org.apache.drill.exec.planner.index

Interface IndexDescriptor

All Superinterfaces:

IndexDefinition

All Known Implementing Classes:

AbstractIndexDescriptor, DrillIndexDescriptor

public interface IndexDescriptor
extends IndexDefinition

IndexDefinition + functions to access materialized index(index table/scan, etc)

Nested Class Summary

Nested classes/interfaces inherited from interface org.apache.drill.exec.planner.index.IndexDefinition

IndexDefinition.IndexType

Method Summary

Modifier and Type	Method and Description
org.apache.calcite.plan.RelOptCost	getCost(IndexProperties indexProps, org.apache.calcite.plan.RelOptPlanner planner, int numProjectedFields, GroupScan primaryGroupScan) Get the total cost of index access (I/O, CPU) in the context of the current query
FunctionalIndexInfo	getFunctionalInfo() Get the functional index information associated with this index (Functional indexes are indexes involving expressions e.g CAST(a as INT).
IndexGroupScan	getIndexGroupScan() Get an instance of the group scan associated with this index descriptor
PluginCost	getPluginCostModel() Get the costing factors associated with the storage/format plugin
double	getRows(org.apache.calcite.rel.RelNode input, org.apache.calcite.rex.RexNode indexCondition) Get the estimated row count for a single index condition
boolean	isAsyncIndex() Whether this index is maintained synchronously (i.e primary table updates are propagated to the index synchronously) or asynchronously with some delay.
boolean	supportsFullTextSearch() Whether or not the index supports full-text search (to allow pushing down such filters)
boolean	supportsRowCountStats() Whether or not the index supports getting row count statistics

Methods inherited from interface org.apache.drill.exec.planner.index.IndexDefinition

allColumnsIndexed, getCollation, getCollationMap, getIndexColumnOrdinal, getIndexColumns, getIndexName, getIndexType, getNonIndexColumns, getNullsOrderingDirection, getRowKeyColumns, getTableName, isCoveringIndex, someColumnsIndexed

Method Detail

getRows

double getRows(org.apache.calcite.rel.RelNode input,
               org.apache.calcite.rex.RexNode indexCondition)

Get the estimated row count for a single index condition

Parameters:

input - The rel node corresponding to the primary table

indexCondition - The index condition (e.g index_col1 < 10 AND index_col2 = 'abc')

Returns:

The estimated row count

supportsRowCountStats

boolean supportsRowCountStats()

Whether or not the index supports getting row count statistics

Returns:

True if index supports getting row count, False otherwise

getIndexGroupScan

IndexGroupScan getIndexGroupScan()

Get an instance of the group scan associated with this index descriptor

Returns:

An instance of group scan for this index

supportsFullTextSearch

boolean supportsFullTextSearch()

Whether or not the index supports full-text search (to allow pushing down such filters)

Returns:

True if index supports full-text search, False otherwise

getFunctionalInfo

FunctionalIndexInfo getFunctionalInfo()

Get the functional index information associated with this index (Functional indexes are indexes involving expressions e.g CAST(a as INT).

getCost

org.apache.calcite.plan.RelOptCost getCost(IndexProperties indexProps,
                                           org.apache.calcite.plan.RelOptPlanner planner,
                                           int numProjectedFields,
                                           GroupScan primaryGroupScan)

Get the total cost of index access (I/O, CPU) in the context of the current query

Parameters:

indexProps - properties (metrics) of a single index in the context of current query

planner - Planner instance

numProjectedFields - Number of projected fields

primaryGroupScan - Primary table's GroupScan instance

Returns:

a RelOptCost instance representing the total cost

getPluginCostModel

PluginCost getPluginCostModel()

Get the costing factors associated with the storage/format plugin

isAsyncIndex

boolean isAsyncIndex()

Whether this index is maintained synchronously (i.e primary table updates are propagated to the index synchronously) or asynchronously with some delay. The latter is more common for distributed NoSQL databases.

Returns:

True if the index is maintained asynchronously, False otherwise