How to convert []int8 to string

Question

What's the best way (fastest performance) to convert from []int8 to string?

For []byte we could do string(byteslice), but for []int8 it gives an error:

cannot convert ba (type []int8) to type string

I got the ba from SliceScan() method of *sqlx.Rows that produces []int8 instead of string

Is this solution the fastest?

func B2S(bs []int8) string {
    ba := []byte{}
    for _, b := range bs {
        ba = append(ba, byte(b))
    }
    return string(ba)
}

EDIT my bad, it's uint8 instead of int8.. so I can do string(ba) directly.

Answer 1

Note beforehand: The asker first stated that input slice is []int8 so that is what the answer is for. Later he realized the input is []uint8 which can be directly converted to string because byte is an alias for uint8 (and []byte => string conversion is supported by the language spec).

---

You can't convert slices of different types, you have to do it manually.

Question is what type of slice should we convert to? We have 2 candidates: []byte and []rune. Strings are stored as UTF-8 encoded byte sequences internally ([]byte), and a string can also be converted to a slice of runes. The language supports converting both of these types ([]byte and []rune) to string.

A rune is a unicode codepoint. And if we try to convert an int8 to a rune in a one-to-one fashion, it will fail (meaning wrong output) if the input contains characters which are encoded to multiple bytes (using UTF-8) because in this case multiple int8 values should end up in one rune.

Let's start from the string "世界" whose bytes are:

fmt.Println([]byte("世界"))
// Output: [228 184 150 231 149 140]

And its runes:

fmt.Println([]rune("世界"))
// [19990 30028]

It's only 2 runes and 6 bytes. So obviously 1-to-1 int8->rune mapping won't work, we have to go with 1-1 int8->byte mapping.

^{byte is alias for uint8 having range 0..255, to convert it to []int8 (having range -128..127) we have to use -256+bytevalue if the byte value is > 127 so the "世界" string in []int8 looks like this:}

[-28 -72 -106 -25 -107 -116]

^{The backward conversion what we want is: bytevalue = 256 + int8value if the int8 is negative but we can't do this as int8 (range -128..127) and neither as byte (range 0..255) so we also have to convert it to int first (and back to byte at the end). This could look something like this:}

if v < 0 {
    b[i] = byte(256 + int(v))
} else {
    b[i] = byte(v)
}

But actually since signed integers are represented using 2's complement, we get the same result if we simply use a byte(v) conversion (which in case of negative numbers this is equivalent to 256 + v).

Note: Since we know the length of the slice, it is much faster to allocate a slice with this length and just set its elements using indexing [] and not calling the built-in append function.

So here is the final conversion:

func B2S(bs []int8) string {
    b := make([]byte, len(bs))
    for i, v := range bs {
        b[i] = byte(v)
    }
    return string(b)
}

Try it on the Go Playground.

Answer 2

Not entirely sure it is the fastest, but I haven't found anything better. Change ba := []byte{} for ba := make([]byte,0, len(bs) so at the end you have:

func B2S(bs []int8) string {
    ba := make([]byte,0, len(bs))
    for _, b := range bs {
        ba = append(ba, byte(b))
    }
    return string(ba)
}

This way the append function will never try to insert more data that it can fit in the slice's underlying array and you will avoid unnecessary copying to a bigger array.