Search in sources :

Example 21 with TupleTypeInfo

use of org.apache.flink.api.java.typeutils.TupleTypeInfo in project flink by apache.

the class LargeRecordHandlerITCase method fileTest.

@Test
public void fileTest() {
    final IOManager ioMan = new IOManagerAsync();
    final int PAGE_SIZE = 4 * 1024;
    final int NUM_PAGES = 4;
    final int NUM_RECORDS = 10;
    FileIOChannel.ID channel = null;
    try {
        final MemoryManager memMan = new MemoryManager(NUM_PAGES * PAGE_SIZE, 1, PAGE_SIZE, MemoryType.HEAP, true);
        final AbstractInvokable owner = new DummyInvokable();
        final List<MemorySegment> memory = memMan.allocatePages(owner, NUM_PAGES);
        final TypeInformation<?>[] types = new TypeInformation<?>[] { BasicTypeInfo.LONG_TYPE_INFO, new ValueTypeInfo<SomeVeryLongValue>(SomeVeryLongValue.class), BasicTypeInfo.BYTE_TYPE_INFO };
        final TupleTypeInfo<Tuple3<Long, SomeVeryLongValue, Byte>> typeInfo = new TupleTypeInfo<Tuple3<Long, SomeVeryLongValue, Byte>>(types);
        final TypeSerializer<Tuple3<Long, SomeVeryLongValue, Byte>> serializer = typeInfo.createSerializer(new ExecutionConfig());
        channel = ioMan.createChannel();
        FileChannelOutputView out = new FileChannelOutputView(ioMan.createBlockChannelWriter(channel), memMan, memory, PAGE_SIZE);
        // add the test data
        Random rnd = new Random();
        List<Long> offsets = new ArrayList<Long>();
        for (int i = 0; i < NUM_RECORDS; i++) {
            offsets.add(out.getWriteOffset());
            long val = rnd.nextLong();
            Tuple3<Long, SomeVeryLongValue, Byte> next = new Tuple3<Long, SomeVeryLongValue, Byte>(val, new SomeVeryLongValue((int) val), (byte) val);
            serializer.serialize(next, out);
        }
        out.close();
        for (int i = 1; i < offsets.size(); i++) {
            assertTrue(offsets.get(i) > offsets.get(i - 1));
        }
        memMan.allocatePages(owner, memory, NUM_PAGES);
        SeekableFileChannelInputView in = new SeekableFileChannelInputView(ioMan, channel, memMan, memory, out.getBytesInLatestSegment());
        for (int i = 0; i < NUM_RECORDS; i++) {
            in.seek(offsets.get(i));
            Tuple3<Long, SomeVeryLongValue, Byte> next = serializer.deserialize(in);
            // key and value must be equal
            assertTrue(next.f0.intValue() == next.f1.val());
            assertTrue(next.f0.byteValue() == next.f2);
        }
        in.closeAndDelete();
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    } finally {
        if (channel != null) {
            try {
                ioMan.deleteChannel(channel);
            } catch (IOException ignored) {
            }
        }
        ioMan.shutdown();
    }
}
Also used : ArrayList(java.util.ArrayList) FileIOChannel(org.apache.flink.runtime.io.disk.iomanager.FileIOChannel) ExecutionConfig(org.apache.flink.api.common.ExecutionConfig) AbstractInvokable(org.apache.flink.runtime.jobgraph.tasks.AbstractInvokable) TypeInformation(org.apache.flink.api.common.typeinfo.TypeInformation) IOManagerAsync(org.apache.flink.runtime.io.disk.iomanager.IOManagerAsync) Random(java.util.Random) DummyInvokable(org.apache.flink.runtime.operators.testutils.DummyInvokable) ValueTypeInfo(org.apache.flink.api.java.typeutils.ValueTypeInfo) SeekableFileChannelInputView(org.apache.flink.runtime.io.disk.SeekableFileChannelInputView) FileChannelOutputView(org.apache.flink.runtime.io.disk.FileChannelOutputView) IOManager(org.apache.flink.runtime.io.disk.iomanager.IOManager) IOException(java.io.IOException) MemoryManager(org.apache.flink.runtime.memory.MemoryManager) MemorySegment(org.apache.flink.core.memory.MemorySegment) TupleTypeInfo(org.apache.flink.api.java.typeutils.TupleTypeInfo) IOException(java.io.IOException) Tuple3(org.apache.flink.api.java.tuple.Tuple3) Test(org.junit.Test)

Example 22 with TupleTypeInfo

use of org.apache.flink.api.java.typeutils.TupleTypeInfo in project flink by apache.

the class LargeRecordHandlerITCase method testRecordHandlerCompositeKey.

