Taking sequence elements fulfilling a predicate then continuing from there in Kotlin - kotlin

In Kotlin sequences have a takeWhile function that will let you take items as long as they adhere to a given predicate. What I'd like to do is take items according to that predicate, use them in some way, then alter the predicate and take the next "batch". So far I haven't really found a way of doing this purely with what sequences and iterators offer.
Following snippet of code illustrates the problem. The primeGenerator() function returns a Sequence of prime (Long) numbers. Suppose that I want to make lists with each list having prime numbers with the same number of digits. On creating each list I'd use it for some purpose. If the list conforms to what I was searching the iteration can end, otherwise move onto the next list.
val primeIt = primeGenerator().iterator()
var digits = 1
var next: Long? = null
val currentList = ArrayList<Long>()
while (digits < 4) {
next?.also { currentList.add(it) }
next = primeIt.next()
if (next.toString().length > digits) {
println("Primes with $digits: $currentList")
currentList.clear()
digits++
}
}
In this case it ends once the number of digits exceeds 3. This works fine, but I was wondering if there is some way to achieve the same with operations chained purely on the sequence or an iterator of it. Basically chunking the sequence but based on a predicate rather than a set size. The prime number example above is just for illustration, I'm after the general principle, not something that'd only work for this case.

There are no such functions in standard library for large (or infinite) sequences, but you may write such function by yourself (although it requires some extra code):
class BufferedIterator<T>(private val iterator: Iterator<T>) : Iterator<T> {
var current: T? = null
private set
var reachedEnd: Boolean = false
private set
override fun hasNext(): Boolean = iterator.hasNext().also { reachedEnd = !it }
override fun next(): T = iterator.next().also { current = it }
}
fun <T> Iterator<T>.buffered() = BufferedIterator(this)
fun <T> BufferedIterator<T>.takeWhile(predicate: (T) -> Boolean): List<T> {
val list = ArrayList<T>()
if (reachedEnd) return list
current?.let {
if (predicate(it)) list += it
}
while (hasNext()) {
val next = next()
if (predicate(next)) list += next
else break
}
return list
}
fun main() {
val sequence = sequence {
var next = 0
while (true) {
yield(next++)
}
}
val iter = sequence.iterator().buffered()
for (i in 0..3) {
println(iter.takeWhile { it.toString().length <= i })
}
}
With this approach you can easily work even with infinite sequences.

I believe there is a way to accomplish what you want using the standard library. Limit the sequence first and then groupBy the number of digits.
val Int.numberOfDigits
get() = this.toString().length
sequenceOf(1,22,333).takeWhile{ it.numberOfDigits < 3 }.groupBy{ it.numberOfDigits }.values
If you want to avoid the eager evaluation of groupBy you could use groupingBy instead and then reduce potentially leaving the accumulator blank.

ardenit's answer seems like the best reusable approach. Since taking "chunks" of a sequence requires some state it doesn't seem likely something easily done in a purely functional manner. Delegating the state to a separate class enveloping the sequence makes sense.
Here's a small snippet showing what I ended up using. This assumes the sequence will not be empty and is (technically) infinite or further results aren't requested at some point.
class ChunkedIterator<T>(seq: Sequence<T>) {
private val it = seq.iterator()
var next: T = it.next()
fun next(predicate: (T) -> Boolean): List<T> {
val result = ArrayList<T>();
while (predicate.invoke(next)) {
result.add(next)
next = it.next();
}
return result
}
}

one way you could achieve this is by getting an iterator from your your original sequence and then building a new sequence out of it for each "take" -
val itr = seq.iterator()
val batch1 = itr.asSequence().takeWhile { predicate1(it) }.toList()
val batch2 = itr.asSequence().takeWhile { predicate2(it) }.toList()

