@Test
@SuppressWarnings("unchecked")
public void testTop() {
Pipeline p = TestPipeline.create();
PCollection<String> input =
p.apply(Create.of(Arrays.asList(COLLECTION)).withCoder(StringUtf8Coder.of()));
PCollection<List<String>> top1 = input.apply(Top.of(1, new OrderByLength()));
PCollection<List<String>> top2 = input.apply(Top.<String>largest(2));
PCollection<List<String>> top3 = input.apply(Top.<String>smallest(3));
PCollection<KV<String, Integer>> inputTable = createInputTable(p);
PCollection<KV<String, List<Integer>>> largestPerKey =
inputTable.apply(Top.<String, Integer>largestPerKey(2));
PCollection<KV<String, List<Integer>>> smallestPerKey =
inputTable.apply(Top.<String, Integer>smallestPerKey(2));
DataflowAssert.thatSingletonIterable(top1).containsInAnyOrder(Arrays.asList("bb"));
DataflowAssert.thatSingletonIterable(top2).containsInAnyOrder("z", "c");
DataflowAssert.thatSingletonIterable(top3).containsInAnyOrder("a", "bb", "c");
DataflowAssert.that(largestPerKey)
.containsInAnyOrder(KV.of("a", Arrays.asList(3, 2)), KV.of("b", Arrays.asList(100, 10)));
DataflowAssert.that(smallestPerKey)
.containsInAnyOrder(KV.of("a", Arrays.asList(1, 2)), KV.of("b", Arrays.asList(1, 10)));
p.run();
}
@Test(dataProvider = "reads")
public void testGATKReadCoding(final List<GATKRead> reads) {
// The simplest way to figure out if a class is coded correctly is to create a PCollection
// of that type and see if it matches the List version.
final Pipeline p = GATKTestPipeline.create();
DataflowUtils.registerGATKCoders(p);
// Need to explicitly set the coder to GATKReadCoder, otherwise Create fails to infer
// a coder properly in the case where the List contains a mix of different GATKRead
// implementations.
final PCollection<GATKRead> dataflowReads =
p.apply(Create.of(reads).withCoder(new GATKReadCoder()));
DataflowAssert.that(dataflowReads).containsInAnyOrder(reads);
final PCollection<GATKRead> dataflowReadsAfterTransform =
dataflowReads
.apply(
ParDo.of(
new DoFn<GATKRead, GATKRead>() {
private static final long serialVersionUID = 1l;
@Override
public void processElement(ProcessContext c) throws Exception {
c.output(c.element());
}
}))
.setCoder(new GATKReadCoder());
DataflowAssert.that(dataflowReadsAfterTransform).containsInAnyOrder(reads);
p.run();
}
@Test
@SuppressWarnings("unchecked")
public void testTopEmpty() {
Pipeline p = TestPipeline.create();
PCollection<String> input =
p.apply(Create.of(Arrays.asList(EMPTY_COLLECTION)).withCoder(StringUtf8Coder.of()));
PCollection<List<String>> top1 = input.apply(Top.of(1, new OrderByLength()));
PCollection<List<String>> top2 = input.apply(Top.<String>largest(2));
PCollection<List<String>> top3 = input.apply(Top.<String>smallest(3));
PCollection<KV<String, Integer>> inputTable = createEmptyInputTable(p);
PCollection<KV<String, List<Integer>>> largestPerKey =
inputTable.apply(Top.<String, Integer>largestPerKey(2));
PCollection<KV<String, List<Integer>>> smallestPerKey =
inputTable.apply(Top.<String, Integer>smallestPerKey(2));
DataflowAssert.thatSingletonIterable(top1).empty();
DataflowAssert.thatSingletonIterable(top2).empty();
DataflowAssert.thatSingletonIterable(top3).empty();
DataflowAssert.that(largestPerKey).empty();
DataflowAssert.that(smallestPerKey).empty();
p.run();
}
@Override
public PDone apply(PCollection<Integer> input) {
// Apply an operation so that this is a composite transform.