@Test
public void testRecordHandlerCompositeKey() {
    final IOManager ioMan = new IOManagerAsync();
    final int PAGE_SIZE = 4 * 1024;
    final int NUM_PAGES = 1000;
    final int NUM_RECORDS = 10;
    try {
        final MemoryManager memMan = new MemoryManager(NUM_PAGES * PAGE_SIZE, 1, PAGE_SIZE, MemoryType.HEAP, true);
        final AbstractInvokable owner = new DummyInvokable();
        final List<MemorySegment> initialMemory = memMan.allocatePages(owner, 6);
        final List<MemorySegment> sortMemory = memMan.allocatePages(owner, NUM_PAGES - 6);
        final TypeInformation<?>[] types = new TypeInformation<?>[] { BasicTypeInfo.LONG_TYPE_INFO, new ValueTypeInfo<SomeVeryLongValue>(SomeVeryLongValue.class), BasicTypeInfo.BYTE_TYPE_INFO };
        final TupleTypeInfo<Tuple3<Long, SomeVeryLongValue, Byte>> typeInfo = new TupleTypeInfo<Tuple3<Long, SomeVeryLongValue, Byte>>(types);
        final TypeSerializer<Tuple3<Long, SomeVeryLongValue, Byte>> serializer = typeInfo.createSerializer(new ExecutionConfig());
        final TypeComparator<Tuple3<Long, SomeVeryLongValue, Byte>> comparator = typeInfo.createComparator(new int[] { 2, 0 }, new boolean[] { true, true }, 0, new ExecutionConfig());
        LargeRecordHandler<Tuple3<Long, SomeVeryLongValue, Byte>> handler = new LargeRecordHandler<Tuple3<Long, SomeVeryLongValue, Byte>>(serializer, comparator, ioMan, memMan, initialMemory, owner, 128);
        assertFalse(handler.hasData());
        // add the test data
        Random rnd = new Random();
        for (int i = 0; i < NUM_RECORDS; i++) {
            long val = rnd.nextLong();
            handler.addRecord(new Tuple3<Long, SomeVeryLongValue, Byte>(val, new SomeVeryLongValue((int) val), (byte) val));
            assertTrue(handler.hasData());
        }
        MutableObjectIterator<Tuple3<Long, SomeVeryLongValue, Byte>> sorted = handler.finishWriteAndSortKeys(sortMemory);
        try {
            handler.addRecord(new Tuple3<Long, SomeVeryLongValue, Byte>(92L, null, (byte) 1));
            fail("should throw an exception");
        } catch (IllegalStateException e) {
        // expected
        }
        Tuple3<Long, SomeVeryLongValue, Byte> previous = null;
        Tuple3<Long, SomeVeryLongValue, Byte> next;
        while ((next = sorted.next(null)) != null) {
            // key and value must be equal
            assertTrue(next.f0.intValue() == next.f1.val());
            assertTrue(next.f0.byteValue() == next.f2);
            // order must be correct
            if (previous != null) {
                assertTrue(previous.f2 <= next.f2);
                assertTrue(previous.f2.byteValue() != next.f2.byteValue() || previous.f0 <= next.f0);
            }
            previous = next;
        }
        handler.close();
        assertFalse(handler.hasData());
        handler.close();
        try {
            handler.addRecord(new Tuple3<Long, SomeVeryLongValue, Byte>(92L, null, (byte) 1));
            fail("should throw an exception");
        } catch (IllegalStateException e) {
        // expected
        }
        assertTrue(memMan.verifyEmpty());
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    } finally {
        ioMan.shutdown();
    }
}
Also used : ExecutionConfig(org.apache.flink.api.common.ExecutionConfig) AbstractInvokable(org.apache.flink.runtime.jobgraph.tasks.AbstractInvokable) TypeInformation(org.apache.flink.api.common.typeinfo.TypeInformation) IOManagerAsync(org.apache.flink.runtime.io.disk.iomanager.IOManagerAsync) Random(java.util.Random) DummyInvokable(org.apache.flink.runtime.operators.testutils.DummyInvokable) ValueTypeInfo(org.apache.flink.api.java.typeutils.ValueTypeInfo) IOManager(org.apache.flink.runtime.io.disk.iomanager.IOManager) MemoryManager(org.apache.flink.runtime.memory.MemoryManager) MemorySegment(org.apache.flink.core.memory.MemorySegment) TupleTypeInfo(org.apache.flink.api.java.typeutils.TupleTypeInfo) IOException(java.io.IOException) Tuple3(org.apache.flink.api.java.tuple.Tuple3) Test(org.junit.Test)

Example 23 with TupleTypeInfo

use of org.apache.flink.api.java.typeutils.TupleTypeInfo in project flink by apache.

the class LargeRecordHandlerTest method testRecordHandlerCompositeKey.

