org.apache.spark.sql.catalyst.expressions.aggregate
TypedImperativeAggregate
Related Doc:
package aggregate
org.apache.spark.sql.catalyst.expressions.aggregate
Returns a Seq of the children of this node.
Returns a Seq of the children of this node. Children should not change. Immutability required for containsChild optimization
Creates an empty aggregation buffer object.
Creates an empty aggregation buffer object. This is called before processing each key group (group by key).
an aggregation buffer object
Returns the DataType of the result of evaluating this expression.
Returns the DataType of the result of evaluating this expression. It is
invalid to query the dataType of an unresolved expression (i.e., when resolved == false).
De-serializes the serialized format Array[Byte], and produces aggregation buffer object T
Generates the final aggregation result value for current key group with the aggregation buffer object.
Generates the final aggregation result value for current key group with the aggregation buffer object.
aggregation buffer object.
The aggregation result of current key group
The offset of this function's start buffer value in the underlying shared input aggregation buffer.
The offset of this function's start buffer value in the underlying shared input aggregation
buffer. An input aggregation buffer is used when we merge two aggregation buffers together in
the update() function and is immutable (we merge an input aggregation buffer and a mutable
aggregation buffer and then store the new buffer values to the mutable aggregation buffer).
An input aggregation buffer may contain extra fields, such as grouping keys, at its start, so mutableAggBufferOffset and inputAggBufferOffset are often different.
For example, say we have a grouping expression, key, and two aggregate functions,
avg(x) and avg(y). In the shared input aggregation buffer, the position of the first
buffer value of avg(x) will be 1 and the position of the first buffer value of avg(y)
will be 3 (position 0 is used for the value of key):
avg(x) inputAggBufferOffset = 1 | v +--------+--------+--------+--------+--------+ | key | sum1 | count1 | sum2 | count2 | +--------+--------+--------+--------+--------+ ^ | avg(y) inputAggBufferOffset = 3
Merges an input aggregation object into aggregation buffer object and returns a new buffer object.
Merges an input aggregation object into aggregation buffer object and returns a new buffer object. For performance, the function may do in-place merge and return it instead of constructing new buffer object.
This is typically called when doing PartialMerge or Final mode aggregation.
the aggregation buffer object used to store the aggregation result.
an input aggregation object. Input aggregation object can be produced by de-serializing the partial aggregate's output from Mapper side.
The offset of this function's first buffer value in the underlying shared mutable aggregation buffer.
The offset of this function's first buffer value in the underlying shared mutable aggregation buffer.
For example, we have two aggregate functions avg(x) and avg(y), which share the same
aggregation buffer. In this shared buffer, the position of the first buffer value of avg(x)
will be 0 and the position of the first buffer value of avg(y) will be 2:
avg(x) mutableAggBufferOffset = 0 | v +--------+--------+--------+--------+ | sum1 | count1 | sum2 | count2 | +--------+--------+--------+--------+ ^ | avg(y) mutableAggBufferOffset = 2
Serializes the aggregation buffer object T to Array[Byte]
Updates the aggregation buffer object with an input row and returns a new buffer object.
Updates the aggregation buffer object with an input row and returns a new buffer object. For performance, the function may do in-place update and return it instead of constructing new buffer object.
This is typically called when doing Partial or Complete mode aggregation.
The aggregation buffer object.
an input row
Returns a copy of this ImperativeAggregate with an updated mutableAggBufferOffset.
Returns a copy of this ImperativeAggregate with an updated mutableAggBufferOffset. This new copy's attributes may have different ids than the original.
Returns a copy of this ImperativeAggregate with an updated mutableAggBufferOffset.
Returns a copy of this ImperativeAggregate with an updated mutableAggBufferOffset. This new copy's attributes may have different ids than the original.
Attributes of fields in aggBufferSchema.
Attributes of fields in aggBufferSchema.
The schema of the aggregation buffer.
The schema of the aggregation buffer.
Returns the tree node at the specified number, used primarily for interactive debugging.
Returns the tree node at the specified number, used primarily for interactive debugging. Numbers for each node can be found in the numberedTreeString.
Note that this cannot return BaseType because logical plan's plan node might return physical plan for innerChildren, e.g. in-memory relation logical plan node has a reference to the physical plan node it is referencing.
Returns a string representing the arguments to this node, minus any children
Returns a string representing the arguments to this node, minus any children
Returns a 'scala code' representation of this TreeNode and its children.
