Examples with TextInputFormat org.apache.flink.api.java.io.TextInputFormat used on opensource projects

Example 6 with TextInputFormat

use of org.apache.flink.api.java.io.TextInputFormat in project flink by apache.

the class ContinuousFileProcessingTest method testFunctionRestore.

@Test
public void testFunctionRestore() throws Exception {
    String testBasePath = hdfsURI + "/" + UUID.randomUUID() + "/";
    org.apache.hadoop.fs.Path path = null;
    long fileModTime = Long.MIN_VALUE;
    for (int i = 0; i < 1; i++) {
        Tuple2<org.apache.hadoop.fs.Path, String> file = createFileAndFillWithData(testBasePath, "file", i, "This is test line.");
        path = file.f0;
        fileModTime = hdfs.getFileStatus(file.f0).getModificationTime();
    }
    TextInputFormat format = new TextInputFormat(new Path(testBasePath));
    final ContinuousFileMonitoringFunction<String> monitoringFunction = createTestContinuousFileMonitoringFunction(format, FileProcessingMode.PROCESS_CONTINUOUSLY);
    StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> src = new StreamSource<>(monitoringFunction);
    final AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0);
    testHarness.open();
    final Throwable[] error = new Throwable[1];
    final OneShotLatch latch = new OneShotLatch();
    final DummySourceContext sourceContext = new DummySourceContext() {

        @Override
        public void collect(TimestampedFileInputSplit element) {
            latch.trigger();
        }
    };
    // run the source asynchronously
    Thread runner = new Thread() {

        @Override
        public void run() {
            try {
                monitoringFunction.run(sourceContext);
            } catch (Throwable t) {
                t.printStackTrace();
                error[0] = t;
            }
        }
    };
    runner.start();
    // first condition for the source to have updated its state: emit at least one element
    if (!latch.isTriggered()) {
        latch.await();
    }
    // this means it has processed all the splits and updated its state.
    synchronized (sourceContext.getCheckpointLock()) {
    }
    OperatorSubtaskState snapshot = testHarness.snapshot(0, 0);
    monitoringFunction.cancel();
    runner.join();
    testHarness.close();
    final ContinuousFileMonitoringFunction<String> monitoringFunctionCopy = createTestContinuousFileMonitoringFunction(format, FileProcessingMode.PROCESS_CONTINUOUSLY);
    StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> srcCopy = new StreamSource<>(monitoringFunctionCopy);
    AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarnessCopy = new AbstractStreamOperatorTestHarness<>(srcCopy, 1, 1, 0);
    testHarnessCopy.initializeState(snapshot);
    testHarnessCopy.open();
    Assert.assertNull(error[0]);
    Assert.assertEquals(fileModTime, monitoringFunctionCopy.getGlobalModificationTime());
    hdfs.delete(path, false);
}

Example 7 with TextInputFormat

use of org.apache.flink.api.java.io.TextInputFormat in project flink by apache.

the class ContinuousFileProcessingTest method testProcessOnce.

@Test
public void testProcessOnce() throws Exception {
    String testBasePath = hdfsURI + "/" + UUID.randomUUID() + "/";
    final OneShotLatch latch = new OneShotLatch();
    // create a single file in the directory
    Tuple2<org.apache.hadoop.fs.Path, String> bootstrap = createFileAndFillWithData(testBasePath, "file", NO_OF_FILES + 1, "This is test line.");
    Assert.assertTrue(hdfs.exists(bootstrap.f0));
    // the source is supposed to read only this file.
    final Set<String> filesToBeRead = new TreeSet<>();
    filesToBeRead.add(bootstrap.f0.getName());
    TextInputFormat format = new TextInputFormat(new Path(testBasePath));
    format.setFilesFilter(FilePathFilter.createDefaultFilter());
    final ContinuousFileMonitoringFunction<String> monitoringFunction = createTestContinuousFileMonitoringFunction(format, FileProcessingMode.PROCESS_ONCE);
    final FileVerifyingSourceContext context = new FileVerifyingSourceContext(latch, monitoringFunction);
    final Thread t = new Thread() {

