String prefix of requested length in golang working with utf-8 symbols

Is there some elegant way to crop string and create pretty string prefixes in golang? I have this function for start:

func prettyCrop(in string, cropLength int) string {
	if len(in) < cropLength {
		return in
	} else {
		in = in[0:cropLength]
		in = strings.TrimRightFunc(in, func(r rune) bool {
			if r == ' ' {
				return true
			}
			return false
		})
		return in + "…"
	}
}

It works good enough for english texts, but has problems with something more complicated. See this example:

prettyCrop("čřč čřč", 8) //čř?…

TrimRightFunc is not working as I expect here. I expect it to return čřč. Why is this function not returning valid utf-8 string? Is there a library for this? How can I fix it? Is there a better solution?

The problem is that slicing a string slices the UTF-8 encoded byte slice that represents the string, not the characters or runes of the string; this also means that if the string contains characters that are represented by multiple bytes in UTF-8 encoding, slicing / cutting the string may result in an invalid UTF-8 encoded sequence.

Assuming cropLength means to be a character limit (and not a byte-count limit), you should first convert the string to a []rune, and operate on that:

func prettyCrop(in string, cropLength int) string {
	in2 := []rune(in)
	if len(in2) < cropLength {
		return in
	} else {
		in2 = in2[:cropLength]
		in = strings.TrimRightFunc(string(in2), func(r rune) bool {
			if r == ' ' {
				return true
			}
			return false
		})
		return in + "…"
	}
}

Testing it:

for i := 0; i < 7; i++ {
	fmt.Println(prettyCrop("čřč čřč", i))
}

Output (try it on the Go Playground):

…
č…
čř…
čřč…
čřč…
čřč č…
čřč čř…

Performance notes:

The above example is not "performance" friendly, because:

• It converts the whole in string to []rune, it would be enough to just get its first cropLength runes with a for range.
• Calling strings.TrimRightFunc() requires to convert the []rune back to string, and then again a string concatenation is performed to generate the result. This could be avoided by manually looping over the []rune, and only create a single string that is returned.