com.hazelcast.org.apache.calcite.plan.volcano.RelSet Maven / Gradle / Ivy
/*
* 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 com.hazelcast.org.apache.calcite.plan.volcano;
import com.hazelcast.org.apache.calcite.plan.Convention;
import com.hazelcast.org.apache.calcite.plan.RelOptCluster;
import com.hazelcast.org.apache.calcite.plan.RelOptListener;
import com.hazelcast.org.apache.calcite.plan.RelOptUtil;
import com.hazelcast.org.apache.calcite.plan.RelTrait;
import com.hazelcast.org.apache.calcite.plan.RelTraitDef;
import com.hazelcast.org.apache.calcite.plan.RelTraitSet;
import com.hazelcast.org.apache.calcite.rel.RelNode;
import com.hazelcast.org.apache.calcite.rel.convert.Converter;
import com.hazelcast.org.apache.calcite.rel.core.CorrelationId;
import com.hazelcast.org.apache.calcite.rel.core.Spool;
import com.hazelcast.org.apache.calcite.util.Pair;
import com.hazelcast.org.apache.calcite.util.trace.CalciteTrace;
import com.hazelcast.com.google.common.collect.ImmutableList;
import com.hazelcast.org.checkerframework.checker.nullness.qual.MonotonicNonNull;
import com.hazelcast.org.checkerframework.checker.nullness.qual.Nullable;
import com.hazelcast.org.slf4j.Logger;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import static com.hazelcast.org.apache.calcite.linq4j.Nullness.castNonNull;
import static java.util.Objects.requireNonNull;
/**
* A RelSet is an equivalence-set of expressions; that is, a set of
* expressions which have identical semantics. We are generally interested in
* using the expression which has the lowest cost.
*
* All of the expressions in an RelSet have the same calling
* convention.
*/
class RelSet {
//~ Static fields/initializers ---------------------------------------------
private static final Logger LOGGER = CalciteTrace.getPlannerTracer();
//~ Instance fields --------------------------------------------------------
final List rels = new ArrayList<>();
/**
* Relational expressions that have a subset in this set as a child. This
* is a multi-set. If multiple relational expressions in this set have the
* same parent, there will be multiple entries.
*/
final List parents = new ArrayList<>();
final List subsets = new ArrayList<>();
/**
* Set to the superseding set when this is found to be equivalent to another
* set.
*/
@MonotonicNonNull RelSet equivalentSet;
@MonotonicNonNull RelNode rel;
/**
* Exploring state of current RelSet.
*/
@Nullable ExploringState exploringState;
/**
* Records conversions / enforcements that have happened on the
* pair of derived and required traitset.
*/
final Set> conversions = new HashSet<>();
/**
* Variables that are set by relational expressions in this set
* and available for use by parent and child expressions.
*/
final Set variablesPropagated;
/**
* Variables that are used by relational expressions in this set.
*/
final Set variablesUsed;
final int id;
/**
* Reentrancy flag.
*/
boolean inMetadataQuery;
//~ Constructors -----------------------------------------------------------
RelSet(
int id,
Set variablesPropagated,
Set variablesUsed) {
this.id = id;
this.variablesPropagated = variablesPropagated;
this.variablesUsed = variablesUsed;
}
//~ Methods ----------------------------------------------------------------
/**
* Returns all of the {@link RelNode}s which reference {@link RelNode}s in
* this set.
*/
public List getParentRels() {
return parents;
}
/**
* Returns the child RelSet for the current set.
*/
public Set getChildSets(VolcanoPlanner planner) {
Set childSets = new HashSet<>();
for (RelNode node : this.rels) {
if (node instanceof Converter) {
continue;
}
for (RelNode child : node.getInputs()) {
RelSet childSet = planner.equivRoot(((RelSubset) child).getSet());
if (childSet.id != this.id) {
childSets.add(childSet);
}
}
}
return childSets;
}
/**
* Returns all of the {@link RelNode}s contained by any subset of this set
* (does not include the subset objects themselves).
*/
public List getRelsFromAllSubsets() {
return rels;
}
public @Nullable RelSubset getSubset(RelTraitSet traits) {
for (RelSubset subset : subsets) {
if (subset.getTraitSet().equals(traits)) {
return subset;
}
}
return null;
}
/**
* Removes all references to a specific {@link RelNode} in both the subsets
* and their parent relationships.
