RelBuilder.java
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to you under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.calcite.tools;
import org.apache.calcite.linq4j.Ord;
import org.apache.calcite.linq4j.function.Experimental;
import org.apache.calcite.plan.Context;
import org.apache.calcite.plan.Contexts;
import org.apache.calcite.plan.Convention;
import org.apache.calcite.plan.RelOptCluster;
import org.apache.calcite.plan.RelOptPredicateList;
import org.apache.calcite.plan.RelOptSamplingParameters;
import org.apache.calcite.plan.RelOptSchema;
import org.apache.calcite.plan.RelOptTable;
import org.apache.calcite.plan.RelOptUtil;
import org.apache.calcite.plan.ViewExpanders;
import org.apache.calcite.prepare.RelOptTableImpl;
import org.apache.calcite.rel.RelCollation;
import org.apache.calcite.rel.RelCollations;
import org.apache.calcite.rel.RelDistribution;
import org.apache.calcite.rel.RelFieldCollation;
import org.apache.calcite.rel.RelNode;
import org.apache.calcite.rel.core.Aggregate;
import org.apache.calcite.rel.core.AggregateCall;
import org.apache.calcite.rel.core.Correlate;
import org.apache.calcite.rel.core.CorrelationId;
import org.apache.calcite.rel.core.Filter;
import org.apache.calcite.rel.core.Intersect;
import org.apache.calcite.rel.core.Join;
import org.apache.calcite.rel.core.JoinRelType;
import org.apache.calcite.rel.core.Match;
import org.apache.calcite.rel.core.Minus;
import org.apache.calcite.rel.core.Project;
import org.apache.calcite.rel.core.RelFactories;
import org.apache.calcite.rel.core.RepeatUnion;
import org.apache.calcite.rel.core.Sample;
import org.apache.calcite.rel.core.Snapshot;
import org.apache.calcite.rel.core.Sort;
import org.apache.calcite.rel.core.Spool;
import org.apache.calcite.rel.core.TableFunctionScan;
import org.apache.calcite.rel.core.TableScan;
import org.apache.calcite.rel.core.TableSpool;
import org.apache.calcite.rel.core.Uncollect;
import org.apache.calcite.rel.core.Union;
import org.apache.calcite.rel.core.Values;
import org.apache.calcite.rel.hint.Hintable;
import org.apache.calcite.rel.hint.RelHint;
import org.apache.calcite.rel.logical.LogicalAsofJoin;
import org.apache.calcite.rel.logical.LogicalFilter;
import org.apache.calcite.rel.logical.LogicalProject;
import org.apache.calcite.rel.metadata.RelColumnMapping;
import org.apache.calcite.rel.metadata.RelMdUtil;
import org.apache.calcite.rel.metadata.RelMetadataQuery;
import org.apache.calcite.rel.rules.AggregateRemoveRule;
import org.apache.calcite.rel.type.RelDataType;
import org.apache.calcite.rel.type.RelDataTypeFactory;
import org.apache.calcite.rel.type.RelDataTypeField;
import org.apache.calcite.rel.type.RelDataTypeFieldImpl;
import org.apache.calcite.rex.RexBuilder;
import org.apache.calcite.rex.RexCall;
import org.apache.calcite.rex.RexCallBinding;
import org.apache.calcite.rex.RexCorrelVariable;
import org.apache.calcite.rex.RexDynamicParam;
import org.apache.calcite.rex.RexExecutor;
import org.apache.calcite.rex.RexFieldCollation;
import org.apache.calcite.rex.RexInputRef;
import org.apache.calcite.rex.RexLiteral;
import org.apache.calcite.rex.RexNode;
import org.apache.calcite.rex.RexShuttle;
import org.apache.calcite.rex.RexSimplify;
import org.apache.calcite.rex.RexSubQuery;
import org.apache.calcite.rex.RexUtil;
import org.apache.calcite.rex.RexWindowBound;
import org.apache.calcite.rex.RexWindowBounds;
import org.apache.calcite.rex.RexWindowExclusion;
import org.apache.calcite.runtime.Hook;
import org.apache.calcite.runtime.ImmutablePairList;
import org.apache.calcite.runtime.PairList;
import org.apache.calcite.schema.TransientTable;
import org.apache.calcite.schema.impl.ListTransientTable;
import org.apache.calcite.sql.SqlAggFunction;
import org.apache.calcite.sql.SqlKind;
import org.apache.calcite.sql.SqlOperator;
import org.apache.calcite.sql.SqlStaticAggFunction;
import org.apache.calcite.sql.SqlUtil;
import org.apache.calcite.sql.SqlWindow;
import org.apache.calcite.sql.fun.SqlCountAggFunction;
import org.apache.calcite.sql.fun.SqlInternalOperators;
import org.apache.calcite.sql.fun.SqlLikeOperator;
import org.apache.calcite.sql.fun.SqlQuantifyOperator;
import org.apache.calcite.sql.fun.SqlStdOperatorTable;
import org.apache.calcite.sql.parser.SqlParserPos;
import org.apache.calcite.sql.type.SqlReturnTypeInference;
import org.apache.calcite.sql.type.SqlTypeName;
import org.apache.calcite.sql.type.TableFunctionReturnTypeInference;
import org.apache.calcite.sql.validate.SqlValidatorUtil;
import org.apache.calcite.sql2rel.SqlToRelConverter;
import org.apache.calcite.util.DateString;
import org.apache.calcite.util.Holder;
import org.apache.calcite.util.ImmutableBitSet;
import org.apache.calcite.util.ImmutableIntList;
import org.apache.calcite.util.ImmutableNullableList;
import org.apache.calcite.util.Litmus;
import org.apache.calcite.util.NlsString;
import org.apache.calcite.util.Optionality;
import org.apache.calcite.util.Pair;
import org.apache.calcite.util.Util;
import org.apache.calcite.util.mapping.Mapping;
import org.apache.calcite.util.mapping.Mappings;
import com.google.common.collect.ImmutableList;
import com.google.common.collect.ImmutableMap;
import com.google.common.collect.ImmutableSet;
import com.google.common.collect.ImmutableSortedMultiset;
import com.google.common.collect.Iterables;
import com.google.common.collect.Lists;
import com.google.common.collect.Multiset;
import com.google.common.collect.Sets;
import org.checkerframework.checker.nullness.qual.Nullable;
import org.immutables.value.Value;
import java.math.BigDecimal;
import java.util.AbstractList;
import java.util.ArrayDeque;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.BitSet;
import java.util.Collections;
import java.util.Deque;
import java.util.EnumSet;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Locale;
import java.util.Map;
import java.util.Objects;
import java.util.Set;
import java.util.SortedSet;
import java.util.TreeSet;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.function.BiFunction;
import java.util.function.Consumer;
import java.util.function.Function;
import java.util.function.UnaryOperator;
import java.util.stream.Collectors;
import static com.google.common.base.Preconditions.checkArgument;
import static com.google.common.collect.ImmutableList.toImmutableList;
import static org.apache.calcite.linq4j.Nullness.castNonNull;
import static org.apache.calcite.rel.rules.AggregateRemoveRule.canFlattenStatic;
import static org.apache.calcite.sql.SqlKind.UNION;
import static org.apache.calcite.util.Static.RESOURCE;
import static org.apache.calcite.util.Util.first;
import static java.util.Objects.requireNonNull;
/**
* Builder for relational expressions.
*
* <p>{@code RelBuilder} does not make possible anything that you could not
* also accomplish by calling the factory methods of the particular relational
* expression. But it makes common tasks more straightforward and concise.
*
* <p>{@code RelBuilder} uses factories to create relational expressions.
* By default, it uses the default factories, which create logical relational
* expressions ({@link LogicalFilter},
* {@link LogicalProject} and so forth).
* But you could override those factories so that, say, {@code filter} creates
* instead a {@code HiveFilter}.
*
* <p>It is not thread-safe.
*/
@Value.Enclosing
public class RelBuilder {
protected final RelOptCluster cluster;
protected final @Nullable RelOptSchema relOptSchema;
private final Deque<Frame> stack = new ArrayDeque<>();
private RexSimplify simplifier;
private final Config config;
private final RelOptTable.ViewExpander viewExpander;
private RelFactories.Struct struct;
protected RelBuilder(@Nullable Context context, RelOptCluster cluster,
@Nullable RelOptSchema relOptSchema) {
this.cluster = cluster;
this.relOptSchema = relOptSchema;
if (context == null) {
context = Contexts.EMPTY_CONTEXT;
}
this.config = getConfig(context);
this.viewExpander = getViewExpander(cluster, context);
this.struct =
requireNonNull(RelFactories.Struct.fromContext(context));
final RexExecutor executor =
context.maybeUnwrap(RexExecutor.class)
.orElse(
first(cluster.getPlanner().getExecutor(),
RexUtil.EXECUTOR));
final RelOptPredicateList predicates = RelOptPredicateList.EMPTY;
this.simplifier =
new RexSimplify(cluster.getRexBuilder(), predicates, executor);
}
/**
* Derives the view expander
* {@link org.apache.calcite.plan.RelOptTable.ViewExpander}
* to be used for this RelBuilder.
*
* <p>The ViewExpander instance is used for expanding views in the default
* table scan factory {@code RelFactories.TableScanFactoryImpl}.
* You can also define a new table scan factory in the {@code struct}
* to override the whole table scan creation.
*
* <p>The default view expander does not support expanding views.
*/
private static RelOptTable.ViewExpander getViewExpander(RelOptCluster cluster,
Context context) {
return context.maybeUnwrap(RelOptTable.ViewExpander.class)
.orElseGet(() -> ViewExpanders.simpleContext(cluster));
}
/** Derives the Config to be used for this RelBuilder.
*
* <p>Overrides {@link RelBuilder.Config#simplify} if
* {@link Hook#REL_BUILDER_SIMPLIFY} is set.
*/
private static Config getConfig(Context context) {
final Config config =
context.maybeUnwrap(Config.class).orElse(Config.DEFAULT);
boolean simplify = Hook.REL_BUILDER_SIMPLIFY.get(config.simplify());
return config.withSimplify(simplify);
}
/** Creates a RelBuilder. */
public static RelBuilder create(FrameworkConfig config) {
return Frameworks.withPrepare(config,
(cluster, relOptSchema, rootSchema, statement) ->
new RelBuilder(config.getContext(), cluster, relOptSchema));
}
/** Creates a copy of this RelBuilder, with the same state as this, applying
* a transform to the config. */
public RelBuilder transform(UnaryOperator<Config> transform) {
final Context context =
Contexts.of(struct, transform.apply(config));
return new RelBuilder(context, cluster, relOptSchema);
}
/** Performs an action on this RelBuilder.
*
* <p>For example, consider the following code:
*
* <blockquote><pre>
* RelNode filterAndRename(RelBuilder relBuilder, RelNode rel,
* RexNode condition, List<String> fieldNames) {
* relBuilder.push(rel)
* .filter(condition);
* if (fieldNames != null) {
* relBuilder.rename(fieldNames);
* }
* return relBuilder
* .build();</pre>
* </blockquote>
*
* <p>The pipeline is disrupted by the 'if'. The {@code let} method
* allows you to perform the flow as a single pipeline:
*
* <blockquote><pre>
* RelNode filterAndRename(RelBuilder relBuilder, RelNode rel,
* RexNode condition, List<String> fieldNames) {
* return relBuilder.push(rel)
* .filter(condition)
* .let(r -> fieldNames == null ? r : r.rename(fieldNames))
* .build();</pre>
* </blockquote>
*
* <p>In pipelined cases such as this one, the lambda must return this
* RelBuilder. But {@code let} return values of other types.
*/
public <R> R let(Function<RelBuilder, R> consumer) {
return consumer.apply(this);
}
/** Converts this RelBuilder to a string.
* The string is the string representation of all of the RelNodes on the stack. */
@Override public String toString() {
return stack.stream()
.map(frame -> RelOptUtil.toString(frame.rel))
.collect(Collectors.joining(""));
}
/** Returns the type factory. */
public RelDataTypeFactory getTypeFactory() {
return cluster.getTypeFactory();
}
/** Returns new RelBuilder that adopts the convention provided.
* RelNode will be created with such convention if corresponding factory is provided. */
public RelBuilder adoptConvention(Convention convention) {
this.struct = convention.getRelFactories();
return this;
}
/** Returns the builder for {@link RexNode} expressions. */
public RexBuilder getRexBuilder() {
return cluster.getRexBuilder();
}
/** Creates a {@link RelBuilderFactory}, a partially-created RelBuilder.
* Just add a {@link RelOptCluster} and a {@link RelOptSchema} */
public static RelBuilderFactory proto(final Context context) {
return (cluster, schema) -> new RelBuilder(context, cluster, schema);
}
/** Creates a {@link RelBuilderFactory} that uses a given set of factories. */
public static RelBuilderFactory proto(Object... factories) {
return proto(Contexts.of(factories));
}
public RelOptCluster getCluster() {
return cluster;
}
public @Nullable RelOptSchema getRelOptSchema() {
return relOptSchema;
}
public RelFactories.TableScanFactory getScanFactory() {
return struct.scanFactory;
}
// Methods for manipulating the stack
/** Adds a relational expression to be the input to the next relational
* expression constructed.
*
* <p>This method is usual when you want to weave in relational expressions
* that are not supported by the builder. If, while creating such expressions,
* you need to use previously built expressions as inputs, call
* {@link #build()} to pop those inputs. */
public RelBuilder push(RelNode node) {
stack.push(new Frame(node));
return this;
}
/** Adds a rel node to the top of the stack while preserving the field names
* and aliases. */
private void replaceTop(RelNode node) {
final Frame frame = stack.pop();
stack.push(new Frame(node, frame.fields));
}
/** Pushes a collection of relational expressions. */
public RelBuilder pushAll(Iterable<? extends RelNode> nodes) {
for (RelNode node : nodes) {
push(node);
}
return this;
}
/** Returns the size of the stack. */
public int size() {
return stack.size();
}
/** Returns the final relational expression.
*
* <p>Throws if the stack is empty.
*/
public RelNode build() {
return stack.pop().rel;
}
/** Returns the relational expression at the top of the stack, but does not
* remove it. */
public RelNode peek() {
return castNonNull(peek_()).rel;
}
private @Nullable Frame peek_() {
return stack.peek();
}
/** Returns the relational expression {@code n} positions from the top of the
* stack, but does not remove it. */
public RelNode peek(int n) {
return peek_(n).rel;
}
private Frame peek_(int n) {
if (n == 0) {
// more efficient than starting an iterator
return requireNonNull(stack.peek(), "stack.peek");
}
return Iterables.get(stack, n);
}
/** Returns the relational expression {@code n} positions from the top of the
* stack, but does not remove it. */
public RelNode peek(int inputCount, int inputOrdinal) {
return peek_(inputCount, inputOrdinal).rel;
}
private Frame peek_(int inputCount, int inputOrdinal) {
return peek_(inputCount - 1 - inputOrdinal);
}
/** Returns the number of fields in all inputs before (to the left of)
* the given input.
*
* @param inputCount Number of inputs
* @param inputOrdinal Input ordinal
*/
private int inputOffset(int inputCount, int inputOrdinal) {
int offset = 0;
for (int i = 0; i < inputOrdinal; i++) {
offset += peek(inputCount, i).getRowType().getFieldCount();
}
return offset;
}
/** Evaluates an expression with a relational expression temporarily on the
* stack. */
public <E> E with(RelNode r, Function<RelBuilder, E> fn) {
try {
push(r);
return fn.apply(this);
} finally {
stack.pop();
}
}
/** Performs an action with a temporary simplifier. */
public <E> E withSimplifier(
BiFunction<RelBuilder, RexSimplify, RexSimplify> simplifierTransform,
Function<RelBuilder, E> fn) {
final RexSimplify previousSimplifier = this.simplifier;
try {
this.simplifier = simplifierTransform.apply(this, previousSimplifier);
return fn.apply(this);
} finally {
this.simplifier = previousSimplifier;
}
}
/** Performs an action using predicates of
* the {@link #peek() current node} to simplify. */
public <E> E withPredicates(RelMetadataQuery mq,
Function<RelBuilder, E> fn) {
final RelOptPredicateList predicates = mq.getPulledUpPredicates(peek());
return withSimplifier((r, s) -> s.withPredicates(predicates), fn);
}
// Methods that return scalar expressions
/** Creates a literal (constant expression). */
public RexLiteral literal(@Nullable Object value) {
final RexBuilder rexBuilder = cluster.getRexBuilder();
if (value == null) {
final RelDataType type = getTypeFactory().createSqlType(SqlTypeName.NULL);
return rexBuilder.makeNullLiteral(type);
} else if (value instanceof Boolean) {
return rexBuilder.makeLiteral((Boolean) value);
} else if (value instanceof Character) {
return rexBuilder.makeLiteral(value,
getTypeFactory().createSqlType(SqlTypeName.CHAR));
} else if (value instanceof BigDecimal) {
return rexBuilder.makeExactLiteral((BigDecimal) value);
} else if (value instanceof Float || value instanceof Double) {
return rexBuilder.makeApproxLiteral(
((Number) value).doubleValue(), getTypeFactory().createSqlType(SqlTypeName.DOUBLE));
} else if (value instanceof Number) {
return rexBuilder.makeExactLiteral(
BigDecimal.valueOf(((Number) value).longValue()));
} else if (value instanceof String) {
return rexBuilder.makeLiteral((String) value);
} else if (value instanceof Enum) {
return rexBuilder.makeLiteral(value,
getTypeFactory().createSqlType(SqlTypeName.SYMBOL));
} else if (value instanceof DateString) {
return rexBuilder.makeDateLiteral((DateString) value);
} else {
throw new IllegalArgumentException("cannot convert " + value
+ " (" + value.getClass() + ") to a constant");
}
}
@Deprecated // to be removed before 2.0
public RelBuilder variable(Holder<RexCorrelVariable> v) {
return variable(v::set);
}
/** Creates a correlation variable for the current input, and writes it into
* a Consumer.
*
* <p>Often the Consumer will write to a {@link Holder}, as follows:
* <blockquote>{@code
* RelBuilder builder;
* builder.scan("EMP")
* .variable(v::set)
* .filter(builder.equals(builder.field(0), v.get()))
* }</blockquote>
*/
public RelBuilder variable(Consumer<RexCorrelVariable> consumer) {
consumer.accept((RexCorrelVariable)
getRexBuilder().makeCorrel(peek().getRowType(),
cluster.createCorrel()));
return this;
}
/** Creates a reference to a field by name.
*
* <p>Equivalent to {@code field(1, 0, fieldName)}.
*
* @param fieldName Field name
*/
public RexInputRef field(String fieldName) {
return field(1, 0, fieldName);
}
/** Creates a reference to a field of given input relational expression
* by name.
*
* @param inputCount Number of inputs
* @param inputOrdinal Input ordinal
* @param fieldName Field name
*/
public RexInputRef field(int inputCount, int inputOrdinal, String fieldName) {
final Frame frame = peek_(inputCount, inputOrdinal);
final List<String> fieldNames = Pair.left(frame.fields());
int i = fieldNames.indexOf(fieldName);
if (i >= 0) {
return field(inputCount, inputOrdinal, i);
} else {
throw new IllegalArgumentException("field [" + fieldName
+ "] not found; input fields are: " + fieldNames);
}
}
/** Creates a reference to an input field by ordinal.
*
* <p>Equivalent to {@code field(1, 0, ordinal)}.
*
* @param fieldOrdinal Field ordinal
*/
public RexInputRef field(int fieldOrdinal) {
return (RexInputRef) field(1, 0, fieldOrdinal, false);
}
/** Creates a reference to a field of a given input relational expression
* by ordinal.
*
* @param inputCount Number of inputs
* @param inputOrdinal Input ordinal
* @param fieldOrdinal Field ordinal within input
*/
public RexInputRef field(int inputCount, int inputOrdinal, int fieldOrdinal) {
return (RexInputRef) field(inputCount, inputOrdinal, fieldOrdinal, false);
}
/** As {@link #field(int, int, int)}, but if {@code alias} is true, the method
* may apply an alias to make sure that the field has the same name as in the
* input frame. If no alias is applied the expression is definitely a
* {@link RexInputRef}. */
private RexNode field(int inputCount, int inputOrdinal, int fieldOrdinal,
boolean alias) {
final Frame frame = peek_(inputCount, inputOrdinal);
final RelNode input = frame.rel;
final RelDataType rowType = input.getRowType();
if (fieldOrdinal < 0 || fieldOrdinal > rowType.getFieldCount()) {
throw new IllegalArgumentException("field ordinal [" + fieldOrdinal
+ "] out of range; input fields are: " + rowType.getFieldNames());
}
final RelDataTypeField field = rowType.getFieldList().get(fieldOrdinal);
final int offset = inputOffset(inputCount, inputOrdinal);
final RexInputRef ref = cluster.getRexBuilder()
.makeInputRef(field.getType(), offset + fieldOrdinal);
final RelDataTypeField aliasField = frame.fields().get(fieldOrdinal);
if (!alias || field.getName().equals(aliasField.getName())) {
return ref;
} else {
return alias(ref, aliasField.getName());
}
}
/** Creates a reference to a field of the current record which originated
* in a relation with a given alias. */
public RexNode field(String alias, String fieldName) {
return field(1, alias, fieldName);
}
/** Creates a reference to a field which originated in a relation with the
* given alias. Searches for the relation starting at the top of the
* stack. */
public RexNode field(int inputCount, String alias, String fieldName) {
requireNonNull(alias, "alias");
requireNonNull(fieldName, "fieldName");
for (int inputOrdinal = 0; inputOrdinal < inputCount; ++inputOrdinal) {
final Frame frame = peek_(inputOrdinal);
final List<ImmutableSet<String>> aliasSets = frame.fields.leftList();
final List<RelDataTypeField> fields = frame.fields.rightList();
for (int i = 0, n = frame.fields.size(); i < n; i++) {
// If alias and field name match, reference that field.
if (aliasSets.get(i).contains(alias)
&& fields.get(i).getName().equals(fieldName)) {
return field(inputCount, inputCount - 1 - inputOrdinal, i);
}
}
}
// Not found. Build a message.