input.apply(Count.<Integer>perElement());
return PDone.in(input.getPipeline());
}
@Override
public PCollectionView<Iterable<T>> apply(PCollection<T> input) {
return input.apply(
CreatePCollectionView.<T, Iterable<T>>of(
PCollectionViews.iterableView(
input.getPipeline(), input.getWindowingStrategy(), input.getCoder())));
}
@Override
public PCollection<T> apply(PInput input) {
if (filepattern == null) {
throw new IllegalStateException(
"need to set the filepattern of an AvroIO.Read transform");
}
if (schema == null) {
throw new IllegalStateException("need to set the schema of an AvroIO.Read transform");
}
if (validate) {
try {
checkState(
!IOChannelUtils.getFactory(filepattern).match(filepattern).isEmpty(),
"Unable to find any files matching %s",
filepattern);
} catch (IOException e) {
throw new IllegalStateException(String.format("Failed to validate %s", filepattern), e);
}
}
@SuppressWarnings("unchecked")
Bounded<T> read =
type == GenericRecord.class
? (Bounded<T>)
com.google.cloud.dataflow.sdk.io.Read.from(
AvroSource.from(filepattern).withSchema(schema))
: com.google.cloud.dataflow.sdk.io.Read.from(
AvroSource.from(filepattern).withSchema(type));
PCollection<T> pcol = input.getPipeline().apply("Read", read);
// Honor the default output coder that would have been used by this PTransform.
pcol.setCoder(getDefaultOutputCoder());
return pcol;
}
/**
* Applies {@code ApproximateUnique(sampleSize)} verifying that the estimation error falls within
* the maximum allowed error of {@code 2/sqrt(sampleSize)}.
*/
private void runApproximateUniquePipeline(int sampleSize) {
Pipeline p = TestPipeline.create();
PCollection<String> collection = readPCollection(p);
final PCollectionView<Long> exact =
collection
.apply(RemoveDuplicates.<String>create())
.apply(Combine.globally(new CountElements<String>()))
.apply(View.<Long>asSingleton());
PCollection<Long> approximate =
collection.apply(ApproximateUnique.<String>globally(sampleSize));
PCollection<KV<Long, Long>> approximateAndExact =
approximate.apply(
ParDo.of(
new DoFn<Long, KV<Long, Long>>() {
@Override
public void processElement(ProcessContext c) {
c.output(KV.of(c.element(), c.sideInput(exact)));
}
})
.withSideInputs(exact));
DataflowAssert.that(approximateAndExact).satisfies(new VerifyEstimatePerKeyFn(sampleSize));
p.run();
}
@Override
public PCollection<Integer> apply(PCollection<Integer> input) {
// Apply an operation so that this is a composite transform.
input.apply(Count.<Integer>perElement());
// Return a value unrelated to the input.
return input.getPipeline().apply(Create.of(1, 2, 3, 4));
}
@Override
public PCollectionTuple apply(PCollection<Integer> input) {
PCollection<Integer> sum = input.apply(Sum.integersGlobally());
// Fails here when attempting to construct a tuple with an unbound object.
