org.modeshape.graph.query.plan
Class CanonicalPlanner

java.lang.Object
  extended by org.modeshape.graph.query.plan.CanonicalPlanner
All Implemented Interfaces:
Planner

public class CanonicalPlanner
extends Object
implements Planner

The planner that produces a canonical query plan given a query command.

A canonical plan always has the same structure:

       LIMIT       if row limit or offset are used
         |
      SORTING      if 'ORDER BY' is used
         |
     DUP_REMOVE    if 'SELECT DISTINCT' is used
         |
      PROJECT      with the list of columns being SELECTed
         |
       GROUP       if 'GROUP BY' is used
         |
      SELECT1
         |         One or more SELECT plan nodes that each have
      SELECT2      a single non-join constraint that are then all AND-ed
         |         together (see separateAndConstraints(Constraint, List))
      SELECTn
         |
    SOURCE or JOIN     A single SOURCE or JOIN node, depending upon the query
              /  \
             /    \
           SOJ    SOJ    A SOURCE or JOIN node for the left and right side of the JOIN
 

There leaves of the tree are always SOURCE nodes, so conceptually data always flows through this plan from the bottom SOURCE nodes, is adjusted/filtered as it trickles up through the plan, and is then ready to be used by the caller as it emerges from the top node of the plan.

This canonical plan, however, is later optimized and rearranged so that it performs faster.


Constructor Summary
CanonicalPlanner()
           
 
Method Summary
protected  void allColumnsFor(Schemata.Table table, SelectorName tableName, List<Column> columns, List<String> columnTypes)
           
protected  PlanNode attachCriteria(QueryContext context, PlanNode plan, Constraint constraint)
          Attach all criteria above the join nodes.
protected  PlanNode attachDuplicateRemoval(QueryContext context, PlanNode plan)
          Attach DUP_REMOVE node at top of tree.
protected  PlanNode attachLimits(QueryContext context, PlanNode plan, Limit limit)
          Attach a LIMIT node at the top of the plan tree.
protected  PlanNode attachProject(QueryContext context, PlanNode plan, List<? extends Column> columns, Map<SelectorName,Schemata.Table> selectors)
          Attach a PROJECT node at the top of the plan tree.
protected  PlanNode attachSorting(QueryContext context, PlanNode plan, List<? extends Ordering> orderings)
          Attach SORT node at top of tree.
protected  PlanNode createCanonicalPlan(QueryContext context, Query query)
          Create a canonical query plan for the given query.
protected  PlanNode createCanonicalPlan(QueryContext context, SetQuery query)
          Create a canonical query plan for the given set query.
 PlanNode createPlan(QueryContext context, QueryCommand query)
          Create a canonical query plan for the given command.
protected  PlanNode createPlanNode(QueryContext context, Source source, Map<SelectorName,Schemata.Table> usedSelectors)
          Create a JOIN or SOURCE node that contain the source information.
protected  void separateAndConstraints(Constraint constraint, List<Constraint> andableConstraints)
          Walk the supplied constraint to extract a list of the constraints that can be AND-ed together.
protected  void validate(QueryContext context, QueryCommand query, Map<SelectorName,Schemata.Table> usedSelectors)
          Validate the supplied query.
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Constructor Detail

CanonicalPlanner

public CanonicalPlanner()
Method Detail

createPlan

public PlanNode createPlan(QueryContext context,
                           QueryCommand query)
Create a canonical query plan for the given command.

Specified by:
createPlan in interface Planner
Parameters:
context - the context in which the query is being planned
query - the query command to be planned
Returns:
the root node of the plan tree representing the canonical plan
See Also:
Planner.createPlan(org.modeshape.graph.query.QueryContext, org.modeshape.graph.query.model.QueryCommand)

createCanonicalPlan

protected PlanNode createCanonicalPlan(QueryContext context,
                                       Query query)
Create a canonical query plan for the given query.