StringBuilder b =
new StringBuilder().append("{alias=").append(alias)
.append(",fieldName=").append(fieldName)
.append("} field not found; fields are: ");
for (int inputOrdinal = 0; inputOrdinal < inputCount; ++inputOrdinal) {
final Frame frame = peek_(inputOrdinal);
frame.fields.forEach((aliasSet, field) ->
b.append(
String.format(Locale.ROOT, "{aliases=%s,fieldName=%s}",
aliasSet, field.getName())));
}
throw new IllegalArgumentException(b.toString());
}
/** Returns a reference to a given field (by name, case-insensitive)
* of a record-valued expression. */
public RexNode field(RexNode e, String name) {
return getRexBuilder().makeFieldAccess(e, name, false);
}
/** Returns a reference to a given field (by ordinal)
* of a record-valued expression. */
public RexNode field(RexNode e, int ordinal) {
return getRexBuilder().makeFieldAccess(e, ordinal);
}
/** Returns references to the fields of the top input. */
public ImmutableList<RexNode> fields() {
return fields(1, 0);
}
/** Returns references to the fields of a given input. */
public ImmutableList<RexNode> fields(int inputCount, int inputOrdinal) {
final RelNode input = peek(inputCount, inputOrdinal);
final RelDataType rowType = input.getRowType();
final ImmutableList.Builder<RexNode> nodes = ImmutableList.builder();
for (int fieldOrdinal : Util.range(rowType.getFieldCount())) {
nodes.add(field(inputCount, inputOrdinal, fieldOrdinal));
}
return nodes.build();
}
/** Returns references to fields for a given collation. */
public ImmutableList<RexNode> fields(RelCollation collation) {
final ImmutableList.Builder<RexNode> nodes = ImmutableList.builder();
for (RelFieldCollation fieldCollation : collation.getFieldCollations()) {
RexNode node = field(fieldCollation.getFieldIndex());
switch (fieldCollation.direction) {
case DESCENDING:
node = desc(node);
break;
default:
break;
}
switch (fieldCollation.nullDirection) {
case FIRST:
node = nullsFirst(node);
break;
case LAST:
node = nullsLast(node);
break;
default:
break;
}
nodes.add(node);
}
return nodes.build();
}
/** Returns references to fields for a given list of input ordinals. */
public ImmutableList<RexNode> fields(List<? extends Number> ordinals) {
final ImmutableList.Builder<RexNode> nodes = ImmutableList.builder();
for (Number ordinal : ordinals) {
RexNode node = field(1, 0, ordinal.intValue(), false);
nodes.add(node);
}
return nodes.build();
}
/** Returns references to fields for a given bit set of input ordinals. */
public ImmutableList<RexNode> fields(ImmutableBitSet ordinals) {
return fields(ordinals.asList());
}
/** Returns references to fields identified by name. */
public ImmutableList<RexNode> fields(Iterable<String> fieldNames) {
final ImmutableList.Builder<RexNode> builder = ImmutableList.builder();
for (String fieldName : fieldNames) {
builder.add(field(fieldName));
}
return builder.build();
}
/** Returns references to fields identified by a mapping. */
public ImmutableList<RexNode> fields(Mappings.TargetMapping mapping) {
return fields(Mappings.asListNonNull(mapping));
}
/** Creates an access to a field by name. */
public RexNode dot(RexNode node, String fieldName) {
final RexBuilder builder = cluster.getRexBuilder();
return builder.makeFieldAccess(node, fieldName, true);
}
/** Creates an access to a field by ordinal. */
public RexNode dot(RexNode node, int fieldOrdinal) {
final RexBuilder builder = cluster.getRexBuilder();
return builder.makeFieldAccess(node, fieldOrdinal);
}
/** Creates a call to a scalar operator. */
public RexNode call(SqlOperator operator, RexNode... operands) {
return call(operator, ImmutableList.copyOf(operands));
}
/** Creates a call to a scalar operator. */
private RexCall call(SqlParserPos pos, SqlOperator operator, List<RexNode> operandList) {
switch (operator.getKind()) {
case LIKE:
case SIMILAR:
final SqlLikeOperator likeOperator = (SqlLikeOperator) operator;
if (likeOperator.isNegated()) {
final SqlOperator notLikeOperator = likeOperator.not();
return (RexCall) not(call(notLikeOperator, operandList));
}
break;
case BETWEEN:
assert operandList.size() == 3;
return (RexCall) between(operandList.get(0), operandList.get(1),
operandList.get(2));
default:
break;
}
final RexBuilder builder = cluster.getRexBuilder();
final RelDataType type = builder.deriveReturnType(operator, operandList);
return (RexCall) builder.makeCall(pos, type, operator, operandList);
}
/** Creates a call to a scalar operator. */
public RexNode call(SqlOperator operator,
Iterable<? extends RexNode> operands) {
return call(SqlParserPos.ZERO, operator, ImmutableList.copyOf(operands));
}
/** Creates an IN predicate with a list of values.
*
* <p>For example,
* <pre>{@code
* b.scan("Emp")
* .filter(b.in(b.field("deptno"), b.literal(10), b.literal(20)))
* }</pre>
* is equivalent to SQL
* <pre>{@code
* SELECT *
* FROM Emp
* WHERE deptno IN (10, 20)
* }</pre> */
public RexNode in(RexNode arg, RexNode... ranges) {
return in(arg, ImmutableList.copyOf(ranges));
}
/** Creates an IN predicate with a list of values.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Emps")
* .filter(
* b.in(b.field("deptno"),
* Arrays.asList(b.literal(10), b.literal(20))))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT *
* FROM Emps
* WHERE deptno IN (10, 20)
* }</pre> */
public RexNode in(RexNode arg, Iterable<? extends RexNode> ranges) {
return getRexBuilder().makeIn(arg, ImmutableList.copyOf(ranges));
}
/** Creates an IN predicate with a sub-query. */
@Experimental
public RexSubQuery in(RelNode rel, Iterable<? extends RexNode> nodes) {
return RexSubQuery.in(rel, ImmutableList.copyOf(nodes));
}
/** Creates an IN predicate with a sub-query.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Emps")
* .filter(
* b.in(b.field("deptno"),
* b2 -> b2.scan("Depts")
* .filter(
* b2.equals(b2.field("location"), b2.literal("Boston")))
* .project(b.field("deptno"))
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT *
* FROM Emps
* WHERE deptno IN (SELECT deptno FROM Dept WHERE location = 'Boston')
* }</pre> */
@Experimental
public RexNode in(RexNode arg, Function<RelBuilder, RelNode> f) {
final RelNode rel = f.apply(this);
return RexSubQuery.in(rel, ImmutableList.of(arg));
}
/** Creates a SOME (or ANY) predicate.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Emps")
* .filter(
* b.some(b.field("commission"),
* SqlStdOperatorTable.GREATER_THAN,
* b2 -> b2.scan("Emps")
* .filter(
* b2.equals(b2.field("job"), b2.literal("Manager")))
* .project(b2.field("sal"))
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT *
* FROM Emps
* WHERE commission > SOME (SELECT sal FROM Emps WHERE job = 'Manager')
* }</pre>
*
* <p>or (since {@code SOME} and {@code ANY} are synonyms) the SQL
*
* <pre>{@code
* SELECT *
* FROM Emps
* WHERE commission > ANY (SELECT sal FROM Emps WHERE job = 'Manager')
* }</pre> */
@Experimental
public RexSubQuery some(RexNode node, SqlOperator op,
Function<RelBuilder, RelNode> f) {
return some_(node, op.kind, f);
}
private RexSubQuery some_(RexNode node, SqlKind kind,
Function<RelBuilder, RelNode> f) {
final RelNode rel = f.apply(this);
final SqlQuantifyOperator quantifyOperator =
SqlStdOperatorTable.some(kind);
return RexSubQuery.some(rel, ImmutableList.of(node), quantifyOperator);
}
/** Creates an ALL predicate.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Emps")
* .filter(
* b.all(b.field("commission"),
* SqlStdOperatorTable.GREATER_THAN,
* b2 -> b2.scan("Emps")
* .filter(
* b2.equals(b2.field("job"), b2.literal("Manager")))
* .project(b2.field("sal"))
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT *
* FROM Emps
* WHERE commission > ALL (SELECT sal FROM Emps WHERE job = 'Manager')
* }</pre>
*
* <p>Calcite translates {@code ALL} predicates to {@code NOT SOME}. The
* following SQL is equivalent to the previous:
*
* <pre>{@code
* SELECT *
* FROM Emps
* WHERE NOT (commission <= SOME (SELECT sal FROM Emps WHERE job = 'Manager'))
* }</pre> */
@Experimental
public RexNode all(RexNode node, SqlOperator op,
Function<RelBuilder, RelNode> f) {
return not(some_(node, op.kind.negateNullSafe(), f));
}
/** Creates an EXISTS predicate.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Depts")
* .filter(
* b.exists(b2 ->
* b2.scan("Emps")
* .filter(
* b2.equals(b2.field("job"), b2.literal("Manager")))
* .project(b2.literal(1))
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT *
* FROM Depts
* WHERE EXISTS (SELECT 1 FROM Emps WHERE job = 'Manager')
* }</pre> */
@Experimental
public RexSubQuery exists(Function<RelBuilder, RelNode> f) {
final RelNode rel = f.apply(this);
return RexSubQuery.exists(rel);
}
/** Creates a UNIQUE predicate.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Depts")
* .filter(
* b.unique(b2 ->
* b2.scan("Emps")
* .filter(
* b2.equals(b2.field("job"), b2.literal("Manager")))
* .project(b2.field("deptno")
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT *
* FROM Depts
* WHERE UNIQUE (SELECT deptno FROM Emps WHERE job = 'Manager')
* }</pre> */
@Experimental
public RexSubQuery unique(Function<RelBuilder, RelNode> f) {
final RelNode rel = f.apply(this);
return RexSubQuery.unique(rel);
}
/** Creates a scalar sub-query.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Depts")
* .project(
* b.field("deptno")
* b.scalarQuery(b2 ->
* b2.scan("Emps")
* .aggregate(
* b2.groupKey(), b2.max(b2.field("sal")))
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT deptno, (SELECT MAX(sal) FROM Emps)
* FROM Depts
* }</pre> */
@Experimental
public RexSubQuery scalarQuery(Function<RelBuilder, RelNode> f) {
return RexSubQuery.scalar(f.apply(this));
}
/** Creates an ARRAY sub-query.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Depts")
* .project(
* b.field("deptno")
* b.arrayQuery(b2 ->
* b2.scan("Emps")
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT deptno, ARRAY (SELECT * FROM Emps)
* FROM Depts
* }</pre> */
@Experimental
public RexSubQuery arrayQuery(Function<RelBuilder, RelNode> f) {
return RexSubQuery.array(f.apply(this));
}
/** Creates a MULTISET sub-query.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Depts")
* .project(
* b.field("deptno")
* b.multisetQuery(b2 ->
* b2.scan("Emps")
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT deptno, MULTISET (SELECT * FROM Emps)
* FROM Depts
* }</pre> */
@Experimental
public RexSubQuery multisetQuery(Function<RelBuilder, RelNode> f) {
return RexSubQuery.multiset(f.apply(this));
}
/** Creates a MAP sub-query.
*
* <p>For example,
*
* <pre>{@code
* b.scan("Depts")
* .project(
* b.field("deptno")
* b.mapQuery(b2 ->
* b2.scan("Emps")
* .project(b2.field("empno"), b2.field("job"))
* .build()))
* }</pre>
*
* <p>is equivalent to the SQL
*
* <pre>{@code
* SELECT deptno, MAP (SELECT empno, job FROM Emps)
* FROM Depts
* }</pre> */
@Experimental
public RexSubQuery mapQuery(Function<RelBuilder, RelNode> f) {
return RexSubQuery.map(f.apply(this));
}
/** Creates an AND. */
public RexNode and(RexNode... operands) {
return and(ImmutableList.copyOf(operands));
}
/** Creates an AND.
*
* <p>Simplifies the expression a little:
* {@code e AND TRUE} becomes {@code e};
* {@code e AND e2 AND NOT e} becomes {@code e2}. */
public RexNode and(Iterable<? extends RexNode> operands) {
return RexUtil.composeConjunction(getRexBuilder(), operands);
}
/** Creates an OR. */
public RexNode or(RexNode... operands) {
return or(ImmutableList.copyOf(operands));
}
/** Creates an OR. */
public RexNode or(Iterable<? extends RexNode> operands) {
return RexUtil.composeDisjunction(cluster.getRexBuilder(), operands);
}
/** Creates a NOT. */
public RexNode not(RexNode operand) {
return call(SqlStdOperatorTable.NOT, operand);
}
/** Creates an {@code =}. */
public RexNode equals(RexNode operand0, RexNode operand1) {
return call(SqlStdOperatorTable.EQUALS, operand0, operand1);
}
/** Creates a {@code >}. */
public RexNode greaterThan(RexNode operand0, RexNode operand1) {
return call(SqlStdOperatorTable.GREATER_THAN, operand0, operand1);
}
/** Creates a {@code >=}. */
public RexNode greaterThanOrEqual(RexNode operand0, RexNode operand1) {
return call(SqlStdOperatorTable.GREATER_THAN_OR_EQUAL, operand0, operand1);
}
/** Creates a {@code <}. */
public RexNode lessThan(RexNode operand0, RexNode operand1) {
return call(SqlStdOperatorTable.LESS_THAN, operand0, operand1);
}
/** Creates a {@code <=}. */
public RexNode lessThanOrEqual(RexNode operand0, RexNode operand1) {
return call(SqlStdOperatorTable.LESS_THAN_OR_EQUAL, operand0, operand1);
}
/** Creates a {@code <>}. */
public RexNode notEquals(RexNode operand0, RexNode operand1) {
return call(SqlStdOperatorTable.NOT_EQUALS, operand0, operand1);
}
/** Creates an expression equivalent to "{@code o0 IS NOT DISTINCT FROM o1}".
* It is also equivalent to
* "{@code o0 = o1 OR (o0 IS NULL AND o1 IS NULL)}". */
public RexNode isNotDistinctFrom(RexNode operand0, RexNode operand1) {
return RelOptUtil.isDistinctFrom(getRexBuilder(), operand0, operand1, true);
}
/** Creates an expression equivalent to {@code o0 IS DISTINCT FROM o1}.
* It is also equivalent to
* "{@code NOT (o0 = o1 OR (o0 IS NULL AND o1 IS NULL))}. */
public RexNode isDistinctFrom(RexNode operand0, RexNode operand1) {
return RelOptUtil.isDistinctFrom(getRexBuilder(), operand0, operand1, false);
}
/** Creates a {@code BETWEEN}. */
public RexNode between(RexNode arg, RexNode lower, RexNode upper) {
return getRexBuilder().makeBetween(arg, lower, upper);
}
/** Creates ab {@code IS NULL}. */
public RexNode isNull(RexNode operand) {
return call(SqlStdOperatorTable.IS_NULL, operand);
}
/** Creates an {@code IS NOT NULL}. */
public RexNode isNotNull(RexNode operand) {
return call(SqlStdOperatorTable.IS_NOT_NULL, operand);
}
/** Creates an expression that casts an expression to a given type. */
public RexNode cast(RexNode expr, SqlTypeName typeName) {
return cast(SqlParserPos.ZERO, expr, typeName);
}
/** Creates an expression that casts an expression to a given type. */
public RexNode cast(SqlParserPos pos, RexNode expr, SqlTypeName typeName) {
final RelDataType type = cluster.getTypeFactory().createSqlType(typeName);
return cluster.getRexBuilder().makeCast(pos, type, expr);
}
/** Creates an expression that casts an expression to a type with a given name
* and precision or length. */
public RexNode cast(RexNode expr, SqlTypeName typeName, int precision) {
return cast(SqlParserPos.ZERO, expr, typeName, precision);
}
/** Creates an expression that casts an expression to a type with a given name
* and precision or length. */
public RexNode cast(SqlParserPos pos, RexNode expr, SqlTypeName typeName, int precision) {
final RelDataType type =
cluster.getTypeFactory().createSqlType(typeName, precision);
return cluster.getRexBuilder().makeCast(pos, type, expr);
}
/** Creates an expression that casts an expression to a type with a given
* name, precision and scale. */
public RexNode cast(RexNode expr, SqlTypeName typeName, int precision,
int scale) {
return cast(SqlParserPos.ZERO, expr, typeName, precision, scale);
}
/** Creates an expression that casts an expression to a type with a given
* name, precision and scale. */
public RexNode cast(SqlParserPos pos, RexNode expr, SqlTypeName typeName, int precision,
int scale) {
final RelDataType type =
cluster.getTypeFactory().createSqlType(typeName, precision, scale);
return cluster.getRexBuilder().makeCast(pos, type, expr);
}
/**
* Returns an expression wrapped in an alias.
*
* <p>This method is idempotent: If the expression is already wrapped in the
* correct alias, does nothing; if wrapped in an incorrect alias, removes
* the incorrect alias and applies the correct alias.
*
* @see #project
*/
public RexNode alias(RexNode expr, String alias) {
final RexNode aliasLiteral = literal(alias);
switch (expr.getKind()) {
case AS:
final RexCall call = (RexCall) expr;
if (call.operands.get(1).equals(aliasLiteral)) {
// current alias is correct
return expr;
}
expr = call.operands.get(0);
// strip current (incorrect) alias, and fall through
default:
return call(SqlStdOperatorTable.AS, expr, aliasLiteral);
}
}
private RexNode aliasMaybe(RexNode node, @Nullable String name) {
return name == null ? node : alias(node, name);
}
/** Converts a sort expression to descending. */
public RexNode desc(RexNode node) {
return call(SqlStdOperatorTable.DESC, node);
}
/** Converts a sort expression to nulls last. */
public RexNode nullsLast(RexNode node) {
return call(SqlStdOperatorTable.NULLS_LAST, node);
}
/** Converts a sort expression to nulls first. */
public RexNode nullsFirst(RexNode node) {
return call(SqlStdOperatorTable.NULLS_FIRST, node);
}
// Methods that create window bounds
/** Creates an {@code UNBOUNDED PRECEDING} window bound,
* for use in methods such as {@link OverCall#rowsFrom(RexWindowBound)}
* and {@link OverCall#rangeBetween(RexWindowBound, RexWindowBound)}. */
public RexWindowBound unboundedPreceding() {
return RexWindowBounds.UNBOUNDED_PRECEDING;
}
/** Creates a {@code bound PRECEDING} window bound,
* for use in methods such as {@link OverCall#rowsFrom(RexWindowBound)}
* and {@link OverCall#rangeBetween(RexWindowBound, RexWindowBound)}. */
public RexWindowBound preceding(RexNode bound) {
return RexWindowBounds.preceding(bound);
}
/** Creates a {@code CURRENT ROW} window bound,
* for use in methods such as {@link OverCall#rowsFrom(RexWindowBound)}
* and {@link OverCall#rangeBetween(RexWindowBound, RexWindowBound)}. */
public RexWindowBound currentRow() {
return RexWindowBounds.CURRENT_ROW;
}
/** Creates a {@code bound FOLLOWING} window bound,
* for use in methods such as {@link OverCall#rowsFrom(RexWindowBound)}
* and {@link OverCall#rangeBetween(RexWindowBound, RexWindowBound)}. */
public RexWindowBound following(RexNode bound) {
return RexWindowBounds.following(bound);
}
/** Creates an {@code UNBOUNDED FOLLOWING} window bound,
* for use in methods such as {@link OverCall#rowsFrom(RexWindowBound)}
* and {@link OverCall#rangeBetween(RexWindowBound, RexWindowBound)}. */
public RexWindowBound unboundedFollowing() {
return RexWindowBounds.UNBOUNDED_FOLLOWING;
}
// Methods that create group keys and aggregate calls
/** Creates an empty group key. */
public GroupKey groupKey() {
return groupKey(ImmutableList.of());
}
/** Creates a group key. */
public GroupKey groupKey(RexNode... nodes) {
return groupKey(ImmutableList.copyOf(nodes));
}
/** Creates a group key. */
public GroupKey groupKey(Iterable<? extends RexNode> nodes) {
return new GroupKeyImpl(ImmutableList.copyOf(nodes), null, null);
}
/** Creates a group key with grouping sets. */
public GroupKey groupKey(Iterable<? extends RexNode> nodes,
Iterable<? extends Iterable<? extends RexNode>> nodeLists) {
return groupKey_(nodes, nodeLists);
}
// CHECKSTYLE: IGNORE 1
/** @deprecated Now that indicator is deprecated, use
* {@link #groupKey(Iterable, Iterable)}, which has the same behavior as
* calling this method with {@code indicator = false}. */
@Deprecated // to be removed before 2.0
public GroupKey groupKey(Iterable<? extends RexNode> nodes, boolean indicator,
Iterable<? extends Iterable<? extends RexNode>> nodeLists) {
Aggregate.checkIndicator(indicator);
return groupKey_(nodes, nodeLists);
}
private static GroupKey groupKey_(Iterable<? extends RexNode> nodes,
Iterable<? extends Iterable<? extends RexNode>> nodeLists) {
final ImmutableList.Builder<ImmutableList<RexNode>> builder =
ImmutableList.builder();
for (Iterable<? extends RexNode> nodeList : nodeLists) {
builder.add(ImmutableList.copyOf(nodeList));
}
return new GroupKeyImpl(ImmutableList.copyOf(nodes), builder.build(), null);
}
/** Creates a group key of fields identified by ordinal. */
public GroupKey groupKey(int... fieldOrdinals) {
return groupKey(fields(ImmutableIntList.of(fieldOrdinals)));
}
/** Creates a group key of fields identified by name. */
public GroupKey groupKey(String... fieldNames) {
return groupKey(fields(ImmutableList.copyOf(fieldNames)));
}
/** Creates a group key, identified by field positions
* in the underlying relational expression.
*
* <p>This method of creating a group key does not allow you to group on new
* expressions, only column projections, but is efficient, especially when you
* are coming from an existing {@link Aggregate}. */
public GroupKey groupKey(ImmutableBitSet groupSet) {
return groupKey_(groupSet, ImmutableList.of(groupSet));
}
/** Creates a group key with grouping sets, both identified by field positions
* in the underlying relational expression.
*
* <p>This method of creating a group key does not allow you to group on new
* expressions, only column projections, but is efficient, especially when you
* are coming from an existing {@link Aggregate}.
*
* <p>It is possible for {@code groupSet} to be strict superset of all
* {@code groupSets}. For example, in the pseudo SQL
*
* <pre>{@code
* GROUP BY 0, 1, 2
* GROUPING SETS ((0, 1), 0)
* }</pre>
*
* <p>column 2 does not appear in either grouping set. This is not valid SQL.
