Examples with Cube io.cdap.cdap.api.dataset.lib.cube.Cube used on opensource projects

Example 1 with Cube

use of io.cdap.cdap.api.dataset.lib.cube.Cube in project cdap by caskdata.

the class AbstractCubeTest method testBasics.

@Test
public void testBasics() throws Exception {
    Aggregation agg1 = new DefaultAggregation(ImmutableList.of("dim1", "dim2", "dim3"), ImmutableList.of("dim1", "dim2"));
    Aggregation agg2 = new DefaultAggregation(ImmutableList.of("dim1", "dim2"), ImmutableList.of("dim1"));
    int resolution = 1;
    Cube cube = getCube("myCube", new int[] { resolution }, ImmutableMap.of("agg1", agg1, "agg2", agg2));
    // write some data
    // NOTE: we mostly use different ts, as we are interested in checking incs not at persist, but rather at query time
    writeInc(cube, "metric1", 1, 1, "1", "1", "1");
    writeInc(cube, "metric1", 1, 1, "1", "1", "1");
    writeInc(cube, "metric1", 2, 2, null, "1", "1");
    writeInc(cube, "metric1", 3, 3, "1", "2", "1");
    writeInc(cube, "metric1", 3, 5, "1", "2", "3");
    writeInc(cube, "metric1", 3, 7, "2", "1", "1");
    writeInc(cube, "metric1", 4, 4, "1", null, "2");
    writeInc(cube, "metric1", 5, 5, null, null, "1");
    writeInc(cube, "metric1", 6, 6, "1", null, null);
    writeInc(cube, "metric1", 7, 3, "1", "1", null);
    // writing using BatchWritable APIs
    writeIncViaBatchWritable(cube, "metric1", 8, 2, null, "1", null);
    writeIncViaBatchWritable(cube, "metric1", 9, 1, null, null, null);
    // writing in batch
    cube.add(ImmutableList.of(getFact("metric1", 10, 2, MeasureType.COUNTER, "1", "1", "1", "1"), getFact("metric1", 11, 3, MeasureType.COUNTER, "1", "1", "1", null), getFact("metric1", 12, 4, MeasureType.COUNTER, "2", "1", "1", "1"), getFact("metric1", 13, 5, MeasureType.COUNTER, null, null, null, "1")));
    writeInc(cube, "metric2", 1, 1, "1", "1", "1");
    // todo: do some write instead of increments - test those as well
    // now let's query!
    verifyCountQuery(cube, 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1"), ImmutableList.of("dim2"), ImmutableList.of(new TimeSeries("metric1", dimensionValues("dim2", "1"), timeValues(1, 2, 7, 3, 10, 2, 11, 3)), new TimeSeries("metric1", dimensionValues("dim2", "2"), timeValues(3, 8))));
    verifyCountQuery(cube, 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1", "dim3", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 2, 10, 2, 11, 3))));
    verifyCountQuery(cube, 0, 15, resolution, "metric1", AggregationFunction.SUM, new HashMap<String, String>(), ImmutableList.of("dim1"), ImmutableList.of(new TimeSeries("metric1", dimensionValues("dim1", "1"), timeValues(1, 2, 3, 8, 4, 4, 6, 6, 7, 3, 10, 2, 11, 3)), new TimeSeries("metric1", dimensionValues("dim1", "2"), timeValues(3, 7, 12, 4))));
    verifyCountQuery(cube, 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim3", "3"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(3, 5))));
    // test querying specific aggregations
    verifyCountQuery(cube, "agg1", 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 2, 3, 8, 7, 3, 10, 2, 11, 3))));
    verifyCountQuery(cube, "agg2", 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 2, 3, 8, 4, 4, 6, 6, 7, 3, 10, 2, 11, 3))));
    // query with different agg functions
    verifyCountQuery(cube, "agg1", 0, 15, resolution, "metric1", AggregationFunction.MAX, ImmutableMap.of("dim1", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 2, 3, 5, 7, 3, 10, 2, 11, 3))));
    verifyCountQuery(cube, "agg1", 0, 15, resolution, "metric1", AggregationFunction.MIN, ImmutableMap.of("dim1", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 2, 3, 3, 7, 3, 10, 2, 11, 3))));
    verifyCountQuery(cube, "agg1", 0, 15, resolution, "metric1", AggregationFunction.LATEST, ImmutableMap.of("dim1", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 2, 3, 5, 7, 3, 10, 2, 11, 3))));
    // delete cube data for "metric1" for dim->1,dim2->1,dim3->1 for timestamp 1 - 8 and
    // check data for other timestamp is available
    Map<String, String> deleteTags = new LinkedHashMap<>();
    deleteTags.put("dim1", "1");
    deleteTags.put("dim2", "1");
    deleteTags.put("dim3", "1");
    Predicate<List<String>> predicate = aggregates -> Collections.indexOfSubList(aggregates, new ArrayList<>(deleteTags.keySet())) == 0;
    CubeDeleteQuery query = new CubeDeleteQuery(0, 8, resolution, deleteTags, Collections.singletonList("metric1"), predicate);
    cube.delete(query);
    verifyCountQuery(cube, 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1", "dim3", "1"), ImmutableList.<String>of(), ImmutableList.of(new TimeSeries("metric1", new HashMap<>(), timeValues(10, 2, 11, 3))));
    // delete cube data for "metric1" for dim1->1 and dim2->1  and check by scanning dim1->1 and dim2->1 is empty,
    deleteTags.remove("dim3");
    query = new CubeDeleteQuery(0, 15, resolution, deleteTags, Collections.singletonList("metric1"), predicate);
    cube.delete(query);
    verifyCountQuery(cube, 0, 15, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1"), ImmutableList.<String>of(), ImmutableList.<TimeSeries>of());
}