*/
void obliterateRelNode(RelNode rel) {
parents.remove(rel);
}
/**
* Adds a relational expression to a set, with its results available under a
* particular calling convention. An expression may be in the set several
* times with different calling conventions (and hence different costs).
*/
public RelSubset add(RelNode rel) {
assert equivalentSet == null : "adding to a dead set";
final RelTraitSet traitSet = rel.getTraitSet().simplify();
final RelSubset subset = getOrCreateSubset(
rel.getCluster(), traitSet, rel.isEnforcer());
subset.add(rel);
return subset;
}
/**
* If the subset is required, convert delivered subsets to this subset.
* Otherwise, convert this subset to required subsets in this RelSet.
* The subset can be both required and delivered.
*/
void addConverters(RelSubset subset, boolean required,
boolean useAbstractConverter) {
RelOptCluster cluster = subset.getCluster();
List others = subsets.stream().filter(
n -> required ? n.isDelivered() : n.isRequired())
.collect(Collectors.toList());
for (RelSubset other : others) {
assert other.getTraitSet().size() == subset.getTraitSet().size();
RelSubset from = subset;
RelSubset to = other;
if (required) {
from = other;
to = subset;
}
if (from == to
|| to.isEnforceDisabled()
|| useAbstractConverter
&& from.getConvention() != null
&& !from.getConvention().useAbstractConvertersForConversion(
from.getTraitSet(), to.getTraitSet())) {
continue;
}
if (!conversions.add(Pair.of(from.getTraitSet(), to.getTraitSet()))) {
continue;
}
final ImmutableList difference =
to.getTraitSet().difference(from.getTraitSet());
boolean needsConverter = false;
for (RelTrait fromTrait : difference) {
RelTraitDef traitDef = fromTrait.getTraitDef();
RelTrait toTrait = to.getTraitSet().getTrait(traitDef);
if (toTrait == null || !traitDef.canConvert(
cluster.getPlanner(), fromTrait, toTrait)) {
needsConverter = false;
break;
}
if (!fromTrait.satisfies(toTrait)) {
needsConverter = true;
}
}
if (needsConverter) {
final RelNode enforcer;
if (useAbstractConverter) {
enforcer = new AbstractConverter(
cluster, from, null, to.getTraitSet());
} else {
Convention convention = requireNonNull(
subset.getConvention(),
() -> "convention is null for " + subset);
enforcer = convention.enforce(from, to.getTraitSet());
}
if (enforcer != null) {
cluster.getPlanner().register(enforcer, to);
}
}
}
}
RelSubset getOrCreateSubset(
RelOptCluster cluster, RelTraitSet traits, boolean required) {
boolean needsConverter = false;
final VolcanoPlanner planner = (VolcanoPlanner) cluster.getPlanner();
RelSubset subset = getSubset(traits);
if (subset == null) {
needsConverter = true;
subset = new RelSubset(cluster, this, traits);
// Need to first add to subset before adding the abstract
// converters (for others->subset), since otherwise during
// register() the planner will try to add this subset again.
subsets.add(subset);
if (planner.getListener() != null) {
postEquivalenceEvent(planner, subset);
}
} else if ((required && !subset.isRequired())
|| (!required && !subset.isDelivered())) {
needsConverter = true;
}
if (subset.getConvention() == Convention.NONE) {
needsConverter = false;
} else if (required) {
subset.setRequired();
} else {
subset.setDelivered();
}
if (needsConverter) {
addConverters(subset, required, !planner.topDownOpt);
}
return subset;
}
private void postEquivalenceEvent(VolcanoPlanner planner, RelNode rel) {
RelOptListener listener = planner.getListener();
if (listener == null) {
return;
}
RelOptListener.RelEquivalenceEvent event =
new RelOptListener.RelEquivalenceEvent(
planner,
rel,
"equivalence class " + id,
false);
listener.relEquivalenceFound(event);
}
/**
* Adds an expression rel to this set, without creating a
* {@link com.hazelcast.org.apache.calcite.plan.volcano.RelSubset}. (Called only from
* {@link com.hazelcast.org.apache.calcite.plan.volcano.RelSubset#add}.
*
* @param rel Relational expression
*/
void addInternal(RelNode rel) {
if (!rels.contains(rel)) {
rels.add(rel);
for (RelTrait trait : rel.getTraitSet()) {
assert trait == trait.getTraitDef().canonize(trait);
}
VolcanoPlanner planner =
(VolcanoPlanner) rel.getCluster().getPlanner();
if (planner.getListener() != null) {
postEquivalenceEvent(planner, rel);
}
}
if (this.rel == null) {
this.rel = rel;
} else {
// Row types must be the same, except for field names.
