use of org.apache.fluo.api.data.Column in project incubator-rya by apache.
the class FluoBinPruner method pruneBindingSetBin.
/**
* This method deletes BindingSets in the specified bin from the BindingSet
* Column of the indicated Fluo nodeId
*
* @param id
* - Fluo nodeId
* @param bin
* - bin id
*/
@Override
public void pruneBindingSetBin(final NodeBin nodeBin) {
final String id = nodeBin.getNodeId();
final long bin = nodeBin.getBin();
try (Transaction tx = client.newTransaction()) {
final Optional<NodeType> type = NodeType.fromNodeId(id);
if (!type.isPresent()) {
log.trace("Unable to determine NodeType from id: " + id);
throw new RuntimeException();
}
final Column batchInfoColumn = type.get().getResultColumn();
final Bytes batchInfoSpanPrefix = BindingHashShardingFunction.getShardedScanPrefix(id, vf.createLiteral(bin));
final SpanBatchDeleteInformation batchInfo = SpanBatchDeleteInformation.builder().setColumn(batchInfoColumn).setSpan(Span.prefix(batchInfoSpanPrefix)).build();
BatchInformationDAO.addBatch(tx, id, batchInfo);
tx.commit();
}
}
use of org.apache.fluo.api.data.Column in project incubator-rya by apache.
the class JoinBatchBindingSetUpdater method processBatch.
/**
* Processes {@link JoinBatchInformation}. Updates the BindingSets
* associated with the specified nodeId. The BindingSets are processed in
* batch fashion, where the number of results is indicated by
* {@link JoinBatchInformation#getBatchSize()}. BindingSets are either
* Added, Deleted, or Updated according to
* {@link JoinBatchInformation#getTask()}. In the event that the number of
* entries that need to be updated exceeds the batch size, the row of the
* first unprocessed BindingSets is used to create a new JoinBatch job to
* process the remaining BindingSets.
* @throws Exception
*/
@Override
public void processBatch(final TransactionBase tx, final Bytes row, final BatchInformation batch) throws Exception {
super.processBatch(tx, row, batch);
final String nodeId = BatchRowKeyUtil.getNodeId(row);
Preconditions.checkArgument(batch instanceof JoinBatchInformation);
final JoinBatchInformation joinBatch = (JoinBatchInformation) batch;
final Task task = joinBatch.getTask();
// Figure out which join algorithm we are going to use.
final IterativeJoin joinAlgorithm;
switch(joinBatch.getJoinType()) {
case NATURAL_JOIN:
joinAlgorithm = new NaturalJoin();
break;
case LEFT_OUTER_JOIN:
joinAlgorithm = new LeftOuterJoin();
break;
default:
throw new RuntimeException("Unsupported JoinType: " + joinBatch.getJoinType());
}
final Set<VisibilityBindingSet> bsSet = new HashSet<>();
final Optional<RowColumn> rowCol = fillSiblingBatch(tx, joinBatch, bsSet);
// Iterates over the resulting BindingSets from the join.
final Iterator<VisibilityBindingSet> newJoinResults;
final VisibilityBindingSet bs = joinBatch.getBs();
if (joinBatch.getSide() == Side.LEFT) {
newJoinResults = joinAlgorithm.newLeftResult(bs, bsSet.iterator());
} else {
newJoinResults = joinAlgorithm.newRightResult(bsSet.iterator(), bs);
}
// Read join metadata, create new join BindingSets and insert them into the Fluo table.
final JoinMetadata joinMetadata = CACHE.readJoinMetadata(tx, nodeId);
final VariableOrder joinVarOrder = joinMetadata.getVariableOrder();
while (newJoinResults.hasNext()) {
final VisibilityBindingSet newJoinResult = newJoinResults.next();
// create BindingSet value
final Bytes bsBytes = BS_SERDE.serialize(newJoinResult);
// make rowId
Bytes rowKey = BindingHashShardingFunction.addShard(nodeId, joinVarOrder, newJoinResult);
final Column col = FluoQueryColumns.JOIN_BINDING_SET;
processTask(tx, task, rowKey, col, bsBytes);
}
// update the span and register updated batch job
if (rowCol.isPresent()) {
final Span newSpan = getNewSpan(rowCol.get(), joinBatch.getSpan());
joinBatch.setSpan(newSpan);
BatchInformationDAO.addBatch(tx, nodeId, joinBatch);
}
}
use of org.apache.fluo.api.data.Column in project incubator-rya by apache.
the class SpanBatchBindingSetUpdater method processBatch.
/**
* Process SpanBatchDeleteInformation objects by deleting all entries indicated by Span until batch limit is met.