Example 2 with Cube

use of io.cdap.cdap.api.dataset.lib.cube.Cube in project cdap by caskdata.

the class AbstractCubeTest method testMetricsAggregationOptionSum.

@Test
public void testMetricsAggregationOptionSum() throws Exception {
    // two aggregation groups
    Aggregation agg1 = new DefaultAggregation(ImmutableList.of("dim1", "dim2", "dim3"), ImmutableList.of("dim1"));
    Aggregation agg2 = new DefaultAggregation(ImmutableList.of("dim1", "dim2"), ImmutableList.of("dim1"));
    int resolution = 1;
    Cube cube = getCube("testAggOptionSum", new int[] { resolution }, ImmutableMap.of("agg1", agg1, "agg2", agg2));
    Map<String, String> agg1Dims = new LinkedHashMap<>();
    agg1Dims.put("dim1", "tag1");
    agg1Dims.put("dim2", "tag2");
    agg1Dims.put("dim3", "tag3");
    Map<String, String> agg2Dims = new LinkedHashMap<>();
    agg2Dims.put("dim1", "tag1");
    agg2Dims.put("dim2", "tag4");
    // write 100 data points to agg1
    for (int i = 1; i <= 100; i++) {
        writeInc(cube, "metric1", i, 1, agg1Dims);
        writeInc(cube, "metric2", i, 2, agg1Dims);
    }
    // write 50 data points to agg2
    for (int i = 1; i <= 50; i++) {
        writeInc(cube, "metric1", i, 3, agg2Dims);
    }
    // test limit must be greater than 0
    CubeQuery query = new CubeQuery(null, 0, 200, 1, 0, ImmutableMap.of("metric1", AggregationFunction.SUM), agg1Dims, Collections.emptyList(), AggregationOption.SUM, null);
    try {
        cube.query(query);
        Assert.fail();
    } catch (Exception e) {
    // expected
    }
    query = new CubeQuery(null, 0, 200, 1, -10, ImmutableMap.of("metric1", AggregationFunction.SUM), agg1Dims, Collections.emptyList(), AggregationOption.SUM, null);
    try {
        cube.query(query);
        Assert.fail();
    } catch (Exception e) {
    // expected
    }
    // test a limit greater than the number data points, all the data points should be returned
    query = new CubeQuery(null, 0, 200, 1, 200, ImmutableMap.of("metric1", AggregationFunction.SUM, "metric2", AggregationFunction.SUM), agg1Dims, Collections.emptyList(), AggregationOption.SUM, null);
    List<TimeSeries> result = new ArrayList<>(cube.query(query));
    Assert.assertEquals(2, result.size());
    verifySumAggregation(result.get(0), "metric1", 100, 1, 1, 0, 0);
    verifySumAggregation(result.get(1), "metric2", 100, 2, 1, 0, 0);
    // test aggregation option with sum for metric1 and metric2 for agg1, 5 data points for agg1 should get returned
    // for both metric1 and metric2
    query = new CubeQuery(null, 0, 200, 1, 5, ImmutableMap.of("metric1", AggregationFunction.SUM, "metric2", AggregationFunction.SUM), agg1Dims, Collections.emptyList(), AggregationOption.SUM, null);
    result = new ArrayList<>(cube.query(query));
    Assert.assertEquals(2, result.size());
    // metric1 increment by 1 per second, so sum will be 100/5=20, metric2 increment by 2 per second, so sum will be
    // 200/5=40
    verifySumAggregation(result.get(0), "metric1", 5, 20, 20, 0, 0);
    verifySumAggregation(result.get(1), "metric2", 5, 40, 20, 0, 0);
    // test aggregation option with sum for metric1 with tag name dim1, it should return two time series for agg1 and
    // agg2 for metric1, each with 5 data points
    query = new CubeQuery(null, 0, 200, 1, 5, ImmutableMap.of("metric1", AggregationFunction.SUM), ImmutableMap.of("dim1", "tag1"), ImmutableList.of("dim2"), AggregationOption.SUM, null);
    result = new ArrayList<>(cube.query(query));
    Assert.assertEquals(2, result.size());
    // agg1 gets increment by 1 for 100 seconds, so sum will be 100/5=20, agg2 gets increment by 3 for 50 seconds, so
    // sum will be 3*50/5=30
    verifySumAggregation(result.get(0), "metric1", 5, 30, 10, 0, 0);
    verifySumAggregation(result.get(1), "metric1", 5, 20, 20, 0, 0);
    // test metric1 with count 9, this will have a remainder 100%9=1, so there will be 9 aggregated data points, each
    // with partition size 11
    query = new CubeQuery(null, 0, 200, 1, 9, ImmutableMap.of("metric1", AggregationFunction.SUM), agg1Dims, Collections.emptyList(), AggregationOption.SUM, null);
    result = new ArrayList<>(cube.query(query));
    Assert.assertEquals(1, result.size());
    // the rest data points have sum 11
    verifySumAggregation(result.get(0), "metric1", 9, 11, 11, 1, 1);
    // test metric1 with count 70, this will have a remainder 100%70=30, so the result will have last 70 data points,
    // the first 30 data points will be ignored
    query = new CubeQuery(null, 0, 200, 1, 70, ImmutableMap.of("metric1", AggregationFunction.SUM), agg1Dims, Collections.emptyList(), AggregationOption.SUM, null);
    result = new ArrayList<>(cube.query(query));
    Assert.assertEquals(1, result.size());
    // the rest data points have sum 11
    verifySumAggregation(result.get(0), "metric1", 70, 1, 1, 30, 30);
}