Returns a 'scala code' representation of this TreeNode and its children. Intended for use
when debugging where the prettier toString function is obfuscating the actual structure. In the
case of 'pure' TreeNodes that only contain primitives and other TreeNodes, the result can be
pasted in the REPL to build an equivalent Tree.
Returns an expression where a best effort attempt has been made to transform this in a way
that preserves the result but removes cosmetic variations (case sensitivity, ordering for
commutative operations, etc.) See Canonicalize for more details.
Returns an expression where a best effort attempt has been made to transform this in a way
that preserves the result but removes cosmetic variations (case sensitivity, ordering for
commutative operations, etc.) See Canonicalize for more details.
deterministic expressions where this.canonicalized == other.canonicalized will always
evaluate to the same result.
Checks the input data types, returns TypeCheckResult.success if it's valid,
or returns a TypeCheckResult with an error message if invalid.
Checks the input data types, returns TypeCheckResult.success if it's valid,
or returns a TypeCheckResult with an error message if invalid.
Note: it's not valid to call this method until childrenResolved == true.
Returns true if all the children of this expression have been resolved to a specific schema and false if any still contains any unresolved placeholders.
Returns true if all the children of this expression have been resolved to a specific schema and false if any still contains any unresolved placeholders.
Returns a Seq containing the result of applying a partial function to all elements in this tree on which the function is defined.
Returns a Seq containing the result of applying a partial function to all elements in this tree on which the function is defined.
Finds and returns the first TreeNode of the tree for which the given partial function is defined (pre-order), and applies the partial function to it.
Returns a Seq containing the leaves in this tree.
Returns a Seq containing the leaves in this tree.
Result of the aggregate function when the input is empty.
Result of the aggregate function when the input is empty. This is currently only used for the proper rewriting of distinct aggregate functions.
Returns true when the current expression always return the same result for fixed inputs from children.
Returns true when the current expression always return the same result for fixed inputs from children.
Note that this means that an expression should be considered as non-deterministic if: - it relies on some mutable internal state, or - it relies on some implicit input that is not part of the children expression list. - it has non-deterministic child or children.
An example would be SparkPartitionID that relies on the partition id returned by TaskContext.
By default leaf expressions are deterministic as Nil.forall(_.deterministic) returns true.
Returns Java source code that can be compiled to evaluate this expression.
Returns Java source code that can be compiled to evaluate this expression. The default behavior is to call the eval method of the expression. Concrete expression implementations should override this to do actual code generation.
an ExprCode with unique terms.
an ExprCode containing the Java source code to generate the given expression
Returns the result of evaluating this expression on a given input Row
Returns the result of evaluating this expression on a given input Row
Faster version of equality which short-circuits when two treeNodes are the same instance.
Faster version of equality which short-circuits when two treeNodes are the same instance.
We don't just override Object.equals, as doing so prevents the scala compiler from
generating case class equals methods
Find the first TreeNode that satisfies the condition specified by f.
Returns a Seq by applying a function to all nodes in this tree and using the elements of the resulting collections.
Returns a Seq by applying a function to all nodes in this tree and using the elements of the resulting collections.
An aggregate function is not foldable.
An aggregate function is not foldable.
Runs the given function on this node and then recursively on children.
Runs the given function recursively on children then on this node.
Returns an ExprCode, that contains the Java source code to generate the result of evaluating the expression on an input row.
Returns an ExprCode, that contains the Java source code to generate the result of evaluating the expression on an input row.
a CodegenContext
ExprCode
Appends the string represent of this node and its children to the given StringBuilder.
Appends the string represent of this node and its children to the given StringBuilder.
The i-th element in lastChildren indicates whether the ancestor of the current node at
depth i + 1 is the last child of its own parent node. The depth of the root node is 0, and
lastChildren for the root node should be empty.
Note that this traversal (numbering) order must be the same as getNodeNumbered.
Initializes the mutable aggregation buffer located in mutableAggBuffer.
Initializes the mutable aggregation buffer located in mutableAggBuffer.
Use fieldNumber + mutableAggBufferOffset to access fields of mutableAggBuffer.
All the nodes that should be shown as a inner nested tree of this node.
All the nodes that should be shown as a inner nested tree of this node. For example, this can be used to show sub-queries.