@Test
public void testRecordHandlerCompositeKey() {
    final IOManager ioMan = new IOManagerAsync();
    final int PAGE_SIZE = 4 * 1024;
    final int NUM_PAGES = 24;
    final int NUM_RECORDS = 25000;
    try {
        final MemoryManager memMan = new MemoryManager(NUM_PAGES * PAGE_SIZE, 1, PAGE_SIZE, MemoryType.HEAP, true);
        final AbstractInvokable owner = new DummyInvokable();
        final List<MemorySegment> initialMemory = memMan.allocatePages(owner, 6);
        final List<MemorySegment> sortMemory = memMan.allocatePages(owner, NUM_PAGES - 6);
        final TupleTypeInfo<Tuple3<Long, String, Byte>> typeInfo = (TupleTypeInfo<Tuple3<Long, String, Byte>>) TypeInfoParser.<Tuple3<Long, String, Byte>>parse("Tuple3<Long, String, Byte>");
        final TypeSerializer<Tuple3<Long, String, Byte>> serializer = typeInfo.createSerializer(new ExecutionConfig());
        final TypeComparator<Tuple3<Long, String, Byte>> comparator = typeInfo.createComparator(new int[] { 2, 0 }, new boolean[] { true, true }, 0, new ExecutionConfig());
        LargeRecordHandler<Tuple3<Long, String, Byte>> handler = new LargeRecordHandler<Tuple3<Long, String, Byte>>(serializer, comparator, ioMan, memMan, initialMemory, owner, 128);
        assertFalse(handler.hasData());
        // add the test data
        Random rnd = new Random();
        for (int i = 0; i < NUM_RECORDS; i++) {
            long val = rnd.nextLong();
            handler.addRecord(new Tuple3<Long, String, Byte>(val, String.valueOf(val), (byte) val));
            assertTrue(handler.hasData());
        }
        MutableObjectIterator<Tuple3<Long, String, Byte>> sorted = handler.finishWriteAndSortKeys(sortMemory);
        try {
            handler.addRecord(new Tuple3<Long, String, Byte>(92L, "peter pepper", (byte) 1));
            fail("should throw an exception");
        } catch (IllegalStateException e) {
        // expected
        }
        Tuple3<Long, String, Byte> previous = null;
        Tuple3<Long, String, Byte> next;
        while ((next = sorted.next(null)) != null) {
            // key and value must be equal
            assertTrue(next.f0.equals(Long.parseLong(next.f1)));
            assertTrue(next.f0.byteValue() == next.f2);
            // order must be correct
            if (previous != null) {
                assertTrue(previous.f2 <= next.f2);
                assertTrue(previous.f2.byteValue() != next.f2.byteValue() || previous.f0 <= next.f0);
            }
            previous = next;
        }
        handler.close();
        assertFalse(handler.hasData());
        handler.close();
        try {
            handler.addRecord(new Tuple3<Long, String, Byte>(92L, "peter pepper", (byte) 1));
            fail("should throw an exception");
        } catch (IllegalStateException e) {
        // expected
        }
        assertTrue(memMan.verifyEmpty());
    } catch (Exception e) {
        e.printStackTrace();
        fail(e.getMessage());
    } finally {
        ioMan.shutdown();
    }
}
Also used : ExecutionConfig(org.apache.flink.api.common.ExecutionConfig) AbstractInvokable(org.apache.flink.runtime.jobgraph.tasks.AbstractInvokable) IOManagerAsync(org.apache.flink.runtime.io.disk.iomanager.IOManagerAsync) Random(java.util.Random) DummyInvokable(org.apache.flink.runtime.operators.testutils.DummyInvokable) IOManager(org.apache.flink.runtime.io.disk.iomanager.IOManager) MemoryManager(org.apache.flink.runtime.memory.MemoryManager) MemorySegment(org.apache.flink.core.memory.MemorySegment) TupleTypeInfo(org.apache.flink.api.java.typeutils.TupleTypeInfo) Tuple3(org.apache.flink.api.java.tuple.Tuple3) Test(org.junit.Test)

Example 24 with TupleTypeInfo

use of org.apache.flink.api.java.typeutils.TupleTypeInfo in project flink by apache.

the class HCatInputFormatBase method asFlinkTuples.