        @Override
        public void run() {
            try {
                monitoringFunction.open(new Configuration());
                monitoringFunction.run(context);
                // we would never arrive here if we were in
                // PROCESS_CONTINUOUSLY mode.
                // this will trigger the latch
                context.close();
            } catch (Exception e) {
                Assert.fail(e.getMessage());
            }
        }
    };
    t.start();
    if (!latch.isTriggered()) {
        latch.await();
    }
    // create some additional files that should be processed in the case of PROCESS_CONTINUOUSLY
    final org.apache.hadoop.fs.Path[] filesCreated = new org.apache.hadoop.fs.Path[NO_OF_FILES];
    for (int i = 0; i < NO_OF_FILES; i++) {
        Tuple2<org.apache.hadoop.fs.Path, String> ignoredFile = createFileAndFillWithData(testBasePath, "file", i, "This is test line.");
        filesCreated[i] = ignoredFile.f0;
    }
    // wait until the monitoring thread exits
    t.join();
    Assert.assertArrayEquals(filesToBeRead.toArray(), context.getSeenFiles().toArray());
    // finally delete the files created for the test.
    hdfs.delete(bootstrap.f0, false);
    for (org.apache.hadoop.fs.Path path : filesCreated) {
        hdfs.delete(path, false);
    }
}

Example 8 with TextInputFormat

use of org.apache.flink.api.java.io.TextInputFormat in project flink by apache.

the class ContinuousFileProcessingTest method testNestedFilesProcessing.

@Test
public void testNestedFilesProcessing() throws Exception {
    String testBasePath = hdfsURI + "/" + UUID.randomUUID() + "/";
    final Set<org.apache.hadoop.fs.Path> filesCreated = new HashSet<>();
    final Set<String> filesToBeRead = new TreeSet<>();
    // create two nested directories
    org.apache.hadoop.fs.Path firstLevelDir = new org.apache.hadoop.fs.Path(testBasePath + "/" + "firstLevelDir");
    org.apache.hadoop.fs.Path secondLevelDir = new org.apache.hadoop.fs.Path(testBasePath + "/" + "firstLevelDir" + "/" + "secondLevelDir");
    Assert.assertFalse(hdfs.exists(firstLevelDir));
    hdfs.mkdirs(firstLevelDir);
    hdfs.mkdirs(secondLevelDir);
    // create files in the base dir, the first level dir and the second level dir
    for (int i = 0; i < NO_OF_FILES; i++) {
        Tuple2<org.apache.hadoop.fs.Path, String> file = createFileAndFillWithData(testBasePath, "firstLevelFile", i, "This is test line.");
        filesCreated.add(file.f0);
        filesToBeRead.add(file.f0.getName());
    }
    for (int i = 0; i < NO_OF_FILES; i++) {
        Tuple2<org.apache.hadoop.fs.Path, String> file = createFileAndFillWithData(firstLevelDir.toString(), "secondLevelFile", i, "This is test line.");
        filesCreated.add(file.f0);
        filesToBeRead.add(file.f0.getName());
    }
    for (int i = 0; i < NO_OF_FILES; i++) {
        Tuple2<org.apache.hadoop.fs.Path, String> file = createFileAndFillWithData(secondLevelDir.toString(), "thirdLevelFile", i, "This is test line.");
        filesCreated.add(file.f0);
        filesToBeRead.add(file.f0.getName());
    }
    TextInputFormat format = new TextInputFormat(new Path(testBasePath));
    format.setFilesFilter(FilePathFilter.createDefaultFilter());
    format.setNestedFileEnumeration(true);
    ContinuousFileMonitoringFunction<String> monitoringFunction = createTestContinuousFileMonitoringFunction(format, FileProcessingMode.PROCESS_ONCE);
    final FileVerifyingSourceContext context = new FileVerifyingSourceContext(new OneShotLatch(), monitoringFunction);
    monitoringFunction.open(new Configuration());
    monitoringFunction.run(context);
    Assert.assertArrayEquals(filesToBeRead.toArray(), context.getSeenFiles().toArray());
    // finally delete the dirs and the files created for the test.
    for (org.apache.hadoop.fs.Path file : filesCreated) {
        hdfs.delete(file, false);
    }
    hdfs.delete(secondLevelDir, false);
    hdfs.delete(firstLevelDir, false);
}

