column

package
v0.1.3 Latest Latest
Warning

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

 Go to latest Published: Sep 10, 2021 License: Apache-2.0 Imports: 13 Imported by: 0

Documentation

Overview

steps to add a new function: 1. add an implementation here and add it to FuncProj 2. specify its return types (return_types.go) 3. test both implementations above

Index

Constants

View Source 
const ALL_ONES = uint64(1<<64 - 1)
View Source 
const ALL_ZEROS = uint64(0)
View Source 
const DATETIME_BYTE_SIZE = 8
View Source 
const DATE_BYTE_SIZE = 4

Variables

View Source 
var FuncProj = map[string]func(...Chunk) (Chunk, error){
	"now":      evalNow,
	"version":  evalVersion,
	"nullif":   evalNullIf,
	"coalesce": evalCoalesce,
	"round":    evalRound,
	"sin":      numFunc(math.Sin),
	"cos":      numFunc(math.Cos),
	"tan":      numFunc(math.Tan),
	"asin":     numFunc(math.Asin),
	"acos":     numFunc(math.Acos),
	"atan":     numFunc(math.Atan),
	"sinh":     numFunc(math.Sinh),
	"cosh":     numFunc(math.Cosh),
	"tanh":     numFunc(math.Tanh),
	"sqrt":     numFunc(math.Sqrt),
	"exp":      numFunc(math.Exp),
	"exp2":     numFunc(math.Exp2),
	"log":      numFunc(math.Log),
	"log2":     numFunc(math.Log2),
	"log10":    numFunc(math.Log10),

	"trim":       stringFunc(strings.TrimSpace),
	"lower":      stringFunc(strings.ToLower),
	"upper":      stringFunc(strings.ToUpper),
	"left":       evalLeft,
	"split_part": evalSplitPart,
}

TODO: this will be hard to cover properly, so let's make sure we test everything explicitly ARCH: we're not treating literals any differently, but since they share the same backing store 

as non-literals, we're okay... is that okay?

Functions

func ChunksEqual 

func ChunksEqual(c1 Chunk, c2 Chunk) bool

ChunksEqual compares two chunks, even if they contain []float64 data consider making this lenient enough to compare only the relevant bits in ChunkBools

func DatesEqual 

func DatesEqual(a, b date) bool

func DatesGreaterThan 

func DatesGreaterThan(a, b date) bool

func DatesGreaterThanEqual 

func DatesGreaterThanEqual(a, b date) bool

func DatesLessThan 

func DatesLessThan(a, b date) bool

func DatesLessThanEqual 

func DatesLessThanEqual(a, b date) bool

func DatesNotEqual 

func DatesNotEqual(a, b date) bool

func DatetimesEqual 

func DatetimesEqual(a, b datetime) bool

func DatetimesGreaterThan 

func DatetimesGreaterThan(a, b datetime) bool

func DatetimesGreaterThanEqual 

func DatetimesGreaterThanEqual(a, b datetime) bool

func DatetimesLessThan 

func DatetimesLessThan(a, b datetime) bool

func DatetimesLessThanEqual 

func DatetimesLessThanEqual(a, b datetime) bool

func DatetimesNotEqual 

func DatetimesNotEqual(a, b datetime) bool

func NewAggregator 

func NewAggregator(function string, distinct bool) (func(...Dtype) (*AggState, error), error)

NewAggregator implements a constructor for various aggregating functions. We got inspired by Postgres' functions https://www.postgresql.org/docs/12/functions-aggregate.html

  • not implemented: xml/json functions (don't have the data types), array_agg (no arrays), every (just an alias), bit_and/bit_or (doesn't seem useful for us)
  • implemented: min, max, sum, avg, count
  • planned: bool_and, bool_or, string_agg
  • thinking: countDistinct, sketch-based approxCountDistinct

ARCH: function string -> uint8 const? dtypes are types of inputs - rename? TODO: check for function existence OPTIM: the switch(function) could be hoisted outside the closure (would work as a function existence validator)

Types

type AggState 

type AggState struct {
	AddChunk func(buckets []uint64, ndistinct int, data Chunk)
	Resolve  func() (Chunk, error)
	// contains filtered or unexported fields
}

type Chunk 

type Chunk interface {
	Dtype() Dtype
	AddValue(string) error
	AddValues([]string) error // consider merging AddValues and AddValue (using varargs)
	WriteTo(io.Writer) (int64, error)
	Prune(*bitmap.Bitmap) Chunk
	Append(Chunk) error
	Hash(int, []uint64)
	Clone() Chunk
	JSONLiteral(int) (string, bool) // the bool stands for 'ok' (not null)
	Compare(bool, bool, int, int) int
	// contains filtered or unexported methods
}

