How to divide String by regex? e.g. `"AAA{foo}BBB{...

# getting-started

[ken_kentan]

How to divide String by regex?
e.g. "AAA{foo}BBB{bar}CCC" to ["AAA", "{foo}", "BBB", "{bar}", "CCC"] if regex is \{.*?}

[Tobias Berger]

A normal split operation won't do becasue the separator is usually not included in the result. you could do something like this for your specific case:

fun customSplit(input: String): List<String> {
    return "([^{]+)(\\{.*?})?".toRegex().findAll(input).flatMap { match ->
        sequence {
            yield(match.groupValues[1])
            match.groupValues[2].takeIf { it.isNotBlank() }?.let { yield(it) }
        }
    }.toList()
}

[ephemient]

zero-width lookahead/lookbehind:

"AAA{foo}BBB{bar}CCC".split("""(?=\{)|(?<=})""".toRegex())

(not all JS RegExp engines support lookbehind though)

[Tobias Berger]

good idea, didn't think of this. I don't know the case where this will be used but it is probaby sufficient, although it will just split before each { and after each }.

[ephemient]

https://caniuse.com/js-regexp-lookbehind

[ephemient]

yeah, both our solutions have some edge cases on different inputs

[ephemient]

(what should happen on {{{foo}? yours drops the leading {, mine separates them)

[ken_kentan]

It seems difficult to support some edge case like AAA{foo}BBB{bar}CCC}DDD{EEE etc..

[Tobias Berger]

sure, mine isn't perfect at all, it was just my first idea. Both approaches require some regex-skills if you want to customize

[Tobias Berger]

Here is a different approach that takes any regex

fun customSplit(input: String, regex: Regex): List<String> {
    var nextSliceIndex = 0
    val result = regex.findAll(input).fold(mutableListOf<String>()) { result, match ->
        val range = match.range
        result += input.slice(nextSliceIndex until range.first)
        result += input.slice(range)
        nextSliceIndex = range.last + 1
        result
    }
    if (nextSliceIndex < input.length) {
        result += input.substring(nextSliceIndex)
    }
    return result
}

[stephanmg]

@ephemient you can use capture groups to handle your edge case

[ken_kentan]

I will try to use Tobias's approach! Thank you everyone!