Example 3 with Cube

use of io.cdap.cdap.api.dataset.lib.cube.Cube in project cdap by caskdata.

the class AbstractCubeTest method testInterpolate.

@Test
public void testInterpolate() throws Exception {
    Aggregation agg1 = new DefaultAggregation(ImmutableList.of("dim1", "dim2", "dim3"), ImmutableList.of("dim1", "dim2", "dim3"));
    int resolution = 1;
    Cube cube = getCube("myInterpolatedCube", new int[] { resolution }, ImmutableMap.of("agg1", agg1));
    // test step interpolation
    long startTs = 1;
    long endTs = 10;
    writeInc(cube, "metric1", startTs, 5, "1", "1", "1");
    writeInc(cube, "metric1", endTs, 3, "1", "1", "1");
    List<TimeValue> expectedTimeValues = Lists.newArrayList();
    for (long i = startTs; i < endTs; i++) {
        expectedTimeValues.add(new TimeValue(i, 5));
    }
    expectedTimeValues.add(new TimeValue(endTs, 3));
    verifyCountQuery(cube, startTs, endTs, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1", "dim3", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), expectedTimeValues)), new Interpolators.Step());
    Map<String, String> deleteTags = new LinkedHashMap<>();
    deleteTags.put("dim1", "1");
    deleteTags.put("dim2", "1");
    deleteTags.put("dim3", "1");
    Predicate<List<String>> predicate = aggregates -> Collections.indexOfSubList(aggregates, new ArrayList<>(deleteTags.keySet())) == 0;
    CubeDeleteQuery query = new CubeDeleteQuery(startTs, endTs, resolution, deleteTags, Collections.singletonList("metric1"), predicate);
    cube.delete(query);
    // test small-slope linear interpolation
    startTs = 1;
    endTs = 5;
    writeInc(cube, "metric1", startTs, 5, "1", "1", "1");
    writeInc(cube, "metric1", endTs, 3, "1", "1", "1");
    verifyCountQuery(cube, startTs, endTs, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1", "dim3", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 5, 2, 5, 3, 4, 4, 4, 5, 3))), new Interpolators.Linear());
    query = new CubeDeleteQuery(startTs, endTs, resolution, deleteTags, Collections.singletonList("metric1"), predicate);
    cube.delete(query);
    // test big-slope linear interpolation
    writeInc(cube, "metric1", startTs, 100, "1", "1", "1");
    writeInc(cube, "metric1", endTs, 500, "1", "1", "1");
    verifyCountQuery(cube, startTs, endTs, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1", "dim3", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 100, 2, 200, 3, 300, 4, 400, 5, 500))), new Interpolators.Linear());
    cube.delete(query);
    // test limit on Interpolate
    long limit = 20;
    writeInc(cube, "metric1", 0, 10, "1", "1", "1");
    writeInc(cube, "metric1", limit + 1, 50, "1", "1", "1");
    expectedTimeValues.clear();
    expectedTimeValues.add(new TimeValue(0, 10));
    for (long i = 1; i <= limit; i++) {
        expectedTimeValues.add(new TimeValue(i, 0));
    }
    expectedTimeValues.add(new TimeValue(limit + 1, 50));
    verifyCountQuery(cube, 0, 21, resolution, "metric1", AggregationFunction.SUM, ImmutableMap.of("dim1", "1", "dim2", "1", "dim3", "1"), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), expectedTimeValues)), new Interpolators.Step(limit));
}