Attributes of fields in input aggregation buffers (immutable aggregation buffers that are merged with mutable aggregation buffers in the merge() function or merge expressions).
Attributes of fields in input aggregation buffers (immutable aggregation buffers that are merged with mutable aggregation buffers in the merge() function or merge expressions). These attributes are created automatically by cloning the aggBufferAttributes.
Creates a copy of this type of tree node after a transformation.
Creates a copy of this type of tree node after a transformation. Must be overridden by child classes that have constructor arguments that are not present in the productIterator.
the new product arguments.
Returns a Seq containing the result of applying the given function to each node in this tree in a preorder traversal.
Returns a Seq containing the result of applying the given function to each node in this tree in a preorder traversal.
the function to be applied.
Returns a copy of this node where f has been applied to all the nodes children.
Returns a copy of this node where f has been applied to all the nodes children.
Efficient alternative to productIterator.map(f).toArray.
Efficient alternative to productIterator.map(f).toArray.
Combines new intermediate results from the inputAggBuffer with the existing intermediate
results in the mutableAggBuffer.
Combines new intermediate results from the inputAggBuffer with the existing intermediate
results in the mutableAggBuffer.
Use fieldNumber + mutableAggBufferOffset to access fields of mutableAggBuffer.
Use fieldNumber + inputAggBufferOffset to access fields of inputAggBuffer.
Returns the name of this type of TreeNode.
Returns the name of this type of TreeNode. Defaults to the class name. Note that we remove the "Exec" suffix for physical operators here.
Returns a string representation of the nodes in this tree, where each operator is numbered.
Returns a string representation of the nodes in this tree, where each operator is numbered. The numbers can be used with TreeNode.apply to easily access specific subtrees.
The numbers are based on depth-first traversal of the tree (with innerChildren traversed first before children).
Args to the constructor that should be copied, but not transformed.
Args to the constructor that should be copied, but not transformed. These are appended to the transformed args automatically by makeCopy
Returns the tree node at the specified number, used primarily for interactive debugging.
Returns the tree node at the specified number, used primarily for interactive debugging. Numbers for each node can be found in the numberedTreeString.
This is a variant of apply that returns the node as BaseType (if the type matches).
Returns a user-facing string representation of this expression's name.
Returns a user-facing string representation of this expression's name. This should usually match the name of the function in SQL.
Returns true if this expression and all its children have been resolved to a specific schema
and input data types checking passed, and false if it still contains any unresolved
placeholders or has data types mismatch.
Returns true if this expression and all its children have been resolved to a specific schema
and input data types checking passed, and false if it still contains any unresolved
placeholders or has data types mismatch.
Implementations of expressions should override this if the resolution of this type of
expression involves more than just the resolution of its children and type checking.
Returns true when two expressions will always compute the same result, even if they differ cosmetically (i.e.
Returns true when two expressions will always compute the same result, even if they differ cosmetically (i.e. capitalization of names in attributes may be different).
See Canonicalize for more details.
Returns a hashCode for the calculation performed by this expression.
Returns a hashCode for the calculation performed by this expression. Unlike the standard
hashCode, an attempt has been made to eliminate cosmetic differences.
See Canonicalize for more details.
In-place replaces the aggregation buffer object stored at buffer's index
mutableAggBufferOffset, with SparkSQL internally supported underlying storage format
(BinaryType).
In-place replaces the aggregation buffer object stored at buffer's index
mutableAggBufferOffset, with SparkSQL internally supported underlying storage format
(BinaryType).
This is only called when doing Partial or PartialMerge mode aggregation, before the framework shuffle out aggregate buffers.
ONE line description of this node.
ONE line description of this node.
Returns SQL representation of this expression.
Returns SQL representation of this expression. For expressions extending NonSQLExpression, this method may return an arbitrary user facing string.
The arguments that should be included in the arg string.
The arguments that should be included in the arg string. Defaults to the productIterator.
String representation used in explain plans.
String representation used in explain plans.
Wraps this AggregateFunction in an AggregateExpression and set isDistinct field of the AggregateExpression to the given value because AggregateExpression is the container of an AggregateFunction, aggregation mode, and the flag indicating if this aggregation is distinct aggregation or not.
Wraps this AggregateFunction in an AggregateExpression and set isDistinct field of the AggregateExpression to the given value because AggregateExpression is the container of an AggregateFunction, aggregation mode, and the flag indicating if this aggregation is distinct aggregation or not. An AggregateFunction should not be used without being wrapped in an AggregateExpression.
