compute

package
v18.0.0-...-f9949aa Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Dec 16, 2024 License: Apache-2.0, BSD-3-Clause Imports: 34 Imported by: 0

Documentation

Overview

Package compute is a native-go implementation of an Acero-like arrow compute engine. It requires go1.18+

While consumers of Arrow that are able to use CGO could utilize the C Data API (using the cdata package) and could link against the acero library directly, there are consumers who cannot use CGO. This is an attempt to provide for those users, and in general create a native-go arrow compute engine.

The overwhelming majority of things in this package require go1.18 as it utilizes generics. The files in this package and its sub-packages are all excluded from being built by go versions lower than 1.18 so that the larger Arrow module itself is still compatible with go1.17.

Everything in this package should be considered Experimental for now.

Example (CustomFunction)

This example demonstrates how to register a custom scalar function. 

package main

import (
	"context"
	"fmt"
	"log"

	"github.com/joechenrh/arrow-go/v18/arrow"
	"github.com/joechenrh/arrow-go/v18/arrow/array"
	"github.com/joechenrh/arrow-go/v18/arrow/compute"
	"github.com/joechenrh/arrow-go/v18/arrow/compute/exec"
	"github.com/joechenrh/arrow-go/v18/arrow/memory"
)

func main() {
	pool := memory.NewGoAllocator()

	ctx := context.Background()
	execCtx := compute.DefaultExecCtx()
	ctx = compute.SetExecCtx(ctx, execCtx)

	add42 := compute.NewScalarFunction("add_42", compute.Arity{
		NArgs: 1,
	}, compute.FunctionDoc{
		Summary:  "Returns the input values plus 42",
		ArgNames: []string{"input"},
	})

	if err := add42.AddNewKernel(
		[]exec.InputType{
			// We accept a single argument (array) of Int8 type.
			{
				Kind: exec.InputExact,
				Type: arrow.PrimitiveTypes.Int8,
			},
		},
		// We'll return a single Int8 array.
		exec.NewOutputType(arrow.PrimitiveTypes.Int8),
		func(ctx *exec.KernelCtx, span *exec.ExecSpan, result *exec.ExecResult) error {
			// The second buffer contains the values. Both for the input and the output arrays.
			for i, x := range span.Values[0].Array.Buffers[1].Buf {
				result.Buffers[1].Buf[i] = x + 42
			}
			return nil
		},
		nil,
	); err != nil {
		log.Fatal(err)
	}
	execCtx.Registry.AddFunction(add42, true)

	inputArrayBuilder := array.NewInt8Builder(pool)
	for i := 0; i < 16; i++ {
		inputArrayBuilder.Append(int8(i))
	}
	inputArray := inputArrayBuilder.NewArray()

	outputArrayDatum, err := compute.CallFunction(
		compute.SetExecCtx(context.Background(), execCtx),
		"add_42",
		nil,
		&compute.ArrayDatum{Value: inputArray.Data()},
	)
	if err != nil {
		log.Fatal(err)
	}

	fmt.Println(array.NewInt8Data(outputArrayDatum.(*compute.ArrayDatum).Value).Int8Values())

}

Output:

[42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57]

Index

Examples

Constants

View Source 
const (
	// Round to nearest integer less than or equal in magnitude (aka "floor")
	RoundDown = kernels.RoundDown
	// Round to nearest integer greater than or equal in magnitude (aka "ceil")
	RoundUp = kernels.RoundUp
	// Get integral part without fractional digits (aka "trunc")
	RoundTowardsZero = kernels.TowardsZero
	// Round negative values with DOWN and positive values with UP
	RoundTowardsInfinity = kernels.AwayFromZero
	// Round ties with DOWN (aka "round half towards negative infinity")
	RoundHalfDown = kernels.HalfDown
	// Round ties with UP (aka "round half towards positive infinity")
	RoundHalfUp = kernels.HalfUp
	// Round ties with TowardsZero (aka "round half away from infinity")
	RoundHalfTowardsZero = kernels.HalfTowardsZero
	// Round ties with AwayFromZero (aka "round half towards infinity")
	RoundHalfTowardsInfinity = kernels.HalfAwayFromZero
	// Round ties to nearest even integer
	RoundHalfToEven = kernels.HalfToEven
	// Round ties to nearest odd integer
	RoundHalfToOdd = kernels.HalfToOdd
)
View Source 
const (
	SelectionEmitNulls = kernels.EmitNulls
	SelectionDropNulls = kernels.DropNulls
)
View Source 
const DefaultMaxChunkSize = math.MaxInt64
View Source 
const UnknownLength int64 = -1

Variables

View Source 
var (
	DefaultRoundOptions           = RoundOptions{NDigits: 0, Mode: RoundHalfToEven}
	DefaultRoundToMultipleOptions = RoundToMultipleOptions{
		Multiple: scalar.NewFloat64Scalar(1), Mode: RoundHalfToEven}
)
View Source 
var (

	// WithAllocator returns a new context with the provided allocator
	// embedded into the context.
	WithAllocator = exec.WithAllocator
	// GetAllocator retrieves the allocator from the context, or returns
	// memory.DefaultAllocator if there was no allocator in the provided
	// context.
	GetAllocator = exec.GetAllocator
)
View Source 
var (
	ErrEmpty           = errors.New("cannot traverse empty field path")
	ErrNoChildren      = errors.New("trying to get child of type with no children")
	ErrIndexRange      = errors.New("index out of range")
	ErrMultipleMatches = errors.New("multiple matches")
	ErrNoMatch         = errors.New("no match")
	ErrInvalid         = errors.New("field ref invalid")
)

Functions

func CanCast 

func CanCast(from, to arrow.DataType) bool

CanCast returns true if there is an implementation for casting an array or scalar value from the specified DataType to the other data type.

func CastArray 

func CastArray(ctx context.Context, val arrow.Array, opts *CastOptions) (arrow.Array, error)

CastArray is a convenience function for casting an Array to another type. It is equivalent to constructing a Datum for the array and using CallFunction(ctx, "cast", ...).

func CastFromExtension 

func CastFromExtension(ctx *exec.KernelCtx, batch *exec.ExecSpan, out *exec.ExecResult) error

func CastList 

func CastList[SrcOffsetT, DestOffsetT int32 | int64](ctx *exec.KernelCtx, batch *exec.ExecSpan, out *exec.ExecResult) error

func CastStruct 

func CastStruct(ctx *exec.KernelCtx, batch *exec.ExecSpan, out *exec.ExecResult) error

func CastToType 

func CastToType(ctx context.Context, val arrow.Array, toType arrow.DataType) (arrow.Array, error)

CastToType is a convenience function equivalent to calling CastArray(ctx, val, compute.SafeCastOptions(toType))

func DatumIsValue 

func DatumIsValue(d Datum) bool

DatumIsValue returns true if the datum passed is a Scalar, Array or ChunkedArray type (e.g. it contains a specific value not a group of values)

func ExecSpanFromBatch 

func ExecSpanFromBatch(batch *ExecBatch) *exec.ExecSpan

ExecSpanFromBatch constructs and returns a new ExecSpan from the values inside of the ExecBatch which could be scalar or arrays.

