Package org.apache.drill.exec.planner.index

Class IndexPlanUtils

java.lang.Object
org.apache.drill.exec.planner.index.IndexPlanUtils
public class IndexPlanUtils extends Object

  • Constructor Details

    • IndexPlanUtils

      public IndexPlanUtils()

  • Method Details

    • conditionIndexed

      public static IndexPlanUtils.ConditionIndexed conditionIndexed(IndexableExprMarker exprMarker, IndexDescriptor indexDesc)
      Check if any of the fields of the index are present in a list of LogicalExpressions supplied as part of IndexableExprMarker
      Parameters:
      exprMarker - the marker that has analyzed original index condition on top of original scan
      indexDesc - the index definition plus functions to access materialized index
      Returns:
      ConditionIndexed.FULL, PARTIAL or NONE depending on whether all, some or no columns of the indexDesc are present in the list of LogicalExpressions supplied as part of exprMarker

    • checkScan

      public static boolean checkScan(DrillScanRel scanRel)
      check if we want to apply index rules on this scan, if group scan is not instance of DbGroupScan, or this DbGroupScan instance does not support secondary index, or this scan is already an index scan or Restricted Scan, do not apply index plan rules on it.
      Parameters:
      scanRel - the current scan rel
      Returns:
      true to indicate that we want to apply index rules on this scan, otherwise false

    • isCoveringIndex

      public static boolean isCoveringIndex(IndexCallContext indexContext, FunctionalIndexInfo functionInfo)
      For a particular table scan for table T1 and an index on that table, find out if it is a covering index
      Returns:
      true if it is a covering index, otherwise false

    • buildCollationLowerProject

      public static org.apache.calcite.rel.RelCollation buildCollationLowerProject(List<org.apache.calcite.rex.RexNode> projectRexs, org.apache.calcite.rel.RelNode input, FunctionalIndexInfo indexInfo)
      Build collation property for the 'lower' project, the one closer to the Scan
      Parameters:
      projectRexs - list of row expressions
      input - input as a relational expression
      indexInfo - collects functional index information
      Returns:
      the output RelCollation

    • buildCollationUpperProject

      public static org.apache.calcite.rel.RelCollation buildCollationUpperProject(List<org.apache.calcite.rex.RexNode> projectRexs, org.apache.calcite.rel.RelCollation inputCollation, FunctionalIndexInfo indexInfo, Map<Integer,List<org.apache.calcite.rex.RexNode>> collationFilterMap)
      Build collation property for the 'upper' project, the one above the filter
      Parameters:
      projectRexs - list of row expressions
      inputCollation - the input collation
      indexInfo - collects functional index information
      collationFilterMap - map for collation filter
      Returns:
      the output RelCollation

    • getIndexFromCollation

      public static int getIndexFromCollation(int refIndex, List<org.apache.calcite.rel.RelFieldCollation> inputFieldCollations)

    • getProjects

      public static List<org.apache.calcite.rex.RexNode> getProjects(DrillProjectRelBase proj)

    • generateLimit

      public static boolean generateLimit(OrderedRel sort)

    • getOffset

      public static org.apache.calcite.rex.RexNode getOffset(OrderedRel sort)

    • getFetch

      public static org.apache.calcite.rex.RexNode getFetch(OrderedRel sort)

    • updateSortExpression

      public static void updateSortExpression(IndexCallContext indexContext, List<org.apache.calcite.rel.RelFieldCollation> coll)
      generate logical expressions for sort rexNodes in SortRel, the result is store to IndexPlanCallContext
      Parameters:
      indexContext - the index call context
      coll - list of field collations

    • updateSortExpression

      public static void updateSortExpression(IndexPhysicalPlanCallContext indexContext, List<org.apache.calcite.rel.RelFieldCollation> coll)
      generate logical expressions for sort rexNodes in SortRel, the result is store to IndexPlanCallContext
      Parameters:
      indexContext - the index call context
      coll - list of field collations

    • buildCollationProject

      public static org.apache.calcite.rel.RelCollation buildCollationProject(List<org.apache.calcite.rex.RexNode> projectRexs, DrillProjectRelBase project, org.apache.calcite.rel.RelNode input, FunctionalIndexInfo indexInfo, IndexCallContext context)
      Build collation property for project, the one closer to the Scan
      Parameters:
      projectRexs - the expressions to project
      project - the project between projectRexs and input, it could be null if no such intermediate project(lower project)
      input - the input RelNode to the project, usually it is the scan operator.
      indexInfo - the index for which we are building index plan
      context - the context of this index planning process
      Returns:
      the output RelCollation

    • buildCollationCoveringIndexScan

      public static org.apache.calcite.rel.RelCollation buildCollationCoveringIndexScan(IndexDescriptor indexDesc, IndexCallContext context)
      Build the collation property for index scan
      Parameters:
      indexDesc - the index for which we are building index plan
      context - the context of this index planning process
      Returns:
      the output RelCollation for the scan on index

    • getProjectExprs

      public static Map<LogicalExpression,Integer> getProjectExprs(List<org.apache.calcite.rex.RexNode> projectRexs, DrillProjectRelBase project, org.apache.calcite.rel.RelNode input)

