I have a list of strings that I want to split into...

# stdlib

[Sam Stone]

I have a list of strings that I want to split into sublists, where the delimiter is a predicate lambda (string matches a certain property). How can I do this?

[jw]

partition or groupBy and then values

[ephemient]

that doesn't exactly work (without hacks) if what Sam wants is contiguous sublists

[ephemient]

assuming that, https://youtrack.jetbrains.com/issue/KT-8280/Group-consecutive-elements-in-an-IterableSequence would help

[ephemient]

the hack could be something like this fwiw

fun <T> Iterable<T>.splitWhen(predicate: (T) -> Boolean): List<List<T>> {
    var counter = 0
    return groupBy { if (predicate(it)) ++counter else counter }.values.map { it.dropWhile(predicate) }
}

listOf(1, 2, 0, 3, 0, 0, 4).splitWhen { it == 0 } // [[1, 2], [3], [], [4]]

[Sam Stone]

Seems O(N^3) or so. Adapted this from SO (~O(N^2)):

fun <E> List<E>.split(includeDelimiter: Boolean = false, predicate: (E) -> Boolean): List<List<E>> {
        return flatMapIndexed { index, x ->
                when {
                    index == 0 || index == lastIndex -> listOf(index)
                    predicate(x) -> listOf(index - 1, index + 1)
                    else -> emptyList()
                }
            }
            .windowed(size = 2, step = 2) { (from, to) -> slice( (if(includeDelimiter) (from - 1).coerceAtLeast(0) else from)..to) }
    }

[ephemient]

what makes you think it's O(N^3)/O(N^2)? they're both O(N)

[ephemient]

if you care about minimizing garbage then you should choose something like

fun <E> Iterable<E>.split(predicate: (E) -> Boolean): List<List<E>> = buildList {
    val last = [e-mail](mutableListOf<E>()) { acc, elem ->
        if (predicate(elem)) {
            add(acc.orEmpty())
            mutableListOf()
        } else {
            acc.add(elem)
            acc
        }
    }
    if (last.isNotEmpty()) add(last)
}

over either of the above

[Sam Stone]

3 loops: groupBy,map, dropWhile.

[ephemient]

so? groupBy iterates N elements once. map iterates the groups and dropWhile iterates over at most 1 element of each group, and even if it iterated the whole group, it would still sum to at most N total.