This is mostly used for tests but is also a convenience method for other cases.

func FilterArray 

func FilterArray(ctx context.Context, values, filter arrow.Array, options FilterOptions) (arrow.Array, error)

FilterArray is a convenience method for calling Filter without having to manually construct the intervening Datum objects (they will be created for you internally here).

func FilterRecordBatch 

func FilterRecordBatch(ctx context.Context, batch arrow.Record, filter arrow.Array, opts *FilterOptions) (arrow.Record, error)

func FilterTable 

func FilterTable(ctx context.Context, tbl arrow.Table, filter Datum, opts *FilterOptions) (arrow.Table, error)

func RegisterScalarArithmetic 

func RegisterScalarArithmetic(reg FunctionRegistry)

func RegisterScalarBoolean 

func RegisterScalarBoolean(reg FunctionRegistry)

func RegisterScalarCast 

func RegisterScalarCast(reg FunctionRegistry)

func RegisterScalarComparisons 

func RegisterScalarComparisons(reg FunctionRegistry)

func RegisterVectorHash 

func RegisterVectorHash(reg FunctionRegistry)

func RegisterVectorRunEndFuncs 

func RegisterVectorRunEndFuncs(reg FunctionRegistry)

func RegisterVectorSelection 

func RegisterVectorSelection(reg FunctionRegistry)

RegisterVectorSelection registers functions that select specific values from arrays such as Take and Filter

func RunEndDecodeArray 

func RunEndDecodeArray(ctx context.Context, input arrow.Array) (arrow.Array, error)

func RunEndEncodeArray 

func RunEndEncodeArray(ctx context.Context, opts RunEndEncodeOptions, input arrow.Array) (arrow.Array, error)

func SerializeExpr 

func SerializeExpr(expr Expression, mem memory.Allocator) (*memory.Buffer, error)

SerializeExpr serializes expressions by converting them to Metadata and storing this in the schema of a Record. Embedded arrays and scalars are stored in its columns. Finally the record is written as an IPC file

func SerializeOptions 

func SerializeOptions(opts FunctionOptions, mem memory.Allocator) (*memory.Buffer, error)

func SetExecCtx 

func SetExecCtx(ctx context.Context, e ExecCtx) context.Context

SetExecCtx returns a new child context containing the passed in ExecCtx

func TakeArray 

func TakeArray(ctx context.Context, values, indices arrow.Array) (arrow.Array, error)

func TakeArrayOpts 

func TakeArrayOpts(ctx context.Context, values, indices arrow.Array, opts TakeOptions) (arrow.Array, error)

func UniqueArray 

func UniqueArray(ctx context.Context, values arrow.Array) (arrow.Array, error)

Types

type ArithmeticOptions 

type ArithmeticOptions struct {
	NoCheckOverflow bool `compute:"check_overflow"`
}

func (ArithmeticOptions) TypeName 

func (ArithmeticOptions) TypeName() string

type Arity 

type Arity struct {
	NArgs     int
	IsVarArgs bool
}

Arity defines the number of required arguments for a function.

Naming conventions are taken from https://en.wikipedia.org/wiki/Arity

func Binary 

func Binary() Arity

func Nullary 

func Nullary() Arity

func Ternary 

func Ternary() Arity

func Unary 

func Unary() Arity

func VarArgs 

func VarArgs(minArgs int) Arity

type ArrayDatum 

type ArrayDatum struct {
	Value arrow.ArrayData
}

ArrayDatum references an array.Data object which can be used to create array instances from if needed.

func (*ArrayDatum) Chunks 

func (d *ArrayDatum) Chunks() []arrow.Array

func (*ArrayDatum) Equals 

func (d *ArrayDatum) Equals(other Datum) bool

func (ArrayDatum) Kind 

func (ArrayDatum) Kind() DatumKind

func (*ArrayDatum) Len 

func (d *ArrayDatum) Len() int64

func (*ArrayDatum) MakeArray 

func (d *ArrayDatum) MakeArray() arrow.Array

func (*ArrayDatum) NullN 

func (d *ArrayDatum) NullN() int64

func (*ArrayDatum) Release 

func (d *ArrayDatum) Release()

func (*ArrayDatum) String 

func (d *ArrayDatum) String() string

func (*ArrayDatum) ToScalar 

func (d *ArrayDatum) ToScalar() (scalar.Scalar, error)

func (*ArrayDatum) Type 

func (d *ArrayDatum) Type() arrow.DataType

type ArrayLikeDatum 

type ArrayLikeDatum interface {
	Datum
	NullN() int64
	Type() arrow.DataType
	Chunks() []arrow.Array
}

ArrayLikeDatum is an interface for treating a Datum similarly to an Array, so that it is easy to differentiate between Record/Table/Collection and Scalar, Array/ChunkedArray for ease of use. Chunks will return an empty slice for Scalar, a slice with 1 element for Array, and the slice of chunks for a chunked array.

type Call deprecated 

type Call struct {
	// contains filtered or unexported fields
}

Call is a function call with specific arguments which are themselves other expressions. A call can also have options that are specific to the function in question. It must be bound to determine the shape and type.

Deprecated: use substrait-go expression functions instead.

func (*Call) Equals 

func (c *Call) Equals(other Expression) bool

func (*Call) FieldRef 

func (c *Call) FieldRef() *FieldRef

func (*Call) Hash 

func (c *Call) Hash() uint64

func (*Call) IsBound 

func (c *Call) IsBound() bool

func (*Call) IsNullLiteral 

func (c *Call) IsNullLiteral() bool

func (*Call) IsSatisfiable 

func (c *Call) IsSatisfiable() bool

func (*Call) IsScalarExpr 

func (c *Call) IsScalarExpr() bool

func (*Call) Release 

func (c *Call) Release()

func (*Call) String 

func (c *Call) String() string

func (*Call) Type 

func (c *Call) Type() arrow.DataType

type CastOptions 

type CastOptions = kernels.CastOptions

func DefaultCastOptions 

func DefaultCastOptions(safe bool) *CastOptions

func NewCastOptions 

func NewCastOptions(dt arrow.DataType, safe bool) *CastOptions

func SafeCastOptions 

func SafeCastOptions(dt arrow.DataType) *CastOptions

func UnsafeCastOptions 

func UnsafeCastOptions(dt arrow.DataType) *CastOptions

type ChunkedDatum 

type ChunkedDatum struct {
	Value *arrow.Chunked
}

ChunkedDatum contains a chunked array for use with expressions and compute.