* We can approximate in actual SQL by adding an extra grouping set and
* filtering out using {@code HAVING}, as follows:
*
* <pre>{@code
* GROUP BY GROUPING SETS ((0, 1, 2), (0, 1), 0)
* HAVING GROUPING_ID(0, 1, 2) <> 0
* }</pre>
*/
public GroupKey groupKey(ImmutableBitSet groupSet,
Iterable<? extends ImmutableBitSet> groupSets) {
return groupKey_(groupSet, ImmutableList.copyOf(groupSets));
}
// CHECKSTYLE: IGNORE 1
/** @deprecated Use {@link #groupKey(ImmutableBitSet, Iterable)}. */
@Deprecated // to be removed before 2.0
public GroupKey groupKey(ImmutableBitSet groupSet, boolean indicator,
@Nullable ImmutableList<ImmutableBitSet> groupSets) {
Aggregate.checkIndicator(indicator);
return groupKey_(groupSet, groupSets == null
? ImmutableList.of(groupSet) : ImmutableList.copyOf(groupSets));
}
private GroupKey groupKey_(ImmutableBitSet groupSet,
ImmutableList<ImmutableBitSet> groupSets) {
if (groupSet.length() > peek().getRowType().getFieldCount()) {
throw new IllegalArgumentException("out of bounds: " + groupSet);
}
requireNonNull(groupSets, "groupSets");
final ImmutableList<RexNode> nodes = fields(groupSet);
return groupKey_(nodes, Util.transform(groupSets, this::fields));
}
@Deprecated // to be removed before 2.0
public AggCall aggregateCall(SqlAggFunction aggFunction, boolean distinct,
RexNode filter, @Nullable String alias, RexNode... operands) {
return aggregateCall(aggFunction, distinct, false, false, filter, null,
ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
@Deprecated // to be removed before 2.0
public AggCall aggregateCall(SqlAggFunction aggFunction, boolean distinct,
boolean approximate, RexNode filter, @Nullable String alias,
RexNode... operands) {
return aggregateCall(aggFunction, distinct, approximate, false, filter,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
@Deprecated // to be removed before 2.0
public AggCall aggregateCall(SqlAggFunction aggFunction, boolean distinct,
RexNode filter, @Nullable String alias,
Iterable<? extends RexNode> operands) {
return aggregateCall(aggFunction, distinct, false, false, filter, null,
ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
@Deprecated // to be removed before 2.0
public AggCall aggregateCall(SqlAggFunction aggFunction, boolean distinct,
boolean approximate, RexNode filter, @Nullable String alias,
Iterable<? extends RexNode> operands) {
return aggregateCall(aggFunction, distinct, approximate, false, filter,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to an aggregate function.
*
* <p>To add other operands, apply
* {@link AggCall#distinct()},
* {@link AggCall#approximate(boolean)},
* {@link AggCall#filter(RexNode...)},
* {@link AggCall#sort},
* {@link AggCall#as} to the result. */
public AggCall aggregateCall(SqlAggFunction aggFunction,
Iterable<? extends RexNode> operands) {
return aggregateCall(aggFunction, false, false, false, null, null,
ImmutableList.of(), null, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to an aggregate function.
*
* <p>To add other operands, apply
* {@link AggCall#distinct()},
* {@link AggCall#approximate(boolean)},
* {@link AggCall#filter(RexNode...)},
* {@link AggCall#sort},
* {@link AggCall#as} to the result. */
public AggCall aggregateCall(SqlParserPos pos, SqlAggFunction aggFunction,
Iterable<? extends RexNode> operands) {
return aggregateCall(pos, aggFunction, false, false, false, null, null,
ImmutableList.of(), null, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to an aggregate function.
*
* <p>To add other operands, apply
* {@link AggCall#distinct()},
* {@link AggCall#approximate(boolean)},
* {@link AggCall#filter(RexNode...)},
* {@link AggCall#sort},
* {@link AggCall#as} to the result. */
public AggCall aggregateCall(SqlParserPos pos, SqlAggFunction aggFunction,
RexNode... operands) {
return aggregateCall(pos, aggFunction, false, false, false, null, null,
ImmutableList.of(), null, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to an aggregate function.
*
* <p>To add other operands, apply
* {@link AggCall#distinct()},
* {@link AggCall#approximate(boolean)},
* {@link AggCall#filter(RexNode...)},
* {@link AggCall#sort},
* {@link AggCall#as} to the result. */
public AggCall aggregateCall(SqlAggFunction aggFunction,
RexNode... operands) {
return aggregateCall(aggFunction, false, false, false, null, null,
ImmutableList.of(), null, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to an aggregate function as a copy of an
* {@link AggregateCall}. */
public AggCall aggregateCall(AggregateCall a) {
return aggregateCall(a.getParserPosition(),
a.getAggregation(), a.isDistinct(), a.isApproximate(),
a.ignoreNulls(), a.filterArg < 0 ? null : field(a.filterArg),
a.distinctKeys == null ? null : fields(a.distinctKeys),
fields(a.collation), a.name, ImmutableList.copyOf(a.rexList),
fields(a.getArgList()));
}
/** Creates a call to an aggregate function as a copy of an
* {@link AggregateCall}, applying a mapping. */
public AggCall aggregateCall(AggregateCall a, Mapping mapping) {
return aggregateCall(a.getParserPosition(),
a.getAggregation(), a.isDistinct(), a.isApproximate(),
a.ignoreNulls(),
a.filterArg < 0 ? null : field(Mappings.apply(mapping, a.filterArg)),
a.distinctKeys == null ? null
: fields(Mappings.apply(mapping, a.distinctKeys)),
fields(RexUtil.apply(mapping, a.collation)), a.name,
ImmutableList.copyOf(a.rexList),
fields(Mappings.apply2(mapping, a.getArgList())));
}
/** Creates a call to an aggregate function with all applicable operands. */
protected AggCall aggregateCall(SqlAggFunction aggFunction, boolean distinct,
boolean approximate, boolean ignoreNulls, @Nullable RexNode filter,
@Nullable ImmutableList<RexNode> distinctKeys,
ImmutableList<RexNode> orderKeys, @Nullable String alias,
ImmutableList<RexNode> preOperands, ImmutableList<RexNode> operands) {
return new AggCallImpl(SqlParserPos.ZERO, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeys);
}
/** Creates a call to an aggregate function with all applicable operands. */
protected AggCall aggregateCall(SqlParserPos pos, SqlAggFunction aggFunction, boolean distinct,
boolean approximate, boolean ignoreNulls, @Nullable RexNode filter,
@Nullable ImmutableList<RexNode> distinctKeys,
ImmutableList<RexNode> orderKeys, @Nullable String alias,
ImmutableList<RexNode> preOperands, ImmutableList<RexNode> operands) {
return new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeys);
}
/** Creates a call to the {@code COUNT} aggregate function. */
public AggCall count(RexNode... operands) {
return count(false, null, operands);
}
/** Creates a call to the {@code COUNT} aggregate function. */
public AggCall count(Iterable<? extends RexNode> operands) {
return count(false, null, operands);
}
/** Creates a call to the {@code COUNT} aggregate function,
* optionally distinct and with an alias. */
public AggCall count(boolean distinct, @Nullable String alias,
RexNode... operands) {
return aggregateCall(SqlStdOperatorTable.COUNT, distinct, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to the {@code COUNT} aggregate function,
* optionally distinct and with an alias. */
public AggCall count(boolean distinct, @Nullable String alias,
Iterable<? extends RexNode> operands) {
return aggregateCall(SqlStdOperatorTable.COUNT, distinct, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands));
}
/** Creates a call to the {@code COUNT(*)} aggregate function. */
public AggCall countStar(@Nullable String alias) {
return count(false, alias);
}
/** Creates a call to the {@code SUM} aggregate function. */
public AggCall sum(RexNode operand) {
return sum(false, null, operand);
}
/** Creates a call to the {@code SUM} aggregate function. */
public AggCall sum(SqlParserPos pos, RexNode operand) {
return sum(pos, false, null, operand);
}
/** Creates a call to the {@code SUM} aggregate function,
* optionally distinct and with an alias. */
public AggCall sum(boolean distinct, @Nullable String alias,
RexNode operand) {
return aggregateCall(SqlStdOperatorTable.SUM, distinct, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.of(operand));
}
/** Creates a call to the {@code SUM} aggregate function,
* optionally distinct and with an alias. */
public AggCall sum(SqlParserPos pos, boolean distinct, @Nullable String alias,
RexNode operand) {
return aggregateCall(pos, SqlStdOperatorTable.SUM, distinct, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.of(operand));
}
/** Creates a call to the {@code AVG} aggregate function. */
public AggCall avg(RexNode operand) {
return avg(false, null, operand);
}
/** Creates a call to the {@code AVG} aggregate function. */
public AggCall avg(SqlParserPos pos, RexNode operand) {
return avg(pos, false, null, operand);
}
/** Creates a call to the {@code AVG} aggregate function,
* optionally distinct and with an alias. */
public AggCall avg(boolean distinct, @Nullable String alias,
RexNode operand) {
return aggregateCall(SqlParserPos.ZERO, SqlStdOperatorTable.AVG, distinct, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.of(operand));
}
/** Creates a call to the {@code AVG} aggregate function,
* optionally distinct and with an alias. */
public AggCall avg(SqlParserPos pos, boolean distinct, @Nullable String alias,
RexNode operand) {
return aggregateCall(pos, SqlStdOperatorTable.AVG, distinct, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.of(operand));
}
/** Creates a call to the {@code MIN} aggregate function. */
public AggCall min(RexNode operand) {
return min(null, operand);
}
/** Creates a call to the {@code MIN} aggregate function,
* optionally with an alias. */
public AggCall min(@Nullable String alias, RexNode operand) {
return aggregateCall(SqlStdOperatorTable.MIN, false, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.of(operand));
}
/** Creates a call to the {@code MAX} aggregate function,
* optionally with an alias. */
public AggCall max(RexNode operand) {
return max(null, operand);
}
/** Creates a call to the {@code MAX} aggregate function. */
public AggCall max(@Nullable String alias, RexNode operand) {
return aggregateCall(SqlStdOperatorTable.MAX, false, false, false, null,
null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.of(operand));
}
/** Creates a call to the {@code LITERAL_AGG} aggregate function. */
public AggCall literalAgg(@Nullable Object value) {
return aggregateCall(SqlInternalOperators.LITERAL_AGG)
.preOperands(literal(value));
}
/** Creates a call to the {@code LITERAL_AGG} aggregate function.
* optionally an alias. */
public AggCall literalAgg(@Nullable Object value, @Nullable String alias, RexNode... operands) {
return aggregateCall(SqlInternalOperators.LITERAL_AGG, false, false, false,
null, null, ImmutableList.of(), alias, ImmutableList.of(),
ImmutableList.copyOf(operands)).preOperands(literal(value));
}
// Methods for patterns
/**
* Creates a reference to a given field of the pattern.
*
* @param alpha the pattern name
* @param type Type of field
* @param i Ordinal of field
* @return Reference to field of pattern
*/
public RexNode patternField(String alpha, RelDataType type, int i) {
return getRexBuilder().makePatternFieldRef(alpha, type, i);
}
/** Creates a call that concatenates patterns;
* for use in {@link #match}. */
public RexNode patternConcat(Iterable<? extends RexNode> nodes) {
final ImmutableList<RexNode> list = ImmutableList.copyOf(nodes);
if (list.size() > 2) {
// Convert into binary calls
return patternConcat(patternConcat(Util.skipLast(list)), Util.last(list));
}
final RelDataType t = getTypeFactory().createSqlType(SqlTypeName.NULL);
return getRexBuilder().makeCall(t, SqlStdOperatorTable.PATTERN_CONCAT,
list);
}
/** Creates a call that concatenates patterns;
* for use in {@link #match}. */
public RexNode patternConcat(RexNode... nodes) {
return patternConcat(ImmutableList.copyOf(nodes));
}
/** Creates a call that creates alternate patterns;
* for use in {@link #match}. */
public RexNode patternAlter(Iterable<? extends RexNode> nodes) {
final RelDataType t = getTypeFactory().createSqlType(SqlTypeName.NULL);
return getRexBuilder().makeCall(t, SqlStdOperatorTable.PATTERN_ALTER,
ImmutableList.copyOf(nodes));
}
/** Creates a call that creates alternate patterns;
* for use in {@link #match}. */
public RexNode patternAlter(RexNode... nodes) {
return patternAlter(ImmutableList.copyOf(nodes));
}
/** Creates a call that creates quantify patterns;
* for use in {@link #match}. */
public RexNode patternQuantify(Iterable<? extends RexNode> nodes) {
final RelDataType t = getTypeFactory().createSqlType(SqlTypeName.NULL);
return getRexBuilder().makeCall(t, SqlStdOperatorTable.PATTERN_QUANTIFIER,
ImmutableList.copyOf(nodes));
}
/** Creates a call that creates quantify patterns;
* for use in {@link #match}. */
public RexNode patternQuantify(RexNode... nodes) {
return patternQuantify(ImmutableList.copyOf(nodes));
}
/** Creates a call that creates permute patterns;
* for use in {@link #match}. */
public RexNode patternPermute(Iterable<? extends RexNode> nodes) {
final RelDataType t = getTypeFactory().createSqlType(SqlTypeName.NULL);
return getRexBuilder().makeCall(t, SqlStdOperatorTable.PATTERN_PERMUTE,
ImmutableList.copyOf(nodes));
}
/** Creates a call that creates permute patterns;
* for use in {@link #match}. */
public RexNode patternPermute(RexNode... nodes) {
return patternPermute(ImmutableList.copyOf(nodes));
}
/** Creates a call that creates an exclude pattern;
* for use in {@link #match}. */
public RexNode patternExclude(RexNode node) {
final RelDataType t = getTypeFactory().createSqlType(SqlTypeName.NULL);
return getRexBuilder().makeCall(t, SqlStdOperatorTable.PATTERN_EXCLUDE,
ImmutableList.of(node));
}
// Methods that create relational expressions
/** Creates a {@link TableScan} of the table
* with a given name.
*
* <p>Throws if the table does not exist.
*
* <p>Returns this builder.
*
* @param tableNames Name of table (can optionally be qualified)
*/
public RelBuilder scan(Iterable<String> tableNames) {
final List<String> names = ImmutableList.copyOf(tableNames);
requireNonNull(relOptSchema, "relOptSchema");
final RelOptTable relOptTable = relOptSchema.getTableForMember(names);
if (relOptTable == null) {
throw RESOURCE.tableNotFound(String.join(".", names)).ex();
}
final RelNode scan =
struct.scanFactory.createScan(
ViewExpanders.toRelContext(viewExpander, cluster),
relOptTable);
push(scan);
rename(relOptTable.getRowType().getFieldNames());
// When the node is not a TableScan but from expansion,
// we need to explicitly add the alias.
if (!(scan instanceof TableScan)) {
as(Util.last(ImmutableList.copyOf(tableNames)));
}
return this;
}
/** Creates a {@link TableScan} of the table
* with a given name.
*
* <p>Throws if the table does not exist.
*
* <p>Returns this builder.
*
* @param tableNames Name of table (can optionally be qualified)
*/
public RelBuilder scan(String... tableNames) {
return scan(ImmutableList.copyOf(tableNames));
}
/** Creates a {@link Snapshot} of a given snapshot period.
*
* <p>Returns this builder.
*
* @param period Name of table (can optionally be qualified)
*/
public RelBuilder snapshot(RexNode period) {
final Frame frame = stack.pop();
final RelNode snapshot =
struct.snapshotFactory.createSnapshot(frame.rel, period);
stack.push(new Frame(snapshot, frame.fields));
return this;
}
/**
* Gets column mappings of the operator.
*
* @param op operator instance
* @return column mappings associated with this function
*/
private static @Nullable Set<RelColumnMapping> getColumnMappings(SqlOperator op) {
SqlReturnTypeInference inference = op.getReturnTypeInference();
if (inference instanceof TableFunctionReturnTypeInference) {
return ((TableFunctionReturnTypeInference) inference).getColumnMappings();
} else {
return null;
}
}
/**
* Creates a RexCall to the {@code CURSOR} function by ordinal.
*
* @param inputCount Number of inputs
* @param ordinal The reference to the relational input
* @return RexCall to CURSOR function
*/
public RexNode cursor(int inputCount, int ordinal) {
if (inputCount <= ordinal || ordinal < 0) {
throw new IllegalArgumentException("bad input count or ordinal");
}
// Refer to the "ordinal"th input as if it were a field
// (because that's how things are laid out inside a TableFunctionScan)
final RelNode input = peek(inputCount, ordinal);
return call(SqlStdOperatorTable.CURSOR,
getRexBuilder().makeInputRef(input.getRowType(), ordinal));
}
/** Creates a {@link TableFunctionScan}. */
public RelBuilder functionScan(SqlOperator operator,
int inputCount, RexNode... operands) {
return functionScan(operator, inputCount, ImmutableList.copyOf(operands));
}
/** Creates a {@link TableFunctionScan}. */
public RelBuilder functionScan(SqlOperator operator,
int inputCount, Iterable<? extends RexNode> operands) {
if (inputCount < 0 || inputCount > stack.size()) {
throw new IllegalArgumentException("bad input count");
}
// Gets inputs.
final List<RelNode> inputs = new ArrayList<>();
for (int i = 0; i < inputCount; i++) {
inputs.add(0, build());
}
final RexCall call = call(SqlParserPos.ZERO, operator, ImmutableList.copyOf(operands));
final RelNode functionScan =
struct.tableFunctionScanFactory.createTableFunctionScan(cluster,
inputs, call, null, getColumnMappings(operator));
push(functionScan);
return this;
}
/** Creates a {@link Filter} of an array of
* predicates.
*
* <p>The predicates are combined using AND,
* and optimized in a similar way to the {@link #and} method.
* If the result is TRUE no filter is created. */
public RelBuilder filter(RexNode... predicates) {
return filter(ImmutableSet.of(), ImmutableList.copyOf(predicates));
}
/** Creates a {@link Filter} of a list of
* predicates.
*
* <p>The predicates are combined using AND,
* and optimized in a similar way to the {@link #and} method.
* If the result is TRUE no filter is created. */
public RelBuilder filter(Iterable<? extends RexNode> predicates) {
return filter(ImmutableSet.of(), predicates);
}
/** Creates a {@link Filter} of a list of correlation variables
* and an array of predicates.
*
* <p>The predicates are combined using AND,
* and optimized in a similar way to the {@link #and} method.
* If the result is TRUE no filter is created. */
public RelBuilder filter(Iterable<CorrelationId> variablesSet,
RexNode... predicates) {
return filter(variablesSet, ImmutableList.copyOf(predicates));
}
/**
* Creates a {@link Filter} of a list of correlation variables
* and a list of predicates.
*
* <p>The predicates are combined using AND,
* and optimized in a similar way to the {@link #and} method.
* If simplification is on and the result is TRUE, no filter is created. */
public RelBuilder filter(Iterable<CorrelationId> variablesSet,
Iterable<? extends RexNode> predicates) {
final RexNode conjunctionPredicates;
if (config.simplify()) {
conjunctionPredicates = simplifier.simplifyFilterPredicates(predicates);
} else {
conjunctionPredicates =
RexUtil.composeConjunction(simplifier.rexBuilder, predicates);
}
if (conjunctionPredicates == null || conjunctionPredicates.isAlwaysFalse()) {
return empty();
}
if (conjunctionPredicates.isAlwaysTrue()) {
return this;
}
final Frame frame = stack.pop();
final RelNode filter =
struct.filterFactory.createFilter(frame.rel,
conjunctionPredicates, ImmutableSet.copyOf(variablesSet));
stack.push(new Frame(filter, frame.fields));
return this;
}
/** Creates a {@link Project} of the given
* expressions. */
public RelBuilder project(RexNode... nodes) {
return project(ImmutableList.copyOf(nodes));
}
/** Creates a {@link Project} of the given list
* of expressions.
*
* <p>Infers names as would {@link #project(Iterable, Iterable)} if all
* suggested names were null.
*
* @param nodes Expressions
*/
public RelBuilder project(Iterable<? extends RexNode> nodes) {
return project(nodes, ImmutableList.of());
}
/** Creates a {@link Project} of the given list
* of expressions and field names.
*
* @param nodes Expressions
* @param fieldNames field names for expressions
*/
public RelBuilder project(Iterable<? extends RexNode> nodes,
Iterable<? extends @Nullable String> fieldNames) {
return project(nodes, fieldNames, false);
}
/** Creates a {@link Project} of the given list
* of expressions, using the given names.
*
* <p>Names are deduced as follows:
* <ul>
* <li>If the length of {@code fieldNames} is greater than the index of
* the current entry in {@code nodes}, and the entry in
* {@code fieldNames} is not null, uses it; otherwise
* <li>If an expression projects an input field,
* or is a cast an input field,
* uses the input field name; otherwise
* <li>If an expression is a call to
* {@link SqlStdOperatorTable#AS}
* (see {@link #alias}), removes the call but uses the intended alias.
* </ul>
*
* <p>After the field names have been inferred, makes the
* field names unique by appending numeric suffixes.
*
* @param nodes Expressions
* @param fieldNames Suggested field names
* @param force create project even if it is identity
*/
public RelBuilder project(Iterable<? extends RexNode> nodes,
Iterable<? extends @Nullable String> fieldNames, boolean force) {
return project(nodes, fieldNames, force, ImmutableSet.of());
}
/**
* The same with {@link #project(Iterable, Iterable, boolean)}, with additional
* variablesSet param.
*
* @param nodes Expressions
* @param fieldNames Suggested field names
* @param force create project even if it is identity
* @param variablesSet Correlating variables that are set when reading a row
* from the input, and which may be referenced from the
* projection expressions
*/
public RelBuilder project(Iterable<? extends RexNode> nodes,
Iterable<? extends @Nullable String> fieldNames, boolean force,
Iterable<CorrelationId> variablesSet) {
return project_(nodes, fieldNames, ImmutableList.of(), force, variablesSet);
}
/** Creates a {@link Project} of all original fields, plus the given
* expressions. */
public RelBuilder projectPlus(RexNode... nodes) {
return projectPlus(ImmutableList.copyOf(nodes));
}
/** Creates a {@link Project} of all original fields, plus the given list of
* expressions. */
public RelBuilder projectPlus(Iterable<? extends RexNode> nodes) {
return project(Iterables.concat(fields(), nodes));
}
/** Creates a {@link Project} of all original fields, except the given
* expressions.
*
* @throws IllegalArgumentException if the given expressions contain duplicates
* or there is an expression that does not match an existing field
*/
public RelBuilder projectExcept(RexNode... expressions) {
return projectExcept(ImmutableList.copyOf(expressions));
}
/** Creates a {@link Project} of all original fields, except the given list of
* expressions.
*
* @throws IllegalArgumentException if the given expressions contain duplicates
* or there is an expression that does not match an existing field
*/
public RelBuilder projectExcept(Iterable<RexNode> expressions) {
List<RexNode> allExpressions = new ArrayList<>(fields());
Set<RexNode> excludeExpressions = new HashSet<>();
for (RexNode excludeExp : expressions) {
if (!excludeExpressions.add(excludeExp)) {
throw new IllegalArgumentException(
"Input list contains duplicates. Expression " + excludeExp + " exists multiple times.");
}
if (!allExpressions.remove(excludeExp)) {
throw new IllegalArgumentException("Expression " + excludeExp.toString() + " not found.");
}
}
return this.project(allExpressions);
}
/** Creates a {@link Project} of the given list
* of expressions, using the given names.
*
* <p>Names are deduced as follows:
* <ul>
* <li>If the length of {@code fieldNames} is greater than the index of
* the current entry in {@code nodes}, and the entry in
* {@code fieldNames} is not null, uses it; otherwise
* <li>If an expression projects an input field,
* or is a cast an input field,
* uses the input field name; otherwise
* <li>If an expression is a call to
* {@link SqlStdOperatorTable#AS}
* (see {@link #alias}), removes the call but uses the intended alias.
* </ul>
*
* <p>After the field names have been inferred, makes the
* field names unique by appending numeric suffixes.