Related

Mapping array elements using a function in Kotlin

New to Kotlin from Python. In Python, I can simply use the code below to pass each element of a List to a multiline function and return an iterator of the result.
countArr = list(map(countReps, arr))
In Kotlin, I found that I had to do the following. Am I missing something?
fun LetterCountI(str: String): String {
val arr = str.split(" ")
var transform:(String) -> Int = {countReps(it)}
val countArr = arr.map(transform)
val mxIndex:Int
var ans:String
if (countArr.max()!=1){
mxIndex = countArr.indexOf(countArr.max())
ans = arr[mxIndex]
} else {
ans = "-1"
}
return ans;
}
fun countReps(str: String): Int {
var m = mutableMapOf<Char, Int>()
var v:Int
for (c in str){
if (c in m.keys){
v = m[c]?:0
m.put(c,v+1)
} else {
m.put(c,1)
}
}
return m.values.max() ?: 0
}```
I'm having a bit of a hard time understanding your code, but one thing I can tell you is that you can replace
var transform:(String) -> Int = {countReps(it)}
val countArr = arr.map(transform)
with
val countArr = arr.map(::countReps)
In addition to the line you ask about, just about all of that code could be rewritten more concisely and idiomatically in Kotlin. For example:
fun String.wordWithMostRepeatedLetters()
= split(" ")
.associateWith{ it.maxRepeatedLetters() }
.filter{ it.value > 1 }
.maxByOrNull{ it.value }?.key ?: "-1"
fun String.maxRepeatedLetters()
= groupBy{ it }.map{ it.value.size }.maxOrNull() ?: 0
I've renamed the functions to try to explain what they give; replaced the countArr list with a map from each word to its count, so that you don't need to re-scan it to find the word resulting; and changed both functions to take a String receiver instead of a parameter. Then, because each variable was only used once, I removed them and made it all in-line, using an expression body for each function.
Some of those things don't always improve clarity, of course, especially for long functions — but I hope it demonstrates how concise Kotlin can be. (Hopefully without sacrificing maintainability. Which version would be easier to read? Which would be more likely to harbour subtle bugs?)
It's still not clear what the hard-coded "-1" return value indicates, though… If no word has any repeated letters, a null return would be more idiomatic. (Or it would be simpler just to return the first word, removing the filter() call, and returning null only if the string is blank.)

Get top N elements in SortedMap?

Is there any correct way to get top N elements from SortedMap? My variant is:
val sortedMap = map.filterValues { it in sortedValues }.toSortedMap()
if (sortedMap.size <= 20) {
return sortedMap
}
var result = mutableMapOf<String, Int>()
for ((key, value) in sortedMap) {
result[key] = value
if (result.size == 20) {
break
}
}
headMap is also a good solution for getting a portion of this map, for example
return when {
sortedMap.size >= 20 -> {
sortedMap.headMap(sortedMap.keys.elementAt(20))
}
else -> {
sortedMap
}
}
more from the docs here
val first20AsList: List<Map.Entry<String, Int>> = sortedMap.asIterable().take(20)
val first20AsMap: Map<String, Int> = first20AsList.associate { it.toPair() }
Solution
In my opinion the best way is to use asSequence() because it is evaluated lazily:
return sortedMap.asSequence().take(20).map{ it.toPair() }.toMap()
Information on Lazy evaluation
In general it is often favorable to use lazy variants of iterable containers, because it means that in many cases not the whole datastructure needs to be evaluated but just the portion of data that is required.
In your case that would be the first N in the SortedMap. If it really happens to result in a performance advantage is questionable, but at least it is possible.
Some Information from stack-overflow and kotlin on lazy sequences:
Kotlin's Iterable and Sequence look exactly same. Why are two types required?
Sequences # kotlinlang.org
Old Answer:
This answer was improved in some details thanks to the comment by #IR42. Before I used the spread operator for conversion to a sorted-map again (via sortedMapOf), and map { it.key to it.value } instead of { it.toPair() }:
return sortedMapOf(
*sortedMap.asSequence().take(20)
.map{ it.key to it.value }
.toList().toTypedArray())
What u could try is flatMap{} from where u can get list of those items and then use take() function on that list to determine amount of it. Example:
val list = mapOf<Int, Boolean>(
1 to true,
2 to false,
3 to true
).toSortedMap().flatMap {
listOf(it)
}.take(2)
println(list[1].key) // <-- 2
take(n)
Returns a list containing first n elements.

Сheck if map contains substring. Kotlin

I have a
val map = Map<String,String>
map.put("Nurseiyt","android")
I want to get a value by subString like:
map["Nurs"] should return "android"
is it possible?
Use kotlin.Collections, there are methods like filter.
Two things - it's better to use regular expression. So, you can even get better control what will be returned. And the second one, there can be more than one elements matched to that regex. So that's why I return list.
fun <T> substringKey(map: Map<String, T>, regex: Regex): List<T> {
return map.filter { it.key.contains(regex) }
.map { it.value }
}
If you want to use that notation you need to create your own map and override proper operator. What's worth to notice, you cannot return list of values then. So, in this case I just return first found value.
class SubstringMap<V> : HashMap<String, V>() {
override operator fun get(key: String): V? {
return this.entries.first { it.key.contains(key) }.value
}
}
fun main() {
val map = SubstringMap<String>()
map["Nurseiyt"] = "android"
println(map["Nurs"]) // "android"
}
And as the last thing - in kotlin you can create your own operator, like withKeyPart. This would be much better than overriding default operator (because I wouldn't expect that [] operator will work in different way than usual.
infix fun <V> Map<String, V>.withKeyPart(keyPart: String): List<V> {
return this.filter { it.key.contains(keyPart) }
.map { it.value }
}
and then call it like this:
fun main() {
val map = HashMap<String, String>()
map withKeyPart "KeyPart" // infix notation
map.withKeyPart("KeyPart") // standard call
}
Filtering the map, as per other answers, is simple and straightforward, but it doesn't scale well; it takes time proportional to the size of the map, so if the map could grow big, it could get very slow.
If you're always going to be searching for a leading substring, i.e. the start of a map key, then a better general solution is a data structure called a trie.  This lets you search efficiently, with just one lookup per character.
Of course, writing one from scratch may not be justified for your project.  But there are third-party implementations you could use, such as this one in Apache Commons.  Or see the answers to this question.
write top level function like this
fun HashMap<String, String>.getContainskeyValue(search: String): String?
{
var returnList = ArrayList<String?>()
this.keys.filter { it.contains(search) }.map {
returnList.add(this[it])
}
return returnList.first()
//if you want all keys 'contains' values just return list
/* Ex
map.put("Nurseiyt", "android")
map.put("Nurseiyt1", "androidone")
map.put("Nurseirt2", "andrrroidtwo")
val isContainsdata = map.getContainskeyValue("N")
println(" result " + containsdata)
output :result [andrrroidtwo, android, androidone]
*/
}
then call like this
val map = HashMap<String, String>()
map.put("Nurseiyt", "android")
val containsdata = map.getContainskeyValue("Nurs")
println(" result " + containsdata)
output
android

