use of org.apache.hyracks.storage.common.IIndexBulkLoader in project asterixdb by apache.
the class OrderedIndexExamplesTest method bulkOrderVerificationExample.
/**
* Bulk load failure example. Repeatedly loads a tree with 1,000 tuples, of
* which one tuple at each possible position does not conform to the
* expected order. We expect the bulk load to fail with an exception.
*/
@Test
public void bulkOrderVerificationExample() throws Exception {
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info("Bulk load order verification example");
}
// Declare fields.
int fieldCount = 2;
ITypeTraits[] typeTraits = new ITypeTraits[fieldCount];
typeTraits[0] = IntegerPointable.TYPE_TRAITS;
typeTraits[1] = IntegerPointable.TYPE_TRAITS;
// declare keys
int keyFieldCount = 1;
IBinaryComparatorFactory[] cmpFactories = new IBinaryComparatorFactory[keyFieldCount];
cmpFactories[0] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
Random rnd = new Random();
ArrayTupleBuilder tb = new ArrayTupleBuilder(fieldCount);
ArrayTupleReference tuple = new ArrayTupleReference();
// This is only used for the LSM-BTree.
int[] bloomFilterKeyFields = new int[keyFieldCount];
bloomFilterKeyFields[0] = 0;
int ins = 1000;
for (int i = 1; i < ins; i++) {
ITreeIndex treeIndex = createTreeIndex(typeTraits, cmpFactories, bloomFilterKeyFields, null, null, null, null);
treeIndex.create();
treeIndex.activate();
// Load sorted records, and expect to fail at tuple i.
IIndexBulkLoader bulkLoader = treeIndex.createBulkLoader(0.7f, true, ins, true);
for (int j = 0; j < ins; j++) {
if (j > i) {
fail("Bulk load failure test unexpectedly succeeded past tuple: " + j);
}
int key = j;
if (j == i) {
int swapElementCase = Math.abs(rnd.nextInt()) % 2;
if (swapElementCase == 0) {
// Element equal to previous element.
key--;
} else {
// Element smaller than previous element.
key -= Math.abs(Math.random() % (ins - 1)) + 1;
}
}
TupleUtils.createIntegerTuple(tb, tuple, key, 5);
try {
bulkLoader.add(tuple);
} catch (HyracksDataException e) {
if (e.getErrorCode() == ErrorCode.UNSORTED_LOAD_INPUT || e.getErrorCode() == ErrorCode.DUPLICATE_KEY || e.getErrorCode() == ErrorCode.DUPLICATE_LOAD_INPUT) {
if (j != i) {
fail("Unexpected exception: " + e.getMessage());
}
// Success.
break;
} else {
throw e;
}
}
}
treeIndex.deactivate();
treeIndex.destroy();
}
}
use of org.apache.hyracks.storage.common.IIndexBulkLoader in project asterixdb by apache.
the class AbstractRTreeExamplesTest method bulkLoadExample.
/**
* Bulk load example. Load a tree with 10,000 tuples.
*/
@Test
public void bulkLoadExample() throws Exception {
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info("Bulk load example");
}
// Declare fields.
int fieldCount = 5;
ITypeTraits[] typeTraits = new ITypeTraits[fieldCount];
typeTraits[0] = IntegerPointable.TYPE_TRAITS;
typeTraits[1] = IntegerPointable.TYPE_TRAITS;
typeTraits[2] = IntegerPointable.TYPE_TRAITS;
typeTraits[3] = IntegerPointable.TYPE_TRAITS;
typeTraits[4] = IntegerPointable.TYPE_TRAITS;
// Declare field serdes.
ISerializerDeserializer[] fieldSerdes = { IntegerSerializerDeserializer.INSTANCE, IntegerSerializerDeserializer.INSTANCE, IntegerSerializerDeserializer.INSTANCE, IntegerSerializerDeserializer.INSTANCE, IntegerSerializerDeserializer.INSTANCE };
// Declare RTree keys.