*
* @param nodes Expressions
* @param fieldNames Suggested field names
* @param hints Hints
* @param force create project even if it is identity
*/
private RelBuilder project_(
Iterable<? extends RexNode> nodes,
Iterable<? extends @Nullable String> fieldNames,
Iterable<RelHint> hints,
boolean force,
Iterable<CorrelationId> variablesSet) {
final Frame frame = requireNonNull(peek_(), "frame stack is empty");
final RelDataType inputRowType = frame.rel.getRowType();
final List<RexNode> nodeList = Lists.newArrayList(nodes);
final Set<CorrelationId> variables = ImmutableSet.copyOf(variablesSet);
// Perform a quick check for identity. We'll do a deeper check
// later when we've derived column names.
if (!force && Iterables.isEmpty(fieldNames)
&& RexUtil.isIdentity(nodeList, inputRowType)) {
return this;
}
final List<@Nullable String> fieldNameList = Lists.newArrayList(fieldNames);
while (fieldNameList.size() < nodeList.size()) {
fieldNameList.add(null);
}
// Do not merge projection when top projection has correlation variables
bloat:
if (frame.rel instanceof Project
&& config.bloat() >= 0
&& variables.isEmpty()) {
final Project project = (Project) frame.rel;
// Populate field names. If the upper expression is an input ref and does
// not have a recommended name, use the name of the underlying field.
for (int i = 0; i < fieldNameList.size(); i++) {
if (fieldNameList.get(i) == null) {
final RexNode node = nodeList.get(i);
if (node instanceof RexInputRef) {
final RexInputRef ref = (RexInputRef) node;
fieldNameList.set(i,
project.getRowType().getFieldNames().get(ref.getIndex()));
}
}
}
final List<RexNode> newNodes =
RelOptUtil.pushPastProjectUnlessBloat(nodeList, project,
config.bloat());
if (newNodes == null) {
// The merged expression is more complex than the input expressions.
// Do not merge.
break bloat;
}
// Carefully build a list of fields, so that table aliases from the input
// can be seen for fields that are based on a RexInputRef.
final Frame frame1 = stack.pop();
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
project.getInput().getRowType().getFieldList()
.forEach(f -> fields.add(ImmutableSet.of(), f));
for (Pair<RexNode, ImmutableSet<String>> pair
: Pair.zip(project.getProjects(), frame1.fields.leftList())) {
switch (pair.left.getKind()) {
case INPUT_REF:
final int i = ((RexInputRef) pair.left).getIndex();
fields.set(i, pair.right, fields.right(i));
break;
default:
break;
}
}
stack.push(new Frame(project.getInput(), fields));
final ImmutableSet.Builder<RelHint> mergedHints = ImmutableSet.builder();
mergedHints.addAll(project.getHints());
mergedHints.addAll(hints);
// Keep bottom projection's variablesSet.
return project_(newNodes, fieldNameList, mergedHints.build(), force,
ImmutableSet.copyOf(project.getVariablesSet()));
}
// Simplify expressions.
if (config.simplify()) {
nodeList.replaceAll(e -> simplifier.simplifyPreservingType(e));
}
// Replace null names with generated aliases.
for (int i = 0; i < fieldNameList.size(); i++) {
if (fieldNameList.get(i) == null) {
fieldNameList.set(i, inferAlias(nodeList, nodeList.get(i), i));
}
}
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
final Set<String> uniqueNameList =
getTypeFactory().getTypeSystem().isSchemaCaseSensitive()
? new HashSet<>()
: new TreeSet<>(String.CASE_INSENSITIVE_ORDER);
// calculate final names and build field list
for (int i = 0; i < fieldNameList.size(); ++i) {
final RexNode node = nodeList.get(i);
String name = fieldNameList.get(i);
String originalName = name;
if (name == null || uniqueNameList.contains(name)) {
int j = 0;
if (name == null) {
j = i;
}
do {
name = SqlValidatorUtil.F_SUGGESTER.apply(originalName, j, j++);
} while (uniqueNameList.contains(name));
fieldNameList.set(i, name);
}
RelDataTypeField fieldType =
new RelDataTypeFieldImpl(name, i, node.getType());
switch (node.getKind()) {
case INPUT_REF:
// preserve rel aliases for INPUT_REF fields
final int index = ((RexInputRef) node).getIndex();
fields.add(frame.fields.left(index), fieldType);
break;
default:
fields.add(ImmutableSet.of(), fieldType);
break;
}
uniqueNameList.add(name);
}
if (!force && RexUtil.isIdentity(nodeList, inputRowType)) {
if (fieldNameList.equals(inputRowType.getFieldNames())) {
// Do not create an identity project if it does not rename any fields
return this;
} else {
// create "virtual" row type for project only rename fields
stack.pop();
// Ignore the hints.
stack.push(new Frame(frame.rel, fields));
}
return this;
}
// If the expressions are all literals, and the input is a Values with N
// rows (or an Aggregate with 1 row), replace with a Values with same tuple
// N times.
final int rowCount;
if (config.simplifyValues()
&& nodeList.stream().allMatch(e -> e instanceof RexLiteral)
&& (rowCount = fixedRowCount(frame)) >= 0) {
RelNode unused = build();
final RelDataTypeFactory.Builder typeBuilder = getTypeFactory().builder();
Pair.forEach(fieldNameList, nodeList, (name, expr) ->
typeBuilder.add(requireNonNull(name, "name"), expr.getType()));
@SuppressWarnings({"unchecked", "rawtypes"})
final List<RexLiteral> tuple = (List<RexLiteral>) (List) nodeList;
return values(Collections.nCopies(rowCount, tuple),
typeBuilder.build());
}
final RelNode project =
struct.projectFactory.createProject(frame.rel,
ImmutableList.copyOf(hints),
ImmutableList.copyOf(nodeList),
fieldNameList,
variables);
stack.pop();
stack.push(new Frame(project, fields));
return this;
}
/** If current frame will return a known, constant number of
* rows, returns that number; otherwise returns -1. */
private static int fixedRowCount(Frame frame) {
if (frame.rel instanceof Values) {
return ((Values) frame.rel).tuples.size();
}
if (frame.rel instanceof Aggregate) {
final Aggregate aggregate = (Aggregate) frame.rel;
if (aggregate.getGroupSet().isEmpty()
&& aggregate.getGroupType() == Aggregate.Group.SIMPLE) {
return 1;
}
}
return -1;
}
/** Creates a {@link Project} of the given
* expressions and field names, and optionally optimizing.
*
* <p>If {@code fieldNames} is null, or if a particular entry in
* {@code fieldNames} is null, derives field names from the input
* expressions.
*
* <p>If {@code force} is false,
* and the input is a {@code Project},
* and the expressions make the trivial projection ($0, $1, ...),
* modifies the input.
*
* @param nodes Expressions
* @param fieldNames Suggested field names, or null to generate
* @param force Whether to create a renaming Project if the
* projections are trivial
*/
public RelBuilder projectNamed(Iterable<? extends RexNode> nodes,
@Nullable Iterable<? extends @Nullable String> fieldNames, boolean force) {
return projectNamed(nodes, fieldNames, force, ImmutableSet.of());
}
/** Creates a {@link Project} of the given
* expressions and field names, and optionally optimizing.
*
* <p>If {@code fieldNames} is null, or if a particular entry in
* {@code fieldNames} is null, derives field names from the input
* expressions.
*
* <p>If {@code force} is false,
* and the input is a {@code Project},
* and the expressions make the trivial projection ($0, $1, ...),
* modifies the input.
*
* @param nodes Expressions
* @param fieldNames Suggested field names, or null to generate
* @param force Whether to create a renaming Project if the
* projections are trivial
* @param variablesSet Correlating variables that are set when reading a row
* from the input, and which may be referenced from the
* projection expressions
*/
public RelBuilder projectNamed(Iterable<? extends RexNode> nodes,
@Nullable Iterable<? extends @Nullable String> fieldNames, boolean force,
Iterable<CorrelationId> variablesSet) {
@SuppressWarnings({"unchecked", "rawtypes"})
final List<? extends RexNode> nodeList =
nodes instanceof List ? (List) nodes : ImmutableList.copyOf(nodes);
final List<@Nullable String> fieldNameList =
fieldNames == null ? null
: fieldNames instanceof List ? (List<@Nullable String>) fieldNames
: ImmutableNullableList.copyOf(fieldNames);
final RelNode input = peek();
final RelDataType rowType =
RexUtil.createStructType(cluster.getTypeFactory(), nodeList,
fieldNameList, SqlValidatorUtil.F_SUGGESTER);
if (!force
&& RexUtil.isIdentity(nodeList, input.getRowType())) {
if (input instanceof Project && fieldNames != null) {
// Rename columns of child projection if desired field names are given.
final Frame frame = stack.pop();
final Project childProject = (Project) frame.rel;
final Project newInput =
childProject.copy(childProject.getTraitSet(),
childProject.getInput(), childProject.getProjects(), rowType);
stack.push(new Frame(newInput.attachHints(childProject.getHints()), frame.fields));
}
if (input instanceof Values && fieldNameList != null) {
// Rename columns of child values if desired field names are given.
final Frame frame = stack.pop();
final Values values = (Values) frame.rel;
final RelDataTypeFactory.Builder typeBuilder =
getTypeFactory().builder();
Pair.forEach(fieldNameList, rowType.getFieldList(), (name, field) ->
typeBuilder.add(requireNonNull(name, "name"), field.getType()));
final RelDataType newRowType = typeBuilder.build();
final RelNode newValues =
struct.valuesFactory.createValues(cluster, newRowType,
values.tuples);
stack.push(new Frame(newValues, frame.fields));
}
} else {
project(nodeList, rowType.getFieldNames(), force, variablesSet);
}
return this;
}
/**
* Creates an {@link Uncollect} with given item aliases.
*
* @param itemAliases Operand item aliases, never null
* @param withOrdinality If {@code withOrdinality}, the output contains an extra
* {@code ORDINALITY} column
*/
public RelBuilder uncollect(List<String> itemAliases, boolean withOrdinality) {
Frame frame = stack.pop();
stack.push(
new Frame(
new Uncollect(
cluster,
cluster.traitSetOf(Convention.NONE),
frame.rel,
withOrdinality,
requireNonNull(itemAliases, "itemAliases"))));
return this;
}
/** Ensures that the field names match those given.
*
* <p>If all fields have the same name, adds nothing;
* if any fields do not have the same name, adds a {@link Project}.
*
* <p>Note that the names can be short-lived. Other {@code RelBuilder}
* operations make no guarantees about the field names of the rows they
* produce.
*
* @param fieldNames List of desired field names; may contain null values or
* have fewer fields than the current row type
*/
public RelBuilder rename(List<? extends @Nullable String> fieldNames) {
final List<String> oldFieldNames = peek().getRowType().getFieldNames();
checkArgument(fieldNames.size() <= oldFieldNames.size(),
"More names than fields");
final List<String> newFieldNames = new ArrayList<>(oldFieldNames);
for (int i = 0; i < fieldNames.size(); i++) {
final String s = fieldNames.get(i);
if (s != null) {
newFieldNames.set(i, s);
}
}
if (oldFieldNames.equals(newFieldNames)) {
return this;
}
if (peek() instanceof Values) {
// Special treatment for VALUES. Re-build it rather than add a project.
final Values v = (Values) build();
final RelDataTypeFactory.Builder b = getTypeFactory().builder();
for (Pair<String, RelDataTypeField> p
: Pair.zip(newFieldNames, v.getRowType().getFieldList())) {
b.add(p.left, p.right.getType());
}
return values(v.tuples, b.build());
}
return project(fields(), newFieldNames, true);
}
/** Infers the alias of an expression.
*
* <p>If the expression was created by {@link #alias}, replaces the expression
* in the project list.
*/
private @Nullable String inferAlias(List<RexNode> exprList, RexNode expr, int i) {
switch (expr.getKind()) {
case INPUT_REF:
final RexInputRef ref = (RexInputRef) expr;
return requireNonNull(stack.peek(), "empty frame stack")
.fields.get(ref.getIndex()).getValue().getName();
case CAST:
return inferAlias(exprList, ((RexCall) expr).getOperands().get(0), -1);
case AS:
final RexCall call = (RexCall) expr;
if (i >= 0) {
exprList.set(i, call.getOperands().get(0));
}
NlsString value = (NlsString) ((RexLiteral) call.getOperands().get(1)).getValue();
return castNonNull(value)
.getValue();
default:
return null;
}
}
/** Creates an {@link Aggregate} that makes the
* relational expression distinct on all fields. */
public RelBuilder distinct() {
return aggregate_((GroupKeyImpl) groupKey(fields()), ImmutableList.of());
}
/** Creates an {@link Aggregate} with an array of
* calls. */
@SuppressWarnings({"unchecked", "rawtypes"})
public RelBuilder aggregate(GroupKey groupKey, AggCall... aggCalls) {
return aggregate_((GroupKeyImpl) groupKey,
(ImmutableList) ImmutableList.copyOf(aggCalls));
}
/** Creates an {@link Aggregate} with an array of
* {@link AggregateCall}s. */
public RelBuilder aggregate(GroupKey groupKey,
List<AggregateCall> aggregateCalls) {
return aggregate_((GroupKeyImpl) groupKey,
aggregateCalls.stream()
.map(aggregateCall ->
new AggCallImpl2(aggregateCall,
aggregateCall.getArgList().stream()
.map(this::field)
.collect(toImmutableList())))
.collect(toImmutableList()));
}
/** Creates an {@link Aggregate} with multiple calls. */
@SuppressWarnings({"unchecked", "rawtypes"})
public RelBuilder aggregate(GroupKey groupKey,
Iterable<? extends AggCall> aggCalls) {
return aggregate_((GroupKeyImpl) groupKey,
ImmutableList.<AggCallPlus>copyOf((Iterable) aggCalls));
}
/** Creates an {@link Aggregate} with multiple calls. */
private RelBuilder aggregate_(GroupKeyImpl groupKey,
final ImmutableList<AggCallPlus> aggCalls) {
if (groupKey.nodes.isEmpty()
&& aggCalls.isEmpty()
&& config.pruneInputOfAggregate()) {
// Query is "SELECT /* no fields */ FROM t GROUP BY ()", which always
// returns one row with zero columns.
if (config.preventEmptyFieldList()) {
// Convert to "VALUES ROW(true)".
return values(new String[] {"dummy"}, true);
} else {
// Convert to "VALUES ROW()".
return values(ImmutableList.of(ImmutableList.of()),
getTypeFactory().builder().build());
}
}
final Registrar registrar =
new Registrar(fields(), peek().getRowType().getFieldNames());
final ImmutableBitSet groupSet =
ImmutableBitSet.of(registrar.registerExpressions(groupKey.nodes));
if (alreadyUnique(aggCalls, groupKey, groupSet, registrar.extraNodes)) {
final List<RexNode> nodes = new ArrayList<>(fields(groupSet));
aggCalls.forEach(c -> {
final AggregateCall call = c.aggregateCall();
final SqlStaticAggFunction staticFun =
call.getAggregation().unwrapOrThrow(SqlStaticAggFunction.class);
final RexNode node =
staticFun.constant(getRexBuilder(), groupSet, ImmutableList.of(), call);
nodes.add(aliasMaybe(requireNonNull(node, "node"), call.getName()));
});
return project(nodes);
}
ImmutableList<ImmutableBitSet> groupSets;
if (groupKey.nodeLists != null) {
final int sizeBefore = registrar.extraNodes.size();
final List<ImmutableBitSet> groupSetList = new ArrayList<>();
for (ImmutableList<RexNode> nodeList : groupKey.nodeLists) {
final ImmutableBitSet groupSet2 =
ImmutableBitSet.of(registrar.registerExpressions(nodeList));
if (!groupSet.contains(groupSet2)) {
throw new IllegalArgumentException("group set element " + nodeList
+ " must be a subset of group key");
}
groupSetList.add(groupSet2);
}
final ImmutableSortedMultiset<ImmutableBitSet> groupSetMultiset =
ImmutableSortedMultiset.copyOf(ImmutableBitSet.COMPARATOR,
groupSetList);
if (aggCalls.stream().anyMatch(RelBuilder::isGroupId)
|| !ImmutableBitSet.ORDERING.isStrictlyOrdered(groupSetMultiset)) {
return rewriteAggregateWithDuplicateGroupSets(groupSet, groupSetMultiset,
aggCalls);
}
groupSets = ImmutableList.copyOf(groupSetMultiset.elementSet());
if (registrar.extraNodes.size() > sizeBefore) {
throw new IllegalArgumentException("group sets contained expressions "
+ "not in group key: "
+ Util.skip(registrar.extraNodes, sizeBefore));
}
} else {
groupSets = ImmutableList.of(groupSet);
}
aggCalls.forEach(aggCall -> aggCall.register(registrar));
project(registrar.extraNodes);
rename(registrar.names);
final Frame frame = stack.pop();
RelNode r = frame.rel;
final List<AggregateCall> aggregateCalls = new ArrayList<>();
for (AggCallPlus aggCall : aggCalls) {
AggregateCall aggregateCall =
aggCall.aggregateCall(registrar, groupSets.contains(ImmutableBitSet.of()), r);
if (groupSets.size() <= 1) {
aggregateCall = removeRedundantAggregateDistinct(aggregateCall, groupSet, r);
}
aggregateCalls.add(aggregateCall);
}
assert ImmutableBitSet.ORDERING.isStrictlyOrdered(groupSets) : groupSets;
for (ImmutableBitSet set : groupSets) {
assert groupSet.contains(set);
}
return pruneAggregateInputFieldsAndDeduplicateAggCalls(r, groupSet, groupSets, aggregateCalls,
frame.fields, registrar.extraNodes);
}
/**
* Prunes unused fields on the input of the aggregate and removes duplicate aggregation calls.
*/
private RelBuilder pruneAggregateInputFieldsAndDeduplicateAggCalls(
RelNode r,
final ImmutableBitSet groupSet,
final ImmutableList<ImmutableBitSet> groupSets,
final List<AggregateCall> aggregateCalls,
PairList<ImmutableSet<String>, RelDataTypeField> inFields,
final List<RexNode> extraNodes) {
final ImmutableBitSet groupSetAfterPruning;
final ImmutableList<ImmutableBitSet> groupSetsAfterPruning;
if (config.pruneInputOfAggregate()
&& r instanceof Project) {
final Set<Integer> fieldsUsed =
RelOptUtil.getAllFields2(groupSet, aggregateCalls);
// Some parts of the system can't handle rows with zero fields, so
// pretend that one field is used.
if (fieldsUsed.isEmpty()) {
r = ((Project) r).getInput();
groupSetAfterPruning = groupSet;
groupSetsAfterPruning = groupSets;
} else if (fieldsUsed.size() < r.getRowType().getFieldCount()) {
// Some fields are computed but not used. Prune them.
final Map<Integer, Integer> sourceFieldToTargetFieldMap = new HashMap<>();
for (int source : fieldsUsed) {
sourceFieldToTargetFieldMap.put(source, sourceFieldToTargetFieldMap.size());
}
groupSetAfterPruning = groupSet.permute(sourceFieldToTargetFieldMap);
groupSetsAfterPruning =
ImmutableBitSet.ORDERING.immutableSortedCopy(
ImmutableBitSet.permute(groupSets, sourceFieldToTargetFieldMap));
final Mappings.TargetMapping targetMapping =
Mappings.target(sourceFieldToTargetFieldMap, r.getRowType().getFieldCount(),
fieldsUsed.size());
final List<AggregateCall> oldAggregateCalls =
new ArrayList<>(aggregateCalls);
aggregateCalls.clear();
for (AggregateCall aggregateCall : oldAggregateCalls) {
aggregateCalls.add(aggregateCall.transform(targetMapping));
}
final PairList<ImmutableSet<String>, RelDataTypeField> newInFields =
PairList.of();
newInFields.addAll(Mappings.permute(inFields, targetMapping.inverse()));
inFields = newInFields;
final Project project = (Project) r;
final List<RexNode> newProjects = new ArrayList<>();
final RelDataTypeFactory.Builder builder =
cluster.getTypeFactory().builder();
for (int i : fieldsUsed) {
newProjects.add(project.getProjects().get(i));
builder.add(project.getRowType().getFieldList().get(i));
}
// This currently does not apply mappings correctly to the RelCollation due to
// https://issues.apache.org/jira/browse/CALCITE-6391
r =
project.copy(project.getTraitSet().apply(targetMapping), project.getInput(),
newProjects, builder.build());
} else {
groupSetAfterPruning = groupSet;
groupSetsAfterPruning = groupSets;
}
} else {
groupSetAfterPruning = groupSet;
groupSetsAfterPruning = groupSets;
}
if (!config.dedupAggregateCalls() || Util.isDistinct(aggregateCalls)) {
return aggregate_(groupSetAfterPruning, groupSetsAfterPruning, r, aggregateCalls,
extraNodes, inFields);
}
// There are duplicate aggregate calls. Rebuild the list to eliminate
// duplicates, then add a Project.
final Set<AggregateCall> callSet = new HashSet<>();
final PairList<Integer, @Nullable String> projects = PairList.of();
Util.range(groupSetAfterPruning.cardinality())
.forEach(i -> projects.add(i, null));
final List<AggregateCall> distinctAggregateCalls = new ArrayList<>();
for (AggregateCall aggregateCall : aggregateCalls) {
final int i;
if (callSet.add(aggregateCall)) {
i = distinctAggregateCalls.size();
distinctAggregateCalls.add(aggregateCall);
} else {
i = distinctAggregateCalls.indexOf(aggregateCall);
assert i >= 0;
}
projects.add(groupSetAfterPruning.cardinality() + i, aggregateCall.name);
}
aggregate_(groupSetAfterPruning, groupSetsAfterPruning, r, distinctAggregateCalls,
extraNodes, inFields);
return project(projects.transform((i, name) -> aliasMaybe(field(i), name)));
}
/**
* Removed redundant distinct if an input is already unique.
*/
private AggregateCall removeRedundantAggregateDistinct(
AggregateCall aggregateCall,
ImmutableBitSet groupSet,
RelNode relNode) {
if (aggregateCall.isDistinct() && config.removeRedundantDistinct()) {
final RelMetadataQuery mq = relNode.getCluster().getMetadataQuery();
final List<Integer> argList = aggregateCall.getArgList();
final ImmutableBitSet distinctArg = ImmutableBitSet.builder()
.addAll(argList)
.build();
final ImmutableBitSet columns = groupSet.union(distinctArg);
final Boolean alreadyUnique =
mq.areColumnsUnique(relNode, columns);
if (alreadyUnique != null && alreadyUnique) {
// columns have been distinct or columns are primary keys
return aggregateCall.withDistinct(false);
}
}
return aggregateCall;
}
/** Returns whether an input is already unique, and therefore a Project
* can be created instead of an Aggregate.