func (*ChunkedDatum) Chunks 

func (d *ChunkedDatum) Chunks() []arrow.Array

func (*ChunkedDatum) Equals 

func (d *ChunkedDatum) Equals(other Datum) bool

func (ChunkedDatum) Kind 

func (ChunkedDatum) Kind() DatumKind

func (*ChunkedDatum) Len 

func (d *ChunkedDatum) Len() int64

func (*ChunkedDatum) NullN 

func (d *ChunkedDatum) NullN() int64

func (*ChunkedDatum) Release 

func (d *ChunkedDatum) Release()

func (*ChunkedDatum) String 

func (d *ChunkedDatum) String() string

func (*ChunkedDatum) Type 

func (d *ChunkedDatum) Type() arrow.DataType

type Datum 

type Datum interface {
	fmt.Stringer
	Kind() DatumKind
	Len() int64
	Equals(Datum) bool
	Release()
	// contains filtered or unexported methods
}

Datum is a variant interface for wrapping the various Arrow data structures for now the various Datum types just hold a Value which is the type they are wrapping, but it might make sense in the future for those types to actually be aliases or embed their types instead. Not sure yet.

func AbsoluteValue 

func AbsoluteValue(ctx context.Context, opts ArithmeticOptions, input Datum) (Datum, error)

AbsoluteValue returns the AbsoluteValue for each element in the input argument. It accepts either a scalar or an array.

ArithmeticOptions specifies whether or not to check for overflows, performance is faster if not explicitly checking for overflows but will error on an overflow if CheckOverflow is true.

func Acos 

func Acos(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Add 

func Add(ctx context.Context, opts ArithmeticOptions, left, right Datum) (Datum, error)

Add performs an addition between the passed in arguments (scalar or array) and returns the result. If one argument is a scalar and the other is an array, the scalar value is added to each value of the array.

ArithmeticOptions specifies whether or not to check for overflows, performance is faster if not explicitly checking for overflows but will error on an overflow if NoCheckOverflow is false (default).

func Asin 

func Asin(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Atan 

func Atan(ctx context.Context, arg Datum) (Datum, error)

func Atan2 

func Atan2(ctx context.Context, x, y Datum) (Datum, error)

func CallFunction 

func CallFunction(ctx context.Context, funcName string, opts FunctionOptions, args ...Datum) (Datum, error)

CallFunction is a one-shot invoker for all types of functions.

It will perform kernel-dispatch, argument checking, iteration of ChunkedArray inputs and wrapping of outputs.

To affect the execution options, you must call SetExecCtx and pass the resulting context in here.

func CastDatum 

func CastDatum(ctx context.Context, val Datum, opts *CastOptions) (Datum, error)

CastDatum is a convenience function for casting a Datum to another type. It is equivalent to calling CallFunction(ctx, "cast", opts, Datum) and should work for Scalar, Array or ChunkedArray Datums.

func Cos 

func Cos(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Divide 

func Divide(ctx context.Context, opts ArithmeticOptions, left, right Datum) (Datum, error)

Divide performs a division between the passed in arguments (scalar or array) and returns the result. If one argument is a scalar and the other is an array, the scalar value is used with each value of the array.

ArithmeticOptions specifies whether or not to check for overflows, performance is faster if not explicitly checking for overflows but will error on an overflow if NoCheckOverflow is false (default).

Will error on divide by zero regardless of whether or not checking for overflows.

func Filter 

func Filter(ctx context.Context, values, filter Datum, options FilterOptions) (Datum, error)

Filter is a wrapper convenience that is equivalent to calling CallFunction(ctx, "filter", &options, values, filter) for filtering an input array (values) by a boolean array (filter). The two inputs must be the same length.

func Ln 

func Ln(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Log10 

func Log10(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Log1p 

func Log1p(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Log2 

func Log2(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Logb 

func Logb(ctx context.Context, opts ArithmeticOptions, x, base Datum) (Datum, error)

func Multiply 

func Multiply(ctx context.Context, opts ArithmeticOptions, left, right Datum) (Datum, error)

Multiply performs a multiplication between the passed in arguments (scalar or array) and returns the result. If one argument is a scalar and the other is an array, the scalar value is multiplied against each value of the array.

ArithmeticOptions specifies whether or not to check for overflows, performance is faster if not explicitly checking for overflows but will error on an overflow if NoCheckOverflow is false (default).

func Negate 

func Negate(ctx context.Context, opts ArithmeticOptions, input Datum) (Datum, error)

Negate returns a result containing the negation of each element in the input argument. It accepts either a scalar or an array.

ArithmeticOptions specifies whether or not to check for overflows, or to throw an error on unsigned types.

func NewDatum 

func NewDatum(value interface{}) Datum

NewDatum will construct the appropriate Datum type based on what is passed in as the argument.

An arrow.Array gets an ArrayDatum An array.Chunked gets a ChunkedDatum An array.Record gets a RecordDatum an array.Table gets a TableDatum a scalar.Scalar gets a ScalarDatum

Anything else is passed to scalar.MakeScalar and receives a scalar datum of that appropriate type.

func NewDatumWithoutOwning 

func NewDatumWithoutOwning(value interface{}) Datum

func Power 

func Power(ctx context.Context, opts ArithmeticOptions, base, exp Datum) (Datum, error)

Power returns base**exp for each element in the input arrays. Should work for both Arrays and Scalars

func Round 

func Round(ctx context.Context, opts RoundOptions, arg Datum) (Datum, error)

func RoundToMultiple 

func RoundToMultiple(ctx context.Context, opts RoundToMultipleOptions, arg Datum) (Datum, error)

func RunEndDecode 

func RunEndDecode(ctx context.Context, arg Datum) (Datum, error)

func RunEndEncode 

func RunEndEncode(ctx context.Context, opts RunEndEncodeOptions, arg Datum) (Datum, error)

func ShiftLeft 

func ShiftLeft(ctx context.Context, opts ArithmeticOptions, lhs, rhs Datum) (Datum, error)

ShiftLeft only accepts integral types and shifts each element of the first argument to the left by the value of the corresponding element in the second argument.

The value to shift by should be >= 0 and < precision of the type.

func ShiftRight 

func ShiftRight(ctx context.Context, opts ArithmeticOptions, lhs, rhs Datum) (Datum, error)

ShiftRight only accepts integral types and shifts each element of the first argument to the right by the value of the corresponding element in the second argument.

