Convert a float64 to an int in Go

package main
import "fmt"
func main() {
  var x float64 = 5.7
  var y int = int(x)
  fmt.Println(y)  // outputs "5"
}

Simply casting to an int truncates the float, which if your system internally represent 2.0 as 1.9999999999, you will not get what you expect. The various printf conversions deal with this and properly round the number when converting. So to get a more accurate value, the conversion is even more complicated than you might first expect:

package main

import (
    "fmt"
    "strconv"
)

func main() {
    floats := []float64{1.9999, 2.0001, 2.0}
    for _, f := range floats {
        t := int(f)
        s := fmt.Sprintf("%.0f", f)
        if i, err := strconv.Atoi(s); err == nil {
            fmt.Println(f, t, i)
        } else {
            fmt.Println(f, t, err)
        }
    }
}

Code on Go Playground

If its simply from float64 to int, this should work

package main

import (
    "fmt"
)

func main() {
    nf := []float64{-1.9999, -2.0001, -2.0, 0, 1.9999, 2.0001, 2.0}

    //round
    fmt.Printf("Round : ")
    for _, f := range nf {
        fmt.Printf("%d ", round(f))
    }
    fmt.Printf("\n")

    //rounddown ie. math.floor
    fmt.Printf("RoundD: ")
    for _, f := range nf {
        fmt.Printf("%d ", roundD(f))
    }
    fmt.Printf("\n")

    //roundup ie. math.ceil
    fmt.Printf("RoundU: ")
    for _, f := range nf {
        fmt.Printf("%d ", roundU(f)) 
    }
    fmt.Printf("\n")

}

func roundU(val float64) int {
    if val > 0 { return int(val+1.0) }
    return int(val)
}

func roundD(val float64) int {
    if val < 0 { return int(val-1.0) }
    return int(val)
}

func round(val float64) int {
    if val < 0 { return int(val-0.5) }
    return int(val+0.5)
}

Outputs:

Round : -2 -2 -2 0 2 2 2 
RoundD: -2 -3 -3 0 1 2 2 
RoundU: -1 -2 -2 0 2 3 3

Here's the code in the playground - https://play.golang.org/p/HmFfM6Grqh

Correct rounding is likely desired.

Therefore math.Round() is your quick(!) friend. Approaches with fmt.Sprintf and strconv.Atois() were 2 orders of magnitude slower according to my tests with a matrix of float64 values that were intended to become correctly rounded int values.

package main
import (
    "fmt"
    "math"
)
func main() {
    var x float64 = 5.51
    var y float64 = 5.50
    var z float64 = 5.49
    fmt.Println(int(math.Round(x)))  // outputs "6"
    fmt.Println(int(math.Round(y)))  // outputs "6"
    fmt.Println(int(math.Round(z)))  // outputs "5"
}

math.Round() does return a float64 value but with int() applied afterwards, I couldn't find any mismatches so far.

Convert a float64 to an int in Go

Related

Recent Posts