Parameters:
context - the context in which the query is being planned
query - the query to be planned
Returns:
the root node of the plan tree representing the canonical plan

validate

protected void validate(QueryContext context,
                        QueryCommand query,
                        Map<SelectorName,Schemata.Table> usedSelectors)
Validate the supplied query.

Parameters:
context - the context in which the query is being planned
query - the set query to be planned
usedSelectors - the map of SelectorNames (aliases or names) used in the query.

createCanonicalPlan

protected PlanNode createCanonicalPlan(QueryContext context,
                                       SetQuery query)
Create a canonical query plan for the given set query.

Parameters:
context - the context in which the query is being planned
query - the set query to be planned
Returns:
the root node of the plan tree representing the canonical plan

createPlanNode

protected PlanNode createPlanNode(QueryContext context,
                                  Source source,
                                  Map<SelectorName,Schemata.Table> usedSelectors)
Create a JOIN or SOURCE node that contain the source information.

Parameters:
context - the execution context
source - the source to be processed; may not be null
usedSelectors - the map of SelectorNames (aliases or names) used in the query.
Returns:
the new plan; never null

attachCriteria

protected PlanNode attachCriteria(QueryContext context,
                                  PlanNode plan,
                                  Constraint constraint)
Attach all criteria above the join nodes. The optimizer will push these criteria down to the appropriate source.

Parameters:
context - the context in which the query is being planned
plan - the existing plan, which joins all source groups
constraint - the criteria or constraint from the query
Returns:
the updated plan, or the existing plan if there were no constraints; never null

separateAndConstraints

protected void separateAndConstraints(Constraint constraint,
                                      List<Constraint> andableConstraints)
Walk the supplied constraint to extract a list of the constraints that can be AND-ed together. For example, given the constraint tree ((C1 AND C2) AND (C3 OR C4)), this method would result in a list of three separate criteria: [C1,C2,(C3 OR C4)]. The resulting andConstraints list will contain Constraint objects that all must be true.

Parameters:
constraint - the input constraint
andableConstraints - the collection into which all non-AND constraints should be placed

attachSorting

protected PlanNode attachSorting(QueryContext context,
                                 PlanNode plan,
                                 List<? extends Ordering> orderings)
Attach SORT node at top of tree. The SORT may be pushed down to a source (or sources) if possible by the optimizer.

Parameters:
context - the context in which the query is being planned
plan - the existing plan
orderings - list of orderings from the query
Returns:
the updated plan, or the existing plan if there were no orderings; never null

attachLimits

protected PlanNode attachLimits(QueryContext context,
                                PlanNode plan,
                                Limit limit)
Attach a LIMIT node at the top of the plan tree.

Parameters:
context - the context in which the query is being planned
plan - the existing plan
limit - the limit definition; may be null
Returns:
the updated plan, or the existing plan if there were no limits

attachProject

protected PlanNode attachProject(QueryContext context,
                                 PlanNode plan,
                                 List<? extends Column> columns,
                                 Map<SelectorName,Schemata.Table> selectors)
Attach a PROJECT node at the top of the plan tree.

Parameters:
context - the context in which the query is being planned
plan - the existing plan
columns - the columns being projected; may be null
selectors - the selectors keyed by their alias or name
Returns:
the updated plan

allColumnsFor

protected void allColumnsFor(Schemata.Table table,
                             SelectorName tableName,
                             List<Column> columns,
                             List<String> columnTypes)

attachDuplicateRemoval

protected PlanNode attachDuplicateRemoval(QueryContext context,
                                          PlanNode plan)
Attach DUP_REMOVE node at top of tree. The DUP_REMOVE may be pushed down to a source (or sources) if possible by the optimizer.

Parameters:
context - the context in which the query is being planned
plan - the existing plan
Returns:
the updated plan


Copyright © 2008-2010 JBoss, a division of Red Hat. All Rights Reserved.