RelOptUtil.verifyTypeEquivalence(
this.rel,
rel,
this);
}
}
/**
* Merges otherSet into this RelSet.
*
* One generally calls this method after discovering that two relational
* expressions are equivalent, and hence the RelSets they
* belong to are equivalent also.
*
*
After this method completes, otherSet is obsolete, its
* {@link #equivalentSet} member points to this RelSet, and this RelSet is
* still alive.
*
* @param planner Planner
* @param otherSet RelSet which is equivalent to this one
*/
void mergeWith(
VolcanoPlanner planner,
RelSet otherSet) {
assert this != otherSet;
assert this.equivalentSet == null;
assert otherSet.equivalentSet == null;
LOGGER.trace("Merge set#{} into set#{}", otherSet.id, id);
otherSet.equivalentSet = this;
RelOptCluster cluster = castNonNull(rel).getCluster();
// remove from table
boolean existed = planner.allSets.remove(otherSet);
assert existed : "merging with a dead otherSet";
Set changedRels = new HashSet<>();
// merge subsets
for (RelSubset otherSubset : otherSet.subsets) {
RelSubset subset = null;
RelTraitSet otherTraits = otherSubset.getTraitSet();
// If it is logical or delivered physical traitSet
if (otherSubset.isDelivered() || !otherSubset.isRequired()) {
subset = getOrCreateSubset(cluster, otherTraits, false);
}
// It may be required only, or both delivered and required,
// in which case, register again.
if (otherSubset.isRequired()) {
subset = getOrCreateSubset(cluster, otherTraits, true);
}
assert subset != null;
if (subset.passThroughCache == null) {
subset.passThroughCache = otherSubset.passThroughCache;
} else if (otherSubset.passThroughCache != null) {
subset.passThroughCache.addAll(otherSubset.passThroughCache);
}
// collect RelSubset instances, whose best should be changed
if (otherSubset.bestCost.isLt(subset.bestCost) && otherSubset.best != null) {
changedRels.add(otherSubset.best);
}
}
Set parentRels = new HashSet<>(parents);
for (RelNode otherRel : otherSet.rels) {
if (!(otherRel instanceof Spool)
&& !otherRel.isEnforcer()
&& parentRels.contains(otherRel)) {
// If otherRel is a enforcing operator e.g.
// Sort, Exchange, do not prune it. Just in
// case it is not marked as an enforcer.
if (otherRel.getInputs().size() != 1
|| otherRel.getInput(0).getTraitSet()
.satisfies(otherRel.getTraitSet())) {
planner.prune(otherRel);
}
}
planner.reregister(this, otherRel);
}
// Has another set merged with this?
assert equivalentSet == null;
// propagate the new best information from changed relNodes.
for (RelNode rel : changedRels) {
planner.propagateCostImprovements(rel);
}
// Update all rels which have a child in the other set, to reflect the
// fact that the child has been renamed.
//
// Copy array to prevent ConcurrentModificationException.
final List previousParents =
ImmutableList.copyOf(otherSet.getParentRels());
for (RelNode parentRel : previousParents) {
planner.rename(parentRel);
}
// Renaming may have caused this set to merge with another. If so,
// this set is now obsolete. There's no need to update the children
// of this set - indeed, it could be dangerous.
if (equivalentSet != null) {
return;
}
// Make sure the cost changes as a result of merging are propagated.
for (RelNode parentRel : getParentRels()) {
planner.propagateCostImprovements(parentRel);
}
assert equivalentSet == null;
// Each of the relations in the old set now has new parents, so
// potentially new rules can fire. Check for rule matches, just as if
// it were newly registered. (This may cause rules which have fired
// once to fire again.)
for (RelNode rel : rels) {
assert planner.getSet(rel) == this;
planner.fireRules(rel);
}
// Fire rule match on subsets as well
for (RelSubset subset : subsets) {
planner.fireRules(subset);
}
}
//~ Inner Classes ----------------------------------------------------------
/**
* An enum representing exploring state of current RelSet.
*/
enum ExploringState {
/**
* The RelSet is exploring.
* It means all possible rule matches are scheduled, but not fully applied.
* This RelSet will refuse to explore again, but cannot provide a valid LB.
*/
EXPLORING,
/**
* The RelSet is fully explored and is able to provide a valid LB.
*/
EXPLORED
}
}