HashedRelation¶

HashedRelation is an extension of the KnownSizeEstimation abstraction for relations with values hashed by some key.

sealed trait

HashedRelation is a Scala sealed trait which means that all possible implementations (HashedRelations) are all in the same compilation unit (file).

Contract¶

asReadOnlyCopy¶

asReadOnlyCopy(): HashedRelation

A read-only copy of this HashedRelation to be safely used in a separate thread

Used when:

BroadcastHashJoinExec physical operator is executed

close¶

close(): Unit

Used when:

ShuffledHashJoinExec physical operator is requested to buildHashedRelation

get¶

get(
  key: InternalRow): Iterator[InternalRow]
get(
  key: Long): Iterator[InternalRow]

Gets internal rows for the given key or null

Used when:

HashJoin is requested to innerJoin, [outerJoin]/HashJoin.md#outerJoin), semiJoin, existenceJoin and antiJoin
LongHashedRelation is requested to get a value for a key

getValue¶

getValue(
  key: InternalRow): InternalRow
getValue(
  key: Long): InternalRow

Gives the value internal row for the given key

Used when:

LongHashedRelation is requested to get a value for a key

keyIsUnique¶

keyIsUnique: Boolean

Used when:

BroadcastHashJoinExec physical operator is requested to multipleOutputForOneInput, codegenInner, codegenOuter, codegenSemi, codegenAnti, codegenExistence

keys¶

keys(): Iterator[InternalRow]

Used when:

SubqueryBroadcastExec physical operator is requested for relationFuture

Implementations¶

Creating HashedRelation¶

apply(
  input: Iterator[InternalRow],
  key: Seq[Expression],
  sizeEstimate: Int = 64,
  taskMemoryManager: TaskMemoryManager = null,
  isNullAware: Boolean = false,
  allowsNullKey: Boolean = false): HashedRelation

With no elements in the input iterator and with the allowsNullKey flag disabled, apply returns the EmptyHashedRelation.

For exactly one key expression of type LongType and with the allowsNullKey flag disabled, apply returns a LongHashedRelation.

For all other cases, apply returns an UnsafeHashedRelation.

The input key expressions can be the following:

Build join keys of ShuffledHashJoinExec physical operator
Canonicalized build-side join keys of HashedRelationBroadcastMode (of BroadcastHashJoinExec physical operator)

apply is used when:

ShuffledHashJoinExec physical operator is requested to build a HashedRelation from InternalRows
HashedRelationBroadcastMode is requested to transform InternalRows into a HashedRelation