use of io.crate.planner.projection.Projection in project crate by crate.
the class AbstractProjectionsPhase method writeTo.
@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeString(name);
assert jobId != null : "jobId must not be null";
out.writeLong(jobId.getMostSignificantBits());
out.writeLong(jobId.getLeastSignificantBits());
out.writeVInt(executionPhaseId);
int numCols = outputTypes.size();
out.writeVInt(numCols);
for (int i = 0; i < numCols; i++) {
DataTypes.toStream(outputTypes.get(i), out);
}
if (hasProjections()) {
out.writeVInt(projections.size());
for (Projection p : projections) {
Projection.toStream(p, out);
}
} else {
out.writeVInt(0);
}
}
use of io.crate.planner.projection.Projection in project crate by crate.
the class Collect method streamOutputs.
@Override
public List<DataType> streamOutputs() {
List<Projection> projections = collectPhase.projections();
if (projections.isEmpty()) {
return Symbols.extractTypes(collectPhase.toCollect());
}
Projection lastProjection = projections.get(projections.size() - 1);
return Symbols.extractTypes(lastProjection.outputs());
}
use of io.crate.planner.projection.Projection in project crate by crate.
the class MergeNodeTest method testSerialization.
@Test
public void testSerialization() throws Exception {
Reference nameRef = TestingHelpers.createReference("name", DataTypes.STRING);
List<Symbol> keys = Collections.singletonList(nameRef);
List<Aggregation> aggregations = Collections.singletonList(Aggregation.finalAggregation(new FunctionInfo(new FunctionIdent(CountAggregation.NAME, ImmutableList.of()), DataTypes.LONG), ImmutableList.of(), Aggregation.Step.PARTIAL));
GroupProjection groupProjection = new GroupProjection(keys, aggregations, RowGranularity.CLUSTER);
TopNProjection topNProjection = new TopNProjection(10, 0, InputColumn.numInputs(keys.size() + aggregations.size()));
List<Projection> projections = Arrays.asList(groupProjection, topNProjection);
MergePhase node = new MergePhase(UUID.randomUUID(), 0, "merge", 2, Collections.emptyList(), Arrays.<DataType>asList(DataTypes.UNDEFINED, DataTypes.STRING), projections, DistributionInfo.DEFAULT_BROADCAST, null);
node.executionNodes(Sets.newHashSet("node1", "node2"));
BytesStreamOutput output = new BytesStreamOutput();
node.writeTo(output);
StreamInput input = StreamInput.wrap(output.bytes());
MergePhase node2 = MergePhase.FACTORY.create();
node2.readFrom(input);
assertThat(node.numUpstreams(), is(node2.numUpstreams()));
assertThat(node.nodeIds(), is(node2.nodeIds()));
assertThat(node.jobId(), is(node2.jobId()));
assertEquals(node.inputTypes(), node2.inputTypes());
assertThat(node.phaseId(), is(node2.phaseId()));
assertThat(node.distributionInfo(), is(node2.distributionInfo()));
}
use of io.crate.planner.projection.Projection in project crate by crate.
the class CopyStatementPlanner method planCopyFrom.
public Plan planCopyFrom(CopyFromAnalyzedStatement analysis, Planner.Context context) {
/**
* copy from has two "modes":
*
* 1: non-partitioned tables or partitioned tables with partition ident --> import into single es index
* -> collect raw source and import as is
*
* 2: partitioned table without partition ident
* -> collect document and partition by values
* -> exclude partitioned by columns from document
* -> insert into es index (partition determined by partition by value)
*/
DocTableInfo table = analysis.table();
int clusteredByPrimaryKeyIdx = table.primaryKey().indexOf(analysis.table().clusteredBy());
List<String> partitionedByNames;
String partitionIdent = null;
List<BytesRef> partitionValues;
if (analysis.partitionIdent() == null) {
if (table.isPartitioned()) {
partitionedByNames = Lists.newArrayList(Lists.transform(table.partitionedBy(), ColumnIdent::fqn));
} else {
partitionedByNames = Collections.emptyList();
}
partitionValues = ImmutableList.of();
} else {
assert table.isPartitioned() : "table must be partitioned if partitionIdent is set";
// partitionIdent is present -> possible to index raw source into concrete es index
partitionValues = PartitionName.decodeIdent(analysis.partitionIdent());
partitionIdent = analysis.partitionIdent();
partitionedByNames = Collections.emptyList();
}
SourceIndexWriterProjection sourceIndexWriterProjection = new SourceIndexWriterProjection(table.ident(), partitionIdent, table.getReference(DocSysColumns.RAW), table.primaryKey(), table.partitionedBy(), partitionValues, table.clusteredBy(), clusteredByPrimaryKeyIdx, analysis.settings(), null, partitionedByNames.size() > 0 ? partitionedByNames.toArray(new String[partitionedByNames.size()]) : null, // autoCreateIndices
table.isPartitioned());
List<Projection> projections = Collections.<Projection>singletonList(sourceIndexWriterProjection);
partitionedByNames.removeAll(Lists.transform(table.primaryKey(), ColumnIdent::fqn));
int referencesSize = table.primaryKey().size() + partitionedByNames.size() + 1;
referencesSize = clusteredByPrimaryKeyIdx == -1 ? referencesSize + 1 : referencesSize;
List<Symbol> toCollect = new ArrayList<>(referencesSize);
// add primaryKey columns
for (ColumnIdent primaryKey : table.primaryKey()) {
toCollect.add(table.getReference(primaryKey));
}
// add partitioned columns (if not part of primaryKey)
Set<Reference> referencedReferences = new HashSet<>();
for (String partitionedColumn : partitionedByNames) {
Reference reference = table.getReference(ColumnIdent.fromPath(partitionedColumn));
Symbol symbol;
if (reference instanceof GeneratedReference) {
symbol = ((GeneratedReference) reference).generatedExpression();
referencedReferences.addAll(((GeneratedReference) reference).referencedReferences());
} else {
symbol = reference;
}
toCollect.add(symbol);
}
// add clusteredBy column (if not part of primaryKey)
if (clusteredByPrimaryKeyIdx == -1 && table.clusteredBy() != null && !DocSysColumns.ID.equals(table.clusteredBy())) {
toCollect.add(table.getReference(table.clusteredBy()));
}
// add _raw or _doc
if (table.isPartitioned() && analysis.partitionIdent() == null) {
toCollect.add(table.getReference(DocSysColumns.DOC));
} else {
toCollect.add(table.getReference(DocSysColumns.RAW));
}
// add columns referenced by generated columns which are used as partitioned by column
for (Reference reference : referencedReferences) {
if (!toCollect.contains(reference)) {
toCollect.add(reference);
}
}
DiscoveryNodes allNodes = clusterService.state().nodes();
FileUriCollectPhase collectPhase = new FileUriCollectPhase(context.jobId(), context.nextExecutionPhaseId(), "copyFrom", getExecutionNodes(allNodes, analysis.settings().getAsInt("num_readers", allNodes.getSize()), analysis.nodePredicate()), analysis.uri(), toCollect, projections, analysis.settings().get("compression", null), analysis.settings().getAsBoolean("shared", null));
Collect collect = new Collect(collectPhase, TopN.NO_LIMIT, 0, 1, 1, null);
return Merge.ensureOnHandler(collect, context, Collections.singletonList(MergeCountProjection.INSTANCE));
}
Aggregations