int rtreeKeyFieldCount = 4;
IBinaryComparatorFactory[] rtreeCmpFactories = new IBinaryComparatorFactory[rtreeKeyFieldCount];
rtreeCmpFactories[0] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
rtreeCmpFactories[1] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
rtreeCmpFactories[2] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
rtreeCmpFactories[3] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
// Declare BTree keys, this will only be used for LSMRTree
int btreeKeyFieldCount;
IBinaryComparatorFactory[] btreeCmpFactories;
int[] btreeFields = null;
if (rTreeType == RTreeType.LSMRTREE) {
//Parameters look different for LSM RTREE from LSM RTREE WITH ANTI MATTER TUPLES
btreeKeyFieldCount = 1;
btreeCmpFactories = new IBinaryComparatorFactory[btreeKeyFieldCount];
btreeCmpFactories[0] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
btreeFields = new int[btreeKeyFieldCount];
for (int i = 0; i < btreeKeyFieldCount; i++) {
btreeFields[i] = rtreeKeyFieldCount + i;
}
} else {
btreeKeyFieldCount = 5;
btreeCmpFactories = new IBinaryComparatorFactory[btreeKeyFieldCount];
btreeCmpFactories[0] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
btreeCmpFactories[1] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
btreeCmpFactories[2] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
btreeCmpFactories[3] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
btreeCmpFactories[4] = PointableBinaryComparatorFactory.of(IntegerPointable.FACTORY);
}
// create value providers
IPrimitiveValueProviderFactory[] valueProviderFactories = RTreeUtils.createPrimitiveValueProviderFactories(rtreeCmpFactories.length, IntegerPointable.FACTORY);
//6
ITreeIndex treeIndex = createTreeIndex(typeTraits, rtreeCmpFactories, btreeCmpFactories, valueProviderFactories, RTreePolicyType.RTREE, null, btreeFields, null, null, null);
treeIndex.create();
treeIndex.activate();
// Load records.
int numInserts = 10000;
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info("Bulk loading " + numInserts + " tuples");
}
long start = System.currentTimeMillis();
IIndexBulkLoader bulkLoader = treeIndex.createBulkLoader(0.7f, false, numInserts, true);
ArrayTupleBuilder tb = new ArrayTupleBuilder(fieldCount);
ArrayTupleReference tuple = new ArrayTupleReference();
for (int i = 0; i < numInserts; i++) {
int p1x = rnd.nextInt();
int p1y = rnd.nextInt();
int p2x = rnd.nextInt();
int p2y = rnd.nextInt();
int pk = 5;
TupleUtils.createIntegerTuple(tb, tuple, Math.min(p1x, p2x), Math.min(p1y, p2y), Math.max(p1x, p2x), Math.max(p1y, p2y), pk);
bulkLoader.add(tuple);
}
bulkLoader.end();
long end = System.currentTimeMillis();
if (LOGGER.isLoggable(Level.INFO)) {
LOGGER.info(numInserts + " tuples loaded in " + (end - start) + "ms");
}
IIndexAccessor indexAccessor = treeIndex.createAccessor(NoOpOperationCallback.INSTANCE, NoOpOperationCallback.INSTANCE);
// Build key.
ArrayTupleBuilder keyTb = new ArrayTupleBuilder(rtreeKeyFieldCount);
ArrayTupleReference key = new ArrayTupleReference();
TupleUtils.createIntegerTuple(keyTb, key, -1000, -1000, 1000, 1000);
rangeSearch(rtreeCmpFactories, indexAccessor, fieldSerdes, key, null, null);
treeIndex.deactivate();
treeIndex.destroy();
}
use of org.apache.hyracks.storage.common.IIndexBulkLoader in project asterixdb by apache.
the class TreeIndexTestUtils method bulkLoadCheckTuples.
public static void bulkLoadCheckTuples(IIndexTestContext ctx, Collection<CheckTuple> checkTuples, boolean filtered) throws HyracksDataException {
int fieldCount = ctx.getFieldCount();
int numTuples = checkTuples.size();
ArrayTupleBuilder tupleBuilder = filtered ? new ArrayTupleBuilder(fieldCount + 1) : new ArrayTupleBuilder(fieldCount);
ArrayTupleReference tuple = new ArrayTupleReference();
// Perform bulk load.
IIndexBulkLoader bulkLoader = ctx.getIndex().createBulkLoader(0.7f, false, numTuples, false);
int c = 1;
for (CheckTuple checkTuple : checkTuples) {
if (LOGGER.isLoggable(Level.INFO)) {
//if (c % (numTuples / 10) == 0) {
LOGGER.info("Bulk Loading Tuple " + c + "/" + numTuples);
//}
}
createTupleFromCheckTuple(checkTuple, tupleBuilder, tuple, ctx.getFieldSerdes(), filtered);
bulkLoader.add(tuple);
c++;
}
bulkLoader.end();
}
use of org.apache.hyracks.storage.common.IIndexBulkLoader in project asterixdb by apache.
the class ExternalRTree method merge.
