Examples with MatchRemovingJoiner org.apache.flink.runtime.operators.testutils.MatchRemovingJoiner used on opensource projects

Example 1 with MatchRemovingJoiner

use of org.apache.flink.runtime.operators.testutils.MatchRemovingJoiner in project flink by apache.

the class NonReusingSortMergeInnerJoinIteratorITCase method testMergeWithHighNumberOfCommonKeys.

@Test
public void testMergeWithHighNumberOfCommonKeys() {
    // the size of the left and right inputs
    final int INPUT_1_SIZE = 200;
    final int INPUT_2_SIZE = 100;
    final int INPUT_1_DUPLICATES = 10;
    final int INPUT_2_DUPLICATES = 4000;
    final int DUPLICATE_KEY = 13;
    try {
        final TupleGenerator generator1 = new TupleGenerator(SEED1, 500, 4096, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TupleGenerator generator2 = new TupleGenerator(SEED2, 500, 2048, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TestData.TupleGeneratorIterator gen1Iter = new TestData.TupleGeneratorIterator(generator1, INPUT_1_SIZE);
        final TestData.TupleGeneratorIterator gen2Iter = new TestData.TupleGeneratorIterator(generator2, INPUT_2_SIZE);
        final TestData.TupleConstantValueIterator const1Iter = new TestData.TupleConstantValueIterator(DUPLICATE_KEY, "LEFT String for Duplicate Keys", INPUT_1_DUPLICATES);
        final TestData.TupleConstantValueIterator const2Iter = new TestData.TupleConstantValueIterator(DUPLICATE_KEY, "RIGHT String for Duplicate Keys", INPUT_2_DUPLICATES);
        final List<MutableObjectIterator<Tuple2<Integer, String>>> inList1 = new ArrayList<MutableObjectIterator<Tuple2<Integer, String>>>();
        inList1.add(gen1Iter);
        inList1.add(const1Iter);
        final List<MutableObjectIterator<Tuple2<Integer, String>>> inList2 = new ArrayList<MutableObjectIterator<Tuple2<Integer, String>>>();
        inList2.add(gen2Iter);
        inList2.add(const2Iter);
        MutableObjectIterator<Tuple2<Integer, String>> input1 = new MergeIterator<Tuple2<Integer, String>>(inList1, comparator1.duplicate());
        MutableObjectIterator<Tuple2<Integer, String>> input2 = new MergeIterator<Tuple2<Integer, String>>(inList2, comparator2.duplicate());
        // collect expected data
        final Map<Integer, Collection<Match>> expectedMatchesMap = matchValues(collectData(input1), collectData(input2));
        // re-create the whole thing for actual processing
        // reset the generators and iterators
        generator1.reset();
        generator2.reset();
        const1Iter.reset();
        const2Iter.reset();
        gen1Iter.reset();
        gen2Iter.reset();
        inList1.clear();
        inList1.add(gen1Iter);
        inList1.add(const1Iter);
        inList2.clear();
        inList2.add(gen2Iter);
        inList2.add(const2Iter);
        input1 = new MergeIterator<Tuple2<Integer, String>>(inList1, comparator1.duplicate());
        input2 = new MergeIterator<Tuple2<Integer, String>>(inList2, comparator2.duplicate());
        final FlatJoinFunction<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> joinFunction = new MatchRemovingJoiner(expectedMatchesMap);
        final Collector<Tuple2<Integer, String>> collector = new DiscardingOutputCollector<Tuple2<Integer, String>>();
        // we create this sort-merge iterator with little memory for the block-nested-loops
        // fall-back to make sure it
        // needs to spill for the duplicate keys
        NonReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> iterator = new NonReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>>(input1, input2, this.serializer1, this.comparator1, this.serializer2, this.comparator2, this.pairComparator, this.memoryManager, this.ioManager, PAGES_FOR_BNLJN, this.parentTask);
        iterator.open();
        while (iterator.callWithNextKey(joinFunction, collector)) ;
        iterator.close();
        // assert that each expected match was seen
        for (Entry<Integer, Collection<Match>> entry : expectedMatchesMap.entrySet()) {
            if (!entry.getValue().isEmpty()) {
                Assert.fail("Collection for key " + entry.getKey() + " is not empty");
            }
        }
    } catch (Exception e) {
        e.printStackTrace();
        Assert.fail("An exception occurred during the test: " + e.getMessage());
    }
}

Example 2 with MatchRemovingJoiner

use of org.apache.flink.runtime.operators.testutils.MatchRemovingJoiner in project flink by apache.

the class ReusingSortMergeInnerJoinIteratorITCase method testMerge.

