/**
* Load a {@link Dataset} or {@link View} for the given {@link URI}.
*
* <p>URIs must begin with {@code dataset:} or {@code view:}. The remainder of the URI is
* implementation specific, depending on the dataset scheme.
*
* <p>If you use a dataset URI, {@code load} returns the unfiltered dataset. If you use a view
* URI, {@code load} returns a {@code View} configured to read a subset of the dataset.
*
* @param uri a {@code Dataset} or {@code View} URI
* @param type a Java class that represents an entity in the dataset
* @param <E> the type used for readers and writers created by this {@code Dataset}
* @param <V> the type of {@code View} expected
* @return a {@code View} for the given URI
* @throws DatasetNotFoundException if there is no dataset for the given URI
* @throws NullPointerException if any arguments are {@code null}
* @throws IllegalArgumentException if {@code uri} is not a dataset or view URI
*/
@SuppressWarnings("unchecked")
public static <E, V extends View<E>> V load(URI uri, Class<E> type) {
boolean isView = URIBuilder.VIEW_SCHEME.equals(uri.getScheme());
Preconditions.checkArgument(
isView || URIBuilder.DATASET_SCHEME.equals(uri.getScheme()),
"Not a dataset or view URI: " + uri);
Preconditions.checkNotNull(
type,
"The entity type can't be null, use Object.class to have the type"
+ " determined by the schema.");
Pair<DatasetRepository, Map<String, String>> pair =
Registration.lookupDatasetUri(URI.create(uri.getRawSchemeSpecificPart()));
DatasetRepository repo = pair.first();
Map<String, String> uriOptions = pair.second();
Dataset<E> dataset =
repo.load(
uriOptions.get(URIBuilder.NAMESPACE_OPTION),
uriOptions.get(URIBuilder.DATASET_NAME_OPTION),
type);
if (isView) {
return Datasets.<E, V>view(dataset, uriOptions);
} else {
// if the URI isn't a view URI, only load the dataset
return (V) dataset;
}
}
/**
* Create a {@link Dataset} for the given dataset or view URI. {@code create} returns an empty
* dataset. You can use {@code DatasetWriter} to populate your dataset.
*
* <p>URIs must begin with {@code dataset:} or {@code view:}. The remainder of the URI is
* implementation specific, depending on the dataset scheme. If the URI is a view URI, this method
* creates the underlying dataset and returns a view of it.
*
* @param uri a {@code Dataset} or {@code View} URI
* @param type a Java class that represents an entity in the dataset
* @param <E> the type used for readers and writers created by this {@code Dataset}
* @param <V> the type of {@code Dataset} or {@code View} expected
* @return a newly created {@code Dataset} responsible for the given URI
* @throws NullPointerException if {@code uri}, {@code descriptor}, or {@code type} is {@code
* null}
* @throws IllegalArgumentException if {@code uri} is not a dataset or view URI
* @throws DatasetExistsException if a {@code Dataset} for the given URI already exists
* @throws IncompatibleSchemaException if the schema is not compatible with existing datasets with
* shared storage (for example, in the same HBase table)
*/
@SuppressWarnings("unchecked")
public static <E, V extends View<E>> V create(
URI uri, DatasetDescriptor descriptor, Class<E> type) {
boolean isView = URIBuilder.VIEW_SCHEME.equals(uri.getScheme());
Preconditions.checkArgument(
isView || URIBuilder.DATASET_SCHEME.equals(uri.getScheme()),
"Not a dataset or view URI: " + uri);
Preconditions.checkNotNull(
type,
"The entity type can't be null, use Object.class to have the type"
+ " determined by the schema.");
Pair<DatasetRepository, Map<String, String>> pair =
Registration.lookupDatasetUri(URI.create(uri.getRawSchemeSpecificPart()));
DatasetRepository repo = pair.first();
Map<String, String> uriOptions = pair.second();
if (descriptor.getLocation() == null && uriOptions.containsKey("location")) {
descriptor =
new DatasetDescriptor.Builder(descriptor).location(uriOptions.get("location")).build();
}
Dataset<E> dataset =
repo.create(
uriOptions.get(URIBuilder.NAMESPACE_OPTION),
uriOptions.get(URIBuilder.DATASET_NAME_OPTION),
descriptor,
type);
if (isView) {
return Datasets.<E, V>view(dataset, uriOptions);
} else {
return (V) dataset;
}
}
/**
* Check whether a {@link Dataset} identified by the given URI exists.
