Convert camel case to snake case in Go http://play.golang.org/p/9ybqJat1Hr

ToSnakeCase.go

import (
	"fmt"
	"strings"
	"regexp"
)

var matchFirstCap = regexp.MustCompile("(.)([A-Z][a-z]+)")
var matchAllCap   = regexp.MustCompile("([a-z0-9])([A-Z])")

func ToSnakeCase(str string) string {
	snake := matchFirstCap.ReplaceAllString(str, "${1}_${2}")
	snake  = matchAllCap.ReplaceAllString(snake, "${1}_${2}")
	return strings.ToLower(snake)
}

Fails for the input already_Snake

Fails for the input already_Snake

Changing the first regex to ([A-Z])([A-Z][a-z]) seems to solve the issue:

import (
	"fmt"
	"strings"
	"regexp"
)

var matchFirstCap = regexp.MustCompile("([A-Z])([A-Z][a-z])")
var matchAllCap   = regexp.MustCompile("([a-z0-9])([A-Z])")

func ToSnakeCase(str string) string {
	snake := matchFirstCap.ReplaceAllString(str, "${1}_${2}")
	snake  = matchAllCap.ReplaceAllString(snake, "${1}_${2}")
	return strings.ToLower(snake)
}

After changing to our projects guidelines and requirements, I have following function:

var matchFirstCap = regexp.MustCompile("([A-Z])([A-Z][a-z])")
var matchAllCap = regexp.MustCompile("([a-z0-9])([A-Z])")

// ToSnakeCase converts the provided string to snake_case.
// Based on https://gist.github.com/stoewer/fbe273b711e6a06315d19552dd4d33e6
func ToSnakeCase(input string) string {
	output := matchFirstCap.ReplaceAllString(input, "${1}_${2}")
	output = matchAllCap.ReplaceAllString(output, "${1}_${2}")
	output = strings.ReplaceAll(output, "-", "_")
	return strings.ToLower(output)
}

This function passes following (extended) test:

import (
	"testing"
	"github.com/stretchr/testify/assert"
)

func Test_ToSnakeCase(t *testing.T) {
	// Based on https://gist.github.com/stoewer/fbe273b711e6a06315d19552dd4d33e6
	tests := []struct {
		input    string
		expected string
	}{
		{"", ""},
		{"camelCase", "camel_case"},
		{"PascalCase", "pascal_case"},
		{"snake_case", "snake_case"},
		{"Pascal_Snake", "pascal_snake"},
		{"SCREAMING_SNAKE", "screaming_snake"},
		{"kebab-case", "kebab_case"},
		{"Pascal-Kebab", "pascal_kebab"},
		{"SCREAMING-KEBAB", "screaming_kebab"},
		{"A", "a"},
		{"AA", "aa"},
		{"AAA", "aaa"},
		{"AAAA", "aaaa"},
		{"AaAa", "aa_aa"},
		{"HTTPRequest", "http_request"},
		{"BatteryLifeValue", "battery_life_value"},
		{"Id0Value", "id0_value"},
		{"ID0Value", "id0_value"},
	}

	for _, test := range tests {
		result := ToSnakeCase(test.input)
		assert.Equal(t, test.expected, result)
	}
}

Jen's function is awesome and covers a lot of different cases. But using regexp when you just need to convert uppercase letters into '_' plus lowercase letter - is overkill.

Here is very simple and much faster solution:

func toSnakeCase(s string) string {
   var res = make([]rune, 0, len(s))
   for i, r := range s {
       if unicode.IsUpper(r) && i > 0 {
           res = append(res, '_', unicode.ToLower(r))
       } else {
           res = append(res, unicode.ToLower(r))
       }
   }
   return string(res)
}

And here is benchmark for Jen's variant and mine:

BenchmarkRegexp-8   	   92520	     11980 ns/op	    1061 B/op	      59 allocs/op
BenchmarkRange-8    	  609658	      1807 ns/op	     744 B/op	      20 allocs/op

6 times faster and 3 times less allocations!

@porfirion thanks for the proposal and comparison! I noticed most of my unit test fail when using your version. Therefore, I tried to make a version which succeeds (nearly) all tests. It can still use some cleanup and unit test for HTTPRequest still fails, but nearly same performance improvements.

Code

func ToSnakeCase(s string) string {
	var res = make([]rune, 0, len(s))
	var p = '_'
	for i, r := range s {
		if !unicode.IsLetter(r) && !unicode.IsDigit(r) {
			res = append(res, '_')
		} else if unicode.IsUpper(r) && i > 0 {
			if unicode.IsLetter(p) && !unicode.IsUpper(p) || unicode.IsDigit(p) {
				res = append(res, '_', unicode.ToLower(r))
			} else {
				res = append(res, unicode.ToLower(r))
			}
		} else {
			res = append(res, unicode.ToLower(r))
		}

		p = r
	}
	return string(res)
}

Benchmarks

func BenchmarkToSnakeCaseRegex(b *testing.B) {
	for i := 0; i < b.N; i++ {
		ToSnakeCaseRegex("BatteryLifeValue")
	}
}

func BenchmarkToSnakeCase(b *testing.B) {
	for i := 0; i < b.N; i++ {
		ToSnakeCase("BatteryLifeValue")
	}
}

Results

BenchmarkToSnakeCaseRegex-12    	  360540	      3282 ns/op	     282 B/op	      11 allocs/op
BenchmarkToSnakeCase-12         	 1726478	       714 ns/op	     224 B/op	       3 allocs/op

@JensSkipr @porfirion

Hi, I make one more, but it pass all test, and better performance.

source file in here: https://gist.github.com/hxsf/7f5392c0153d3a8607c42eefed02b8cd

Bench

> go test -benchmem -bench 'One'
goos: darwin
goarch: amd64
pkg: mesh-sidecar/cmd/vars
cpu: Intel(R) Core(TM) i7-7700 CPU @ 3.60GHz
BenchmarkOneCase/ToSnakeCaseRegex-8               490156              2403 ns/op             305 B/op         12 allocs/op
BenchmarkOneCase/ToSnakeCaseByJensSkipr-8        4163742               286.1 ns/op           216 B/op          3 allocs/op
BenchmarkOneCase/ToSnakeCase-8                   8867581               131.4 ns/op            24 B/op          1 allocs/op

BenchmarkAllInOne/ToSnakeCaseRegex-8               51538             22480 ns/op            2307 B/op        137 allocs/op
BenchmarkAllInOne/ToSnakeCaseByJensSkipr-8        395026              3051 ns/op            1352 B/op         41 allocs/op
BenchmarkAllInOne/ToSnakeCase-8                   857940              1491 ns/op             232 B/op         17 allocs/op
PASS
ok      mesh-sidecar/cmd/vars   7.930s

@hxsf looks awesome! Yeah, I forgot about additional capacity for underscores %)

@hxsf: That looks great! Do you mind adding a license to your gist?

@adombeck add MIT license.

@hxsf: Great, thanks a lot!

@hxsf Test failed for

MyLIFEIsAwesomE should be my_life_is_awesom_e but got mylife_is_awesome
Japan125Canada130Australia150 should be japan125_canada130_australia150 but got japan1_2_5_canada1_3_0_australia1_5_0

@elvizlai I update the code. and it works

good guys ! keep it up!
kebab case should be concluded here as well

@abdennour just replace '_' with '-' in the code, and change the function name. I think it will work.