The value to shift by should be >= 0 and < precision of the type.

func Sign 

func Sign(ctx context.Context, input Datum) (Datum, error)

Sign returns -1, 0, or 1 depending on the sign of each element in the input. For x in the input: 

	if x > 0: 1
	if x < 0: -1
    if x == 0: 0

func Sin 

func Sin(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Subtract 

func Subtract(ctx context.Context, opts ArithmeticOptions, left, right Datum) (Datum, error)

Sub performs a subtraction between the passed in arguments (scalar or array) and returns the result. If one argument is a scalar and the other is an array, the scalar value is subtracted from each value of the array.

ArithmeticOptions specifies whether or not to check for overflows, performance is faster if not explicitly checking for overflows but will error on an overflow if NoCheckOverflow is false (default).

func Take 

func Take(ctx context.Context, opts TakeOptions, values, indices Datum) (Datum, error)

func Tan 

func Tan(ctx context.Context, opts ArithmeticOptions, arg Datum) (Datum, error)

func Unique 

func Unique(ctx context.Context, values Datum) (Datum, error)

type DatumKind 

type DatumKind int

DatumKind is an enum used for denoting which kind of type a datum is encapsulating 

const (
	KindNone    DatumKind = iota // none
	KindScalar                   // scalar
	KindArray                    // array
	KindChunked                  // chunked_array
	KindRecord                   // record_batch
	KindTable                    // table
)

func (DatumKind) String 

func (i DatumKind) String() string

type EmptyDatum 

type EmptyDatum struct{}

EmptyDatum is the null case, a Datum with nothing in it.

func (EmptyDatum) Equals 

func (EmptyDatum) Equals(other Datum) bool

func (EmptyDatum) Kind 

func (EmptyDatum) Kind() DatumKind

func (EmptyDatum) Len 

func (EmptyDatum) Len() int64

func (EmptyDatum) Release 

func (EmptyDatum) Release()

func (EmptyDatum) String 

func (EmptyDatum) String() string

type ExecBatch 

type ExecBatch struct {
	Values []Datum
	// Guarantee is a predicate Expression guaranteed to evaluate to true for
	// all rows in this batch.
	// Guarantee Expression
	// Len is the semantic length of this ExecBatch. When the values are
	// all scalars, the length should be set to 1 for non-aggregate kernels.
	// Otherwise the length is taken from the array values. Aggregate kernels
	// can have an ExecBatch formed by projecting just the partition columns
	// from a batch in which case it would have scalar rows with length > 1
	//
	// If the array values are of length 0, then the length is 0 regardless of
	// whether any values are Scalar.
	Len int64
}

ExecBatch is a unit of work for kernel execution. It contains a collection of Array and Scalar values.

ExecBatch is semantically similar to a RecordBatch but for a SQL-style execution context. It represents a collection or records, but constant "columns" are represented by Scalar values rather than having to be converted into arrays with repeated values.

func (ExecBatch) NumValues 

func (e ExecBatch) NumValues() int

type ExecCtx 

type ExecCtx struct {
	// ChunkSize is the size used when iterating batches for execution
	// ChunkSize elements will be operated on as a time unless an argument
	// is a chunkedarray with a chunk that is smaller
	ChunkSize int64
	// PreallocContiguous determines whether preallocating memory for
	// execution of compute attempts to preallocate a full contiguous
	// buffer for all of the chunks beforehand.
	PreallocContiguous bool
	// Registry allows specifying the Function Registry to utilize
	// when searching for kernel implementations.
	Registry FunctionRegistry
	// ExecChannelSize is the size of the channel used for passing
	// exec results to the WrapResults function.
	ExecChannelSize int
	// NumParallel determines the number of parallel goroutines
	// allowed for parallel executions.
	NumParallel int
}

ExecCtx holds simple contextual information for execution such as the default ChunkSize for batch iteration, whether or not to ensure contiguous preallocations for kernels that want preallocation, and a reference to the desired function registry to use.

An ExecCtx should be placed into a context.Context by using SetExecCtx and GetExecCtx to pass it along for execution.

func DefaultExecCtx 

func DefaultExecCtx() ExecCtx

DefaultExecCtx returns the default exec context which will be used if there is no ExecCtx set into the context for execution.

This can be called to get a copy of the default values which can then be modified to set into a context.

The default exec context uses the following values:

  • ChunkSize = DefaultMaxChunkSize (MaxInt64)
  • PreallocContiguous = true
  • Registry = GetFunctionRegistry()
  • ExecChannelSize = 10
  • NumParallel = runtime.NumCPU()

func GetExecCtx 

func GetExecCtx(ctx context.Context) ExecCtx

GetExecCtx returns an embedded ExecCtx from the provided context. If it does not contain an ExecCtx, then the default one is returned.

type Expression deprecated 

type Expression interface {
	fmt.Stringer
	// IsBound returns true if this expression has been bound to a particular
	// Datum and/or Schema.
	IsBound() bool
	// IsScalarExpr returns true if this expression is composed only of scalar
	// literals, field references and calls to scalar functions.
	IsScalarExpr() bool
	// IsNullLiteral returns true if this expression is a literal and entirely
	// null.
	IsNullLiteral() bool
	// IsSatisfiable returns true if this expression could evaluate to true
	IsSatisfiable() bool
	// FieldRef returns a pointer to the underlying field reference, or nil if
	// this expression is not a field reference.
	FieldRef() *FieldRef
	// Type returns the datatype this expression will evaluate to.
	Type() arrow.DataType

	Hash() uint64
	Equals(Expression) bool

	// Release releases the underlying bound C++ memory that is allocated when
	// a Bind is performed. Any bound expression should get released to ensure
	// no memory leaks.
	Release()
}

Expression is an interface for mapping one datum to another. An expression is one of: 

A literal Datum
A reference to a single (potentially nested) field of an input Datum
A call to a compute function, with arguments specified by other Expressions

Deprecated: use substrait-go expressions instead.

func And 

func And(lhs, rhs Expression, ops ...Expression) Expression

And constructs a tree of calls to and_kleene for boolean And logic taking an arbitrary number of values.

func Cast 

func Cast(ex Expression, dt arrow.DataType) Expression

func DeserializeExpr 

func DeserializeExpr(mem memory.Allocator, buf *memory.Buffer) (Expression, error)

func Equal 

func Equal(lhs, rhs Expression) Expression

Equal is a convenience function for the equal function

func Greater 

func Greater(lhs, rhs Expression) Expression

Greater is shorthand for NewCall("greater",....)