Traverse a list of filter to appy them

I have some filter on a list(10..1000).
I would like to have a list of function objects, traverse this list and apply that filter to (10..1000).
Maybe later choose some filters.
fun main(args: Array<String>) {
var sol = (10..1000).toList().filter(lastDigitIsLength ).filter(no7andNo1 ).filter(isEvenAndGreater1).filter(first2DigitsOddCrossSumLess10 ).filter(isPrime )
println("The number is $sol")
}
/* The functions exist
There are few ways to do that.
I've listed them in my example, along with some general comments.
// No need for varargs since Kotlin 1.3
fun main() {
// Prefer vals over vars
val sol = (10..1000).
asSequence().// Use asSequence instead of toList
filter { lastDigitIsLength(it) }. // Block
filter(::no7andNo1). // Method reference
toList()
println("The number is $sol")
}
// Don't know what it means, so lets just always return true
fun lastDigitIsLength(i: Int) = true // Short notation
fun no7andNo1(i: Int): Boolean {
return i % 10 == 7 || i % 10 == 1
}
Sorry, maybe I misunderstand your question. Do you want to apply a list of filters to list of ints? If so, it can be done like this:
fun filter(list: List<Int>, filters: List<(Int) -> Boolean>): List<Int> {
var result = list
for (filter in filters) {
result = result.filter(filter)
}
return result
}
so you can rewrite your function as:
fun main() {
val sol = filter((10..1000).toList(), listOf(::lastDigitIsLength, ::no7andNo1, ::isEvenAndGreater1))
println("The number is $sol")
}

What's the fastest/simplest way to calculate a moving Average in Kotlin?

I can think on some dirty ways to calculate a moving average on Kotlin, but I'm not sure which one is the best. I know that kotlin has a lot of interesting features to work with collections and list. What do you think is the most efficient (or simplest) way to calculate a moving average?
Kotlin 1.2 will introduce a sliding window which you can combine with average obviously.
val data = listOf(1,2,5,6,2,7,8,5,9)
// 3 "period" moving average
val movingAverage = data.windowed(3,1,List<Int>::average)
// OR
val movingAverage = data.windowed(3,1) { it.average() }
Until then you would have to introduce your own sliding sequence.
class SlidingSequence<out T>(val source: Iterable<T>,
val slideSize: Int,
val slideStep: Int) : Sequence<List<T>> {
override fun iterator(): Iterator<List<T>> = object : AbstractIterator<List<T>>() {
private val iterator = if (slideSize > 0) source.iterator() else emptyList<T>().iterator()
private var buffer = listOf<T>()
override fun computeNext() = when {
iterator.hasNext() -> {
buffer = buffer.drop(slideStep).let {
it + iterator.asSequence().take(slideSize - it.size)
}
setNext(buffer)
}
else -> done()
}
}
}
fun <T> Iterable<T>.windowed(size: Int,
step: Int = 1): Sequence<List<T>> {
return SlidingSequence(this, size, step)
}
// and then you can do
val data = listOf(1,2,5,6,2,7,8,5,9)
// 3 "period" moving average
val movingAverage = data.windowed(3).map(List<Int>::average)
PS. I haven't looked at the code of Kotlin 1.2 windowed implementation, but since the function takes an immediate transform, I'm guessing the result is not lazy, where in the self implemented case above it's a lazy result, so you need to actually enumerate the sequence with something like .toList() to get the actual values.
Another one-line given period > 0 is:
data?.takeLast(period)?.reduce { v, d -> v + d}?: 0 / period
This also works if the data is empty or null due to takeLast() `s behaviour.