Implement tester

[blahami2]

Implement testing environment for the parametrization evaluation - evaluate and compare several patameter setups (automatically with large data)

Update

• [x] Determine goals
• [x] Design and implements result format
• [x] Design and implement result export tools
  • [x] Implement table export tool
  • [x] Implement plot export tool
• [x] Determine ranges and increments
• [x] Implement runtime environment
  • [x] Get settings
  • [x] Run for the obtained settings
  • [x] Save result
  • [x] Table
  • [ ] Plot
  • [x] Increment and repeat
• [x] Create dataset
• [x] Design result format (for comparison)
• [x] Find routes using regular Dijkstra
• [ ] Add multilevel Dijkstra to testing (for each set of parameters, run Dijkstra on dataset, log results)
• [x] Compare route-search times and consider them as another quality measurement
• [x] Export an overall result
• [x] Report

[blahami2]

• Search for a plot library,
• determine ranges and increments for all the possible settings,
• create an environment for testing,
• export result tables and plots.

[blahami2]

1. Determine goals
   • What is the best result?
   • The lowest cut edge amount -> single partition
   • The fastest Dijkstra inside a partition -> single node in a partition
   • Probably create a size limit based on time requirements and then create partitions roughly of that size
2. Design and implements result format
3. Design and implement result export tools
4. Determine ranges and increments
5. Implement runtime environment
   • Get settings
   • Run for the obtained settings
   • Save result (table, plot)
   • Increment and repeat
6. Export an overall result
7. Report

[blahami2]

Settings

• Cell size: 1000; 50000; *2
• Cell ratio: 0.1; 1; +0.1
• Core ratio: 0.1; 1; +0.1
• Low interval probability: 0.01; 0.5; *2
• Low interval <0,?>: 0.1; 0.9; +0.1
• Number of assembly runs: 1; 1000; +50

Default

• Cell size: 10000
• Cell ratio: 1
• Core ratio: 0.1
• Low interval probability: 0.03
• Low interval <0,?>: 0.6
• Number of assembly runs: 1000

[blahami2]

:white_check_mark: Determine goals :white_check_mark: Design and implements result format :white_check_mark: Design and implement result export tools (table) ✅ Design and implement result export tools (plot) :white_check_mark: Determine ranges and increments :white_check_mark: Implement runtime environment :white_check_mark: Get settings :white_check_mark: Run for the obtained settings :white_check_mark: Save result (table) ❌ Save result (plot) :white_check_mark: Increment and repeat :white_check_mark: Export an overall result (table) :x: Export an overall result (plot) :white_check_mark: Report

[blahami2]

Update

✅ Create dataset ✅ Design result format (for comparison) ✅ Find routes using regular Dijkstra ✅ Add multilevel Dijkstra to testing (for each set of parameters, run Dijkstra on dataset, log results) ✅ Compare route-search times and consider them as another quality measurement

[blahami2]

Update

:white_check_mark: Define plot output :x: Create table for each plot :x: Export

[blahami2]

Plot output

Cell size

• cell size / cut edges
• cell size / routing time

Cell reatio

• cell ratio / cut edges
• cell ratio / routing time

X; X from |Settings|

• X / cut edges
• X / routing time

[blahami2]

MLD usage

public static void demo() {

        OverlayCreator creator = new OverlayCreator();
        OverlayCreator.SaraSetup setup = creator.setup;

        setup.spatialModulePath = "c:/ProgWin/SqLite64/mod_spatialite.dll";
        setup.dbFolder = "D:/";
        setup.setRandomSeed(123);
        setup.layerCount = 5;
        setup.maxCellSize = 20;
        setup.numberOfAssemblyRuns = 1;

       // D://prog-20-5.sqlite
        setup.dbName = "prg-20-5"; //L3

        // punch and save
        //setup.runPunch = true;

       //no punch, load only
        setup.runPunch = false;

        OverlayBuilder builder = creator.createBuilder();
        builder.buildOverlays();

        SaraGraph sara = builder.graph;

        DijkstraAlgorithm oneAlg = new DijkstraAlgorithm();
        MultilevelDijkstraAlgorithm mldAlg = new MultilevelDijkstraAlgorithm();

        long sourceId = 27793;
        long targetId = 28194;
        SaraNode source = sara.getNodeById(sourceId);
        SaraNode target = sara.getNodeById(targetId);

        Optional<Route<SaraNode, SaraEdge>> oneRoute
                = oneAlg.route(sara, Metric.LENGTH, source, target);

        MLDRecursiveRouteUnpacker unpacker = new MLDRecursiveRouteUnpacker();
        Optional<Route<SaraNode, SaraEdge>> mldRoute
                = mldAlg.route(sara, builder, Metric.LENGTH, source, target, unpacker);

        // 114 edges expected
        System.out.println("one=" + oneRoute.get().getEdgeList().size());
        System.out.println("mld=" + mldRoute.get().getEdgeList().size());
    }