@Test
public void testMerge() {
    try {
        final TupleGenerator generator1 = new TupleGenerator(SEED1, 500, 4096, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TupleGenerator generator2 = new TupleGenerator(SEED2, 500, 2048, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TestData.TupleGeneratorIterator input1 = new TestData.TupleGeneratorIterator(generator1, INPUT_1_SIZE);
        final TestData.TupleGeneratorIterator input2 = new TestData.TupleGeneratorIterator(generator2, INPUT_2_SIZE);
        // collect expected data
        final Map<Integer, Collection<Match>> expectedMatchesMap = matchValues(collectData(input1), collectData(input2));
        final FlatJoinFunction<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> joinFunction = new MatchRemovingJoiner(expectedMatchesMap);
        final Collector<Tuple2<Integer, String>> collector = new DiscardingOutputCollector<Tuple2<Integer, String>>();
        // reset the generators
        generator1.reset();
        generator2.reset();
        input1.reset();
        input2.reset();
        // compare with iterator values
        ReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> iterator = new ReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>>(input1, input2, this.serializer1, this.comparator1, this.serializer2, this.comparator2, this.pairComparator, this.memoryManager, this.ioManager, PAGES_FOR_BNLJN, this.parentTask);
        iterator.open();
        while (iterator.callWithNextKey(joinFunction, collector)) ;
        iterator.close();
        // assert that each expected match was seen
        for (Entry<Integer, Collection<Match>> entry : expectedMatchesMap.entrySet()) {
            Assert.assertTrue("Collection for key " + entry.getKey() + " is not empty", entry.getValue().isEmpty());
        }
    } catch (Exception e) {
        e.printStackTrace();
        Assert.fail("An exception occurred during the test: " + e.getMessage());
    }
}

Example 3 with MatchRemovingJoiner

use of org.apache.flink.runtime.operators.testutils.MatchRemovingJoiner in project flink by apache.

the class AbstractSortMergeOuterJoinIteratorITCase method testOuterJoinWithHighNumberOfCommonKeys.