Chunk defines a part of a column - constant type, stored contiguously

func Deserialize 

func Deserialize(r io.Reader, Dtype Dtype) (Chunk, error)

Deserialize reads a chunk from a reader this shouldn't really accept a Dtype - at this point we're requiring it, because we don't serialize Dtypes into the binary representation - but that's just because we always have the schema at hand... but will we always have it? shouldn't the files be readable as standalone files? OPTIM: shouldn't we deserialize based on a byte slice instead? We already have it, so we're just duplicating it using a byte buffer OPTIM: we may be able to safely cast these byte slice in the future - see https://github.com/golang/go/issues/19367

func EvalAdd 

func EvalAdd(c1 Chunk, c2 Chunk) (Chunk, error)

a solid case for generics?

func EvalAnd 

func EvalAnd(c1 Chunk, c2 Chunk) (Chunk, error)

EvalAnd produces a bitwise operation on two bool chunks

func EvalDivide 

func EvalDivide(c1 Chunk, c2 Chunk) (Chunk, error)

different return type for ints! should we perhaps cast to make this more systematic? check for division by zero (gives +- infty, which will break json?)

func EvalEq 

func EvalEq(c1 Chunk, c2 Chunk) (Chunk, error)

EvalEq compares values from two different chunks

func EvalGt 

func EvalGt(c1 Chunk, c2 Chunk) (Chunk, error)

EvalGt checks if values in c1 are greater than in c2

func EvalGte 

func EvalGte(c1 Chunk, c2 Chunk) (Chunk, error)

EvalGte checks if values in c1 are greater than or equal to those in c2

func EvalLt 

func EvalLt(c1 Chunk, c2 Chunk) (Chunk, error)

EvalLt checks if values in c1 are lower than in c2

func EvalLte 

func EvalLte(c1 Chunk, c2 Chunk) (Chunk, error)

EvalLte checks if values in c1 are lower than or equal to those in c2

func EvalMultiply 

func EvalMultiply(c1 Chunk, c2 Chunk) (Chunk, error)

func EvalNeq 

func EvalNeq(c1 Chunk, c2 Chunk) (Chunk, error)

EvalNeq compares values from two different chunks for inequality

func EvalNot 

func EvalNot(c Chunk) (Chunk, error)

func EvalOr 

func EvalOr(c1 Chunk, c2 Chunk) (Chunk, error)

EvalOr produces a bitwise operation on two bool chunks

func EvalSubtract 

func EvalSubtract(c1 Chunk, c2 Chunk) (Chunk, error)

func NewChunkFromSchema 

func NewChunkFromSchema(schema Schema) Chunk

NewChunkFromSchema creates a new Chunk based a column schema provided

func NewChunkLiteralAuto 

func NewChunkLiteralAuto(s string, length int) (Chunk, error)

NewChunkLiteral creates a chunk that only contains a single value in the whole chunk it's useful in e.g. 'foo > 1', where can convert the '1' to a whole chunk OPTIM: we're using single-value slices, should we perhaps have a value specific for each literal to avoid working with slices (stack allocation etc.)

func NewChunkLiteralTyped 

func NewChunkLiteralTyped(s string, dtype Dtype, length int) (Chunk, error)

type ChunkBools 

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

ChunkBools defines a backing struct for a chunk of boolean values

func NewChunkBoolsFromBitmap 

func NewChunkBoolsFromBitmap(bm *bitmap.Bitmap) *ChunkBools

NewChunkBoolsFromBitmap creates a new bool chunk, but in doing so, doesn't clone the incoming bitmap, it uses it as is - the caller might want to clone it aims to mutate it in the future

func NewChunkLiteralBools 

func NewChunkLiteralBools(value bool, length int) *ChunkBools

func (*ChunkBools) AddValue 

func (rc *ChunkBools) AddValue(s string) error

AddValue takes in a string representation of a value and converts it into a value suited for this chunk

func (*ChunkBools) AddValues 

func (rc *ChunkBools) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkBools) Append 

func (rc *ChunkBools) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkBools) Base 

func (bc *ChunkBools) Base() *baseChunk

func (*ChunkBools) Clone 

func (rc *ChunkBools) Clone() Chunk

func (*ChunkBools) Compare 

func (rc *ChunkBools) Compare(asc, nullsFirst bool, i, j int) int

func (*ChunkBools) Dtype 

func (rc *ChunkBools) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkBools) Hash 

func (rc *ChunkBools) Hash(position int, hashes []uint64)

