@Override
public int run() throws IOException {
DatasetRepository repo = getDatasetRepository();
if (targets == null || targets.isEmpty()) {
throw new IllegalArgumentException("No views or datasets were specified.");
}
for (String uriOrName : targets) {
if (isViewUri(uriOrName)) {
View view = Datasets.load(uriOrName);
Preconditions.checkArgument(
viewMatches(view.getUri(), uriOrName),
"Resolved view does not match requested view: " + view.getUri());
view.deleteAll();
} else if (isDatasetUri(uriOrName)) {
Datasets.delete(uriOrName);
} else {
repo.delete(namespace, uriOrName);
}
console.debug("Deleted {}", uriOrName);
}
return 0;
}
/**
* Delete a {@link Dataset} identified by the given dataset URI.
*
* <p>When you call this method using a dataset URI, both data and metadata are deleted. After you
* call this method, the dataset no longer exists, unless an exception is thrown.
*
* <p>When you call this method using a view URI, data in that view is deleted. The dataset's
* metadata is not changed. This can throw an {@code UnsupportedOperationException} if the delete
* requires additional work. For example, if some, but not all, of the data in an underlying data
* file must be removed, then the implementation is allowed to reject the deletion rather than
* copy the remaining records to a new file. An implementation must document under what conditions
* it accepts deletes, and under what conditions it rejects them.
*
* <p>URIs must begin with {@code dataset:}. The remainder of the URI is implementation specific,
* depending on the dataset scheme.
*
* @param uri a {@code Dataset} URI
* @return {@code true} if any data or metadata is removed, {@code false} otherwise
* @throws NullPointerException if {@code uri} is null
* @throws IllegalArgumentException if {@code uri} is not a dataset URI
*/
public static boolean delete(URI uri) {
Preconditions.checkArgument(
URIBuilder.DATASET_SCHEME.equals(uri.getScheme()), "Not a dataset URI: " + uri);
Pair<DatasetRepository, Map<String, String>> pair =
Registration.lookupDatasetUri(URI.create(uri.getRawSchemeSpecificPart()));
DatasetRepository repo = pair.first();
Map<String, String> uriOptions = pair.second();
return repo.delete(
uriOptions.get(URIBuilder.NAMESPACE_OPTION),
uriOptions.get(URIBuilder.DATASET_NAME_OPTION));
}
@Test
public void testWriteModeCheckpoint() throws Exception {
Dataset<Record> inputDataset =
repo.create("ns", "in", new DatasetDescriptor.Builder().schema(USER_SCHEMA).build());
Dataset<Record> outputDataset =
repo.create("ns", "out", new DatasetDescriptor.Builder().schema(USER_SCHEMA).build());
writeTestUsers(inputDataset, 1, 0);
Thread.sleep(
1000); // ensure output is newer than input on local filesystems with 1s granularity
runCheckpointPipeline(inputDataset, outputDataset);
// under hadoop1 the issues with LocalJobRunner (MAPREDUCE-2350) require that we
// manually ready the output dataset
if (Hadoop.isHadoop1()) {
((Signalable) outputDataset).signalReady();
}
checkTestUsers(outputDataset, 1);
long lastModified = ((LastModifiedAccessor) outputDataset).getLastModified();
// re-run without changing input and output should not change
runCheckpointPipeline(inputDataset, outputDataset);
checkTestUsers(outputDataset, 1);
Assert.assertEquals(lastModified, ((LastModifiedAccessor) outputDataset).getLastModified());
// re-write input then re-run and output should be re-written
Thread.sleep(1000); // ensure new input is newer than output
repo.delete("ns", "in");
inputDataset =
repo.create("ns", "in", new DatasetDescriptor.Builder().schema(USER_SCHEMA).build());
writeTestUsers(inputDataset, 1, 0);
runCheckpointPipeline(inputDataset, outputDataset);
checkTestUsers(outputDataset, 1);
Assert.assertTrue(((LastModifiedAccessor) outputDataset).getLastModified() > lastModified);
}
@Before
public void setUp() throws Exception {
this.repo = newRepo();
repo.delete("ns", "in");
repo.delete("ns", "out");
}