func GreaterEqual 

func GreaterEqual(lhs, rhs Expression) Expression

GreaterEqual is shorthand for NewCall("greater_equal",....)

func IsNull 

func IsNull(lhs Expression, nanIsNull bool) Expression

IsNull creates an expression that returns true if the passed in expression is null. Optionally treating NaN as null if desired.

func IsValid 

func IsValid(lhs Expression) Expression

IsValid is the inverse of IsNull

func Less 

func Less(lhs, rhs Expression) Expression

Less is shorthand for NewCall("less",....)

func LessEqual 

func LessEqual(lhs, rhs Expression) Expression

LessEqual is shorthand for NewCall("less_equal",....)

func NewCall 

func NewCall(name string, args []Expression, opts FunctionOptions) Expression

NewCall constructs an expression that represents a specific function call with the given arguments and options.

func NewFieldRef 

func NewFieldRef(field string) Expression

NewFieldRef is shorthand for NewRef(FieldRefName(field))

func NewLiteral 

func NewLiteral(arg interface{}) Expression

NewLiteral constructs a new literal expression from any value. It is passed to NewDatum which will construct the appropriate Datum and/or scalar value for the type provided.

func NewRef 

func NewRef(ref FieldRef) Expression

NewRef constructs a parameter expression which refers to a specific field

func Not 

func Not(expr Expression) Expression

Not creates a call to "invert" for the value specified.

func NotEqual 

func NotEqual(lhs, rhs Expression) Expression

NotEqual creates a call to not_equal

func NullLiteral 

func NullLiteral(dt arrow.DataType) Expression

func Or 

func Or(lhs, rhs Expression, ops ...Expression) Expression

Or constructs a tree of calls to or_kleene for boolean Or logic taking an arbitrary number of values.

func Project 

func Project(values []Expression, names []string) Expression

Project is shorthand for `make_struct` to produce a record batch output from a group of expressions.

type FieldPath 

type FieldPath []int

FieldPath represents a path to a nested field using indices of child fields. For example, given the indices {5, 9, 3} the field could be retrieved with: schema.Field(5).Type().(*arrow.StructType).Field(9).Type().(*arrow.StructType).Field(3)

Attempting to retrieve a child field using a FieldPath which is not valid for a given schema will get an error such as an out of range index, or an empty path.

FieldPaths provide for drilling down to potentially nested children for convenience of accepting a slice of fields, a schema or a datatype (which should contain child fields).

A fieldpath can also be used to retrieve a child arrow.Array or column from a record batch.

func (FieldPath) Get 

func (f FieldPath) Get(s *arrow.Schema) (*arrow.Field, error)

Get retrieves the corresponding nested child field by drilling through the schema's fields as per the field path.

func (FieldPath) GetColumn 

func (f FieldPath) GetColumn(batch arrow.Record) (arrow.Array, error)

GetColumn will return the correct child array by traversing the fieldpath going to the nested arrays of the columns in the record batch.

func (FieldPath) GetField 

func (f FieldPath) GetField(field arrow.Field) (*arrow.Field, error)

GetField is equivalent to GetFieldFromType(field.Type)

func (FieldPath) GetFieldFromSlice 

func (f FieldPath) GetFieldFromSlice(fields []arrow.Field) (*arrow.Field, error)

GetFieldFromSlice treats the slice as the top layer of fields, so the first value in the field path will index into the slice, and then drill down from there.

func (FieldPath) GetFieldFromType 

func (f FieldPath) GetFieldFromType(typ arrow.DataType) (*arrow.Field, error)

GetFieldFromType returns the nested field from a datatype by drilling into it's child fields.

func (FieldPath) String 

func (f FieldPath) String() string

type FieldRef 

type FieldRef struct {
	// contains filtered or unexported fields
}

FieldRef is a descriptor of a (potentially nested) field within a schema.

Unlike FieldPath (which is exclusively indices of child fields), FieldRef may reference a field by name. It can be constructed from either a field index, field name, or field path.

Nested fields can be referenced as well, given the schema: 

		arrow.NewSchema([]arrow.Field{
			{Name: "a", Type: arrow.StructOf(arrow.Field{Name: "n", Type: arrow.Null})},
 		{Name: "b", Type: arrow.PrimitiveTypes.Int32},
		})

the following all indicate the nested field named "n": 

FieldRefPath(FieldPath{0, 0})
FieldRefList("a", 0)
FieldRefList("a", "n")
FieldRefList(0, "n")
NewFieldRefFromDotPath(".a[0]")

FieldPaths matching a FieldRef are retrieved with the FindAll* functions Multiple matches are possible because field names may be duplicated within a schema. For example: 

aIsAmbiguous := arrow.NewSchema([]arrow.Field{
	{Name: "a", Type: arrow.PrimitiveTypes.Int32},
	{Name: "a", Type: arrow.PrimitiveTypes.Float32},
})
matches := FieldRefName("a").FindAll(aIsAmbiguous)
assert.Len(matches, 2)
assert.True(matches[0].Get(aIsAmbiguous).Equals(aIsAmbiguous.Field(0))
assert.True(matches[1].Get(aIsAmbiguous).Equals(aIsAmbiguous.Field(1))

func FieldRefIndex 

func FieldRefIndex(i int) FieldRef

FieldRefIndex is a convenience function to construct a FieldPath reference of a single index

func FieldRefList 

func FieldRefList(elems ...interface{}) FieldRef

FieldRefList takes an arbitrary number of arguments which can be either strings or ints. This will panic if anything other than a string or int is passed in.

func FieldRefName 

func FieldRefName(n string) FieldRef

FieldRefName constructs a FieldRef by name

func FieldRefPath 

func FieldRefPath(p FieldPath) FieldRef

FieldRefPath constructs a FieldRef from a given FieldPath

func NewFieldRefFromDotPath 

func NewFieldRefFromDotPath(dotpath string) (out FieldRef, err error)

NewFieldRefFromDotPath parses a dot path into a field ref.

dot_path = '.' name 

| '[' digit+ ']'
| dot_path+

Examples 

".alpha" => FieldRefName("alpha")
"[2]" => FieldRefIndex(2)
".beta[3]" => FieldRefList("beta", 3)
"[5].gamma.delta[7]" => FieldRefList(5, "gamma", "delta", 7)
".hello world" => FieldRefName("hello world")
`.\[y\]\\tho\.\` => FieldRef(`[y]\tho.\`)

Note: when parsing a name, a '\' preceding any other character will be dropped from the resulting name. therefore if a name must contain the characters '.', '\', '[' or ']' then they must be escaped with a preceding '\'.