Example 9 with TextInputFormat

use of org.apache.flink.api.java.io.TextInputFormat in project flink by apache.

the class ContinuousFileProcessingMigrationTest method writeMonitoringSourceSnapshot.

/**
 * Manually run this to write binary snapshot data. Remove @Ignore to run.
 */
@Ignore
@Test
public void writeMonitoringSourceSnapshot() throws Exception {
    File testFolder = tempFolder.newFolder();
    long fileModTime = Long.MIN_VALUE;
    for (int i = 0; i < 1; i++) {
        Tuple2<File, String> file = createFileAndFillWithData(testFolder, "file", i, "This is test line.");
        fileModTime = file.f0.lastModified();
    }
    TextInputFormat format = new TextInputFormat(new Path(testFolder.getAbsolutePath()));
    final ContinuousFileMonitoringFunction<String> monitoringFunction = new ContinuousFileMonitoringFunction<>(format, FileProcessingMode.PROCESS_CONTINUOUSLY, 1, INTERVAL);
    StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> src = new StreamSource<>(monitoringFunction);
    final AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0);
    testHarness.open();
    final Throwable[] error = new Throwable[1];
    final OneShotLatch latch = new OneShotLatch();
    // run the source asynchronously
    Thread runner = new Thread() {

        @Override
        public void run() {
            try {
                monitoringFunction.run(new DummySourceContext() {

                    @Override
                    public void collect(TimestampedFileInputSplit element) {
                        latch.trigger();
                    }

                    @Override
                    public void markAsTemporarilyIdle() {
                    }
                });
            } catch (Throwable t) {
                t.printStackTrace();
                error[0] = t;
            }
        }
    };
    runner.start();
    if (!latch.isTriggered()) {
        latch.await();
    }
    final OperatorSubtaskState snapshot;
    synchronized (testHarness.getCheckpointLock()) {
        snapshot = testHarness.snapshot(0L, 0L);
    }
    OperatorSnapshotUtil.writeStateHandle(snapshot, "src/test/resources/monitoring-function-migration-test-" + fileModTime + "-flink" + flinkGenerateSavepointVersion + "-snapshot");
    monitoringFunction.cancel();
    runner.join();
    testHarness.close();
}

Example 10 with TextInputFormat

use of org.apache.flink.api.java.io.TextInputFormat in project flink by apache.

the class ContinuousFileProcessingMigrationTest method testMonitoringSourceRestore.

@Test
public void testMonitoringSourceRestore() throws Exception {
    File testFolder = tempFolder.newFolder();
    TextInputFormat format = new TextInputFormat(new Path(testFolder.getAbsolutePath()));
    final ContinuousFileMonitoringFunction<String> monitoringFunction = new ContinuousFileMonitoringFunction<>(format, FileProcessingMode.PROCESS_CONTINUOUSLY, 1, INTERVAL);
    StreamSource<TimestampedFileInputSplit, ContinuousFileMonitoringFunction<String>> src = new StreamSource<>(monitoringFunction);
    final AbstractStreamOperatorTestHarness<TimestampedFileInputSplit> testHarness = new AbstractStreamOperatorTestHarness<>(src, 1, 1, 0);
    testHarness.setup();
    testHarness.initializeState(OperatorSnapshotUtil.getResourceFilename("monitoring-function-migration-test-" + expectedModTime + "-flink" + testMigrateVersion + "-snapshot"));
    testHarness.open();
    Assert.assertEquals((long) expectedModTime, monitoringFunction.getGlobalModificationTime());
}