Wraps this AggregateFunction in an AggregateExpression because AggregateExpression is the container of an AggregateFunction, aggregation mode, and the flag indicating if this aggregation is distinct aggregation or not.
Wraps this AggregateFunction in an AggregateExpression because AggregateExpression is the container of an AggregateFunction, aggregation mode, and the flag indicating if this aggregation is distinct aggregation or not. An AggregateFunction should not be used without being wrapped in an AggregateExpression.
Returns a copy of this node where rule has been recursively applied to the tree.
Returns a copy of this node where rule has been recursively applied to the tree.
When rule does not apply to a given node it is left unchanged.
Users should not expect a specific directionality. If a specific directionality is needed,
transformDown or transformUp should be used.
the function use to transform this nodes children
Returns a copy of this node where rule has been recursively applied to it and all of its
children (pre-order).
Returns a copy of this node where rule has been recursively applied to it and all of its
children (pre-order). When rule does not apply to a given node it is left unchanged.
the function used to transform this nodes children
Returns a copy of this node where rule has been recursively applied first to all of its
children and then itself (post-order).
Returns a copy of this node where rule has been recursively applied first to all of its
children and then itself (post-order). When rule does not apply to a given node, it is left
unchanged.
the function use to transform this nodes children
Returns a string representation of the nodes in this tree
Returns a string representation of the nodes in this tree
Updates its aggregation buffer, located in mutableAggBuffer, based on the given inputRow.
Updates its aggregation buffer, located in mutableAggBuffer, based on the given inputRow.
Use fieldNumber + mutableAggBufferOffset to access fields of mutableAggBuffer.
ONE line description of this node with more information
ONE line description of this node with more information
ONE line description of this node with some suffix information
ONE line description of this node with some suffix information
Returns a copy of this node with the children replaced.
Returns a copy of this node with the children replaced. TODO: Validate somewhere (in debug mode?) that children are ordered correctly.
Aggregation function which allows **arbitrary** user-defined java object to be used as internal aggregation buffer.
General work flow:
Stage 1: initialize aggregate buffer object.
initialize(buffer: MutableRow)to set up the empty aggregate buffer. 2. Ininitialize, we callcreateAggregationBuffer(): Tto get the initial buffer object, and set it to the global buffer row.Stage 2: process input rows.
If the aggregate mode is
PartialorComplete:update(buffer: MutableRow, input: InternalRow)to process the input row. 2. Inupdate, we get the buffer object from the global buffer row and callupdate(buffer: T, input: InternalRow): Unit.If the aggregate mode is
PartialMergeorFinal:merge(buffer: MutableRow, inputBuffer: InternalRow)to process the input row, which are serialized buffer objects shuffled from other nodes. 2. Inmerge, we get the buffer object from the global buffer row, and get the binary data from input row and deserialize it to buffer object, then we callmerge(buffer: T, input: T): Unitto merge these 2 buffer objects.Stage 3: output results.
If the aggregate mode is
PartialorPartialMerge:serializeAggregateBufferInPlaceto replace the buffer object in the global buffer row with binary data. 2. InserializeAggregateBufferInPlace, we get the buffer object from the global buffer row and callserialize(buffer: T): Array[Byte]to serialize the buffer object to binary. 3. The framework outputs buffer attributes and shuffle them to other nodes.If the aggregate mode is
FinalorComplete:eval(buffer: InternalRow)to calculate the final result. 2. Ineval, we get the buffer object from the global buffer row and calleval(buffer: T): Anyto get the final result. 3. The framework outputs these final results.Window function work flow: The framework calls
update(buffer: MutableRow, input: InternalRow)several times and then calleval(buffer: InternalRow), so there is no need for window operator to callserializeAggregateBufferInPlace.NOTE: SQL with TypedImperativeAggregate functions is planned in sort based aggregation, instead of hash based aggregation, as TypedImperativeAggregate use BinaryType as aggregation buffer's storage format, which is not supported by hash based aggregation. Hash based aggregation only support aggregation buffer of mutable types (like LongType, IntType that have fixed length and can be mutated in place in UnsafeRow). NOTE: The newly added ObjectHashAggregateExec supports TypedImperativeAggregate functions in hash based aggregation under some constraints.