*
* <p>{@link AggregateRemoveRule} does something similar, but also handles
* {@link org.apache.calcite.sql.SqlSingletonAggFunction} calls. */
private boolean alreadyUnique(List<AggCallPlus> aggCallList,
GroupKeyImpl groupKey, ImmutableBitSet groupSet,
List<RexNode> extraNodes) {
final RelMetadataQuery mq = peek().getCluster().getMetadataQuery();
if (aggCallList.isEmpty() && groupSet.isEmpty()) {
// We can't remove "GROUP BY ()" if there's a chance the rel could be
// empty.
if (RelMdUtil.isRelDefinitelyEmpty(mq, peek())) {
return false;
}
}
// If there are aggregate functions, we must be able to flatten them
if (!aggCallList.stream()
.allMatch(c -> canFlattenStatic(c.aggregateCall()))) {
return false;
}
if (extraNodes.size() == fields().size()) {
final Boolean unique = mq.areColumnsUnique(peek(), groupSet);
if (unique != null && unique
&& !config.aggregateUnique()
&& groupKey.isSimple()) {
// Rel is already unique.
return true;
}
}
// If there is at most one row, rel is already unique.
final Double maxRowCount = mq.getMaxRowCount(peek());
return maxRowCount != null && maxRowCount <= 1D
&& !config.aggregateUnique()
&& groupKey.isSimple();
}
/** Creates an {@link Aggregate} with a set of hybrid expressions represented
* as {@link RexNode}. */
public RelBuilder aggregateRex(GroupKey groupKey,
RexNode... nodes) {
return aggregateRex(groupKey, false, ImmutableList.copyOf(nodes));
}
/** Creates an {@link Aggregate} with a set of hybrid expressions represented
* as {@link RexNode}, optionally projecting the {@code groupKey} columns. */
public RelBuilder aggregateRex(GroupKey groupKey, boolean projectKey,
Iterable<? extends RexNode> nodes) {
final GroupKeyImpl groupKeyImpl = (GroupKeyImpl) groupKey;
final AggBuilder aggBuilder = new AggBuilder(groupKeyImpl.nodes);
// First pass. Call convert on each expression to ensure that aggCalls
// gets populated.
aggBuilder.registerExpressions(nodes);
// Create the Aggregate on the stack.
aggregate(groupKey, aggBuilder.aggCalls);
// Second pass. Call convert on each expression so that it references the
// actual aggCalls in the Aggregate that was just pushed onto the stack.
final List<RexNode> projects = new ArrayList<>();
if (projectKey) {
projects.addAll(fields(Util.range(groupKey.groupKeyCount())));
}
aggBuilder.convertExpressions(projects::add, nodes);
return project(projects);
}
/** Finishes the implementation of {@link #aggregate} by creating an
* {@link Aggregate} and pushing it onto the stack. */
private RelBuilder aggregate_(ImmutableBitSet groupSet,
ImmutableList<ImmutableBitSet> groupSets, RelNode input,
List<AggregateCall> aggregateCalls, List<RexNode> extraNodes,
PairList<ImmutableSet<String>, RelDataTypeField> inFields) {
final RelNode aggregate =
struct.aggregateFactory.createAggregate(input,
ImmutableList.of(), groupSet, groupSets, aggregateCalls);
// build field list
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
final List<RelDataTypeField> aggregateFields =
aggregate.getRowType().getFieldList();
int i = 0;
// first, group fields
for (Integer groupField : groupSet.asList()) {
RexNode node = extraNodes.get(groupField);
final SqlKind kind = node.getKind();
switch (kind) {
case INPUT_REF:
fields.add(inFields.get(((RexInputRef) node).getIndex()));
break;
default:
String name = aggregateFields.get(i).getName();
RelDataTypeField fieldType =
new RelDataTypeFieldImpl(name, i, node.getType());
fields.add(ImmutableSet.of(), fieldType);
break;
}
i++;
}
// second, aggregate fields. retain `i' as field index
for (int j = 0; j < aggregateCalls.size(); ++j) {
final AggregateCall call = aggregateCalls.get(j);
final RelDataTypeField fieldType =
new RelDataTypeFieldImpl(aggregateFields.get(i + j).getName(), i + j,
call.getType());
fields.add(ImmutableSet.of(), fieldType);
}
stack.push(new Frame(aggregate, fields));
return this;
}
/**
* The {@code GROUP_ID()} function is used to distinguish duplicate groups.
* However, as Aggregate normalizes group sets to canonical form (i.e.,
* flatten, sorting, redundancy removal), this information is lost in RelNode.
* Therefore, it is impossible to implement the function in runtime.
*
* <p>To fill this gap, an aggregation query that contains duplicate group
* sets is rewritten into a Union of Aggregate operators whose group sets are
* distinct. The number of inputs to the Union is equal to the maximum number
* of duplicates. In the {@code N}th input to the Union, calls to the
* {@code GROUP_ID} aggregate function are replaced by the integer literal
* {@code N}.
*
* <p>This method also handles the case where group sets are distinct but
* there is a call to {@code GROUP_ID}. That call is replaced by the integer
* literal {@code 0}.
*
* <p>Also see the discussion in
* <a href="https://issues.apache.org/jira/browse/CALCITE-1824">[CALCITE-1824]
* GROUP_ID returns wrong result</a>,
* <a href="https://issues.apache.org/jira/browse/CALCITE-4748">[CALCITE-4748]
* If there are duplicate GROUPING SETS, Calcite should return duplicate
* rows</a> and
* <a href="https://issues.apache.org/jira/browse/CALCITE-7126">[CALCITE-7126]
* The calculation result of grouping function is wrong</a>.
*/
private RelBuilder rewriteAggregateWithDuplicateGroupSets(
ImmutableBitSet groupSet,
ImmutableSortedMultiset<ImmutableBitSet> groupSets,
List<AggCallPlus> aggregateCalls) {
final List<String> fieldNamesIfNoRewrite =
Aggregate.deriveRowType(getTypeFactory(), peek().getRowType(), false,
groupSet, groupSets.asList(),
aggregateCalls.stream().map(AggCallPlus::aggregateCall)
.collect(toImmutableList())).getFieldNames();
final Frame input = stack.pop();
final Map<Integer, Set<ImmutableBitSet>> groupIdToGroupSets = new HashMap<>();
for (Multiset.Entry<ImmutableBitSet> entry : groupSets.entrySet()) {
int groupId = entry.getCount() - 1;
for (int i = 0; i <= groupId; i++) {
groupIdToGroupSets.computeIfAbsent(i,
k -> Sets.newTreeSet(ImmutableBitSet.COMPARATOR))
.add(entry.getElement());
}
}
// Create aggregate for each GROUP_ID value
for (Map.Entry<Integer, Set<ImmutableBitSet>> entry : groupIdToGroupSets.entrySet()) {
// If n duplicates exist for a particular grouping, the {@code GROUP_ID()}
// function produces values in the range 0 to n-1. For each value,
// we need to figure out the corresponding group sets.
//
// For example, "... GROUPING SETS (a, a, b, c, c, c, c)"
// (i) The max value of the GROUP_ID() function returns is 3
// (ii) GROUPING SETS (a, b, c) produces value 0,
// GROUPING SETS (a, c) produces value 1,
// GROUPING SETS (c) produces value 2
// GROUPING SETS (c) produces value 3
int groupId = entry.getKey();
Set<ImmutableBitSet> newGroupSets = entry.getValue();
rewriteGroupAggCalls(newGroupSets, groupSet, aggregateCalls,
fieldNamesIfNoRewrite, input, groupId);
}
return union(true, groupIdToGroupSets.size());
}
private void rewriteGroupAggCalls(
Set<ImmutableBitSet> subGroupSets,
ImmutableBitSet oriGroupSet,
List<AggCallPlus> oriAggCalls,
List<String> fieldNames,
Frame input,
int groupId) {
stack.push(input);
List<AggCallPlus> subAggCalls = new ArrayList<>();
ImmutableBitSet subGroupSet = ImmutableBitSet.union(subGroupSets);
List<RexNode> subProjects = new ArrayList<>();
// 1. For GroupSet
RelDataType subAggregateGroupSetType =
Aggregate.deriveRowType(getTypeFactory(), peek().getRowType(), false,
subGroupSet, ImmutableList.copyOf(subGroupSets), ImmutableList.of());
for (int i = 0; i < oriGroupSet.cardinality(); i++) {
int groupKey = oriGroupSet.nth(i);
if (subGroupSet.get(groupKey)) {
subProjects.add(
RexInputRef.of(
subGroupSet.indexOf(groupKey),
subAggregateGroupSetType));
} else {
// If the groupKey is not in the GroupSet, use null as a placeholder.
subProjects.add(getRexBuilder().makeNullLiteral(field(groupKey).getType()));
}
}
// 2. For AggregateCalls
int newGroupCount = subGroupSet.cardinality();
for (AggCallPlus oriAggCall : oriAggCalls) {
AggregateCall aggCall = oriAggCall.aggregateCall();
switch (oriAggCall.op().getKind()) {
case GROUPING:
if (!subGroupSet.contains(ImmutableBitSet.of(aggCall.getArgList()))) {
// If the parameters of the GROUPING function cannot be fully covered by the GroupSet,
// the GROUPING function needs to be split.
splitGrouping(subAggCalls, subProjects, aggCall, subGroupSet);
} else {
subProjects.add(
new RexInputRef(
newGroupCount + subAggCalls.size(),
aggCall.getType()));
subAggCalls.add(oriAggCall);
}
break;
case GROUP_ID:
subProjects.add(
getRexBuilder().makeLiteral(
groupId,
aggCall.getType()));
break;
default:
subProjects.add(
new RexInputRef(
newGroupCount + subAggCalls.size(),
aggCall.getType()));
subAggCalls.add(oriAggCall);
break;
}
}
aggregate(groupKey(subGroupSet, subGroupSets), subAggCalls);
project(subProjects, fieldNames);
}
/**
* This method is used to expand the SQL GROUPING operator
* into a set of expressions. For example, it expands GROUPING(x, y, z)
* into 2^2 * GROUPING(x) + 2^1 * GROUPING(y) + 2^0 * GROUPING(z)
*
* <p>For example, in "GROUP BY GROUPING SETS ((a, b), (a), (), (b))":
* In the grouping set (a, b), GROUPING(a) = 0, GROUPING(b) = 0,
* GROUPING(a, b) = 2^1 * 0 + 2^0 * 0 => 0.
* In the grouping set (a), GROUPING(a) = 0, GROUPING(b) = 1,
* GROUPING(a, b) = 2^1 * 0 + 2^0 * 1 => 1.
* In the grouping set (), GROUPING(a) = 1, GROUPING(b) = 1,
* GROUPING(a, b) = 2^1 * 1 + 2^0 * 1 = 3.
* In the grouping set (b), GROUPING(a) = 1, GROUPING(b) = 0,
* GROUPING(a, b) = 2^1 * 1 + 2^0 * 0 = 2.
* Thus, GROUPING(a, b) produces 0, 1, 3, and 2 for each grouping set respectively.
*/
private void splitGrouping(
List<AggCallPlus> aggCalls,
List<RexNode> projects,
AggregateCall grouping,
ImmutableBitSet groupSet) {
List<RexNode> splitOperands = new ArrayList<>();
List<Integer> groupingArgs = grouping.getArgList();
RelDataType groupingType = getTypeFactory().createSqlType(SqlTypeName.BIGINT);
for (int i = 0; i < groupingArgs.size(); i++) {
int groupingArg = groupingArgs.get(i);
if (groupSet.get(groupingArg)) {
RexInputRef groupingRef =
new RexInputRef(groupSet.cardinality() + aggCalls.size(), groupingType);
splitOperands.add(
call(SqlStdOperatorTable.MULTIPLY,
literal(safeShift(groupingArgs.size() - 1 - i)),
groupingRef));
aggCalls.add((AggCallPlus) aggregateCall(SqlStdOperatorTable.GROUPING, field(groupingArg)));
} else {
// Sets GROUPING function value to 1 for parameters not in GroupSet and calculates offset.
splitOperands.add(literal(safeShift(groupingArgs.size() - 1 - i)));
}
}
// Plus all expanded expressions (including GROUPING functions and constants).
RexNode plus = splitOperands.get(0);
for (int i = 1; i < splitOperands.size(); i++) {
plus = call(SqlStdOperatorTable.PLUS, plus, splitOperands.get(i));
}
projects.add(plus);
}
private static int safeShift(int shift) {
if (shift < 0 || shift >= Integer.SIZE) {
throw new IllegalArgumentException(
"Too many grouping keys. Maximum is " + (Integer.SIZE - 1) + " for grouping functions.");
}
return 1 << shift;
}
private static boolean isGroupId(AggCall c) {
return ((AggCallPlus) c).op().kind == SqlKind.GROUP_ID;
}
/** Given a list of literals and a target row type, make the literals
* respectively match the fields types of the row.
*
* @param rowType Type expected for values
* @param values A list of literals that should match the rowType */
private List<RexLiteral> convertLiteralTypes(
RelDataType rowType, List<RexLiteral> values) {
assert values.size() == rowType.getFieldCount()
: "List of literals of size " + values.size() + " does not match expected type " + rowType;
List<RelDataTypeField> fields = rowType.getFieldList();
List<RexLiteral> constants = new ArrayList<>();
for (int i = 0; i < values.size(); i++) {
RexLiteral vi = values.get(i);
RelDataType type = fields.get(i).getType();
RexNode e = cluster.getRexBuilder().makeAbstractCast(type, vi, false);
RexNode simplified = simplifier.simplify(e);
assert simplified instanceof RexLiteral
: "Could not simplify expression to literal" + simplified;
constants.add((RexLiteral) simplified);
}
return ImmutableList.copyOf(constants);
}
private RelBuilder setOp(boolean all, SqlKind kind, int n) {
List<RelNode> inputs = new ArrayList<>();
for (int i = 0; i < n; i++) {
inputs.add(0, build());
}
switch (kind) {
case UNION:
case INTERSECT:
case EXCEPT:
if (n < 1) {
throw new IllegalArgumentException(
"bad INTERSECT/UNION/EXCEPT input count");
}
break;
default:
throw new AssertionError("bad setOp " + kind);
}
if (n == 1) {
return push(inputs.get(0));
}
if (config.simplifyValues()
&& kind == UNION
&& inputs.stream().allMatch(r -> r instanceof Values)) {
List<RelDataType> inputTypes = Util.transform(inputs, RelNode::getRowType);
RelDataType rowType = getTypeFactory()
.leastRestrictive(inputTypes);
requireNonNull(rowType, () -> "leastRestrictive(" + inputTypes + ")");
final List<List<RexLiteral>> tuples = new ArrayList<>();
for (RelNode input : inputs) {
ImmutableList<ImmutableList<RexLiteral>> literals = ((Values) input).tuples;
for (ImmutableList<RexLiteral> l : literals) {
List<RexLiteral> converted = convertLiteralTypes(rowType, l);
tuples.add(converted);
}
}
final List<List<RexLiteral>> tuples2 =
all ? tuples : Util.distinctList(tuples);
return values(tuples2, rowType);
}
return push(struct.setOpFactory.createSetOp(kind, inputs, all));
}
/** Creates a {@link Union} of the two most recent
* relational expressions on the stack.
*
* @param all Whether to create UNION ALL
*/
public RelBuilder union(boolean all) {
return union(all, 2);
}
/** Creates a {@link Union} of the {@code n}
* most recent relational expressions on the stack.
*
* @param all Whether to create UNION ALL
* @param n Number of inputs to the UNION operator
*/
public RelBuilder union(boolean all, int n) {
return setOp(all, UNION, n);
}
/** Creates an {@link Intersect} of the two most
* recent relational expressions on the stack.
*
* @param all Whether to create INTERSECT ALL
*/
public RelBuilder intersect(boolean all) {
return intersect(all, 2);
}
/** Creates an {@link Intersect} of the {@code n}
* most recent relational expressions on the stack.
*
* @param all Whether to create INTERSECT ALL
* @param n Number of inputs to the INTERSECT operator
*/
public RelBuilder intersect(boolean all, int n) {
return setOp(all, SqlKind.INTERSECT, n);
}
/** Creates a {@link Minus} of the two most recent
* relational expressions on the stack.
*
* @param all Whether to create EXCEPT ALL
*/
public RelBuilder minus(boolean all) {
return minus(all, 2);
}
/** Creates a {@link Minus} of the {@code n}
* most recent relational expressions on the stack.
*
* @param all Whether to create EXCEPT ALL
*/
public RelBuilder minus(boolean all, int n) {
return setOp(all, SqlKind.EXCEPT, n);
}
/**
* Creates a {@link TableScan} on a {@link TransientTable} with the given name, using as type
* the top of the stack's type.
*
* @param tableName table name
*/
@Experimental
public RelBuilder transientScan(String tableName) {
return this.transientScan(tableName, this.peek().getRowType());
}
/**
* Creates a {@link TableScan} on a {@link TransientTable} with the given name and type.
*
* @param tableName table name
* @param rowType row type of the table
*/
@Experimental
public RelBuilder transientScan(String tableName, RelDataType rowType) {
TransientTable transientTable = new ListTransientTable(tableName, rowType);
requireNonNull(relOptSchema, "relOptSchema");
RelOptTable relOptTable =
RelOptTableImpl.create(relOptSchema, rowType, transientTable,
ImmutableList.of(tableName));
RelNode scan =
struct.scanFactory.createScan(
ViewExpanders.toRelContext(viewExpander, cluster),
relOptTable);
push(scan);
rename(rowType.getFieldNames());
return this;
}
/**
* Creates a {@link TableSpool} for the most recent relational expression.
*
* @param readType Spool's read type (as described in {@link Spool.Type})
* @param writeType Spool's write type (as described in {@link Spool.Type})
* @param table Table to write into
*/
private RelBuilder tableSpool(Spool.Type readType, Spool.Type writeType,
RelOptTable table) {
RelNode spool =
struct.spoolFactory.createTableSpool(peek(), readType, writeType,
table);
replaceTop(spool);
return this;
}
/**
* Creates a {@link RepeatUnion} associated to a {@link TransientTable} without a maximum number
* of iterations, i.e. repeatUnion(tableName, all, -1).
*
* @param tableName name of the {@link TransientTable} associated to the {@link RepeatUnion}
* @param all whether duplicates will be considered or not
*/
@Experimental
public RelBuilder repeatUnion(String tableName, boolean all) {
return repeatUnion(tableName, all, -1);
}
/**
* Creates a {@link RepeatUnion} associated to a {@link TransientTable} of the
* two most recent relational expressions on the stack.
*
* <p>Warning: if these relational expressions are not
* correctly defined, this operation might lead to an infinite loop.
*
* <p>The generated {@link RepeatUnion} operates as follows:
*
* <ul>
* <li>Evaluate its left term once, propagating the results into the
* {@link TransientTable};
* <li>Evaluate its right term (which may contain a {@link TableScan} on the
* {@link TransientTable}) over and over until it produces no more results
* (or until an optional maximum number of iterations is reached). On each
* iteration, the results are propagated into the {@link TransientTable},
* overwriting the results from the previous one.
* </ul>
*
* @param tableName Name of the {@link TransientTable} associated to the
* {@link RepeatUnion}
* @param all Whether duplicates are considered
* @param iterationLimit Maximum number of iterations; negative value means no limit
*/
@Experimental
public RelBuilder repeatUnion(String tableName, boolean all, int iterationLimit) {
RelOptTable table = null;
for (int i = 0; i < stack.size(); i++) { // search scan(tableName) in the stack
table = RelOptUtil.findTable(peek(i), tableName);
if (table != null) { // found
break;
}
}
if (table == null) {
throw RESOURCE.tableNotFound(tableName).ex();
}
RelNode iterative = tableSpool(Spool.Type.LAZY, Spool.Type.LAZY, table).build();
RelNode seed = tableSpool(Spool.Type.LAZY, Spool.Type.LAZY, table).build();
RelNode repeatUnion =
struct.repeatUnionFactory.createRepeatUnion(seed, iterative, all,
iterationLimit, table);
return push(repeatUnion);
}
/** Creates a {@link LogicalAsofJoin} with the specified conditions. */
public RelBuilder asofJoin(JoinRelType joinType, RexNode condition, RexNode matchCondition) {
// Implementation based on the 'join' method
assert joinType == JoinRelType.ASOF || joinType == JoinRelType.LEFT_ASOF;
final Frame right = stack.pop();
final Frame left = stack.pop();
if (config.simplify()) {
// Normalize expanded versions IS NOT DISTINCT FROM so that simplifier does not
// transform the expression to something unrecognizable
if (condition instanceof RexCall) {
condition =
RelOptUtil.collapseExpandedIsNotDistinctFromExpr((RexCall) condition,
getRexBuilder());
}
condition = simplifier.simplifyUnknownAsFalse(condition);
}
final RelNode join;
RelNode join0 =
struct.asofJoinFactory.createAsofJoin(left.rel, right.rel,
ImmutableList.of(), condition, matchCondition, joinType);
if (join0 instanceof Join && config.pushJoinCondition()) {
join = RelOptUtil.pushDownJoinConditions((Join) join0, this);
} else {
join = join0;
}
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
fields.addAll(left.fields);
fields.addAll(right.fields);
stack.push(new Frame(join, fields));
return this;
}
/** Creates a {@link Join} with an array of conditions. */
public RelBuilder join(JoinRelType joinType, RexNode condition0,
RexNode... conditions) {
return join(joinType, Lists.asList(condition0, conditions));
}
/** Creates a {@link Join} with multiple
* conditions. */
public RelBuilder join(JoinRelType joinType,
Iterable<? extends RexNode> conditions) {
return join(joinType, and(conditions),
ImmutableSet.of());
}
/** Creates a {@link Join} with one condition. */
public RelBuilder join(JoinRelType joinType, RexNode condition) {
return join(joinType, condition, ImmutableSet.of());
}
/** Creates a {@link Join} with correlating variables. */
public RelBuilder join(JoinRelType joinType, RexNode condition,
Set<CorrelationId> variablesSet) {
Frame right = stack.pop();
final Frame left = stack.pop();
final RelNode join;
final boolean correlate = checkIfCorrelated(variablesSet, joinType, left.rel, right.rel);
RexNode postCondition = literal(true);
if (config.simplify()) {
// Normalize expanded versions IS NOT DISTINCT FROM so that simplifier does not
// transform the expression to something unrecognizable
if (condition instanceof RexCall) {
condition =
RelOptUtil.collapseExpandedIsNotDistinctFromExpr((RexCall) condition,
getRexBuilder());
}
condition = simplifier.simplifyUnknownAsFalse(condition);
}
if (correlate) {
final CorrelationId id = Iterables.getOnlyElement(variablesSet);
// Correlate does not have an ON clause.
switch (joinType) {
case LEFT:
case SEMI:
case ANTI:
// For a LEFT/SEMI/ANTI, predicate must be evaluated first.
stack.push(right);
filter(condition.accept(new Shifter(left.rel, id, right.rel)));
right = stack.pop();
break;
case INNER:
// For INNER, we can defer.
postCondition = condition;
break;
default:
throw new IllegalArgumentException("Correlated " + joinType + " join is not supported");
}
final ImmutableBitSet requiredColumns = RelOptUtil.correlationColumns(id, right.rel);
join =
struct.correlateFactory.createCorrelate(left.rel, right.rel, ImmutableList.of(), id,
requiredColumns, joinType);
} else {
RelNode join0 =
struct.joinFactory.createJoin(left.rel, right.rel,
ImmutableList.of(), condition, variablesSet, joinType, false);
if (join0 instanceof Join && config.pushJoinCondition()) {
join = RelOptUtil.pushDownJoinConditions((Join) join0, this);
} else {
join = join0;
}
}
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
fields.addAll(left.fields);
fields.addAll(right.fields);
stack.push(new Frame(join, fields));
filter(postCondition);
return this;
}
/** Creates a {@link Correlate}
* with a {@link CorrelationId} and an array of fields that are used by correlation. */
public RelBuilder correlate(JoinRelType joinType,
CorrelationId correlationId, RexNode... requiredFields) {
return correlate(joinType, correlationId, ImmutableList.copyOf(requiredFields));
}
/** Creates a {@link Correlate}
* with a {@link CorrelationId} and a list of fields that are used by correlation. */
public RelBuilder correlate(JoinRelType joinType,
CorrelationId correlationId, Iterable<? extends RexNode> requiredFields) {
Frame right = stack.pop();
final Registrar registrar =
new Registrar(fields(), peek().getRowType().getFieldNames());
List<Integer> requiredOrdinals =
registrar.registerExpressions(ImmutableList.copyOf(requiredFields));
project(registrar.extraNodes);
rename(registrar.names);
Frame left = stack.pop();
final RelNode correlate =
struct.correlateFactory.createCorrelate(left.rel, right.rel, ImmutableList.of(),
correlationId, ImmutableBitSet.of(requiredOrdinals), joinType);
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
fields.addAll(left.fields);
fields.addAll(right.fields);
stack.push(new Frame(correlate, fields));
return this;
}
/** Creates a {@link Join} using USING syntax.