func (*FieldRef) Equals 

func (f *FieldRef) Equals(other FieldRef) bool

func (*FieldRef) FieldPath 

func (f *FieldRef) FieldPath() FieldPath

FieldPath returns the fieldpath that this FieldRef uses, otherwise an empty FieldPath if it's not a FieldPath reference

func (FieldRef) FindAll 

func (f FieldRef) FindAll(fields []arrow.Field) []FieldPath

FindAll returns all the fieldpaths which this FieldRef matches in the given slice of fields.

func (FieldRef) FindAllField 

func (f FieldRef) FindAllField(field arrow.Field) []FieldPath

FindAllField returns all the fieldpaths that this FieldRef matches against the type of the given field.

func (FieldRef) FindOne 

func (f FieldRef) FindOne(schema *arrow.Schema) (FieldPath, error)

FindOne returns an error if the field isn't matched or if there are multiple matches otherwise it returns the path to the single valid match.

func (FieldRef) FindOneOrNone 

func (f FieldRef) FindOneOrNone(schema *arrow.Schema) (FieldPath, error)

FindOneOrNone is a convenience helper that will either return 1 fieldpath, or an empty fieldpath, and will return an error if there are multiple matches.

func (FieldRef) FindOneOrNoneRecord 

func (f FieldRef) FindOneOrNoneRecord(root arrow.Record) (FieldPath, error)

FindOneOrNoneRecord is like FindOneOrNone but for the schema of a record, returning an error only if there are multiple matches.

func (FieldRef) GetAllColumns 

func (f FieldRef) GetAllColumns(root arrow.Record) ([]arrow.Array, error)

GetAllColumns gets all the matching column arrays from the given record that this FieldRef references.

func (FieldRef) GetOneColumnOrNone 

func (f FieldRef) GetOneColumnOrNone(root arrow.Record) (arrow.Array, error)

GetOneColumnOrNone returns either a nil or the referenced array if it can be found, erroring only if there is an ambiguous multiple matches.

func (FieldRef) GetOneField 

func (f FieldRef) GetOneField(schema *arrow.Schema) (*arrow.Field, error)

GetOneField will return a pointer to a field or an error if it is not found or if there are multiple matches.

func (FieldRef) GetOneOrNone 

func (f FieldRef) GetOneOrNone(schema *arrow.Schema) (*arrow.Field, error)

GetOneOrNone will return a field or a nil if the field is found or not, and only errors if there are multiple matches.

func (FieldRef) Hash 

func (f FieldRef) Hash(seed maphash.Seed) uint64

Hash produces a hash of this field reference and takes in a seed so that it can maintain consistency across multiple places / processes /etc.

func (*FieldRef) IsFieldPath 

func (f *FieldRef) IsFieldPath() bool

IsFieldPath returns true if this FieldRef uses a fieldpath

func (*FieldRef) IsName 

func (f *FieldRef) IsName() bool

IsName returns true if this fieldref is a name reference

func (*FieldRef) IsNested 

func (f *FieldRef) IsNested() bool

IsNested returns true if this FieldRef expects to represent a nested field.

func (*FieldRef) Name 

func (f *FieldRef) Name() string

Name returns the name of the field this references if it is a Name reference, otherwise the empty string

func (FieldRef) String 

func (f FieldRef) String() string

type FilterOptions 

type FilterOptions = kernels.FilterOptions

func DefaultFilterOptions 

func DefaultFilterOptions() *FilterOptions

type FuncKind 

type FuncKind int8

FuncKind is an enum representing the type of a function 

const (
	// A function that performs scalar data operations on whole arrays
	// of data. Can generally process Array or Scalar values. The size
	// of the output will be the same as the size (or broadcasted size,
	// in the case of mixing Array and Scalar inputs) of the input.
	FuncScalar FuncKind = iota // Scalar
	// A function with array input and output whose behavior depends on
	// the values of the entire arrays passed, rather than the value of
	// each scalar value.
	FuncVector // Vector
	// A function that computes a scalar summary statistic from array input.
	FuncScalarAgg // ScalarAggregate
	// A function that computes grouped summary statistics from array
	// input and an array of group identifiers.
	FuncHashAgg // HashAggregate
	// A function that dispatches to other functions and does not contain
	// its own kernels.
	FuncMeta // Meta
)

func (FuncKind) String 

func (i FuncKind) String() string

type Function 

type Function interface {
	Name() string
	Kind() FuncKind
	Arity() Arity
	Doc() FunctionDoc
	NumKernels() int
	Execute(context.Context, FunctionOptions, ...Datum) (Datum, error)
	DispatchExact(...arrow.DataType) (exec.Kernel, error)
	DispatchBest(...arrow.DataType) (exec.Kernel, error)
	DefaultOptions() FunctionOptions
	Validate() error
}

type FunctionDoc 

type FunctionDoc struct {
	// A one-line summary of the function, using a verb.
	//
	// For example, "Add two numeric arrays or scalars"
	Summary string
	// A detailed description of the function, meant to follow the summary.
	Description string
	// Symbolic names (identifiers) for the function arguments.
	//
	// Can be used to generate nicer function signatures.
	ArgNames []string
	// Name of the options struct type, if any
	OptionsType string
	// Whether or not options are required for function execution.
	//
	// If false, then either there are no options for this function,
	// or there is a usable default options value.
	OptionsRequired bool
}
var EmptyFuncDoc FunctionDoc

EmptyFuncDoc is a reusable empty function doc definition for convenience.

type FunctionOptions 

type FunctionOptions interface {
	TypeName() string
}

FunctionOptions can be any type which has a TypeName function. The fields of the type will be used (via reflection) to determine the information to propagate when serializing to pass to the C++ for execution.

type FunctionOptionsCloneable 

type FunctionOptionsCloneable interface {
	Clone() FunctionOptions
}

type FunctionOptionsEqual 

type FunctionOptionsEqual interface {
	Equals(FunctionOptions) bool
}

type FunctionRegistry 

type FunctionRegistry interface {
	CanAddFunction(fn Function, allowOverwrite bool) bool
	AddFunction(fn Function, allowOverwrite bool) bool
	CanAddAlias(target, source string) bool
	AddAlias(target, source string) bool
	GetFunction(name string) (Function, bool)
	GetFunctionNames() []string
	NumFunctions() int
	// contains filtered or unexported methods
}