// The only reason for overriding the merge method is the way to determine
// the need to keep deleted tuples
// This can be done in a better way by creating a method boolean
// keepDeletedTuples(mergedComponents);
@Override
public ILSMDiskComponent merge(ILSMIOOperation operation) throws HyracksDataException {
LSMRTreeMergeOperation mergeOp = (LSMRTreeMergeOperation) operation;
IIndexCursor cursor = mergeOp.getCursor();
ISearchPredicate rtreeSearchPred = new SearchPredicate(null, null);
ILSMIndexOperationContext opCtx = ((LSMRTreeSortedCursor) cursor).getOpCtx();
opCtx.getComponentHolder().addAll(mergeOp.getMergingComponents());
search(opCtx, cursor, rtreeSearchPred);
LSMRTreeDiskComponent mergedComponent = createDiskComponent(componentFactory, mergeOp.getTarget(), mergeOp.getBTreeTarget(), mergeOp.getBloomFilterTarget(), true);
// In case we must keep the deleted-keys BTrees, then they must be
// merged *before* merging the r-trees so that
// lsmHarness.endSearch() is called once when the r-trees have been
// merged.
boolean keepDeleteTuples = false;
if (version == 0) {
keepDeleteTuples = mergeOp.getMergingComponents().get(mergeOp.getMergingComponents().size() - 1) != diskComponents.get(diskComponents.size() - 1);
} else {
keepDeleteTuples = mergeOp.getMergingComponents().get(mergeOp.getMergingComponents().size() - 1) != secondDiskComponents.get(secondDiskComponents.size() - 1);
}
if (keepDeleteTuples) {
// Keep the deleted tuples since the oldest disk component is not
// included in the merge operation
LSMRTreeDeletedKeysBTreeMergeCursor btreeCursor = new LSMRTreeDeletedKeysBTreeMergeCursor(opCtx);
search(opCtx, btreeCursor, rtreeSearchPred);
BTree btree = mergedComponent.getBTree();
IIndexBulkLoader btreeBulkLoader = btree.createBulkLoader(1.0f, true, 0L, false);
long numElements = 0L;
for (int i = 0; i < mergeOp.getMergingComponents().size(); ++i) {
numElements += ((LSMRTreeDiskComponent) mergeOp.getMergingComponents().get(i)).getBloomFilter().getNumElements();
}
int maxBucketsPerElement = BloomCalculations.maxBucketsPerElement(numElements);
BloomFilterSpecification bloomFilterSpec = BloomCalculations.computeBloomSpec(maxBucketsPerElement, bloomFilterFalsePositiveRate);
IIndexBulkLoader builder = mergedComponent.getBloomFilter().createBuilder(numElements, bloomFilterSpec.getNumHashes(), bloomFilterSpec.getNumBucketsPerElements());
try {
while (btreeCursor.hasNext()) {
btreeCursor.next();
ITupleReference tuple = btreeCursor.getTuple();
btreeBulkLoader.add(tuple);
builder.add(tuple);
}
} finally {
btreeCursor.close();
builder.end();
}
btreeBulkLoader.end();
}
IIndexBulkLoader bulkLoader = mergedComponent.getRTree().createBulkLoader(1.0f, false, 0L, false);
try {
while (cursor.hasNext()) {
cursor.next();
ITupleReference frameTuple = cursor.getTuple();
bulkLoader.add(frameTuple);
}
} finally {
cursor.close();
}
bulkLoader.end();
return mergedComponent;
}
use of org.apache.hyracks.storage.common.IIndexBulkLoader in project asterixdb by apache.
the class ExternalFilesIndexCreateOperatorDescriptor method createPushRuntime.
@Override
public IOperatorNodePushable createPushRuntime(IHyracksTaskContext ctx, IRecordDescriptorProvider recordDescProvider, int partition, int nPartitions) throws HyracksDataException {
return new AbstractOperatorNodePushable() {
@Override
public void initialize() throws HyracksDataException {
IIndexBuilder indexBuilder = indexBuilderFactory.create(ctx, partition);
IIndexDataflowHelper indexHelper = dataflowHelperFactory.create(ctx, partition);
FileIndexTupleTranslator filesTupleTranslator = new FileIndexTupleTranslator();
// Build the index
indexBuilder.build();
// Open the index
indexHelper.open();
try {
IIndex index = indexHelper.getIndexInstance();
// Create bulk loader
IIndexBulkLoader bulkLoader = index.createBulkLoader(BTree.DEFAULT_FILL_FACTOR, false, files.size(), false);
// Load files
for (ExternalFile file : files) {
bulkLoader.add(filesTupleTranslator.getTupleFromFile(file));
}
bulkLoader.end();
} finally {
indexHelper.close();
}
}
@Override
public void deinitialize() throws HyracksDataException {
}
@Override
public int getInputArity() {
return 0;
}
@Override
public void setOutputFrameWriter(int index, IFrameWriter writer, RecordDescriptor recordDesc) throws HyracksDataException {
}
@Override
public IFrameWriter getInputFrameWriter(int index) {
return null;
}
};
}
Aggregations