*
* <p>For each of the field names, both left and right inputs must have a
* field of that name. Constructs a join condition that the left and right
* fields are equal.
*
* @param joinType Join type
* @param fieldNames Field names
*/
public RelBuilder join(JoinRelType joinType, String... fieldNames) {
final List<RexNode> conditions = new ArrayList<>();
for (String fieldName : fieldNames) {
conditions.add(
equals(field(2, 0, fieldName),
field(2, 1, fieldName)));
}
return join(joinType, conditions);
}
/** Creates a {@link Join} with {@link JoinRelType#SEMI}.
*
* <p>A semi-join is a form of join that combines two relational expressions
* according to some condition, and outputs only rows from the left input for
* which at least one row from the right input matches. It only outputs
* columns from the left input, and ignores duplicates on the right.
*
* <p>For example, {@code EMP semi-join DEPT} finds all {@code EMP} records
* that do not have a corresponding {@code DEPT} record, similar to the
* following SQL:
*
* <blockquote><pre>
* SELECT * FROM EMP
* WHERE EXISTS (SELECT 1 FROM DEPT
* WHERE DEPT.DEPTNO = EMP.DEPTNO)</pre>
* </blockquote>
*/
public RelBuilder semiJoin(Iterable<? extends RexNode> conditions) {
final Frame right = stack.pop();
final RelNode semiJoin =
struct.joinFactory.createJoin(peek(),
right.rel,
ImmutableList.of(),
and(conditions),
ImmutableSet.of(),
JoinRelType.SEMI,
false);
replaceTop(semiJoin);
return this;
}
/** Creates a {@link Join} with {@link JoinRelType#SEMI}.
*
* @see #semiJoin(Iterable) */
public RelBuilder semiJoin(RexNode... conditions) {
return semiJoin(ImmutableList.copyOf(conditions));
}
/** Creates an anti-join.
*
* <p>An anti-join is a form of join that combines two relational expressions
* according to some condition, but outputs only rows from the left input
* for which no rows from the right input match.
*
* <p>For example, {@code EMP anti-join DEPT} finds all {@code EMP} records
* that do not have a corresponding {@code DEPT} record, similar to the
* following SQL:
*
* <blockquote><pre>
* SELECT * FROM EMP
* WHERE NOT EXISTS (SELECT 1 FROM DEPT
* WHERE DEPT.DEPTNO = EMP.DEPTNO)</pre>
* </blockquote>
*/
public RelBuilder antiJoin(Iterable<? extends RexNode> conditions) {
final Frame right = stack.pop();
final RelNode antiJoin =
struct.joinFactory.createJoin(peek(),
right.rel,
ImmutableList.of(),
and(conditions),
ImmutableSet.of(),
JoinRelType.ANTI,
false);
replaceTop(antiJoin);
return this;
}
/** Creates an anti-join.
*
* @see #antiJoin(Iterable) */
public RelBuilder antiJoin(RexNode... conditions) {
return antiJoin(ImmutableList.copyOf(conditions));
}
/** Assigns a table alias to the top entry on the stack. */
public RelBuilder as(final String alias) {
final Frame pair = stack.pop();
final PairList<ImmutableSet<String>, RelDataTypeField> newFields =
PairList.of();
pair.fields.forEach((aliases, field) -> {
final ImmutableSet<String> aliasList =
aliases.contains(alias)
? aliases
: ImmutableSet.<String>builder().addAll(aliases).add(alias)
.build();
newFields.add(aliasList, field);
});
stack.push(new Frame(pair.rel, newFields));
return this;
}
/** Creates a {@link Values}.
*
* <p>The {@code values} array must have the same number of entries as
* {@code fieldNames}, or an integer multiple if you wish to create multiple
* rows.
*
* <p>The {@code fieldNames} array must not be null or empty, but may contain
* null values.
*
* <p>If there are zero rows, or if all values of any column are
* null, this method cannot deduce the type of columns. For these cases,
* call {@link #values(Iterable, RelDataType)}.
*
* @param fieldNames Field names
* @param values Values
*/
public RelBuilder values(@Nullable String[] fieldNames, @Nullable Object... values) {
requireNonNull(fieldNames, "fieldNames");
if (fieldNames.length == 0
|| values.length % fieldNames.length != 0
|| values.length < fieldNames.length) {
throw new IllegalArgumentException(
"Value count must be a positive multiple of field count");
}
final int rowCount = values.length / fieldNames.length;
for (Ord<@Nullable String> fieldName : Ord.zip(fieldNames)) {
if (allNull(values, fieldName.i, fieldNames.length)) {
throw new IllegalArgumentException("All values of field '" + fieldName.e
+ "' (field index " + fieldName.i + ")"
+ " are null; cannot deduce type");
}
}
final ImmutableList<ImmutableList<RexLiteral>> tupleList =
tupleList(fieldNames.length, values);
assert tupleList.size() == rowCount;
final List<String> fieldNameList =
Util.transformIndexed(Arrays.asList(fieldNames), (name, i) ->
name != null ? name : SqlUtil.deriveAliasFromOrdinal(i));
return values(tupleList, fieldNameList);
}
private RelBuilder values(List<? extends List<RexLiteral>> tupleList,
List<String> fieldNames) {
final RelDataTypeFactory typeFactory = cluster.getTypeFactory();
final RelDataTypeFactory.Builder builder = typeFactory.builder();
Ord.forEach(fieldNames, (fieldName, i) -> {
final RelDataType type =
typeFactory.leastRestrictive(new AbstractList<RelDataType>() {
@Override public RelDataType get(int index) {
return tupleList.get(index).get(i).getType();
}
@Override public int size() {
return tupleList.size();
}
});
if (type == null) {
throw new AssertionError("can't infer type for field " + i + ", "
+ fieldName);
}
builder.add(fieldName, type);
});
final RelDataType rowType = builder.build();
return values(tupleList, rowType);
}
private ImmutableList<ImmutableList<RexLiteral>> tupleList(int columnCount,
@Nullable Object[] values) {
final ImmutableList.Builder<ImmutableList<RexLiteral>> listBuilder =
ImmutableList.builder();
final List<RexLiteral> valueList = new ArrayList<>();
for (int i = 0; i < values.length; i++) {
Object value = values[i];
valueList.add(literal(value));
if ((i + 1) % columnCount == 0) {
listBuilder.add(ImmutableList.copyOf(valueList));
valueList.clear();
}
}
return listBuilder.build();
}
/** Returns whether all values for a given column are null. */
private static boolean allNull(@Nullable Object[] values, int column, int columnCount) {
for (int i = column; i < values.length; i += columnCount) {
if (values[i] != null) {
return false;
}
}
return true;
}
/** Creates a relational expression that reads from an input and throws
* all of the rows away.
*
* <p>Note that this method always pops one relational expression from the
* stack. {@code values}, in contrast, does not pop any relational
* expressions, and always produces a leaf.
*
* <p>The default implementation creates a {@link Values} with the same
* specified row type and aliases as the input, and ignores the input entirely.
* But schema-on-query systems such as Drill might override this method to
* create a relation expression that retains the input, just to read its
* schema.
*/
public RelBuilder empty() {
final Frame frame = stack.pop();
final RelNode values =
struct.valuesFactory.createValues(cluster, frame.rel.getRowType(),
ImmutableList.of());
stack.push(new Frame(values, frame.fields));
return this;
}
/** Creates a {@link Values} with a specified row type.
*
* <p>This method can handle cases that {@link #values(String[], Object...)}
* cannot, such as all values of a column being null, or there being zero
* rows.
*
* @param rowType Row type
* @param columnValues Values
*/
public RelBuilder values(RelDataType rowType, Object... columnValues) {
final ImmutableList<ImmutableList<RexLiteral>> tupleList =
tupleList(rowType.getFieldCount(), columnValues);
RelNode values =
struct.valuesFactory.createValues(cluster, rowType,
ImmutableList.copyOf(tupleList));
push(values);
return this;
}
/** Creates a {@link Values} with a specified row type.
*
* <p>This method can handle cases that {@link #values(String[], Object...)}
* cannot, such as all values of a column being null, or there being zero
* rows.
*
* @param tupleList Tuple list
* @param rowType Row type
*/
public RelBuilder values(Iterable<? extends List<RexLiteral>> tupleList,
RelDataType rowType) {
RelNode values =
struct.valuesFactory.createValues(cluster, rowType,
copy(tupleList));
push(values);
return this;
}
/** Creates a {@link Values} with a specified row type and
* zero rows.
*
* @param rowType Row type
*/
public RelBuilder values(RelDataType rowType) {
return values(ImmutableList.<ImmutableList<RexLiteral>>of(), rowType);
}
/** Converts an iterable of lists into an immutable list of immutable lists
* with the same contents. Returns the same object if possible. */
private static <E> ImmutableList<ImmutableList<E>> copy(
Iterable<? extends List<E>> tupleList) {
final ImmutableList.Builder<ImmutableList<E>> builder =
ImmutableList.builder();
int changeCount = 0;
for (List<E> literals : tupleList) {
final ImmutableList<E> literals2 =
ImmutableList.copyOf(literals);
builder.add(literals2);
if (literals != literals2) {
++changeCount;
}
}
if (changeCount == 0 && tupleList instanceof ImmutableList) {
// don't make a copy if we don't have to
//noinspection unchecked
return (ImmutableList<ImmutableList<E>>) tupleList;
}
return builder.build();
}
/** Creates a limit and/or offset without a sort.
*
* @param offset Number of rows to skip; non-positive means don't skip any
* @param fetch Maximum number of rows to fetch; negative means no limit
*/
public RelBuilder limit(int offset, int fetch) {
return sortLimit(offset, fetch, ImmutableList.of());
}
/** Creates an Exchange by distribution. */
public RelBuilder exchange(RelDistribution distribution) {
RelNode exchange =
struct.exchangeFactory.createExchange(peek(), distribution);
replaceTop(exchange);
return this;
}
/** Creates a SortExchange by distribution and collation. */
public RelBuilder sortExchange(RelDistribution distribution,
RelCollation collation) {
RelNode exchange =
struct.sortExchangeFactory.createSortExchange(peek(), distribution,
collation);
replaceTop(exchange);
return this;
}
/** Creates a {@link Sort} by field ordinals.
*
* <p>Negative fields mean descending: -1 means field(0) descending,
* -2 means field(1) descending, etc.
*/
public RelBuilder sort(int... fields) {
final ImmutableList.Builder<RexNode> builder = ImmutableList.builder();
for (int field : fields) {
builder.add(field < 0 ? desc(field(-field - 1)) : field(field));
}
return sortLimit(-1, -1, builder.build());
}
/** Creates a {@link Sort} by expressions. */
public RelBuilder sort(RexNode... nodes) {
return sortLimit(-1, -1, ImmutableList.copyOf(nodes));
}
/** Creates a {@link Sort} by expressions. */
public RelBuilder sort(Iterable<? extends RexNode> nodes) {
return sortLimit(-1, -1, nodes);
}
/** Creates a {@link Sort} by expressions, with limit and offset. */
public RelBuilder sortLimit(int offset, int fetch, RexNode... nodes) {
return sortLimit(offset, fetch, ImmutableList.copyOf(nodes));
}
/** Creates a {@link Sort} by specifying collations.
*/
public RelBuilder sort(RelCollation collation) {
final RelNode sort =
struct.sortFactory.createSort(peek(), collation, null, null);
replaceTop(sort);
return this;
}
/** Creates a {@link Sort} by a list of expressions, with limit and offset.
*
* @param offset Number of rows to skip; non-positive means don't skip any
* @param fetch Maximum number of rows to fetch; negative means no limit
* @param nodes Sort expressions
*/
public RelBuilder sortLimit(int offset, int fetch,
Iterable<? extends RexNode> nodes) {
final @Nullable RexNode offsetNode = offset <= 0 ? null : literal(offset);
final @Nullable RexNode fetchNode = fetch < 0 ? null : literal(fetch);
return sortLimit(offsetNode, fetchNode, nodes);
}
/** Creates a {@link Sort} by a list of expressions, with limitNode and offsetNode.
*
* @param offsetNode RexLiteral means number of rows to skip is deterministic,
* RexDynamicParam means number of rows to skip is dynamic.
* @param fetchNode RexLiteral means maximum number of rows to fetch is deterministic,
* RexDynamicParam mean maximum number is dynamic.
* @param nodes Sort expressions
*/
public RelBuilder sortLimit(@Nullable RexNode offsetNode, @Nullable RexNode fetchNode,
Iterable<? extends RexNode> nodes) {
if (offsetNode != null) {
if (!(offsetNode instanceof RexLiteral || offsetNode instanceof RexDynamicParam)) {
throw new IllegalArgumentException("OFFSET node must be RexLiteral or RexDynamicParam");
}
}
if (fetchNode != null) {
if (!(fetchNode instanceof RexLiteral || fetchNode instanceof RexDynamicParam)) {
throw new IllegalArgumentException("FETCH node must be RexLiteral or RexDynamicParam");
}
}
final Registrar registrar = new Registrar(fields(), ImmutableList.of());
final List<RelFieldCollation> fieldCollations =
registrar.registerFieldCollations(nodes);
final int fetch = fetchNode instanceof RexLiteral
? RexLiteral.intValue(fetchNode) : -1;
if (offsetNode == null && fetch == 0 && config.simplifyLimit()) {
return empty();
}
if (offsetNode == null && fetchNode == null && fieldCollations.isEmpty()) {
return this; // sort is trivial
}
if (fieldCollations.isEmpty()) {
assert registrar.addedFieldCount() == 0;
RelNode top = peek();
if (top instanceof Sort) {
final Sort sort2 = (Sort) top;
if ((offsetNode == null || sort2.offset == null)
&& (fetchNode == null || sort2.fetch == null)) {
// We're not trying to replace something that's already set - an
// offset in a sort that already has an offset, or a fetch in a sort
// that already has a fetch - and so we can merge them.
replaceTop(sort2.getInput());
final RelNode sort =
struct.sortFactory.createSort(peek(), sort2.collation,
first(offsetNode, sort2.offset),
first(fetchNode, sort2.fetch));
replaceTop(sort);
return this;
}
}
if (top instanceof Project) {
final Project project = (Project) top;
if (!project.containsOver() && project.getInput() instanceof Sort) {
final Sort sort2 = (Sort) project.getInput();
if (sort2.offset == null && sort2.fetch == null) {
final RelNode sort =
struct.sortFactory.createSort(sort2.getInput(),
sort2.collation, offsetNode, fetchNode);
replaceTop(
struct.projectFactory.createProject(sort,
project.getHints(),
project.getProjects(),
Pair.right(project.getNamedProjects()),
project.getVariablesSet()));
return this;
}
}
}
}
if (registrar.addedFieldCount() > 0) {
project(registrar.extraNodes);
}
final RelNode sort =
struct.sortFactory.createSort(peek(),
RelCollations.of(fieldCollations), offsetNode, fetchNode);
replaceTop(sort);
if (registrar.addedFieldCount() > 0) {
project(registrar.originalExtraNodes);
}
return this;
}
private static RelFieldCollation collation(RexNode node,
RelFieldCollation.Direction direction,
RelFieldCollation.@Nullable NullDirection nullDirection,
List<RexNode> extraNodes) {
switch (node.getKind()) {
case INPUT_REF:
return new RelFieldCollation(((RexInputRef) node).getIndex(), direction,
first(nullDirection, direction.defaultNullDirection()));
case DESCENDING:
return collation(((RexCall) node).getOperands().get(0),
RelFieldCollation.Direction.DESCENDING,
nullDirection, extraNodes);
case NULLS_FIRST:
return collation(((RexCall) node).getOperands().get(0), direction,
RelFieldCollation.NullDirection.FIRST, extraNodes);
case NULLS_LAST:
return collation(((RexCall) node).getOperands().get(0), direction,
RelFieldCollation.NullDirection.LAST, extraNodes);
default:
final int fieldIndex = extraNodes.size();
extraNodes.add(node);
return new RelFieldCollation(fieldIndex, direction,
first(nullDirection, direction.defaultNullDirection()));
}
}
private static RexFieldCollation rexCollation(RexNode node,
RelFieldCollation.Direction direction,
RelFieldCollation.@Nullable NullDirection nullDirection) {
switch (node.getKind()) {
case DESCENDING:
return rexCollation(((RexCall) node).operands.get(0),
RelFieldCollation.Direction.DESCENDING, nullDirection);
case NULLS_LAST:
return rexCollation(((RexCall) node).operands.get(0),
direction, RelFieldCollation.NullDirection.LAST);
case NULLS_FIRST:
return rexCollation(((RexCall) node).operands.get(0),
direction, RelFieldCollation.NullDirection.FIRST);
default:
final Set<SqlKind> flags = EnumSet.noneOf(SqlKind.class);
if (direction == RelFieldCollation.Direction.DESCENDING) {
flags.add(SqlKind.DESCENDING);
}
if (nullDirection == RelFieldCollation.NullDirection.FIRST) {
flags.add(SqlKind.NULLS_FIRST);
}
if (nullDirection == RelFieldCollation.NullDirection.LAST) {
flags.add(SqlKind.NULLS_LAST);
}
return new RexFieldCollation(node, flags);
}
}
/**
* Creates a projection that converts the current relational expression's
* output to a desired row type.
*
* <p>The desired row type and the row type to be converted must have the
* same number of fields.
*
* @param castRowType row type after cast
* @param rename if true, use field names from castRowType; if false,
* preserve field names from rel
*/
public RelBuilder convert(RelDataType castRowType, boolean rename) {
final RelNode r = build();
final RelNode r2 =
RelOptUtil.createCastRel(r, castRowType, rename,
struct.projectFactory);
push(r2);
return this;
}
public RelBuilder permute(Mapping mapping) {
assert mapping.getMappingType().isSingleSource();
assert mapping.getMappingType().isMandatorySource();
if (mapping.isIdentity()) {
return this;
}
final List<RexNode> exprList = new ArrayList<>();
for (int i = 0; i < mapping.getTargetCount(); i++) {
exprList.add(field(mapping.getSource(i)));
}
return project(exprList);
}
/** Creates a {@link Sample}. (Repeatable if seed is not null.) */
public RelBuilder sample(boolean bernoulli, BigDecimal sampleRate,
@Nullable Integer repeatableSeed) {
boolean repeatable;
int seed;
if (repeatableSeed != null) {
repeatable = true;
seed = repeatableSeed;
} else {
repeatable = false;
seed = 0;
}
return sample(bernoulli, sampleRate, repeatable, seed);
}
/** Creates a {@link Sample}. */
private RelBuilder sample(boolean bernoulli, BigDecimal sampleRate,
boolean repeatable, int repeatableSeed) {
if (sampleRate.compareTo(BigDecimal.ZERO) == 0) {
// The sample rate is 0%; the query should return empty.
return empty();
} else if (sampleRate.compareTo(BigDecimal.ONE) == 0) {
// The table sample rate is 100%; the query should return the contents
// of the underlying table.
return this;
} else {
final Frame frame = stack.pop();
final RelNode r = frame.rel;
final RelOptSamplingParameters param =
new RelOptSamplingParameters(bernoulli, sampleRate, repeatable,
repeatableSeed);
return push(struct.sampleFactory.createSample(r, param));
}
}
/** Creates a {@link Match}. */
public RelBuilder match(RexNode pattern, boolean strictStart,
boolean strictEnd, Map<String, RexNode> patternDefinitions,
Iterable<? extends RexNode> measureList, RexNode after,
Map<String, ? extends SortedSet<String>> subsets, boolean allRows,
Iterable<? extends RexNode> partitionKeys,
Iterable<? extends RexNode> orderKeys, RexNode interval) {
final Registrar registrar =
new Registrar(fields(), peek().getRowType().getFieldNames());
final List<RelFieldCollation> fieldCollations =
registrar.registerFieldCollations(orderKeys);
final ImmutableBitSet partitionBitSet =
ImmutableBitSet.of(registrar.registerExpressions(partitionKeys));
final RelDataTypeFactory.Builder typeBuilder = cluster.getTypeFactory().builder();
for (RexNode partitionKey : partitionKeys) {
typeBuilder.add(partitionKey.toString(), partitionKey.getType());
}
if (allRows) {
for (RexNode orderKey : orderKeys) {
if (!typeBuilder.nameExists(orderKey.toString())) {
typeBuilder.add(orderKey.toString(), orderKey.getType());
}
}
final RelDataType inputRowType = peek().getRowType();
for (RelDataTypeField fs : inputRowType.getFieldList()) {
if (!typeBuilder.nameExists(fs.getName())) {
typeBuilder.add(fs);
}
}
}
final ImmutableMap.Builder<String, RexNode> measures = ImmutableMap.builder();
for (RexNode measure : measureList) {
List<RexNode> operands = ((RexCall) measure).getOperands();
String alias = operands.get(1).toString();
typeBuilder.add(alias, operands.get(0).getType());
measures.put(alias, operands.get(0));
}
final RelNode match =
struct.matchFactory.createMatch(peek(), pattern,
typeBuilder.build(), strictStart, strictEnd, patternDefinitions,
measures.build(), after, subsets, allRows,
partitionBitSet, RelCollations.of(fieldCollations), interval);
stack.push(new Frame(match));
return this;
}
/** Creates a Pivot.