func GetFunctionRegistry 

func GetFunctionRegistry() FunctionRegistry

func NewChildRegistry 

func NewChildRegistry(parent FunctionRegistry) FunctionRegistry

func NewRegistry 

func NewRegistry() FunctionRegistry

type KernelExecutor 

type KernelExecutor interface {
	// Init must be called *after* the kernel's init method and any
	// KernelState must be set into the KernelCtx *before* calling
	// this Init method. This is to facilitate the case where
	// Init may be expensive and does not need to be called
	// again for each execution of the kernel. For example,
	// the same lookup table can be re-used for all scanned batches
	// in a dataset filter.
	Init(*exec.KernelCtx, exec.KernelInitArgs) error
	// Execute the kernel for the provided batch and pass the resulting
	// Datum values to the provided channel.
	Execute(context.Context, *ExecBatch, chan<- Datum) error
	// WrapResults exists for the case where an executor wants to post process
	// the batches of result datums. Such as creating a ChunkedArray from
	// multiple output batches or so on. Results from individual batch
	// executions should be read from the out channel, and WrapResults should
	// return the final Datum result.
	WrapResults(ctx context.Context, out <-chan Datum, chunkedArgs bool) Datum
	// CheckResultType checks the actual result type against the resolved
	// output type. If the types don't match an error is returned
	CheckResultType(out Datum) error
	// Clear resets the state in the executor so that it can be reused.
	Clear()
}

KernelExecutor is the interface for all executors to initialize and call kernel execution functions on batches.

func NewScalarExecutor 

func NewScalarExecutor() KernelExecutor

type Literal deprecated 

type Literal struct {
	Literal Datum
}

Literal is an expression denoting a literal Datum which could be any value as a scalar, an array, or so on.

Deprecated: use substrait-go expressions Literal instead.

func (*Literal) Equals 

func (l *Literal) Equals(other Expression) bool

func (Literal) FieldRef 

func (Literal) FieldRef() *FieldRef

func (*Literal) Hash 

func (l *Literal) Hash() uint64

func (*Literal) IsBound 

func (l *Literal) IsBound() bool

func (*Literal) IsNullLiteral 

func (l *Literal) IsNullLiteral() bool

func (*Literal) IsSatisfiable 

func (l *Literal) IsSatisfiable() bool

func (*Literal) IsScalarExpr 

func (l *Literal) IsScalarExpr() bool

func (*Literal) Release 

func (l *Literal) Release()

func (*Literal) String 

func (l *Literal) String() string

func (*Literal) Type 

func (l *Literal) Type() arrow.DataType

type MakeStructOptions 

type MakeStructOptions struct {
	FieldNames       []string          `compute:"field_names"`
	FieldNullability []bool            `compute:"field_nullability"`
	FieldMetadata    []*arrow.Metadata `compute:"field_metadata"`
}

func (MakeStructOptions) TypeName 

func (MakeStructOptions) TypeName() string

type MetaFunction 

type MetaFunction struct {
	// contains filtered or unexported fields
}

MetaFunction is a function which dispatches to other functions, the impl must not be nil.

For Array, ChunkedArray and Scalar datums, this may rely on the execution of concrete function types, but this must handle other Datum kinds on its own.

func NewMetaFunction 

func NewMetaFunction(name string, arity Arity, doc FunctionDoc, impl MetaFunctionImpl) *MetaFunction

NewMetaFunction constructs a new MetaFunction which will call the provided impl for dispatching with the expected arity.

Will panic if impl is nil.

func (*MetaFunction) Arity 

func (b *MetaFunction) Arity() Arity

func (*MetaFunction) DefaultOptions 

func (b *MetaFunction) DefaultOptions() FunctionOptions

func (*MetaFunction) DispatchBest 

func (m *MetaFunction) DispatchBest(...arrow.DataType) (exec.Kernel, error)

func (*MetaFunction) DispatchExact 

func (m *MetaFunction) DispatchExact(...arrow.DataType) (exec.Kernel, error)

func (*MetaFunction) Doc 

func (b *MetaFunction) Doc() FunctionDoc

func (*MetaFunction) Execute 

func (m *MetaFunction) Execute(ctx context.Context, opts FunctionOptions, args ...Datum) (Datum, error)

func (*MetaFunction) Kind 

func (b *MetaFunction) Kind() FuncKind

func (*MetaFunction) Name 

func (b *MetaFunction) Name() string

func (MetaFunction) NumKernels 

func (MetaFunction) NumKernels() int

func (*MetaFunction) Validate 

func (b *MetaFunction) Validate() error

type MetaFunctionImpl 

type MetaFunctionImpl func(context.Context, FunctionOptions, ...Datum) (Datum, error)

MetaFunctionImpl is the signature needed for implementing a MetaFunction which is a function that dispatches to another function instead.

type NullOptions 

type NullOptions struct {
	NanIsNull bool `compute:"nan_is_null"`
}

func (NullOptions) TypeName 

func (NullOptions) TypeName() string

type NullSelectionBehavior 

type NullSelectionBehavior = kernels.NullSelectionBehavior

type Parameter deprecated 

type Parameter struct {
	// contains filtered or unexported fields
}

Parameter represents a field reference and needs to be bound in order to determine its type and shape.

Deprecated: use substrait-go field references instead.

func (*Parameter) Equals 

func (p *Parameter) Equals(other Expression) bool

func (*Parameter) FieldRef 

func (p *Parameter) FieldRef() *FieldRef

func (*Parameter) Hash 

func (p *Parameter) Hash() uint64

func (*Parameter) IsBound 

func (p *Parameter) IsBound() bool

func (Parameter) IsNullLiteral 

func (Parameter) IsNullLiteral() bool

func (*Parameter) IsSatisfiable 

func (p *Parameter) IsSatisfiable() bool

func (*Parameter) IsScalarExpr 

func (p *Parameter) IsScalarExpr() bool

func (*Parameter) Release 

func (p *Parameter) Release()

func (*Parameter) String 

func (p *Parameter) String() string

func (*Parameter) Type 

func (p *Parameter) Type() arrow.DataType

type RecordDatum 

type RecordDatum struct {
	Value arrow.Record
}

RecordDatum contains an array.Record for passing a full record to an expression or to compute.