return PCollectionTuple.of(sumTag, sum)
.and(
doneTag,
PCollection.<Void>createPrimitiveOutputInternal(
input.getPipeline(), WindowingStrategy.globalDefault(), input.isBounded()));
}
@Test
public void testPartiallyBoundFailure() throws IOException {
Pipeline p = DataflowPipeline.create(buildPipelineOptions());
PCollection<Integer> input = p.begin().apply(Create.of(1, 2, 3));
thrown.expect(IllegalStateException.class);
input.apply(new PartiallyBoundOutputCreator());
Assert.fail("Failure expected from use of partially bound output");
}
@Override
public PCollectionView<T> apply(PCollection<T> input) {
return input.apply(
CreatePCollectionView.<T, T>of(
PCollectionViews.singletonView(
input.getPipeline(),
input.getWindowingStrategy(),
hasDefault,
defaultValue,
input.getCoder())));
}
@Test
public void testCountConstraint() {
Pipeline p = TestPipeline.create();
PCollection<String> input =
p.apply(Create.of(Arrays.asList(COLLECTION)).withCoder(StringUtf8Coder.of()));
expectedEx.expect(IllegalArgumentException.class);
expectedEx.expectMessage(Matchers.containsString(">= 0"));
input.apply(Top.of(-1, new OrderByLength()));
}
@Test
@Category(RunnableOnService.class)
public void testApproximateUniqueWithSmallInput() {
Pipeline p = TestPipeline.create();
PCollection<Integer> input = p.apply(Create.of(Arrays.asList(1, 2, 3, 3)));
PCollection<Long> estimate = input.apply(ApproximateUnique.<Integer>globally(1000));
DataflowAssert.thatSingleton(estimate).isEqualTo(3L);
p.run();
}
@Override
protected void testProgram() throws Exception {
Pipeline p = FlinkTestPipeline.create();
PCollection<TableRow> input1 = p.apply(Create.of(EVENT_ARRAY));
PCollection<TableRow> input2 = p.apply(Create.of(CC_ARRAY));
PCollection<String> output = JoinExamples.joinEvents(input1, input2);
output.apply(TextIO.Write.to(resultPath));
p.run();
}
@Test
public void testUnsupportedFilePattern() throws IOException {
File outFolder = tmpFolder.newFolder();
// Windows doesn't like resolving paths with * in them.
String filename = outFolder.toPath().resolve("output@5").toString();
Pipeline p = TestPipeline.create();
PCollection<String> input =
p.apply(Create.of(Arrays.asList(LINES_ARRAY)).withCoder(StringUtf8Coder.of()));
expectedException.expect(IllegalArgumentException.class);
expectedException.expectMessage("Output name components are not allowed to contain");
input.apply(TextIO.Write.to(filename));
}
public static void main(String[] args) {
PipelineOptionsFactory.register(KafkaStreamingWordCountOptions.class);
KafkaStreamingWordCountOptions options =
PipelineOptionsFactory.fromArgs(args).as(KafkaStreamingWordCountOptions.class);
options.setJobName("KafkaExample - WindowSize: " + options.getWindowSize() + " seconds");
options.setStreaming(true);
options.setCheckpointingInterval(1000L);
options.setNumberOfExecutionRetries(5);
options.setExecutionRetryDelay(3000L);
options.setRunner(FlinkPipelineRunner.class);
System.out.println(
options.getKafkaTopic()
+ " "
+ options.getZookeeper()
+ " "
+ options.getBroker()
+ " "
+ options.getGroup());
Pipeline pipeline = Pipeline.create(options);
Properties p = new Properties();
p.setProperty("zookeeper.connect", options.getZookeeper());
p.setProperty("bootstrap.servers", options.getBroker());
p.setProperty("group.id", options.getGroup());
// this is the Flink consumer that reads the input to
// the program from a kafka topic.
FlinkKafkaConsumer08<String> kafkaConsumer =
new FlinkKafkaConsumer08<>(options.getKafkaTopic(), new SimpleStringSchema(), p);
PCollection<String> words =
pipeline
.apply(Read.from(new UnboundedFlinkSource<>(kafkaConsumer)).named("StreamingWordCount"))
.apply(ParDo.of(new ExtractWordsFn()))
.apply(
Window.<String>into(
FixedWindows.of(Duration.standardSeconds(options.getWindowSize())))
.triggering(AfterWatermark.pastEndOfWindow())
.withAllowedLateness(Duration.ZERO)
.discardingFiredPanes());
PCollection<KV<String, Long>> wordCounts = words.apply(Count.<String>perElement());
wordCounts.apply(ParDo.of(new FormatAsStringFn())).apply(TextIO.Write.to("./outputKafka.txt"));
pipeline.run();
}
@Test
public void testMultiGraphPipelineSerialization() throws IOException {
Pipeline p = DataflowPipeline.create(buildPipelineOptions());
PCollection<Integer> input = p.begin().apply(Create.of(1, 2, 3));
input.apply(new UnrelatedOutputCreator());
input.apply(new UnboundOutputCreator());
DataflowPipelineTranslator t =
DataflowPipelineTranslator.fromOptions(
PipelineOptionsFactory.as(DataflowPipelineOptions.class));
// Check that translation doesn't fail.
t.translate(p, Collections.<DataflowPackage>emptyList());
}
/**
* Merge the statistics from each block. The resulting "collection" contains a single element,
* with the answer.