Hash hashes this chunk's values into a provded container

func (*ChunkBools) JSONLiteral 

func (rc *ChunkBools) JSONLiteral(n int) (string, bool)

func (*ChunkBools) Len 

func (bc *ChunkBools) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkBools) Nullify 

func (bc *ChunkBools) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkBools) Prune 

func (rc *ChunkBools) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkBools) Truths 

func (rc *ChunkBools) Truths() *bitmap.Bitmap

Truths returns only true values in this boolean column's bitmap - remove those that are null - we use this for filtering, when we're interested in non-null true values (to select given rows)

func (*ChunkBools) WriteTo 

func (rc *ChunkBools) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type ChunkDates 

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

ChunkDates defines a backing struct for a chunk of date values

func NewChunkLiteralDates 

func NewChunkLiteralDates(value date, length int) *ChunkDates

func (*ChunkDates) AddValue 

func (rc *ChunkDates) AddValue(s string) error

func (*ChunkDates) AddValues 

func (rc *ChunkDates) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkDates) Append 

func (rc *ChunkDates) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkDates) Base 

func (bc *ChunkDates) Base() *baseChunk

func (*ChunkDates) Clone 

func (rc *ChunkDates) Clone() Chunk

func (*ChunkDates) Compare 

func (rc *ChunkDates) Compare(asc, nullsFirst bool, i, j int) int

func (*ChunkDates) Dtype 

func (rc *ChunkDates) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkDates) Hash 

func (rc *ChunkDates) Hash(position int, hashes []uint64)

func (*ChunkDates) JSONLiteral 

func (rc *ChunkDates) JSONLiteral(n int) (string, bool)

func (*ChunkDates) Len 

func (bc *ChunkDates) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkDates) Nullify 

func (bc *ChunkDates) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkDates) Prune 

func (rc *ChunkDates) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkDates) WriteTo 

func (rc *ChunkDates) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type ChunkDatetimes 

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

func NewChunkLiteralDatetimes added in v0.1.3 

func NewChunkLiteralDatetimes(value datetime, length int) *ChunkDatetimes

func (*ChunkDatetimes) AddValue 

func (rc *ChunkDatetimes) AddValue(s string) error

func (*ChunkDatetimes) AddValues 

func (rc *ChunkDatetimes) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkDatetimes) Append 

func (rc *ChunkDatetimes) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkDatetimes) Base 

func (bc *ChunkDatetimes) Base() *baseChunk

func (*ChunkDatetimes) Clone 

func (rc *ChunkDatetimes) Clone() Chunk

func (*ChunkDatetimes) Compare 

func (rc *ChunkDatetimes) Compare(asc, nullsFirst bool, i, j int) int

func (*ChunkDatetimes) Dtype 

func (rc *ChunkDatetimes) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkDatetimes) Hash 

func (rc *ChunkDatetimes) Hash(position int, hashes []uint64)

func (*ChunkDatetimes) JSONLiteral 

func (rc *ChunkDatetimes) JSONLiteral(n int) (string, bool)

func (*ChunkDatetimes) Len 

func (bc *ChunkDatetimes) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkDatetimes) Nullify 

func (bc *ChunkDatetimes) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkDatetimes) Prune 

func (rc *ChunkDatetimes) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkDatetimes) WriteTo 

func (rc *ChunkDatetimes) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type ChunkFloats 

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

ChunkFloats defines a backing struct for a chunk of floating point values

func NewChunkFloatsFromSlice 

func NewChunkFloatsFromSlice(data []float64, nulls *bitmap.Bitmap) *ChunkFloats

func NewChunkLiteralFloats 

func NewChunkLiteralFloats(value float64, length int) *ChunkFloats

func (*ChunkFloats) AddValue 

func (rc *ChunkFloats) AddValue(s string) error

AddValue takes in a string representation of a value and converts it into a value suited for this chunk let's really consider adding standard nulls here, it will probably make our lives a lot easier

func (*ChunkFloats) AddValues 

func (rc *ChunkFloats) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkFloats) Append 

func (rc *ChunkFloats) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkFloats) Base 

func (bc *ChunkFloats) Base() *baseChunk

func (*ChunkFloats) Clone 

func (rc *ChunkFloats) Clone() Chunk

func (*ChunkFloats) Compare 

func (rc *ChunkFloats) Compare(asc, nullsFirst bool, i, j int) int

func (*ChunkFloats) Dtype 

func (rc *ChunkFloats) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkFloats) Hash 

func (rc *ChunkFloats) Hash(position int, hashes []uint64)