*
* <p>To achieve the same effect as the SQL
*
* <blockquote><pre>{@code
* SELECT *
* FROM (SELECT mgr, deptno, job, sal FROM emp)
* PIVOT (SUM(sal) AS ss, COUNT(*) AS c
* FOR (job, deptno)
* IN (('CLERK', 10) AS c10, ('MANAGER', 20) AS m20))
* }</pre></blockquote>
*
* <p>use the builder as follows:
*
* <blockquote><pre>{@code
* RelBuilder b;
* b.scan("EMP");
* final RelBuilder.GroupKey groupKey = b.groupKey("MGR");
* final List<RelBuilder.AggCall> aggCalls =
* Arrays.asList(b.sum(b.field("SAL")).as("SS"),
* b.count().as("C"));
* final List<RexNode> axes =
* Arrays.asList(b.field("JOB"),
* b.field("DEPTNO"));
* final ImmutableMap.Builder<String, List<RexNode>> valueMap =
* ImmutableMap.builder();
* valueMap.put("C10",
* Arrays.asList(b.literal("CLERK"), b.literal(10)));
* valueMap.put("M20",
* Arrays.asList(b.literal("MANAGER"), b.literal(20)));
* b.pivot(groupKey, aggCalls, axes, valueMap.build().entrySet());
* }</pre></blockquote>
*
* <p>Note that the SQL uses a sub-query to project away columns (e.g.
* {@code HIREDATE}) that it does not reference, so that they do not appear in
* the {@code GROUP BY}. You do not need to do that in this API, because the
* {@code groupKey} parameter specifies the keys.
*
* <p>Pivot is implemented by desugaring. The above example becomes the
* following:
*
* <blockquote><pre>{@code
* SELECT mgr,
* SUM(sal) FILTER (WHERE job = 'CLERK' AND deptno = 10) AS c10_ss,
* COUNT(*) FILTER (WHERE job = 'CLERK' AND deptno = 10) AS c10_c,
* SUM(sal) FILTER (WHERE job = 'MANAGER' AND deptno = 20) AS m20_ss,
* COUNT(*) FILTER (WHERE job = 'MANAGER' AND deptno = 20) AS m20_c
* FROM emp
* GROUP BY mgr
* }</pre></blockquote>
*
* @param groupKey Key columns
* @param aggCalls Aggregate expressions to compute for each value
* @param axes Columns to pivot
* @param values Values to pivot, and the alias for each column group
*
* @return this RelBuilder
*/
public RelBuilder pivot(GroupKey groupKey,
Iterable<? extends AggCall> aggCalls,
Iterable<? extends RexNode> axes,
Iterable<? extends Map.Entry<String,
? extends Iterable<? extends RexNode>>> values) {
final List<RexNode> axisList = ImmutableList.copyOf(axes);
final List<AggCall> multipliedAggCalls = new ArrayList<>();
Pair.forEach(values, (alias, expressions) -> {
final List<RexNode> expressionList = ImmutableList.copyOf(expressions);
if (expressionList.size() != axisList.size()) {
throw new IllegalArgumentException("value count must match axis count ["
+ expressionList + "], [" + axisList + "]");
}
aggCalls.forEach(aggCall -> {
final String alias2 = alias + "_" + ((AggCallPlus) aggCall).alias();
final List<RexNode> filters = new ArrayList<>();
Pair.forEach(axisList, expressionList, (axis, expression) ->
filters.add(equals(axis, expression)));
multipliedAggCalls.add(aggCall.filter(and(filters)).as(alias2));
});
});
return aggregate(groupKey, multipliedAggCalls);
}
/**
* Creates an Unpivot.
*
* <p>To achieve the same effect as the SQL
*
* <blockquote><pre>{@code
* SELECT *
* FROM (SELECT deptno, job, sal, comm FROM emp)
* UNPIVOT INCLUDE NULLS (remuneration
* FOR remuneration_type IN (comm AS 'commission',
* sal AS 'salary'))
* }</pre></blockquote>
*
* <p>use the builder as follows:
*
* <blockquote><pre>{@code
* RelBuilder b;
* b.scan("EMP");
* final List<String> measureNames = Arrays.asList("REMUNERATION");
* final List<String> axisNames = Arrays.asList("REMUNERATION_TYPE");
* final Map<List<RexLiteral>, List<RexNode>> axisMap =
* ImmutableMap.<List<RexLiteral>, List<RexNode>>builder()
* .put(Arrays.asList(b.literal("commission")),
* Arrays.asList(b.field("COMM")))
* .put(Arrays.asList(b.literal("salary")),
* Arrays.asList(b.field("SAL")))
* .build();
* b.unpivot(false, measureNames, axisNames, axisMap);
* }</pre></blockquote>
*
* <p>The query generates two columns: {@code remuneration_type} (an axis
* column) and {@code remuneration} (a measure column). Axis columns contain
* values to indicate the source of the row (in this case, {@code 'salary'}
* if the row came from the {@code sal} column, and {@code 'commission'}
* if the row came from the {@code comm} column).
*
* @param includeNulls Whether to include NULL values in the output
* @param measureNames Names of columns to be generated to hold pivoted
* measures
* @param axisNames Names of columns to be generated to hold qualifying values
* @param axisMap Mapping from the columns that hold measures to the values
* that the axis columns will hold in the generated rows
* @return This RelBuilder
*/
public RelBuilder unpivot(boolean includeNulls,
Iterable<String> measureNames, Iterable<String> axisNames,
Iterable<? extends Map.Entry<? extends List<? extends RexLiteral>,
? extends List<? extends RexNode>>> axisMap) {
// Make immutable copies of all arguments.
final List<String> measureNameList = ImmutableList.copyOf(measureNames);
final List<String> axisNameList = ImmutableList.copyOf(axisNames);
final PairList<List<RexLiteral>, List<RexNode>> map = PairList.of();
Pair.forEach(axisMap, (valueList, inputMeasureList) ->
map.add(ImmutableList.copyOf(valueList),
ImmutableList.copyOf(inputMeasureList)));
// Check that counts match.
map.forEach((valueList, inputMeasureList) -> {
if (inputMeasureList.size() != measureNameList.size()) {
throw new IllegalArgumentException("Number of measures ("
+ inputMeasureList.size() + ") must match number of measure names ("
+ measureNameList.size() + ")");
}
if (valueList.size() != axisNameList.size()) {
throw new IllegalArgumentException("Number of axis values ("
+ valueList.size() + ") match match number of axis names ("
+ axisNameList.size() + ")");
}
});
final RelDataType leftRowType = peek().getRowType();
final BitSet usedFields = new BitSet();
map.forEach((aliases, nodes) ->
nodes.forEach(node -> {
if (node instanceof RexInputRef) {
usedFields.set(((RexInputRef) node).getIndex());
}
}));
// Create "VALUES (('commission'), ('salary')) AS t (remuneration_type)"
values(ImmutableList.copyOf(map.leftList()), axisNameList);
join(JoinRelType.INNER);
final ImmutableBitSet unusedFields =
ImmutableBitSet.range(leftRowType.getFieldCount())
.except(ImmutableBitSet.fromBitSet(usedFields));
final List<RexNode> projects = new ArrayList<>(fields(unusedFields));
Ord.forEach(axisNameList, (dimensionName, d) ->
projects.add(
alias(field(leftRowType.getFieldCount() + d),
dimensionName)));
final List<RexNode> conditions = new ArrayList<>();
Ord.forEach(measureNameList, (measureName, m) -> {
final List<RexNode> caseOperands = new ArrayList<>();
map.forEach((literals, nodes) -> {
Ord.forEach(literals, (literal, d) ->
conditions.add(
equals(field(leftRowType.getFieldCount() + d), literal)));
caseOperands.add(and(conditions));
conditions.clear();
caseOperands.add(nodes.get(m));
});
caseOperands.add(literal(null));
projects.add(
alias(call(SqlStdOperatorTable.CASE, caseOperands),
measureName));
});
project(projects);
if (!includeNulls) {
// Add 'WHERE m1 IS NOT NULL OR m2 IS NOT NULL'
final BitSet notNullFields = new BitSet();
Ord.forEach(measureNameList, (measureName, m) -> {
final int f = unusedFields.cardinality() + axisNameList.size() + m;
conditions.add(isNotNull(field(f)));
notNullFields.set(f);
});
filter(or(conditions));
if (measureNameList.size() == 1) {
// If there is one field, EXCLUDE NULLS will have converted it to NOT
// NULL.
final RelDataTypeFactory.Builder builder = getTypeFactory().builder();
peek().getRowType().getFieldList().forEach(field -> {
final RelDataType type = field.getType();
builder.add(field.getName(),
notNullFields.get(field.getIndex())
? getTypeFactory().createTypeWithNullability(type, false)
: type);
});
convert(builder.build(), false);
}
conditions.clear();
}
return this;
}
/**
* Attaches an array of hints to the stack top relational expression.
*
* <p>The redundant hints would be eliminated.
*
* @param hints Hints
*
* @throws AssertionError if the top relational expression does not implement
* {@link org.apache.calcite.rel.hint.Hintable}
*/
public RelBuilder hints(RelHint... hints) {
return hints(ImmutableList.copyOf(hints));
}
/**
* Attaches multiple hints to the stack top relational expression.
*
* <p>The redundant hints would be eliminated.
*
* @param hints Hints
*
* @throws AssertionError if the top relational expression does not implement
* {@link org.apache.calcite.rel.hint.Hintable}
*/
public RelBuilder hints(Iterable<RelHint> hints) {
requireNonNull(hints, "hints");
final List<RelHint> relHintList =
hints instanceof List ? (List<RelHint>) hints
: Lists.newArrayList(hints);
if (relHintList.isEmpty()) {
return this;
}
final Frame frame = peek_();
if (frame == null) {
throw new IllegalArgumentException(
"There is no relational expression to attach the hints");
}
if (!(frame.rel instanceof Hintable)) {
throw new IllegalArgumentException(
"The top relational expression is not a Hintable");
}
Hintable hintable = (Hintable) frame.rel;
replaceTop(hintable.attachHints(relHintList));
return this;
}
/** Clears the stack.
*
* <p>The builder's state is now the same as when it was created. */
public void clear() {
stack.clear();
}
/** Information necessary to create a call to an aggregate function.
*
* @see RelBuilder#aggregateCall */
public interface AggCall {
SqlParserPos getPosition();
/** Returns a copy of this AggCall that applies a filter before aggregating
* values. */
AggCall filter(@Nullable RexNode condition);
/** Returns a copy of this AggCall that sorts its input values by
* {@code orderKeys} before aggregating, as in SQL's {@code WITHIN GROUP}
* clause. */
AggCall sort(Iterable<RexNode> orderKeys);
/** Returns a copy of this AggCall that sorts its input values by
* {@code orderKeys} before aggregating, as in SQL's {@code WITHIN GROUP}
* clause. */
default AggCall sort(RexNode... orderKeys) {
return sort(ImmutableList.copyOf(orderKeys));
}
/** Returns a copy of this AggCall with the given pre-operands.
*
* <p>Pre-operands apply at the start of aggregation and are constant for
* the whole query. They do not reference input columns and are typically
* {@link RexLiteral}. An example is
* {@link org.apache.calcite.sql.fun.SqlInternalOperators#LITERAL_AGG};
* most aggregate functions do not take pre-operands. */
AggCall preOperands(Iterable<? extends RexNode> preOperands);
default AggCall preOperands(RexNode... preOperands) {
return preOperands(ImmutableList.copyOf(preOperands));
}
/** Returns a copy of this AggCall that makes its input values unique by
* {@code distinctKeys} before aggregating, as in SQL's
* {@code WITHIN DISTINCT} clause. */
AggCall unique(@Nullable Iterable<RexNode> distinctKeys);
/** Returns a copy of this AggCall that makes its input values unique by
* {@code distinctKeys} before aggregating, as in SQL's
* {@code WITHIN DISTINCT} clause. */
default AggCall unique(RexNode... distinctKeys) {
return unique(ImmutableList.copyOf(distinctKeys));
}
/** Returns a copy of this AggCall that may return approximate results
* if {@code approximate} is true. */
AggCall approximate(boolean approximate);
/** Returns a copy of this AggCall that ignores nulls. */
AggCall ignoreNulls(boolean ignoreNulls);
/** Returns a copy of this AggCall with a given alias. */
AggCall as(@Nullable String alias);
/** Returns a copy of this AggCall that is optionally distinct. */
AggCall distinct(boolean distinct);
/** Returns a copy of this AggCall that is distinct. */
default AggCall distinct() {
return distinct(true);
}
/** Converts this aggregate call to a windowed aggregate call. */
OverCall over();
}
/** Internal methods shared by all implementations of {@link AggCall}. */
private interface AggCallPlus extends AggCall {
/** Returns the aggregate function. */
SqlAggFunction op();
/** Returns the alias. */
@Nullable String alias();
/** Returns an {@link AggregateCall} that is approximately equivalent
* to this {@code AggCall} and is good for certain things, such as deriving
* field names. */
AggregateCall aggregateCall();
/** Converts this {@code AggCall} to a good {@link AggregateCall}. */
AggregateCall aggregateCall(Registrar registrar, boolean hasEmptyGroup,
RelNode r);
/** Registers expressions in operands and filters. */
void register(Registrar registrar);
}
/** Information necessary to create the GROUP BY clause of an Aggregate.
*
* @see RelBuilder#groupKey */
public interface GroupKey {
/** Assigns an alias to this group key.
*
* <p>Used to assign field names in the {@code group} operation. */
GroupKey alias(@Nullable String alias);
/** Returns the number of columns in the group key. */
int groupKeyCount();
}
/** Implementation of {@link RelBuilder.GroupKey}. */
static class GroupKeyImpl implements GroupKey {
final ImmutableList<RexNode> nodes;
final @Nullable ImmutableList<ImmutableList<RexNode>> nodeLists;
final @Nullable String alias;
GroupKeyImpl(ImmutableList<RexNode> nodes,
@Nullable ImmutableList<ImmutableList<RexNode>> nodeLists,
@Nullable String alias) {
this.nodes = requireNonNull(nodes, "nodes");
this.nodeLists = nodeLists;
this.alias = alias;
}
@Override public String toString() {
return alias == null ? nodes.toString() : nodes + " as " + alias;
}
@Override public int groupKeyCount() {
return nodes.size();
}
@Override public GroupKey alias(@Nullable String alias) {
return Objects.equals(this.alias, alias)
? this
: new GroupKeyImpl(nodes, nodeLists, alias);
}
boolean isSimple() {
return nodeLists == null || nodeLists.size() == 1;
}
}
/**
* Checks for {@link CorrelationId}, then validates the id is not used on left,
* and finally checks if id is actually used on right.
*
* @return true if a correlate id is present and used
*
* @throws IllegalArgumentException if the {@link CorrelationId} is used by left side or if the a
* {@link CorrelationId} is present and the {@link JoinRelType} is FULL or RIGHT.
*/
private boolean checkIfCorrelated(Set<CorrelationId> variablesSet,
JoinRelType joinType, RelNode leftNode, RelNode rightRel) {
if (variablesSet.size() != 1) {
return false;
}
if (!config.convertCorrelateToJoin()) {
return true;
}
CorrelationId id = Iterables.getOnlyElement(variablesSet);
if (!RelOptUtil.notContainsCorrelation(leftNode, id, Litmus.IGNORE)) {
throw new IllegalArgumentException("variable " + id
+ " must not be used by left input to correlation");
}
switch (joinType) {
case LEFT_ASOF:
case ASOF:
case RIGHT:
case FULL:
throw new IllegalArgumentException("Correlated " + joinType + " join is not supported");
default:
return !RelOptUtil.correlationColumns(
Iterables.getOnlyElement(variablesSet),
rightRel).isEmpty();
}
}
/** Implementation of {@link AggCall}. */
private class AggCallImpl implements AggCallPlus {
private final SqlParserPos pos;
private final SqlAggFunction aggFunction;
private final boolean distinct;
private final boolean approximate;
private final boolean ignoreNulls;
private final @Nullable RexNode filter;
private final @Nullable String alias;
private final ImmutableList<RexNode> preOperands; // may be empty
private final ImmutableList<RexNode> operands; // may be empty
private final @Nullable ImmutableList<RexNode> distinctKeys; // may be empty or null
private final ImmutableList<RexNode> orderKeys; // may be empty
AggCallImpl(SqlParserPos pos, SqlAggFunction aggFunction, boolean distinct,
boolean approximate, boolean ignoreNulls, @Nullable RexNode filter,
@Nullable String alias, ImmutableList<RexNode> preOperands,
ImmutableList<RexNode> operands,
@Nullable ImmutableList<RexNode> distinctKeys,
ImmutableList<RexNode> orderKeys) {
this.pos = pos;
this.aggFunction = requireNonNull(aggFunction, "aggFunction");
// If the aggregate function ignores DISTINCT,
// make the DISTINCT flag FALSE.
this.distinct = distinct
&& aggFunction.getDistinctOptionality() != Optionality.IGNORED;
this.approximate = approximate;
this.ignoreNulls = ignoreNulls;
this.alias = alias;
this.preOperands = requireNonNull(preOperands, "preOperands");
this.operands = requireNonNull(operands, "operands");
this.distinctKeys = distinctKeys;
this.orderKeys = requireNonNull(orderKeys, "orderKeys");
if (filter != null) {
if (filter.getType().getSqlTypeName() != SqlTypeName.BOOLEAN) {
throw RESOURCE.filterMustBeBoolean().ex();
}
if (filter.getType().isNullable()) {
filter = call(SqlStdOperatorTable.IS_TRUE, filter);
}
}
this.filter = filter;
}
@Override public SqlParserPos getPosition() {
return pos;
}
@Override public String toString() {
final StringBuilder b = new StringBuilder();
b.append(aggFunction.getName())
.append('(');
if (distinct) {
b.append("DISTINCT ");
}
if (!preOperands.isEmpty()) {
b.append(preOperands.get(0));
for (int i = 1; i < preOperands.size(); i++) {
b.append(", ");
b.append(preOperands.get(i));
}
b.append(operands.isEmpty() ? ";" : "; ");
}
if (!operands.isEmpty()) {
b.append(operands.get(0));
for (int i = 1; i < operands.size(); i++) {
b.append(", ");
b.append(operands.get(i));
}
}
b.append(')');
if (filter != null) {
b.append(" FILTER (WHERE ").append(filter).append(')');
}
if (distinctKeys != null) {
b.append(" WITHIN DISTINCT (").append(distinctKeys).append(')');
}
return b.toString();
}
@Override public SqlAggFunction op() {
return aggFunction;
}
@Override public @Nullable String alias() {
return alias;
}
@Override public AggregateCall aggregateCall() {
// Use dummy values for collation and type. This method only promises to
// return a call that is "approximately equivalent ... and is good for
// deriving field names", so dummy values are good enough.
final RelCollation collation = RelCollations.EMPTY;
final RelDataType type;
if (aggFunction.getKind() == SqlKind.GROUP_ID) {
// The return type of GROUP_ID function is SqlTypeName.BIGINT,
// see SqlGroupIdFunction.
type = getTypeFactory().createSqlType(SqlTypeName.BIGINT);
} else {
type = getTypeFactory().createSqlType(SqlTypeName.BOOLEAN);
}
return AggregateCall.create(pos, aggFunction, distinct, approximate,
ignoreNulls, preOperands, ImmutableList.of(), -1,
null, collation, type, alias);
}
@Override public AggregateCall aggregateCall(Registrar registrar,
boolean hasEmptyGroup, RelNode r) {
List<Integer> args =
registrar.registerExpressions(this.operands);
final int filterArg =
this.filter == null ? -1
: registrar.registerExpression(this.filter);
if (this.distinct && !this.aggFunction.isQuantifierAllowed()) {
throw new IllegalArgumentException("DISTINCT not allowed");
}
if (this.filter != null && !this.aggFunction.allowsFilter()) {
throw new IllegalArgumentException("FILTER not allowed");
}
final @Nullable ImmutableBitSet distinctKeys =
this.distinctKeys == null
? null
: ImmutableBitSet.of(
registrar.registerExpressions(this.distinctKeys));
final RelCollation collation =
RelCollations.of(this.orderKeys
.stream()
.map(orderKey ->
collation(orderKey, RelFieldCollation.Direction.ASCENDING,
null, Collections.emptyList()))
.collect(Collectors.toList()));
if (aggFunction instanceof SqlCountAggFunction && !distinct) {
args = args.stream()
.filter(r::fieldIsNullable)
.collect(toImmutableList());
}
return AggregateCall.create(pos, aggFunction, distinct, approximate,
ignoreNulls, preOperands, args, filterArg, distinctKeys,
collation, hasEmptyGroup, r, null, alias);
}
@Override public void register(Registrar registrar) {
registrar.registerExpressions(operands);
if (filter != null) {
registrar.registerExpression(filter);
}
if (distinctKeys != null) {
registrar.registerExpressions(distinctKeys);
}
registrar.registerExpressions(orderKeys);
}
@Override public AggCall preOperands(
Iterable<? extends RexNode> preOperands) {
final ImmutableList<RexNode> preOperandList =
ImmutableList.copyOf(preOperands);
return preOperandList.equals(this.preOperands)
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperandList, operands, distinctKeys, orderKeys);
}
@Override public OverCall over() {
return new OverCallImpl(aggFunction, distinct, operands, ignoreNulls,
alias);
}
@Override public AggCall sort(Iterable<RexNode> orderKeys) {
final ImmutableList<RexNode> orderKeyList =
ImmutableList.copyOf(orderKeys);
return orderKeyList.equals(this.orderKeys)
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeyList);
}
@Override public AggCall sort(RexNode... orderKeys) {
return sort(ImmutableList.copyOf(orderKeys));
}
@Override public AggCall unique(@Nullable Iterable<RexNode> distinctKeys) {
final @Nullable ImmutableList<RexNode> distinctKeyList =
distinctKeys == null ? null : ImmutableList.copyOf(distinctKeys);
return Objects.equals(distinctKeyList, this.distinctKeys)
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeyList, orderKeys);
}
@Override public AggCall approximate(boolean approximate) {
return approximate == this.approximate
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeys);
}
@Override public AggCall filter(@Nullable RexNode condition) {
return Objects.equals(condition, this.filter)
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
condition, alias, preOperands, operands, distinctKeys, orderKeys);
}
@Override public AggCall as(@Nullable String alias) {
return Objects.equals(alias, this.alias)
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeys);
}
@Override public AggCall distinct(boolean distinct) {
return distinct == this.distinct
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeys);
}
@Override public AggCall ignoreNulls(boolean ignoreNulls) {
return ignoreNulls == this.ignoreNulls
? this
: new AggCallImpl(pos, aggFunction, distinct, approximate, ignoreNulls,
filter, alias, preOperands, operands, distinctKeys, orderKeys);
}
}
/** Implementation of {@link AggCall} that wraps an
* {@link AggregateCall}. */
private class AggCallImpl2 implements AggCallPlus {
private final AggregateCall aggregateCall;
private final ImmutableList<RexNode> operands;
AggCallImpl2(AggregateCall aggregateCall, ImmutableList<RexNode> operands) {
this.aggregateCall = requireNonNull(aggregateCall, "aggregateCall");
this.operands = requireNonNull(operands, "operands");
}
@Override public OverCall over() {
return new OverCallImpl(aggregateCall.getAggregation(),
aggregateCall.isDistinct(), operands, aggregateCall.ignoreNulls(),
aggregateCall.name);
}
@Override public SqlParserPos getPosition() {
return this.aggregateCall.getParserPosition();
}
@Override public String toString() {
return aggregateCall.toString();
}
@Override public SqlAggFunction op() {
return aggregateCall.getAggregation();
}
@Override public @Nullable String alias() {
return aggregateCall.name;
}
@Override public AggregateCall aggregateCall() {
return aggregateCall;
}
@Override public AggregateCall aggregateCall(Registrar registrar,
boolean hasEmptyGroup, RelNode r) {
return AggregateCall.create(aggregateCall.getParserPosition(), aggregateCall.getAggregation(),
aggregateCall.isDistinct(), aggregateCall.isApproximate(), aggregateCall.ignoreNulls(),
aggregateCall.rexList, aggregateCall.getArgList(), aggregateCall.filterArg,
aggregateCall.distinctKeys, aggregateCall.getCollation(), hasEmptyGroup, r, null,
aggregateCall.name);
}
@Override public void register(Registrar registrar) {
// nothing to do
}
@Override public AggCall preOperands(Iterable<? extends RexNode> preOperands) {
throw new UnsupportedOperationException();
}
@Override public AggCall sort(Iterable<RexNode> orderKeys) {
throw new UnsupportedOperationException();
}
@Override public AggCall sort(RexNode... orderKeys) {
throw new UnsupportedOperationException();
}
@Override public AggCall unique(@Nullable Iterable<RexNode> distinctKeys) {
throw new UnsupportedOperationException();
}
@Override public AggCall approximate(boolean approximate) {
throw new UnsupportedOperationException();
}
@Override public AggCall filter(@Nullable RexNode condition) {
throw new UnsupportedOperationException();
}
@Override public AggCall as(@Nullable String alias) {
throw new UnsupportedOperationException();
}
@Override public AggCall distinct(boolean distinct) {
throw new UnsupportedOperationException();
}
@Override public AggCall ignoreNulls(boolean ignoreNulls) {
throw new UnsupportedOperationException();
}
}
/** Call to a windowed aggregate function.