*/
private static PCollection<RecalibrationTables> aggregateStatistics(
final PCollection<RecalibrationTables> tables) {
return tables
// aggregate
.apply(Combine.globally(new RecalibrationTablesMerger()))
// call finalize on the result
.apply(
ParDo.named("finalizeRecalTables")
.of(
new DoFnWLog<RecalibrationTables, RecalibrationTables>("finalizeRecalTables") {
private static final long serialVersionUID = 1L;
@Override
public void processElement(ProcessContext c) throws Exception {
RecalibrationTables tables = c.element();
if (null == tables) {
// the merger may return null when there are no inputs at all. In that
// case we don't want to
// crash (though it's really an edge case).
log.warn("No recalibration tables!");
} else {
// normal case: recalibrate
BaseRecalibrationEngine.finalizeRecalibrationTables(tables);
}
c.output(tables);
}
}));
}
@Override
public PDone apply(PCollection<T> input) {
if (topic == null) {
throw new IllegalStateException("need to set the topic of a PubsubIO.Write transform");
}
return PDone.in(input.getPipeline());
}
@Test
public void testTopEmptyWithIncompatibleWindows() {
Pipeline p = TestPipeline.create();
Bound<String> windowingFn = Window.<String>into(FixedWindows.of(Duration.standardDays(10L)));
PCollection<String> input =
p.apply(Create.timestamped(Collections.<String>emptyList(), Collections.<Long>emptyList()))
.apply(windowingFn);
expectedEx.expect(IllegalStateException.class);
expectedEx.expectMessage("Top");
expectedEx.expectMessage("GlobalWindows");
expectedEx.expectMessage("withoutDefaults");
expectedEx.expectMessage("asSingletonView");
input.apply(Top.of(1, new OrderByLength()));
}
private void runApproximateUniqueWithDuplicates(
int elementCount, int uniqueCount, int sampleSize) {
assert elementCount >= uniqueCount;
List<Double> elements = Lists.newArrayList();
for (int i = 0; i < elementCount; i++) {
elements.add(1.0 / (i % uniqueCount + 1));
}
Collections.shuffle(elements);
Pipeline p = TestPipeline.create();
PCollection<Double> input = p.apply(Create.of(elements));
PCollection<Long> estimate = input.apply(ApproximateUnique.<Double>globally(sampleSize));
DataflowAssert.thatSingleton(estimate).satisfies(new VerifyEstimateFn(uniqueCount, sampleSize));
p.run();
}
@Override
public PCollection<RecalibrationTables> apply(
PCollection<AddContextDataToReadOptimized.ContextShard> input) {
PCollection<RecalibrationTables> oneStatPerWorker =
input.apply(
ParDo.named("BaseRecalibrator")
.withSideInputs(headerView, refDictionary)
.of(new BaseRecalibratorOptimizedFn(headerView, refDictionary, recalArgs)));
return aggregateStatistics(oneStatPerWorker);
}
<T> void runTestWrite(T[] elems, Coder<T> coder) throws Exception {
File tmpFile = tmpFolder.newFile("file.txt");
String filename = tmpFile.getPath();
Pipeline p = TestPipeline.create();
PCollection<T> input = p.apply(Create.of(Arrays.asList(elems)).withCoder(coder));
TextIO.Write.Bound<T> write;
if (coder.equals(StringUtf8Coder.of())) {
TextIO.Write.Bound<String> writeStrings = TextIO.Write.to(filename).withoutSharding();
// T==String
write = (TextIO.Write.Bound<T>) writeStrings;
} else {
write = TextIO.Write.to(filename).withCoder(coder).withoutSharding();
}
input.apply(write);
p.run();
List<String> actual = new ArrayList<>();
try (BufferedReader reader = new BufferedReader(new FileReader(tmpFile))) {
for (; ; ) {
String line = reader.readLine();
if (line == null) {
break;
}
actual.add(line);
}
}
String[] expected = new String[elems.length];
for (int i = 0; i < elems.length; i++) {
T elem = elems[i];
byte[] encodedElem = CoderUtils.encodeToByteArray(coder, elem);
String line = new String(encodedElem);
expected[i] = line;
}
assertThat(actual, containsInAnyOrder(expected));
}
/**
* Takes a few Reads and will write them to a BAM file. The Reads don't have to be sorted
* initially, the BAM file will be. All the reads must fit into a single worker's memory, so this
* won't go well if you have too many.