Hash hashes this chunk's values into a provded container OPTIM/TODO(next): do we need a fnv hasher for ints/floats/dates? We can just take the uint64 representation of these values... or not?

func (*ChunkFloats) JSONLiteral 

func (rc *ChunkFloats) JSONLiteral(n int) (string, bool)

func (*ChunkFloats) Len 

func (bc *ChunkFloats) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkFloats) Nullify 

func (bc *ChunkFloats) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkFloats) Prune 

func (rc *ChunkFloats) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkFloats) WriteTo 

func (rc *ChunkFloats) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type ChunkInts 

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

ChunkInts defines a backing struct for a chunk of integer values

func NewChunkIntsFromSlice 

func NewChunkIntsFromSlice(data []int64, nulls *bitmap.Bitmap) *ChunkInts

the next few functions could use some generics

func NewChunkLiteralInts 

func NewChunkLiteralInts(value int64, length int) *ChunkInts

func (*ChunkInts) AddValue 

func (rc *ChunkInts) AddValue(s string) error

AddValue takes in a string representation of a value and converts it into a value suited for this chunk

func (*ChunkInts) AddValues 

func (rc *ChunkInts) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkInts) Append 

func (rc *ChunkInts) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkInts) Base 

func (bc *ChunkInts) Base() *baseChunk

func (*ChunkInts) Clone 

func (rc *ChunkInts) Clone() Chunk

func (*ChunkInts) Compare 

func (rc *ChunkInts) Compare(asc, nullsFirst bool, i, j int) int

ARCH: this could be made entirely generic by allowing an interface `nthValue(int) T` to genericise v1/v2 

EXCEPT for bools :-( (not comparable)

func (*ChunkInts) Dtype 

func (rc *ChunkInts) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkInts) Hash 

func (rc *ChunkInts) Hash(position int, hashes []uint64)

Hash hashes this chunk's values into a provded container OPTIM: maphash might be faster than fnv or maphash? test it and if it is so, implement everywhere, but be careful about the seed (needs to be the same for all chunks) careful about maphash: "The hash value of a given byte sequence is consistent within a single process, but will be different in different processes." oh and I rebenchmarked maphash and fnv and found maphash to be much slower (despite no allocs) also, check this https://github.com/segmentio/fasthash/ (via https://segment.com/blog/allocation-efficiency-in-high-performance-go-services/) they reimplement fnv using stack allocation only

  • we tested it and got a 90% speedup (no allocs, shorter code) - so let's consider it, it's in the fasthash branch

func (*ChunkInts) JSONLiteral 

func (rc *ChunkInts) JSONLiteral(n int) (string, bool)

func (*ChunkInts) Len 

func (bc *ChunkInts) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkInts) Nullify 

func (bc *ChunkInts) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkInts) Prune 

func (rc *ChunkInts) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkInts) WriteTo 

func (rc *ChunkInts) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type ChunkNulls 

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

ChunkNulls defines a backing struct for a chunk of null values Since it's all nulls, we only need to know how many there are

func (*ChunkNulls) AddValue 

func (rc *ChunkNulls) AddValue(s string) error

AddValue takes in a string representation of a value and converts it into a value suited for this chunk

func (*ChunkNulls) AddValues 

func (rc *ChunkNulls) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkNulls) Append 

func (rc *ChunkNulls) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkNulls) Base 

func (bc *ChunkNulls) Base() *baseChunk

func (*ChunkNulls) Clone 

func (rc *ChunkNulls) Clone() Chunk

func (*ChunkNulls) Compare 

func (rc *ChunkNulls) Compare(asc, nullsFirst bool, i, j int) int

func (*ChunkNulls) Dtype 

func (rc *ChunkNulls) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkNulls) Hash 

func (rc *ChunkNulls) Hash(position int, hashes []uint64)

Hash hashes this chunk's values into a provded container

func (*ChunkNulls) JSONLiteral 

func (rc *ChunkNulls) JSONLiteral(n int) (string, bool)

func (*ChunkNulls) Len 

func (bc *ChunkNulls) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkNulls) Nullify 

func (bc *ChunkNulls) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkNulls) Prune 

func (rc *ChunkNulls) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkNulls) WriteTo 

func (rc *ChunkNulls) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type ChunkStrings 