@SuppressWarnings("unchecked, rawtypes")
protected void testOuterJoinWithHighNumberOfCommonKeys(OuterJoinType outerJoinType, int input1Size, int input1Duplicates, int input1ValueLength, float input1KeyDensity, int input2Size, int input2Duplicates, int input2ValueLength, float input2KeyDensity) {
    TypeSerializer<Tuple2<Integer, String>> serializer1 = new TupleSerializer<>((Class<Tuple2<Integer, String>>) (Class<?>) Tuple2.class, new TypeSerializer<?>[] { IntSerializer.INSTANCE, StringSerializer.INSTANCE });
    TypeSerializer<Tuple2<Integer, String>> serializer2 = new TupleSerializer<>((Class<Tuple2<Integer, String>>) (Class<?>) Tuple2.class, new TypeSerializer<?>[] { IntSerializer.INSTANCE, StringSerializer.INSTANCE });
    TypeComparator<Tuple2<Integer, String>> comparator1 = new TupleComparator<>(new int[] { 0 }, new TypeComparator<?>[] { new IntComparator(true) }, new TypeSerializer<?>[] { IntSerializer.INSTANCE });
    TypeComparator<Tuple2<Integer, String>> comparator2 = new TupleComparator<>(new int[] { 0 }, new TypeComparator<?>[] { new IntComparator(true) }, new TypeSerializer<?>[] { IntSerializer.INSTANCE });
    TypePairComparator<Tuple2<Integer, String>, Tuple2<Integer, String>> pairComparator = new GenericPairComparator<>(comparator1, comparator2);
    final int DUPLICATE_KEY = 13;
    try {
        final TupleGenerator generator1 = new TupleGenerator(SEED1, 500, input1KeyDensity, input1ValueLength, KeyMode.SORTED_SPARSE, ValueMode.RANDOM_LENGTH, null);
        final TupleGenerator generator2 = new TupleGenerator(SEED2, 500, input2KeyDensity, input2ValueLength, KeyMode.SORTED_SPARSE, ValueMode.RANDOM_LENGTH, null);
        final TupleGeneratorIterator gen1Iter = new TupleGeneratorIterator(generator1, input1Size);
        final TupleGeneratorIterator gen2Iter = new TupleGeneratorIterator(generator2, input2Size);
        final TupleConstantValueIterator const1Iter = new TupleConstantValueIterator(DUPLICATE_KEY, "LEFT String for Duplicate Keys", input1Duplicates);
        final TupleConstantValueIterator const2Iter = new TupleConstantValueIterator(DUPLICATE_KEY, "RIGHT String for Duplicate Keys", input2Duplicates);
        final List<MutableObjectIterator<Tuple2<Integer, String>>> inList1 = new ArrayList<>();
        inList1.add(gen1Iter);
        inList1.add(const1Iter);
        final List<MutableObjectIterator<Tuple2<Integer, String>>> inList2 = new ArrayList<>();
        inList2.add(gen2Iter);
        inList2.add(const2Iter);
        MutableObjectIterator<Tuple2<Integer, String>> input1 = new MergeIterator<>(inList1, comparator1.duplicate());
        MutableObjectIterator<Tuple2<Integer, String>> input2 = new MergeIterator<>(inList2, comparator2.duplicate());
        // collect expected data
        final Map<Integer, Collection<Match>> expectedMatchesMap = joinValues(collectData(input1), collectData(input2), outerJoinType);
        // re-create the whole thing for actual processing
        // reset the generators and iterators
        generator1.reset();
        generator2.reset();
        const1Iter.reset();
        const2Iter.reset();
        gen1Iter.reset();
        gen2Iter.reset();
        inList1.clear();
        inList1.add(gen1Iter);
        inList1.add(const1Iter);
        inList2.clear();
        inList2.add(gen2Iter);
        inList2.add(const2Iter);
        input1 = new MergeIterator<>(inList1, comparator1.duplicate());
        input2 = new MergeIterator<>(inList2, comparator2.duplicate());
        final FlatJoinFunction<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> joinFunction = new MatchRemovingJoiner(expectedMatchesMap);
        final Collector<Tuple2<Integer, String>> collector = new DiscardingOutputCollector<>();
        // we create this sort-merge iterator with little memory for the block-nested-loops
        // fall-back to make sure it
        // needs to spill for the duplicate keys
        AbstractMergeOuterJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> iterator = createOuterJoinIterator(outerJoinType, input1, input2, serializer1, comparator1, serializer2, comparator2, pairComparator, this.memoryManager, this.ioManager, PAGES_FOR_BNLJN, this.parentTask);
        iterator.open();
        while (iterator.callWithNextKey(joinFunction, collector)) ;
        iterator.close();
        // assert that each expected match was seen
        for (Entry<Integer, Collection<Match>> entry : expectedMatchesMap.entrySet()) {
            if (!entry.getValue().isEmpty()) {
                Assert.fail("Collection for key " + entry.getKey() + " is not empty");
            }
        }
    } catch (Exception e) {
        e.printStackTrace();
        Assert.fail("An exception occurred during the test: " + e.getMessage());
    }
}

Example 4 with MatchRemovingJoiner

use of org.apache.flink.runtime.operators.testutils.MatchRemovingJoiner in project flink by apache.

the class NonReusingSortMergeInnerJoinIteratorITCase method testMerge.

@Test
public void testMerge() {
    try {
        final TupleGenerator generator1 = new TupleGenerator(SEED1, 500, 4096, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TupleGenerator generator2 = new TupleGenerator(SEED2, 500, 2048, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TestData.TupleGeneratorIterator input1 = new TestData.TupleGeneratorIterator(generator1, INPUT_1_SIZE);
        final TestData.TupleGeneratorIterator input2 = new TestData.TupleGeneratorIterator(generator2, INPUT_2_SIZE);
        // collect expected data
        final Map<Integer, Collection<Match>> expectedMatchesMap = matchValues(collectData(input1), collectData(input2));
        final FlatJoinFunction<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> joinFunction = new MatchRemovingJoiner(expectedMatchesMap);
        final Collector<Tuple2<Integer, String>> collector = new DiscardingOutputCollector<Tuple2<Integer, String>>();
        // reset the generators
        generator1.reset();
        generator2.reset();
        input1.reset();
        input2.reset();
        // compare with iterator values
        NonReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> iterator = new NonReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>>(input1, input2, this.serializer1, this.comparator1, this.serializer2, this.comparator2, this.pairComparator, this.memoryManager, this.ioManager, PAGES_FOR_BNLJN, this.parentTask);
        iterator.open();
        while (iterator.callWithNextKey(joinFunction, collector)) ;
        iterator.close();
        // assert that each expected match was seen
        for (Entry<Integer, Collection<Match>> entry : expectedMatchesMap.entrySet()) {
            Assert.assertTrue("Collection for key " + entry.getKey() + " is not empty", entry.getValue().isEmpty());
        }
    } catch (Exception e) {
        e.printStackTrace();
        Assert.fail("An exception occurred during the test: " + e.getMessage());
    }
}

Example 5 with MatchRemovingJoiner

use of org.apache.flink.runtime.operators.testutils.MatchRemovingJoiner in project flink by apache.

the class ReusingSortMergeInnerJoinIteratorITCase method testMergeWithHighNumberOfCommonKeys.