*
* <p>To create an {@code OverCall}, start with an {@link AggCall} (created
* by a method such as {@link #aggregateCall}, {@link #sum} or {@link #count})
* and call its {@link AggCall#over()} method. For example,
*
* <pre>{@code
* b.scan("EMP")
* .project(b.field("DEPTNO"),
* b.aggregateCall(SqlStdOperatorTable.ROW_NUMBER)
* .over()
* .partitionBy()
* .orderBy(b.field("EMPNO"))
* .rowsUnbounded()
* .allowPartial(true)
* .nullWhenCountZero(false)
* .as("x"))
* }</pre>
*
* <p>Unlike an aggregate call, a windowed aggregate call is an expression
* that you can use in a {@link Project} or {@link Filter}. So, to finish,
* call {@link OverCall#toRex()} to convert the {@code OverCall} to a
* {@link RexNode}; the {@link OverCall#as} method (used in the above example)
* does the same but also assigns an column alias.
*/
public interface OverCall {
/** Performs an action on this OverCall. */
default <R> R let(Function<OverCall, R> consumer) {
return consumer.apply(this);
}
/** Sets the PARTITION BY clause to an array of expressions. */
OverCall partitionBy(RexNode... expressions);
/** Sets the PARTITION BY clause to a list of expressions. */
OverCall partitionBy(Iterable<? extends RexNode> expressions);
/** Sets the ORDER BY BY clause to an array of expressions.
*
* <p>Use {@link #desc(RexNode)}, {@link #nullsFirst(RexNode)},
* {@link #nullsLast(RexNode)} to control the sort order. */
OverCall orderBy(RexNode... expressions);
/** Sets the ORDER BY BY clause to a list of expressions.
*
* <p>Use {@link #desc(RexNode)}, {@link #nullsFirst(RexNode)},
* {@link #nullsLast(RexNode)} to control the sort order. */
OverCall orderBy(Iterable<? extends RexNode> expressions);
/** Sets an unbounded ROWS window,
* equivalent to SQL {@code ROWS BETWEEN UNBOUNDED PRECEDING AND
* UNBOUNDED FOLLOWING}. */
default OverCall rowsUnbounded() {
return rowsBetween(RexWindowBounds.UNBOUNDED_PRECEDING,
RexWindowBounds.UNBOUNDED_FOLLOWING);
}
/** Sets a ROWS window with a lower bound,
* equivalent to SQL {@code ROWS BETWEEN lower AND CURRENT ROW}. */
default OverCall rowsFrom(RexWindowBound lower) {
return rowsBetween(lower, RexWindowBounds.UNBOUNDED_FOLLOWING);
}
/** Sets a ROWS window with an upper bound,
* equivalent to SQL {@code ROWS BETWEEN CURRENT ROW AND upper}. */
default OverCall rowsTo(RexWindowBound upper) {
return rowsBetween(RexWindowBounds.UNBOUNDED_PRECEDING, upper);
}
/** Sets a RANGE window with lower and upper bounds,
* equivalent to SQL {@code ROWS BETWEEN lower ROW AND upper}. */
OverCall rowsBetween(RexWindowBound lower, RexWindowBound upper);
/** Sets an unbounded RANGE window,
* equivalent to SQL {@code RANGE BETWEEN UNBOUNDED PRECEDING AND
* UNBOUNDED FOLLOWING}. */
default OverCall rangeUnbounded() {
return rangeBetween(RexWindowBounds.UNBOUNDED_PRECEDING,
RexWindowBounds.UNBOUNDED_FOLLOWING);
}
/** Sets a RANGE window with a lower bound,
* equivalent to SQL {@code RANGE BETWEEN lower AND CURRENT ROW}. */
default OverCall rangeFrom(RexWindowBound lower) {
return rangeBetween(lower, RexWindowBounds.CURRENT_ROW);
}
/** Sets a RANGE window with an upper bound,
* equivalent to SQL {@code RANGE BETWEEN CURRENT ROW AND upper}. */
default OverCall rangeTo(RexWindowBound upper) {
return rangeBetween(RexWindowBounds.UNBOUNDED_PRECEDING, upper);
}
/** Sets the frame to EXCLUDE rows; default to EXCLUDE_NO_OTHER. */
OverCall exclude(RexWindowExclusion exclude);
/** Sets a RANGE window with lower and upper bounds,
* equivalent to SQL {@code RANGE BETWEEN lower ROW AND upper}. */
OverCall rangeBetween(RexWindowBound lower, RexWindowBound upper);
/** Sets whether to allow partial width windows; default true. */
OverCall allowPartial(boolean allowPartial);
/** Sets whether the aggregate function should evaluate to null if no rows
* are in the window; default false. */
OverCall nullWhenCountZero(boolean nullWhenCountZero);
/** Sets the alias of this expression, and converts it to a {@link RexNode};
* default is the alias that was set via {@link AggCall#as(String)}. */
RexNode as(String alias);
/** Converts this expression to a {@link RexNode}. */
RexNode toRex();
}
/** Implementation of {@link OverCall}. */
private class OverCallImpl implements OverCall {
private final ImmutableList<RexNode> operands;
private final boolean ignoreNulls;
private final @Nullable String alias;
private final boolean nullWhenCountZero;
private final boolean allowPartial;
private final boolean rows;
private final RexWindowBound lowerBound;
private final RexWindowBound upperBound;
private final RexWindowExclusion exclude;
private final ImmutableList<RexNode> partitionKeys;
private final ImmutableList<RexFieldCollation> sortKeys;
private final SqlAggFunction op;
private final boolean distinct;
private OverCallImpl(SqlAggFunction op, boolean distinct,
ImmutableList<RexNode> operands, boolean ignoreNulls,
@Nullable String alias, ImmutableList<RexNode> partitionKeys,
ImmutableList<RexFieldCollation> sortKeys, boolean rows,
RexWindowBound lowerBound, RexWindowBound upperBound,
boolean nullWhenCountZero, boolean allowPartial, RexWindowExclusion exclude) {
this.op = op;
this.distinct = distinct;
this.operands = operands;
this.ignoreNulls = ignoreNulls;
this.alias = alias;
this.partitionKeys = partitionKeys;
this.sortKeys = sortKeys;
this.nullWhenCountZero = nullWhenCountZero;
this.allowPartial = allowPartial;
this.rows = rows;
this.lowerBound = lowerBound;
this.upperBound = upperBound;
this.exclude = exclude;
}
/** Creates an OverCallImpl with default settings. */
OverCallImpl(SqlAggFunction op, boolean distinct,
ImmutableList<RexNode> operands, boolean ignoreNulls,
@Nullable String alias) {
this(op, distinct, operands, ignoreNulls, alias, ImmutableList.of(),
ImmutableList.of(), true, RexWindowBounds.UNBOUNDED_PRECEDING,
RexWindowBounds.UNBOUNDED_FOLLOWING, false, true, RexWindowExclusion.EXCLUDE_NO_OTHER);
}
@Override public OverCall partitionBy(
Iterable<? extends RexNode> expressions) {
return partitionBy_(ImmutableList.copyOf(expressions));
}
@Override public OverCall partitionBy(RexNode... expressions) {
return partitionBy_(ImmutableList.copyOf(expressions));
}
private OverCall partitionBy_(ImmutableList<RexNode> partitionKeys) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, rows, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
private OverCall orderBy_(ImmutableList<RexFieldCollation> sortKeys) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, rows, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
@Override public OverCall orderBy(Iterable<? extends RexNode> sortKeys) {
ImmutableList.Builder<RexFieldCollation> fieldCollations =
ImmutableList.builder();
sortKeys.forEach(sortKey ->
fieldCollations.add(
rexCollation(sortKey, RelFieldCollation.Direction.ASCENDING,
RelFieldCollation.NullDirection.UNSPECIFIED)));
return orderBy_(fieldCollations.build());
}
@Override public OverCall orderBy(RexNode... sortKeys) {
return orderBy(Arrays.asList(sortKeys));
}
@Override public OverCall rowsBetween(RexWindowBound lowerBound,
RexWindowBound upperBound) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, true, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
@Override public OverCall rangeBetween(RexWindowBound lowerBound,
RexWindowBound upperBound) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, false, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
@Override public OverCall exclude(RexWindowExclusion exclude) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, rows, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
@Override public OverCall allowPartial(boolean allowPartial) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, rows, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
@Override public OverCall nullWhenCountZero(boolean nullWhenCountZero) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, rows, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude);
}
@Override public RexNode as(String alias) {
return new OverCallImpl(op, distinct, operands, ignoreNulls, alias,
partitionKeys, sortKeys, rows, lowerBound, upperBound,
nullWhenCountZero, allowPartial, exclude).toRex();
}
@Override public RexNode toRex() {
final RexCallBinding bind =
new RexCallBinding(getTypeFactory(), op, operands,
ImmutableList.of()) {
@Override public boolean hasEmptyGroup() {
return !SqlWindow.isAlwaysNonEmpty(lowerBound, upperBound);
}
};
final RelDataType type = op.inferReturnType(bind);
final RexNode over = getRexBuilder()
.makeOver(type, op, operands, partitionKeys, sortKeys,
lowerBound, upperBound, exclude, rows, allowPartial, nullWhenCountZero,
distinct, ignoreNulls);
return aliasMaybe(over, alias);
}
}
/** Collects the extra expressions needed for {@link #aggregate}.
*
* <p>The extra expressions come from the group key and as arguments to
* aggregate calls, and later there will be a {@link #project} or a
* {@link #rename(List)} if necessary. */
private static class Registrar {
final List<RexNode> originalExtraNodes;
final List<RexNode> extraNodes;
final List<@Nullable String> names;
Registrar(Iterable<RexNode> fields, List<String> fieldNames) {
originalExtraNodes = ImmutableList.copyOf(fields);
extraNodes = new ArrayList<>(originalExtraNodes);
names = new ArrayList<>(fieldNames);
}
int registerExpression(RexNode node) {
switch (node.getKind()) {
case AS:
final List<RexNode> operands = ((RexCall) node).operands;
final int i = registerExpression(operands.get(0));
names.set(i, RexLiteral.stringValue(operands.get(1)));
return i;
case DESCENDING:
case NULLS_FIRST:
case NULLS_LAST:
return registerExpression(((RexCall) node).operands.get(0));
default:
final int i2 = extraNodes.indexOf(node);
if (i2 >= 0) {
return i2;
}
extraNodes.add(node);
names.add(null);
return extraNodes.size() - 1;
}
}
List<Integer> registerExpressions(Iterable<? extends RexNode> nodes) {
final List<Integer> builder = new ArrayList<>();
for (RexNode node : nodes) {
builder.add(registerExpression(node));
}
return builder;
}
List<RelFieldCollation> registerFieldCollations(
Iterable<? extends RexNode> orderKeys) {
final List<RelFieldCollation> fieldCollations = new ArrayList<>();
for (RexNode orderKey : orderKeys) {
final RelFieldCollation collation =
collation(orderKey, RelFieldCollation.Direction.ASCENDING, null,
extraNodes);
if (!RelCollations.ordinals(fieldCollations)
.contains(collation.getFieldIndex())) {
fieldCollations.add(collation);
}
}
return ImmutableList.copyOf(fieldCollations);
}
/** Returns the number of fields added. */
int addedFieldCount() {
return extraNodes.size() - originalExtraNodes.size();
}
}
/** Builder stack frame.
*
* <p>Describes a previously created relational expression and
* information about how table aliases map into its row type. */
private static class Frame {
final RelNode rel;
final ImmutablePairList<ImmutableSet<String>, RelDataTypeField> fields;
private Frame(RelNode rel,
PairList<ImmutableSet<String>, RelDataTypeField> fields) {
this.rel = rel;
this.fields = fields.immutable();
}
private Frame(RelNode rel) {
String tableAlias = deriveAlias(rel);
final PairList<ImmutableSet<String>, RelDataTypeField> fields =
PairList.of();
final ImmutableSet<String> aliases =
tableAlias == null
? ImmutableSet.of()
: ImmutableSet.of(tableAlias);
for (RelDataTypeField field : rel.getRowType().getFieldList()) {
fields.add(aliases, field);
}
this.rel = rel;
this.fields = fields.immutable();
}
@Override public String toString() {
return rel + ": " + fields;
}
private static @Nullable String deriveAlias(RelNode rel) {
if (rel instanceof TableScan) {
TableScan scan = (TableScan) rel;
final List<String> names = scan.getTable().getQualifiedName();
if (!names.isEmpty()) {
return Util.last(names);
}
}
return null;
}
List<RelDataTypeField> fields() {
return fields.rightList();
}
}
/** Shuttle that shifts a predicate's inputs to the left, replacing early
* ones with references to a
* {@link RexCorrelVariable}. */
private class Shifter extends RexShuttle {
private final RelNode left;
private final CorrelationId id;
private final RelNode right;
Shifter(RelNode left, CorrelationId id, RelNode right) {
this.left = left;
this.id = id;
this.right = right;
}
@Override public RexNode visitInputRef(RexInputRef inputRef) {
final RelDataType leftRowType = left.getRowType();
final RexBuilder rexBuilder = getRexBuilder();
final int leftCount = leftRowType.getFieldCount();
if (inputRef.getIndex() < leftCount) {
final RexNode v = rexBuilder.makeCorrel(leftRowType, id);
return rexBuilder.makeFieldAccess(v, inputRef.getIndex());
} else {
return rexBuilder.makeInputRef(right, inputRef.getIndex() - leftCount);
}
}
}
/** Configuration of RelBuilder.
*
* <p>It is immutable, and all fields are public.
*
* <p>Start with the {@link #DEFAULT} instance,
* and call {@code withXxx} methods to set its properties. */
@Value.Immutable
public interface Config {
/** Default configuration. */
Config DEFAULT = ImmutableRelBuilder.Config.of();
/** Controls whether to merge two {@link Project} operators when inlining
* expressions causes complexity to increase.
*
* <p>Usually merging projects is beneficial, but occasionally the
* result is more complex than the original projects. Consider:
*
* <pre>
* P: Project(a+b+c AS x, d+e+f AS y, g+h+i AS z) # complexity 15
* Q: Project(x*y*z AS p, x-y-z AS q) # complexity 10
* R: Project((a+b+c)*(d+e+f)*(g+h+i) AS s,
* (a+b+c)-(d+e+f)-(g+h+i) AS t) # complexity 34
* </pre>
*
* <p>The complexity of an expression is the number of nodes (leaves and
* operators). For example, {@code a+b+c} has complexity 5 (3 field
* references and 2 calls):
*
* <pre>
* +
* / \
* + c
* / \
* a b
* </pre>
*
* <p>A negative value never allows merges.
*
* <p>A zero or positive value, {@code bloat}, allows a merge if complexity
* of the result is less than or equal to the sum of the complexity of the
* originals plus {@code bloat}.
*
* <p>The default value, 100, allows a moderate increase in complexity but
* prevents cases where complexity would run away into the millions and run
* out of memory. Moderate complexity is OK; the implementation, say via
* {@link org.apache.calcite.adapter.enumerable.EnumerableCalc}, will often
* gather common sub-expressions and compute them only once.
*/
@Value.Default default int bloat() {
return RelOptUtil.DEFAULT_BLOAT;
}
/** Sets {@link #bloat}. */
Config withBloat(int bloat);
/** Whether {@link RelBuilder#aggregate} should eliminate duplicate
* aggregate calls; default true. */
@Value.Default default boolean dedupAggregateCalls() {
return true;
}
/** Sets {@link #dedupAggregateCalls}. */
Config withDedupAggregateCalls(boolean dedupAggregateCalls);
/** Whether {@link RelBuilder#aggregate} should prune unused
* input columns; default true. */
@Value.Default default boolean pruneInputOfAggregate() {
return true;
}
/** Sets {@link #pruneInputOfAggregate}. */
Config withPruneInputOfAggregate(boolean pruneInputOfAggregate);
/** Whether to ensure that relational operators always have at least one
* column. */
@Value.Default default boolean preventEmptyFieldList() {
return true;
}
/** Sets {@link #preventEmptyFieldList()}. */
Config withPreventEmptyFieldList(boolean preventEmptyFieldList);
/** Whether to push down join conditions; default false (but
* {@link SqlToRelConverter#config()} by default sets this to true). */
@Value.Default default boolean pushJoinCondition() {
return false;
}
/** Sets {@link #pushJoinCondition()}. */
Config withPushJoinCondition(boolean pushJoinCondition);
/** Whether to simplify expressions; default true. */
@Value.Default default boolean simplify() {
return true;
}
/** Sets {@link #simplify}. */
Config withSimplify(boolean simplify);
/** Whether to simplify LIMIT 0 to an empty relation; default true. */
@Value.Default default boolean simplifyLimit() {
return true;
}
/** Sets {@link #simplifyLimit()}. */
Config withSimplifyLimit(boolean simplifyLimit);
/** Whether to simplify {@code Union(Values, Values)} or
* {@code Union(Project(Values))} to {@code Values}; default true. */
@Value.Default default boolean simplifyValues() {
return true;
}
/** Sets {@link #simplifyValues()}. */
Config withSimplifyValues(boolean simplifyValues);
/** Whether to create an Aggregate even if we know that the input is
* already unique; default false. */
@Value.Default default boolean aggregateUnique() {
return false;
}
/** Sets {@link #aggregateUnique()}. */
Config withAggregateUnique(boolean aggregateUnique);
/** Whether to convert Correlate to Join if correlation variable is unused. */
@Value.Default default boolean convertCorrelateToJoin() {
return true;
}
/** Sets {@link #convertCorrelateToJoin()}. */
Config withConvertCorrelateToJoin(boolean convertCorrelateToJoin);
/** Whether to remove the distinct that in aggregate if we know that the input is
* already unique; default false. */
@Value.Default
default boolean removeRedundantDistinct() {
return false;
}
/**
* Sets {@link #removeRedundantDistinct()}.
*/
Config withRemoveRedundantDistinct(boolean removeRedundantDistinct);
}
/** Working state for {@link #aggregateRex}. */
private class AggBuilder {
final ImmutableList<RexNode> groupKeys;
final List<AggCall> aggCalls = new ArrayList<>();
private AggBuilder(ImmutableList<RexNode> groupKeys) {
this.groupKeys = groupKeys;
}
/** Adds a node that we know to contain an aggregate function, and returns
* an expression whose input row type is the output row type of the
* aggregate layer ({@link #groupKeys} and {@link #aggCalls}). */
private RexNode convert(RegisterAgg registrar, RexNode node,
@Nullable String name) {
switch (node.getKind()) {
case AS:
final ImmutableList<RexNode> asOperands = ((RexCall) node).operands;
final String name2;
if (name != null) {
name2 = name;
} else {
final RexLiteral literal = (RexLiteral) asOperands.get(1);
name2 = requireNonNull(literal.getValueAs(String.class));
}
final RexNode node2 = convert(registrar, asOperands.get(0), name2);
return alias(node2, name2);
case INPUT_REF:
final int j = groupKeys.indexOf(node);
if (j < 0) {
throw new IllegalArgumentException("not a group key: " + node);
}
return field(j);
default:
if (node instanceof RexCall) {
final RexCall call = (RexCall) node;
if (call.getOperator().isAggregator()) {
// return a reference to the i'th agg call
return registrar.registerAgg((SqlAggFunction) call.op,
call.operands, call.type, name);
} else {
return call.clone(call.type,
Util.transform(call.operands, operand ->
convert(registrar, operand, null)));
}
}
return node;
}
}
void registerExpressions(Iterable<? extends RexNode> nodes) {
for (RexNode node : nodes) {
convert(this::registerAgg, node, null);
}
}
RexInputRef registerAgg(SqlAggFunction op, List<RexNode> operands,
RelDataType type, @Nullable String name) {
final int i = groupKeys.size() + aggCalls.size();
aggCalls.add(aggregateCall(op, operands).as(name));
return getRexBuilder().makeInputRef(type, i);
}
void convertExpressions(Consumer<RexNode> projects,
Iterable<? extends RexNode> nodes) {
final AtomicInteger j = new AtomicInteger(groupKeys.size());
for (RexNode node : nodes) {
projects.accept(
convert((op, operands, type, name) -> field(j.getAndIncrement()),
node, null));
}
}
}
/** Callback to handle creation of an aggregate call in
* {@link AggBuilder#convert}. */
private interface RegisterAgg {
RexInputRef registerAgg(SqlAggFunction op, List<RexNode> operands,
RelDataType type, @Nullable String name);
}
}