org.apache.drill.exec.physical.base

Interface GroupScan

All Superinterfaces:

FragmentLeaf, GraphValue<PhysicalOperator>, HasAffinity, Iterable<PhysicalOperator>, Leaf, PhysicalOperator, Scan

All Known Subinterfaces:

FileGroupScan

All Known Implementing Classes:

AbstractFileGroupScan, AbstractGroupScan, DirectGroupScan, EasyGroupScan, InfoSchemaGroupScan, MockGroupScanPOP, ParquetGroupScan, SystemTableScan

public interface GroupScan
extends Scan, HasAffinity

A GroupScan operator represents all data which will be scanned by a given physical plan. It is the superset of all SubScans for the plan.

Field Summary

Fields

Modifier and Type	Field and Description
static List<SchemaPath>	ALL_COLUMNS columns list in GroupScan : 1) empty_column is for skipAll query.
static long	NO_COLUMN_STATS

Fields inherited from interface org.apache.drill.common.graph.GraphValue

logger

Method Summary

Methods

Modifier and Type	Method and Description
void	applyAssignments(List<CoordinationProtos.DrillbitEndpoint> endpoints)
GroupScan	applyLimit(long maxRecords) Apply rowcount based prune for "LIMIT n" query.
boolean	canPushdownProjects(List<SchemaPath> columns) GroupScan should check the list of columns, and see if it could support all the columns in the list.
GroupScan	clone(List<SchemaPath> columns) Returns a clone of GroupScan instance, except that the new GroupScan will use the provided list of columns .
boolean	enforceWidth() Deprecated. Use getMinParallelizationWidth() to determine whether this GroupScan spans more than one fragment.
long	getColumnValueCount(SchemaPath column) Return the number of non-null value in the specified column.
String	getDigest() Returns a signature of the GroupScan which should usually be composed of all its attributes which could describe it uniquely.
Collection<String>	getFiles() Returns a collection of file names associated with this GroupScan.
int	getMaxParallelizationWidth()
int	getMinParallelizationWidth() At minimum, the GroupScan requires these many fragments to run.
List<SchemaPath>	getPartitionColumns() Returns a list of columns that can be used for partition pruning
ScanStats	getScanStats(PlannerSettings settings)
SubScan	getSpecificScan(int minorFragmentId)
boolean	hasFiles() Return true if this GroupScan can return its selection as a list of file names (retrieved by getFiles()).
boolean	supportsLimitPushdown() Whether or not this GroupScan supports limit pushdown
boolean	supportsPartitionFilterPushdown() Whether or not this GroupScan supports pushdown of partition filters (directories for filesystems)

Methods inherited from interface org.apache.drill.exec.physical.base.PhysicalOperator

accept, getCost, getInitialAllocation, getMaxAllocation, getNewWithChildren, getOperatorId, getOperatorType, getSVMode, getUserName, isExecutable, setCost, setOperatorId

Methods inherited from interface org.apache.drill.common.graph.GraphValue

accept

Methods inherited from interface java.lang.Iterable

iterator

Methods inherited from interface org.apache.drill.exec.physical.base.HasAffinity

getOperatorAffinity

Field Detail

ALL_COLUMNS

static final List<SchemaPath> ALL_COLUMNS

columns list in GroupScan : 1) empty_column is for skipAll query. 2) NULL is interpreted as ALL_COLUMNS. How to handle skipAll query is up to each storage plugin, with different policy in corresponding RecordReader.

NO_COLUMN_STATS

static final long NO_COLUMN_STATS

See Also:

Constant Field Values

Method Detail

applyAssignments

void applyAssignments(List<CoordinationProtos.DrillbitEndpoint> endpoints)
                      throws PhysicalOperatorSetupException

Throws:

PhysicalOperatorSetupException

getSpecificScan

SubScan getSpecificScan(int minorFragmentId)
                        throws ExecutionSetupException

Throws:

ExecutionSetupException

getMaxParallelizationWidth

int getMaxParallelizationWidth()

getMinParallelizationWidth

int getMinParallelizationWidth()

At minimum, the GroupScan requires these many fragments to run. Currently, this is used in SimpleParallelizer

Returns:

the minimum number of fragments that should run

enforceWidth

@Deprecated
boolean enforceWidth()

Deprecated. Use getMinParallelizationWidth() to determine whether this GroupScan spans more than one fragment.

Check if GroupScan enforces width to be maximum parallelization width. Currently, this is used in ExcessiveExchangeIdentifier

Returns:

if maximum width should be enforced

getDigest

String getDigest()

Returns a signature of the GroupScan which should usually be composed of all its attributes which could describe it uniquely.

getScanStats

ScanStats getScanStats(PlannerSettings settings)

clone

GroupScan clone(List<SchemaPath> columns)

Returns a clone of GroupScan instance, except that the new GroupScan will use the provided list of columns .

canPushdownProjects

boolean canPushdownProjects(List<SchemaPath> columns)

GroupScan should check the list of columns, and see if it could support all the columns in the list.

getColumnValueCount

long getColumnValueCount(SchemaPath column)

Return the number of non-null value in the specified column. Raise exception, if groupscan does not have exact column row count.

supportsPartitionFilterPushdown

boolean supportsPartitionFilterPushdown()

Whether or not this GroupScan supports pushdown of partition filters (directories for filesystems)

getPartitionColumns

List<SchemaPath> getPartitionColumns()

Returns a list of columns that can be used for partition pruning

supportsLimitPushdown

boolean supportsLimitPushdown()

Whether or not this GroupScan supports limit pushdown

applyLimit

GroupScan applyLimit(long maxRecords)

Apply rowcount based prune for "LIMIT n" query.

Parameters:

maxRecords - : the number of rows requested from group scan.

Returns:

a new instance of group scan if the prune is successful. null when either if row-based prune is not supported, or if prune is not successful.

hasFiles

boolean hasFiles()

Return true if this GroupScan can return its selection as a list of file names (retrieved by getFiles()).

getFiles

Collection<String> getFiles()

Returns a collection of file names associated with this GroupScan. This should be called after checking hasFiles(). If this GroupScan cannot provide file names, it returns null.