@Test
public void testMergeWithHighNumberOfCommonKeys() {
    // the size of the left and right inputs
    final int INPUT_1_SIZE = 200;
    final int INPUT_2_SIZE = 100;
    final int INPUT_1_DUPLICATES = 10;
    final int INPUT_2_DUPLICATES = 4000;
    final int DUPLICATE_KEY = 13;
    try {
        final TupleGenerator generator1 = new TupleGenerator(SEED1, 500, 4096, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TupleGenerator generator2 = new TupleGenerator(SEED2, 500, 2048, KeyMode.SORTED, ValueMode.RANDOM_LENGTH);
        final TestData.TupleGeneratorIterator gen1Iter = new TestData.TupleGeneratorIterator(generator1, INPUT_1_SIZE);
        final TestData.TupleGeneratorIterator gen2Iter = new TestData.TupleGeneratorIterator(generator2, INPUT_2_SIZE);
        final TestData.TupleConstantValueIterator const1Iter = new TestData.TupleConstantValueIterator(DUPLICATE_KEY, "LEFT String for Duplicate Keys", INPUT_1_DUPLICATES);
        final TestData.TupleConstantValueIterator const2Iter = new TestData.TupleConstantValueIterator(DUPLICATE_KEY, "RIGHT String for Duplicate Keys", INPUT_2_DUPLICATES);
        final List<MutableObjectIterator<Tuple2<Integer, String>>> inList1 = new ArrayList<MutableObjectIterator<Tuple2<Integer, String>>>();
        inList1.add(gen1Iter);
        inList1.add(const1Iter);
        final List<MutableObjectIterator<Tuple2<Integer, String>>> inList2 = new ArrayList<MutableObjectIterator<Tuple2<Integer, String>>>();
        inList2.add(gen2Iter);
        inList2.add(const2Iter);
        MutableObjectIterator<Tuple2<Integer, String>> input1 = new MergeIterator<Tuple2<Integer, String>>(inList1, comparator1.duplicate());
        MutableObjectIterator<Tuple2<Integer, String>> input2 = new MergeIterator<Tuple2<Integer, String>>(inList2, comparator2.duplicate());
        // collect expected data
        final Map<Integer, Collection<Match>> expectedMatchesMap = matchValues(collectData(input1), collectData(input2));
        // re-create the whole thing for actual processing
        // reset the generators and iterators
        generator1.reset();
        generator2.reset();
        const1Iter.reset();
        const2Iter.reset();
        gen1Iter.reset();
        gen2Iter.reset();
        inList1.clear();
        inList1.add(gen1Iter);
        inList1.add(const1Iter);
        inList2.clear();
        inList2.add(gen2Iter);
        inList2.add(const2Iter);
        input1 = new MergeIterator<Tuple2<Integer, String>>(inList1, comparator1.duplicate());
        input2 = new MergeIterator<Tuple2<Integer, String>>(inList2, comparator2.duplicate());
        final FlatJoinFunction<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> matcher = new MatchRemovingJoiner(expectedMatchesMap);
        final Collector<Tuple2<Integer, String>> collector = new DiscardingOutputCollector<Tuple2<Integer, String>>();
        // we create this sort-merge iterator with little memory for the block-nested-loops
        // fall-back to make sure it
        // needs to spill for the duplicate keys
        ReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>> iterator = new ReusingMergeInnerJoinIterator<Tuple2<Integer, String>, Tuple2<Integer, String>, Tuple2<Integer, String>>(input1, input2, this.serializer1, this.comparator1, this.serializer2, this.comparator2, this.pairComparator, this.memoryManager, this.ioManager, PAGES_FOR_BNLJN, this.parentTask);
        iterator.open();
        while (iterator.callWithNextKey(matcher, collector)) ;
        iterator.close();
        // assert that each expected match was seen
        for (Entry<Integer, Collection<Match>> entry : expectedMatchesMap.entrySet()) {
            if (!entry.getValue().isEmpty()) {
                Assert.fail("Collection for key " + entry.getKey() + " is not empty");
            }
        }
    } catch (Exception e) {
        e.printStackTrace();
        Assert.fail("An exception occurred during the test: " + e.getMessage());
    }
}