    • getExprsFromRowType

      public static Map<LogicalExpression,Integer> getExprsFromRowType(org.apache.calcite.rel.type.RelDataType indexScanRowType)

    • buildCollationForExpressions

      public static org.apache.calcite.rel.RelCollation buildCollationForExpressions(Map<LogicalExpression,Integer> projectExprs, IndexDescriptor indexDesc, IndexCallContext context)
      Given index, compute the collations for a list of projected expressions(from Scan's rowType or Project's ) in the context
      Parameters:
      projectExprs - the output expression list of a RelNode
      indexDesc - the index for which we are building index plan
      context - the context of this index planning process
      Returns:
      the collation provided by index that will be exposed by the expression list

    • pathOnlyInIndexedFunction

      public static boolean pathOnlyInIndexedFunction(SchemaPath path)

    • buildCollationNonCoveringIndexScan

      public static org.apache.calcite.rel.RelCollation buildCollationNonCoveringIndexScan(IndexDescriptor indexDesc, org.apache.calcite.rel.type.RelDataType indexScanRowType, org.apache.calcite.rel.type.RelDataType restrictedScanRowType, IndexCallContext context)

    • scanIsPartition

      public static boolean scanIsPartition(GroupScan scan)

    • getGroupScan

      public static GroupScan getGroupScan(DrillScanRelBase relNode)

    • getCollation

      public static org.apache.calcite.rel.RelCollation getCollation(org.apache.calcite.rel.core.Sort sort)

    • getDistributionField

      public static List<DrillDistributionTrait.DistributionField> getDistributionField(org.apache.calcite.rel.core.Sort rel)

    • buildCoveringIndexScan

      public static ScanPrel buildCoveringIndexScan(DrillScanRelBase origScan, IndexGroupScan indexGroupScan, IndexCallContext indexContext, IndexDescriptor indexDesc)

    • rewriteFunctionColumn

      public static List<SchemaPath> rewriteFunctionColumn(List<SchemaPath> paths, FunctionalIndexInfo functionInfo, List<SchemaPath> addedPaths)
      For IndexGroupScan, if a column is only appeared in the should-be-renamed function, this column is to-be-replaced column, we replace that column(schemaPath) from 'a.b' to '$1' in the list of SchemaPath.
      Parameters:
      paths - list of paths
      functionInfo - functional index information that may impact rewrite
      addedPaths - list of paths added
      Returns:
      list of new paths

    • rewriteFunctionalRex

      public static org.apache.calcite.rex.RexNode rewriteFunctionalRex(IndexCallContext indexContext, DrillParseContext parseContext, DrillProjectRelBase project, org.apache.calcite.rel.RelNode scan, org.apache.calcite.rex.RexNode toRewriteRex, org.apache.calcite.rel.type.RelDataType newRowType, FunctionalIndexInfo functionInfo)
      A RexNode forest with three RexNodes for expressions "cast(a.q as int) * 2, b+c, concat(a.q, " world")" on Scan RowType('a', 'b', 'c') will be like this: (0)Call:"*" Call:"concat" / \ / \ (1)Call:CAST 2 Call:"+" (5)Call:ITEM ' world' / \ / \ / \ (2)Call:ITEM TYPE:INT (3)$1 (4)$2 $0 'q' / \ $0 'q' So for above expressions, when visiting the RexNode trees using PathInExpr, we could mark indexed expressions in the trees, as shown in the diagram above are the node (1), then collect the schema paths in the indexed expression but found out of the indexed expression -- node (5), and other regular schema paths (3) (4)
      Parameters:
      indexContext - the index call context
      parseContext - the drill parse context
      project - the drill base class for logical and physical projects
      scan - a rel node scan
      toRewriteRex - the RexNode to be converted if it contains a functional index expression
      newRowType - data type for new row
      functionInfo - functional index information that may impact rewrite
      Returns:
      rewritten functional row expression

    • getLeadingPrefixMap

      public static org.apache.calcite.rex.RexNode getLeadingPrefixMap(Map<LogicalExpression,org.apache.calcite.rex.RexNode> leadingPrefixMap, List<LogicalExpression> indexCols, IndexConditionInfo.Builder builder, org.apache.calcite.rex.RexNode condition)

    • getLeadingFilters

      public static List<org.apache.calcite.rex.RexNode> getLeadingFilters(Map<LogicalExpression,org.apache.calcite.rex.RexNode> leadingPrefixMap, List<LogicalExpression> indexCols)

    • getLeadingColumnsFilter

      public static org.apache.calcite.rex.RexNode getLeadingColumnsFilter(List<org.apache.calcite.rex.RexNode> leadingFilters, org.apache.calcite.rex.RexBuilder rexBuilder)

    • getTotalRemainderFilter

      public static org.apache.calcite.rex.RexNode getTotalRemainderFilter(org.apache.calcite.rex.RexNode indexColsRemFilter, org.apache.calcite.rex.RexNode incColsRemFilter, org.apache.calcite.rex.RexBuilder rexBuilder)

    • getTotalFilter

      public static org.apache.calcite.rex.RexNode getTotalFilter(org.apache.calcite.rex.RexNode leadColsFilter, org.apache.calcite.rex.RexNode totRemColsFilter, org.apache.calcite.rex.RexBuilder rexBuilder)