*
* @param tx - Fluo Transaction
* @param row - Byte row identifying BatchInformation
* @param batch - SpanBatchDeleteInformation object to be processed
*/
@Override
public void processBatch(TransactionBase tx, Bytes row, BatchInformation batch) throws Exception {
super.processBatch(tx, row, batch);
Preconditions.checkArgument(batch instanceof SpanBatchDeleteInformation);
SpanBatchDeleteInformation spanBatch = (SpanBatchDeleteInformation) batch;
Optional<String> nodeId = spanBatch.getNodeId();
Task task = spanBatch.getTask();
int batchSize = spanBatch.getBatchSize();
Span span = spanBatch.getSpan();
Column column = batch.getColumn();
Optional<RowColumn> rowCol = Optional.empty();
switch(task) {
case Add:
log.trace("The Task Add is not supported for SpanBatchBindingSetUpdater. Batch " + batch + " will not be processed.");
break;
case Delete:
rowCol = deleteBatch(tx, nodeId, span, column, batchSize);
break;
case Update:
log.trace("The Task Update is not supported for SpanBatchBindingSetUpdater. Batch " + batch + " will not be processed.");
break;
default:
log.trace("Invalid Task type. Aborting batch operation.");
break;
}
if (rowCol.isPresent()) {
Span newSpan = getNewSpan(rowCol.get(), spanBatch.getSpan());
log.trace("Batch size met. There are remaining results that need to be deleted. Creating a new batch of size: " + spanBatch.getBatchSize() + " with Span: " + newSpan + " and Column: " + column);
spanBatch.setSpan(newSpan);
BatchInformationDAO.addBatch(tx, BatchRowKeyUtil.getNodeId(row), spanBatch);
}
}
use of org.apache.fluo.api.data.Column in project incubator-rya by apache.
the class ConstructQueryResultUpdater method updateConstructQueryResults.
/**
* Updates the Construct Query results by applying the {@link ConnstructGraph} to
* create a {@link RyaSubGraph} and then writing the subgraph to {@link FluoQueryColumns#CONSTRUCT_STATEMENTS}.
* @param tx - transaction used to write the subgraph
* @param bs - BindingSet that the ConstructProjection expands into a subgraph
* @param metadata - metadata that the ConstructProjection is extracted from
*/
public void updateConstructQueryResults(TransactionBase tx, VisibilityBindingSet bs, ConstructQueryMetadata metadata) {
String nodeId = metadata.getNodeId();
VariableOrder varOrder = metadata.getVariableOrder();
Column column = FluoQueryColumns.CONSTRUCT_STATEMENTS;
ConstructGraph graph = metadata.getConstructGraph();
String parentId = metadata.getParentNodeId();
// Create the Row Key for the emitted binding set. It does not contain visibilities.
final Bytes resultRow = makeRowKey(nodeId, varOrder, bs);
// If this is a new binding set, then emit it.
if (tx.get(resultRow, column) == null || varOrder.getVariableOrders().size() < bs.size()) {
Set<RyaStatement> statements = graph.createGraphFromBindingSet(bs);
RyaSubGraph subgraph = new RyaSubGraph(parentId, statements);
final Bytes nodeValueBytes = Bytes.of(serializer.toBytes(subgraph));
log.trace("Transaction ID: " + tx.getStartTimestamp() + "\n" + "New Binding Set: " + subgraph + "\n");
tx.set(resultRow, column, nodeValueBytes);
}
}
use of org.apache.fluo.api.data.Column in project incubator-rya by apache.
the class BatchIT method createSpanBatches.
private void createSpanBatches(FluoClient fluoClient, List<String> ids, List<String> prefixes, int batchSize) {
Preconditions.checkArgument(ids.size() == prefixes.size());
try (Transaction tx = fluoClient.newTransaction()) {
for (int i = 0; i < ids.size(); i++) {
String id = ids.get(i);
String bsPrefix = prefixes.get(i);
URI uri = vf.createURI(bsPrefix);
Bytes prefixBytes = BindingHashShardingFunction.getShardedScanPrefix(id, uri);
NodeType type = NodeType.fromNodeId(id).get();
Column bsCol = type.getResultColumn();
SpanBatchDeleteInformation.Builder builder = SpanBatchDeleteInformation.builder().setBatchSize(batchSize).setColumn(bsCol);
if (type == NodeType.JOIN) {
builder.setSpan(Span.prefix(type.getNodeTypePrefix()));
builder.setNodeId(java.util.Optional.of(id));
} else {
builder.setSpan(Span.prefix(prefixBytes));
}
BatchInformationDAO.addBatch(tx, id, builder.build());
}
tx.commit();
}
}
Aggregations