*
* <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 the dataset exists, {@code false} otherwise
* @throws NullPointerException if {@code uri} is null
* @throws IllegalArgumentException if {@code uri} is not a dataset URI
*/
public static boolean exists(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.exists(
uriOptions.get(URIBuilder.NAMESPACE_OPTION),
uriOptions.get(URIBuilder.DATASET_NAME_OPTION));
}
@Override
public void load() {
OptionBuilder<DatasetRepository> builder = new URIBuilder();
Registration.register(
new URIPattern("file:/*path?absolute=true"),
new URIPattern("file:/*path/:namespace/:dataset?absolute=true"),
builder);
Registration.register(
new URIPattern("file:*path"), new URIPattern("file:*path/:namespace/:dataset"), builder);
Registration.register(
new URIPattern("hdfs:/*path?absolute=true"),
new URIPattern("hdfs:/*path/:namespace/:dataset?absolute=true"),
builder);
Registration.register(
new URIPattern("hdfs:*path"), new URIPattern("hdfs:*path/:namespace/:dataset"), builder);
Registration.register(
new URIPattern("webhdfs:/*path?absolute=true"),
new URIPattern("webhdfs:/*path/:namespace/:dataset?absolute=true"),
builder);
}
/**
* List the {@link Dataset} URIs in the repository identified by the URI.
*
* <p>URI formats are defined by {@code Dataset} implementations. The repository URIs you pass to
* this method must begin with {@code repo:}. For example, to list the {@code Dataset} URIs for
* the Hive repository, provide the URI {@code repo:hive}.
*
* @param uri a {@code DatasetRepository} URI
* @return the URIs present in the {@code DatasetRepository}
* @throws NullPointerException if {@code uri} is null
* @throws IllegalArgumentException if {@code uri} is not a repository URI
*/
public static Collection<URI> list(URI uri) {
boolean isRepo = URIBuilder.REPO_SCHEME.equals(uri.getScheme());
Preconditions.checkArgument(isRepo, "Not a repository URI: " + uri);
DatasetRepository repo = Registration.open(URI.create(uri.getRawSchemeSpecificPart()));
// build a URI for each dataset name
URI repoUri = repo.getUri();
List<URI> datasets = Lists.newArrayList();
for (String namespace : repo.namespaces()) {
for (String dataset : repo.datasets(namespace)) {
datasets.add(new URIBuilder(repoUri, namespace, dataset).build());
}
}
return datasets;
}
/**
* Update a {@link Dataset} for the given dataset or view URI.
*
* <p>You can add columns, remove columns, or change the data type of columns in your dataset,
* provided you don't attempt a change that is incompatible with written data. Avro defines rules
* for compatible schema evolution. See <a
* href="http://kitesdk.org/docs/current/Schema-Evolution.html">Schema Evolution</a>.
*
* <p>This method updates the dataset descriptor, so you can also add or change properties.
*
* <p>The recommended way to update a dataset descriptor is to build it based on an existing
* descriptor. Use {@link DatasetDescriptor.Builder(DatasetDescriptor)} to build a
* DatasetDescriptor based on an existing instance.
*
* <p>You cannot change a dataset format or partition strategy.
*
* <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
* @param type a Java class that represents an entity in the dataset
* @param <E> the type used for readers and writers created by this {@code Dataset}
* @param <D> the type of {@code Dataset} expected
* @return a {@code Dataset} for the given URI
* @throws NullPointerException if {@code uri}, {@code descriptor}, or {@code type} is {@code
* null}
* @throws IllegalArgumentException if {@code uri} is not a dataset URI
* @throws DatasetNotFoundException if there is no dataset for the given URI
* @throws UnsupportedOperationException if descriptor updates are not supported by the
* implementation
* @throws ConcurrentSchemaModificationException if the {@code Dataset} schema is updated
* concurrently
* @throws IncompatibleSchemaException if the schema is not compatible with previous schemas, or
* with existing datasets with shared storage (for example, in the same HBase table)
*/
@SuppressWarnings("unchecked")
public static <E, D extends Dataset<E>> D update(
URI uri, DatasetDescriptor descriptor, Class<E> type) {
Preconditions.checkArgument(
URIBuilder.DATASET_SCHEME.equals(uri.getScheme()), "Not a dataset or view URI: " + uri);
Preconditions.checkNotNull(
type,
"The entity type can't be null, use Object.class to have the type"
+ " determined by the schema.");
Pair<DatasetRepository, Map<String, String>> pair =
Registration.lookupDatasetUri(URI.create(uri.getRawSchemeSpecificPart()));
DatasetRepository repo = pair.first();
Map<String, String> uriOptions = pair.second();
return (D)
repo.update(
uriOptions.get(URIBuilder.NAMESPACE_OPTION),
uriOptions.get(URIBuilder.DATASET_NAME_OPTION),
descriptor,
type);
}