distance

Documentation

Index

• Constants
• Variables
• func CalcFFBatch(dim int64, left []float32, lIndex int64, right []float32, metric string, ...)
• func CalcFloatDistance(dim int64, left, right []float32, metric string) ([]float32, error)
• func CosineImplPure(a []float32, b []float32) float32
• func IPImplPure(a []float32, b []float32) float32
• func L2ImplPure(a []float32, b []float32) float32
• func ValidateFloatArrayLength(dim int64, length int) error
• func ValidateMetricType(metric string) (string, error)

Constants

const (
	// L2 represents the Euclidean distance
	L2 = "L2"
	// IP represents the inner product distance
	IP = "IP"
	// COSINE represents the cosine distance
	COSINE = "COSINE"
)

*

• Delete in #25663 Remove calc_distance
• Add back partially as clustering feature needs to calculate distance between search vector and clustering center

Variables

var (
	L2Impl     func(a []float32, b []float32) float32 = L2ImplPure
	IPImpl     func(a []float32, b []float32) float32 = IPImplPure
	CosineImpl func(a []float32, b []float32) float32 = CosineImplPure
)

Functions

func CalcFFBatch

func CalcFFBatch(dim int64, left []float32, lIndex int64, right []float32, metric string, result *[]float32)

CalcFFBatch calculate the distance of @left & @right vectors in batch by given @metic, store result in @result

func CalcFloatDistance

func CalcFloatDistance(dim int64, left, right []float32, metric string) ([]float32, error)

CalcFloatDistance calculate float distance by given metric it will checks input, and calculate the distance concurrently

func CosineImplPure

func CosineImplPure(a []float32, b []float32) float32

func IPImplPure

func IPImplPure(a []float32, b []float32) float32

func L2ImplPure

func L2ImplPure(a []float32, b []float32) float32

func ValidateFloatArrayLength

func ValidateFloatArrayLength(dim int64, length int) error

ValidateFloatArrayLength is used validate float vector length

func ValidateMetricType

func ValidateMetricType(metric string) (string, error)

ValidateMetricType returns metric text or error