func (*RecordDatum) Equals 

func (r *RecordDatum) Equals(other Datum) bool

func (RecordDatum) Kind 

func (RecordDatum) Kind() DatumKind

func (*RecordDatum) Len 

func (r *RecordDatum) Len() int64

func (*RecordDatum) Release 

func (r *RecordDatum) Release()

func (*RecordDatum) Schema 

func (r *RecordDatum) Schema() *arrow.Schema

func (RecordDatum) String 

func (RecordDatum) String() string

type RoundMode 

type RoundMode = kernels.RoundMode

type RoundOptions 

type RoundOptions = kernels.RoundOptions

type RoundToMultipleOptions 

type RoundToMultipleOptions = kernels.RoundToMultipleOptions

type RunEndEncodeOptions 

type RunEndEncodeOptions = kernels.RunEndEncodeState

type ScalarDatum 

type ScalarDatum struct {
	Value scalar.Scalar
}

ScalarDatum contains a scalar value

func (ScalarDatum) Chunks 

func (ScalarDatum) Chunks() []arrow.Array

func (*ScalarDatum) Equals 

func (d *ScalarDatum) Equals(other Datum) bool

func (ScalarDatum) Kind 

func (ScalarDatum) Kind() DatumKind

func (ScalarDatum) Len 

func (ScalarDatum) Len() int64

func (*ScalarDatum) NullN 

func (d *ScalarDatum) NullN() int64

func (*ScalarDatum) Release 

func (d *ScalarDatum) Release()

func (*ScalarDatum) String 

func (d *ScalarDatum) String() string

func (*ScalarDatum) ToScalar 

func (d *ScalarDatum) ToScalar() (scalar.Scalar, error)

func (*ScalarDatum) Type 

func (d *ScalarDatum) Type() arrow.DataType

type ScalarFunction 

type ScalarFunction struct {
	// contains filtered or unexported fields
}

A ScalarFunction is a function that executes element-wise operations on arrays or scalars, and therefore whose results generally do not depend on the order of the values in the arguments. Accepts and returns arrays that are all of the same size. These functions roughly correspond to the functions used in most SQL expressions.

func NewScalarFunction 

func NewScalarFunction(name string, arity Arity, doc FunctionDoc) *ScalarFunction

NewScalarFunction constructs a new ScalarFunction object with the passed in name, arity and function doc.

func (*ScalarFunction) AddKernel 

func (s *ScalarFunction) AddKernel(k exec.ScalarKernel) error

AddKernel adds the provided kernel to the list of kernels this function has. A copy of the kernel is added to the slice of kernels, which means that a given kernel object can be created, added and then reused to add other kernels.

func (*ScalarFunction) AddNewKernel 

func (s *ScalarFunction) AddNewKernel(inTypes []exec.InputType, outType exec.OutputType, execFn exec.ArrayKernelExec, init exec.KernelInitFn) error

AddNewKernel constructs a new kernel with the provided signature and execution/init functions and then adds it to the function's list of kernels. This assumes default null handling (intersection of validity bitmaps)

func (*ScalarFunction) DispatchBest 

func (s *ScalarFunction) DispatchBest(vals ...arrow.DataType) (exec.Kernel, error)

func (*ScalarFunction) DispatchExact 

func (s *ScalarFunction) DispatchExact(vals ...arrow.DataType) (exec.Kernel, error)

func (*ScalarFunction) Execute 

func (s *ScalarFunction) Execute(ctx context.Context, opts FunctionOptions, args ...Datum) (Datum, error)

Execute uses the passed in context, function options and arguments to eagerly execute the function using kernel dispatch, batch iteration and memory allocation details as defined by the kernel.

If opts is nil, then the DefaultOptions() will be used.

func (*ScalarFunction) Kernels 

func (fi *ScalarFunction) Kernels() []*KT

func (*ScalarFunction) NumKernels 

func (fi *ScalarFunction) NumKernels() int

func (*ScalarFunction) SetDefaultOptions 

func (s *ScalarFunction) SetDefaultOptions(opts FunctionOptions)

type SetLookupOptions 

type SetLookupOptions struct {
	ValueSet  Datum `compute:"value_set"`
	SkipNulls bool  `compute:"skip_nulls"`
}

func (*SetLookupOptions) Equals 

func (s *SetLookupOptions) Equals(other FunctionOptions) bool

func (*SetLookupOptions) FromStructScalar 

func (s *SetLookupOptions) FromStructScalar(sc *scalar.Struct) error

func (*SetLookupOptions) Release 

func (s *SetLookupOptions) Release()

func (SetLookupOptions) TypeName 

func (SetLookupOptions) TypeName() string

type StrptimeOptions 

type StrptimeOptions struct {
	Format string         `compute:"format"`
	Unit   arrow.TimeUnit `compute:"unit"`
}

func (StrptimeOptions) TypeName 

func (StrptimeOptions) TypeName() string

type TableDatum 

type TableDatum struct {
	Value arrow.Table
}

TableDatum contains a table so that multiple record batches can be worked with together as a single table for being passed to compute and expression handling.

func (*TableDatum) Equals 

func (d *TableDatum) Equals(other Datum) bool

func (TableDatum) Kind 

func (TableDatum) Kind() DatumKind

func (*TableDatum) Len 

func (d *TableDatum) Len() int64

func (*TableDatum) Release 

func (d *TableDatum) Release()

func (*TableDatum) Schema 

func (d *TableDatum) Schema() *arrow.Schema

func (TableDatum) String 

func (TableDatum) String() string

type TableLikeDatum 

type TableLikeDatum interface {
	Datum
	Schema() *arrow.Schema
}

TableLikeDatum is an interface type for specifying either a RecordBatch or a Table as both contain a schema as opposed to just a single data type.

type TakeOptions 

type TakeOptions = kernels.TakeOptions

func DefaultTakeOptions 

func DefaultTakeOptions() *TakeOptions

type VectorFunction 

type VectorFunction struct {
	// contains filtered or unexported fields
}

func NewVectorFunction 

func NewVectorFunction(name string, arity Arity, doc FunctionDoc) *VectorFunction

func (*VectorFunction) AddKernel 

func (f *VectorFunction) AddKernel(kernel exec.VectorKernel) error

func (*VectorFunction) AddNewKernel 

func (f *VectorFunction) AddNewKernel(inTypes []exec.InputType, outType exec.OutputType, execFn exec.ArrayKernelExec, init exec.KernelInitFn) error

func (*VectorFunction) DispatchBest 

func (f *VectorFunction) DispatchBest(vals ...arrow.DataType) (exec.Kernel, error)

func (*VectorFunction) DispatchExact 

func (f *VectorFunction) DispatchExact(vals ...arrow.DataType) (exec.Kernel, error)

func (*VectorFunction) Execute 

func (f *VectorFunction) Execute(ctx context.Context, opts FunctionOptions, args ...Datum) (Datum, error)

func (*VectorFunction) Kernels 

func (fi *VectorFunction) Kernels() []*KT

func (*VectorFunction) NumKernels 

func (fi *VectorFunction) NumKernels() int

func (*VectorFunction) SetDefaultOptions 

func (f *VectorFunction) SetDefaultOptions(opts FunctionOptions)

Source Files

View all Source files

Directories

Path Synopsis
internal
kernels
Package kernels defines all of the computation kernels for the compute library.
 Package kernels defines all of the computation kernels for the compute library.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.