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

ChunkStrings defines a backing struct for a chunk of string values

func NewChunkLiteralStrings 

func NewChunkLiteralStrings(value string, length int) *ChunkStrings

func (*ChunkStrings) AddValue 

func (rc *ChunkStrings) AddValue(s string) error

AddValue takes in a string representation of a value and converts it into a value suited for this chunk

func (*ChunkStrings) AddValues 

func (rc *ChunkStrings) AddValues(vals []string) error

AddValues is a helper method, it just calls AddValue repeatedly

func (*ChunkStrings) Append 

func (rc *ChunkStrings) Append(tc Chunk) error

Append adds a chunk of the same type at the end of this one (in place update)

func (*ChunkStrings) Base 

func (bc *ChunkStrings) Base() *baseChunk

func (*ChunkStrings) Clone 

func (rc *ChunkStrings) Clone() Chunk

func (*ChunkStrings) Compare 

func (rc *ChunkStrings) Compare(asc, nullsFirst bool, i, j int) int

func (*ChunkStrings) Dtype 

func (rc *ChunkStrings) Dtype() Dtype

Dtype returns the type of this chunk

func (*ChunkStrings) Hash 

func (rc *ChunkStrings) Hash(position int, hashes []uint64)

Hash hashes this chunk's values into a provded container

func (*ChunkStrings) JSONLiteral 

func (rc *ChunkStrings) JSONLiteral(n int) (string, bool)

func (*ChunkStrings) Len 

func (bc *ChunkStrings) Len() int

ARCH: we sometimes use this, sometimes we access the struct field directly... perhaps remove this?

func (*ChunkStrings) Nullify 

func (bc *ChunkStrings) Nullify(bm *bitmap.Bitmap)

ARCH: Nullify does NOT switch the data values to be nulls/empty as well

func (*ChunkStrings) Prune 

func (rc *ChunkStrings) Prune(bm *bitmap.Bitmap) Chunk

Prune filter this chunk and only preserves values for which the bitmap is set

func (*ChunkStrings) WriteTo 

func (rc *ChunkStrings) WriteTo(w io.Writer) (int64, error)

WriteTo converts a chunk into its binary representation

type Dtype 

type Dtype uint8

Dtype denotes the data type of a given object (e.g. int or string) 

const (
	DtypeInvalid Dtype = iota
	DtypeNull
	DtypeString
	DtypeInt
	DtypeFloat
	DtypeBool
	DtypeDate
	DtypeDatetime
	// more to be added
	DtypeMax
)

individual dtypes defined as a sequence

func (Dtype) MarshalJSON 

func (dt Dtype) MarshalJSON() ([]byte, error)

MarshalJSON returns the JSON representation of a dtype (stringified + json string) we want Dtypes to be marshaled within Schema correctly without this they'd be returned as an integer (even with ",string" tags)

func (Dtype) String 

func (dt Dtype) String() string

func (*Dtype) UnmarshalJSON 

func (dt *Dtype) UnmarshalJSON(data []byte) error

UnmarshalJSON deserialises a given dtype from a JSON value

type Schema 

type Schema struct {
	Name     string `json:"name"`
	Dtype    Dtype  `json:"dtype"`
	Nullable bool   `json:"nullable"`
}

Schema defines all the necessary properties of column

type TableSchema 

type TableSchema []Schema

TableSchema is a collection of column schemas

func (*TableSchema) LocateColumn 

func (schema *TableSchema) LocateColumn(s string) (int, Schema, error)

LocateColumn returns a column within a schema - its position and definition; error is triggered if this column is not found or the schema is nil

func (*TableSchema) LocateColumnCaseInsensitive 

func (schema *TableSchema) LocateColumnCaseInsensitive(s string) (int, Schema, error)

LocateColumnCaseInsensitive works just like LocateColumn, but it ignores casing ARCH: we could have used strings.EqualFold, but a) we have one static input (s), so we can 

      amortise the case lowering, b) the extra correctness in EqualFold is irrelevant here,
		 because of our column naming restrictions

type TypeGuesser 

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

TypeGuesser contains state necessary for inferring types from a stream of strings

func NewTypeGuesser 

func NewTypeGuesser() *TypeGuesser

NewTypeGuesser creates a new type guesser

func (*TypeGuesser) AddValue 

func (tg *TypeGuesser) AddValue(s string)

AddValue feeds a new value to a type guesser

func (*TypeGuesser) InferredType 

func (tg *TypeGuesser) InferredType() Schema

InferredType returns the best guess of a type for a given stream of strings

Source Files

View all Source files

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.