimiter('\n')
.fieldDelimiter(' ')
.field(LongValue.class, 0)
.field(LongValue.class, 1);
// return the PACT plan
Plan plan = new Plan(result, "Workset Connected Components");
plan.setDefaultParallelism(numSubTasks);
return plan;
}