*
* @param pipeline the pipeline to add this operation to.
* @param reads the reads to write (they don't need to be sorted).
* @param header the header that corresponds to the reads.
* @param destPath the GCS or local path to write to (must start with "gs://" if writing to GCS).
* @param parquet whether to write out BAM or Parquet data (BDG AlignmentRecords); only applies
* when writing to Hadoop
*/
public static void writeToFile(
Pipeline pipeline,
PCollection<GATKRead> reads,
final SAMFileHeader header,
final String destPath,
final boolean parquet) {
if (BucketUtils.isHadoopUrl(destPath)
|| pipeline.getRunner().getClass().equals(SparkPipelineRunner.class)) {
writeToHadoop(pipeline, reads, header, destPath, parquet);
} else {
PCollectionView<Iterable<GATKRead>> iterableView = reads.apply(View.<GATKRead>asIterable());
PCollection<String> dummy = pipeline.apply("output file name", Create.<String>of(destPath));
dummy.apply(
ParDo.named("save to BAM file")
.withSideInputs(iterableView)
.of(new SaveToBAMFile(header, iterableView)));
}
}
@Test
public void testWriteSharded() throws IOException {
File outFolder = tmpFolder.newFolder();
String filename = outFolder.toPath().resolve("output").toString();
Pipeline p = TestPipeline.create();
PCollection<String> input =
p.apply(Create.of(Arrays.asList(LINES_ARRAY)).withCoder(StringUtf8Coder.of()));
input.apply(TextIO.Write.to(filename).withNumShards(2).withSuffix(".txt"));
p.run();
String[] files = outFolder.list();
assertThat(
Arrays.asList(files),
containsInAnyOrder("output-00000-of-00002.txt", "output-00001-of-00002.txt"));
}
private void runApproximateUniqueWithSkewedDistributions(
int elementCount, final int uniqueCount, final int sampleSize) {
List<Integer> elements = Lists.newArrayList();
// Zipf distribution with approximately elementCount items.
double s = 1 - 1.0 * uniqueCount / elementCount;
double maxCount = Math.pow(uniqueCount, s);
for (int k = 0; k < uniqueCount; k++) {
int count = Math.max(1, (int) Math.round(maxCount * Math.pow(k, -s)));
// Element k occurs count times.
for (int c = 0; c < count; c++) {
elements.add(k);
}
}
Pipeline p = TestPipeline.create();
PCollection<Integer> input = p.apply(Create.of(elements));
PCollection<Long> estimate = input.apply(ApproximateUnique.<Integer>globally(sampleSize));
DataflowAssert.thatSingleton(estimate).satisfies(new VerifyEstimateFn(uniqueCount, sampleSize));
p.run();
}
@Override
public PCollection<GATKRead> apply(PCollection<GATKRead> input) {
return input.apply(
ParDo.named("ApplyBQSR")
.of(
new DoFnWLog<GATKRead, GATKRead>("ApplyBQSRStub") {
private static final long serialVersionUID = 1L;
@Override
public void processElement(ProcessContext c) throws Exception {
c.output(c.element());
}
})
.withSideInputs(header, recalibrationReport));
}
@Override
public PCollection<T> apply(PInput input) {
if (topic == null && subscription == null) {
throw new IllegalStateException(
"need to set either the topic or the subscription for "
+ "a PubsubIO.Read transform");
}
if (topic != null && subscription != null) {
throw new IllegalStateException(
"Can't set both the topic and the subscription for a " + "PubsubIO.Read transform");
}
return PCollection.<T>createPrimitiveOutputInternal(
input.getPipeline(), WindowingStrategy.globalDefault(), IsBounded.UNBOUNDED)
.setCoder(coder);
}
private static void writeToHadoop(
Pipeline pipeline,
PCollection<GATKRead> reads,