Example 4 with Cube

use of io.cdap.cdap.api.dataset.lib.cube.Cube in project cdap by caskdata.

the class CubeDatasetTest method getCubeInternal.

private Cube getCubeInternal(String name, int[] resolutions, Map<String, ? extends Aggregation> aggregations) throws Exception {
    DatasetProperties props = configureProperties(resolutions, aggregations);
    DatasetId id = DatasetFrameworkTestUtil.NAMESPACE_ID.dataset(name);
    if (dsFrameworkUtil.getInstance(id) == null) {
        dsFrameworkUtil.createInstance(Cube.class.getName(), id, props);
    }
    return dsFrameworkUtil.getInstance(id);
}

Example 5 with Cube

use of io.cdap.cdap.api.dataset.lib.cube.Cube in project cdap by caskdata.

the class CubeDatasetTest method testTxRetryOnFailure.

@Test
public void testTxRetryOnFailure() throws Exception {
    // This test ensures that there's no non-transactional cache used in cube dataset. For that, it
    // 1) simulates transaction conflict for the first write to cube
    // 2) attempts to write again, writes successfully
    // 3) uses second cube instance to read the result
    // 
    // In case there's a non-transactional cache used in cube, it would fill entity mappings in the first tx, and only
    // use them to write data. Hence, when reading - there will be no mapping in entity table to decode, as first tx
    // that wrote it is not visible (was aborted on conflict).
    Aggregation agg1 = new DefaultAggregation(ImmutableList.of("dim1", "dim2", "dim3"));
    int resolution = 1;
    Cube cube1 = getCubeInternal("concurrCube", new int[] { resolution }, ImmutableMap.of("agg1", agg1));
    Cube cube2 = getCubeInternal("concurrCube", new int[] { resolution }, ImmutableMap.of("agg1", agg1));
    Configuration txConf = HBaseConfiguration.create();
    TransactionManager txManager = new TransactionManager(txConf);
    txManager.startAndWait();
    try {
        TransactionSystemClient txClient = new InMemoryTxSystemClient(txManager);
        // 1) write and abort after commit to simlate conflict
        Transaction tx = txClient.startShort();
        ((TransactionAware) cube1).startTx(tx);
        writeInc(cube1, "metric1", 1, 1, "1", "1", "1");
        ((TransactionAware) cube1).commitTx();
        txClient.abort(tx);
        ((TransactionAware) cube1).rollbackTx();
        // 2) write successfully
        tx = txClient.startShort();
        ((TransactionAware) cube1).startTx(tx);
        writeInc(cube1, "metric1", 1, 1, "1", "1", "1");
        // let's pretend we had conflict and rollback it
        ((TransactionAware) cube1).commitTx();
        txClient.commitOrThrow(tx);
        ((TransactionAware) cube1).postTxCommit();
        // 3) read using different cube instance
        tx = txClient.startShort();
        ((TransactionAware) cube2).startTx(tx);
        verifyCountQuery(cube2, 0, 2, resolution, "metric1", AggregationFunction.SUM, new HashMap<String, String>(), new ArrayList<String>(), ImmutableList.of(new TimeSeries("metric1", new HashMap<String, String>(), timeValues(1, 1))));
        // let's pretend we had conflict and rollback it
        ((TransactionAware) cube2).commitTx();
        txClient.commitOrThrow(tx);
        ((TransactionAware) cube2).postTxCommit();
    } finally {
        txManager.stopAndWait();
    }
}