/**
	 * Specifies that the InputFormat returns Flink tuples instead of
	 * {@link org.apache.hive.hcatalog.data.HCatRecord}.
	 *
	 * Note: Flink tuples might only support a limited number of fields (depending on the API).
	 *
	 * @return This InputFormat.
	 * @throws org.apache.hive.hcatalog.common.HCatException
	 */
public HCatInputFormatBase<T> asFlinkTuples() throws HCatException {
    // build type information
    int numFields = outputSchema.getFields().size();
    if (numFields > this.getMaxFlinkTupleSize()) {
        throw new IllegalArgumentException("Only up to " + this.getMaxFlinkTupleSize() + " fields can be returned as Flink tuples.");
    }
    TypeInformation[] fieldTypes = new TypeInformation[numFields];
    fieldNames = new String[numFields];
    for (String fieldName : outputSchema.getFieldNames()) {
        HCatFieldSchema field = outputSchema.get(fieldName);
        int fieldPos = outputSchema.getPosition(fieldName);
        TypeInformation fieldType = getFieldType(field);
        fieldTypes[fieldPos] = fieldType;
        fieldNames[fieldPos] = fieldName;
    }
    this.resultType = new TupleTypeInfo(fieldTypes);
    return this;
}
Also used : TypeInformation(org.apache.flink.api.common.typeinfo.TypeInformation) TupleTypeInfo(org.apache.flink.api.java.typeutils.TupleTypeInfo) HCatFieldSchema(org.apache.hive.hcatalog.data.schema.HCatFieldSchema)

Example 25 with TupleTypeInfo

use of org.apache.flink.api.java.typeutils.TupleTypeInfo in project flink by apache.

the class ExpressionKeysTest method testAreCompatible9.

@Test
public void testAreCompatible9() throws Keys.IncompatibleKeysException {
    TypeInformation<Tuple3<String, Long, Integer>> t1 = new TupleTypeInfo<>(BasicTypeInfo.STRING_TYPE_INFO, BasicTypeInfo.LONG_TYPE_INFO, BasicTypeInfo.INT_TYPE_INFO);
    TypeInformation<PojoWithMultiplePojos> t2 = TypeExtractor.getForClass(PojoWithMultiplePojos.class);
    ExpressionKeys<Tuple3<String, Long, Integer>> ek1 = new ExpressionKeys<>(new int[] { 2, 0 }, t1);
    Keys<PojoWithMultiplePojos> ek2 = new Keys.SelectorFunctionKeys<>(new KeySelector3(), t2, new TupleTypeInfo<Tuple2<Integer, String>>(BasicTypeInfo.INT_TYPE_INFO, BasicTypeInfo.STRING_TYPE_INFO));
    Assert.assertTrue(ek1.areCompatible(ek2));
}
Also used : KeySelector3(org.apache.flink.api.common.operators.SelectorFunctionKeysTest.KeySelector3) ExpressionKeys(org.apache.flink.api.common.operators.Keys.ExpressionKeys) Tuple2(org.apache.flink.api.java.tuple.Tuple2) Tuple3(org.apache.flink.api.java.tuple.Tuple3) TupleTypeInfo(org.apache.flink.api.java.typeutils.TupleTypeInfo) Test(org.junit.Test)

Aggregations

TupleTypeInfo (org.apache.flink.api.java.typeutils.TupleTypeInfo)40 Test (org.junit.Test)25 Tuple2 (org.apache.flink.api.java.tuple.Tuple2)24 ExecutionConfig (org.apache.flink.api.common.ExecutionConfig)20 TypeInformation (org.apache.flink.api.common.typeinfo.TypeInformation)10 StringValue (org.apache.flink.types.StringValue)9 IOException (java.io.IOException)8 Random (java.util.Random)8 Tuple3 (org.apache.flink.api.java.tuple.Tuple3)8 IOManager (org.apache.flink.runtime.io.disk.iomanager.IOManager)7 IOManagerAsync (org.apache.flink.runtime.io.disk.iomanager.IOManagerAsync)7 MemoryManager (org.apache.flink.runtime.memory.MemoryManager)7 DummyInvokable (org.apache.flink.runtime.operators.testutils.DummyInvokable)7 ArrayList (java.util.ArrayList)6 ValueTypeInfo (org.apache.flink.api.java.typeutils.ValueTypeInfo)6 GroupReduceFunction (org.apache.flink.api.common.functions.GroupReduceFunction)5 RichGroupReduceFunction (org.apache.flink.api.common.functions.RichGroupReduceFunction)5 MemorySegment (org.apache.flink.core.memory.MemorySegment)5 AbstractInvokable (org.apache.flink.runtime.jobgraph.tasks.AbstractInvokable)5 IntValue (org.apache.flink.types.IntValue)5