final SAMFileHeader header,
final String destPath,
final boolean parquet) {
if (destPath.equals("/dev/null")) {
return;
}
String headerString =
Base64.getEncoder().encodeToString(SerializableUtils.serializeToByteArray(header));
@SuppressWarnings("unchecked")
Class<? extends FileOutputFormat<NullWritable, SAMRecordWritable>> outputFormatClass =
(Class<? extends FileOutputFormat<NullWritable, SAMRecordWritable>>)
(Class<?>) TemplatedKeyIgnoringBAMOutputFormat.class;
@SuppressWarnings("unchecked")
HadoopIO.Write.Bound<NullWritable, SAMRecordWritable> write =
HadoopIO.Write.to(destPath, outputFormatClass, NullWritable.class, SAMRecordWritable.class)
.withConfigurationProperty(
TemplatedKeyIgnoringBAMOutputFormat.SAM_HEADER_PROPERTY_NAME, headerString);
PCollection<KV<NullWritable, SAMRecordWritable>> samReads =
reads
.apply(
ParDo.of(
new DoFn<GATKRead, KV<NullWritable, SAMRecordWritable>>() {
private static final long serialVersionUID = 1L;
@Override
public void processElement(ProcessContext c) throws Exception {
SAMRecord samRecord = c.element().convertToSAMRecord(header);
SAMRecordWritable samRecordWritable = new SAMRecordWritable();
samRecordWritable.set(samRecord);
c.output(KV.of(NullWritable.get(), samRecordWritable));
}
}))
.setCoder(
KvCoder.of(
WritableCoder.of(NullWritable.class),
WritableCoder.of(SAMRecordWritable.class)));
// write as a single (unsharded) file
samReads.apply(write.withoutSharding());
}
/**
* addQuantizationInfo takes the computed RecalibrationTable and adds the QuantizationInfo and
* RequestedCovariates objects. We call this triplet "BaseRecalOutput". It contains everything we
* need from phase 1 to continue onto phase 2 of BQSR.
*/
private static PCollection<BaseRecalOutput> addQuantizationInfo(
PCollectionView<SAMFileHeader> headerView,
RecalibrationArgumentCollection recalArgs,
PCollection<RecalibrationTables> recal) {
return recal.apply(
ParDo.named("addQuantizationInfo")
.withSideInputs(headerView)
.of(
new DoFnWLog<RecalibrationTables, BaseRecalOutput>("addQuantizationInfo") {
private static final long serialVersionUID = 1L;
@Override
public void processElement(ProcessContext c) throws IOException {
RecalibrationTables rt = c.element();
SAMFileHeader header = c.sideInput(headerView);
// BaseRecalOutput ret = new BaseRecalOutput(rt,
// baseRecalibratorWorker.getQuantizationInfo(rt),
// baseRecalibratorWorker.getRequestedCovariates());
// Saving and loading back the report actually changes it. So we have to do it.
// TODO(issue#799): Figure out what it changes, and just do that instead of
// doing the whole rigamarole.
File temp = IOUtils.createTempFile("temp-recalibrationtable-", ".tmp");
if (null == rt) {
// special case where we have zero reads in the input. Create a valid empty
// report.
log.debug("Special case: zero reads in input.");
BaseRecalibrationEngine recalibrationEngine =
new BaseRecalibrationEngine(recalArgs, header);
rt = recalibrationEngine.getRecalibrationTables();
BaseRecalibrationEngine.finalizeRecalibrationTables(rt);
}
try {
BaseRecalibratorOptimizedFn.saveTextualReport(temp, header, rt, recalArgs);
BaseRecalOutput ret = new BaseRecalOutput(temp);
c.output(ret);
} catch (FileNotFoundException e) {
throw new GATKException("can't find my own temporary file", e);
} catch (IOException e) {
throw new GATKException(
"unable to save temporary report to " + temp.getPath(), e);
}
}
}));
}
