I have come across the concept called destructuring declarations - when you can return multiple values from a function at once. It seems very convenient, but at the same time it looks like a tricky workaround. Each time when I think about that feature in Java, I understand that it's a hole in my architecture - there should probably be a class then, not just a couple of variables.
What do you think?
The concept allows having classes that clearly identify a few of their primary properties, the components.
Then you can access these components by using a destructuring declaration, without syntactic noise of accessing the properties.
Compare:
val point = clickEvent.getPointOnScreen()
val x = point.xCoordinate
val y = point.yCoordinate
// Use `x` and `y` in some calculations
and, assuming that the type has component1 and component2, just:
val (x, y) = clickEvent.getPointOnScreen()
Basically, it is not necessary to use this sort of syntactic sugar, and the concept itself does not harm any of the abstractions, it only provides a convenient way to access properties of a class instance in some cases when you don't need the instance itself.
Another example is working with map entries, e.g:
for ((key, value) in myMap) { /* ... */ }
There's still a Map.Entry<K, V> behind the (key, value) destructuring, and you can replace it by for (entry in myMap) ..., but usually it's the two properties that you need. This is where destructuring saves you from a little syntactic noise.
You can also define componentN function as extension for non data classes like this:
operator fun Location.component1() = latitude
operator fun Location.component2() = longitude
and when you want to process on list of locations, you can write this:
for ((lat, lon) in locations) {
......
}
What's the point of destructuring declarations in Kotlin?
Structuring, or construction, is creating an object from values in different variables. Destructuring is the opposite, to extract values into variables from within an existing object.
Part of the Kotlin philosophy is to be concise since the simpler and more concise the code is, the faster you’ll understand what’s going on. Destructuring improves readability which is part of being concise. Compare the following two snippets (let's consider the class Triple)
Without using destructuring
fun getFullName() = Triple("Thomas", "Alva", "Edison")
val result = getFullName()
val first = result.first
val middle = result.second
val last = result.third
Using destructuring
fun getFullName() = Triple("Thomas", "Alva", "Edison")
val (first, middle, last) = getFullName()
It is also possible to take advantage of destructuring to extract key and value from Map's entries.
for ((key, value) in aMap) {
/* ... */
}
Destructuring is the most useful when dealing with built-in data structures. Their fields have names making sense in the context of a data structure (handy when you're writing your own hashmap), but completely cryptic when you're dealing with the data contained there (which is 100% of the time, nobody writes their own hashmaps). Eg. Pair with it's first and second or Map.Entry with key and value.
Consider transforming Map values:
val myMap = mapOf("apples" to 0, "oranges" to 1, "bananas" to 2)
myMap
.asIterable()
.filter { it.value > 0 }
.sortedBy { it.key.length }
.joinToString(prefix = "We have ", postfix = " in the warehouse") {
"{$it.value} of ${it.key}"
}
To make it readable, you'd have to define intermediate variables:
myMap
.asIterable()
.filter {
val count = it.value
count > 0
}
.sortedBy {
val fruit = it.key
fruit.length
}
.joinToString(prefix = "We have ", postfix = " in the warehouse") {
val count = it.value
val fruit = it.key
"$count of $fruit"
}
Now it's readable, but at what cost?!?
Destructuring makes this cost more beareable:
myMap
.asIterable()
.filter { (fruit, count) -> count > 0 }
.sortedBy { (fruit, count) -> fruit.length }
.joinToString(prefix = "We have ", postfix = " in the warehouse") { (fruit, count) ->
"$count of $fruit"
}
That's the point.
Related
I'm trying my hands on Kotlin. Being from a Python background is really giving me a tough time to get the knack of the Kotlin syntax. I'm trying to do a simple dictionary (Mutable Map) operation. However, its giving me exceptions.
This is what I tried. Kotlin compiler
Adding the code snippet for reference.
fun main() {
val openActivityMap = mutableMapOf<String, MutableMap<String, Any>>()
val packageName = "amazon"
val currentTime = 23454321234
if(openActivityMap.containsKey(packageName)){
if(openActivityMap[packageName]?.get("isAlreadyApplied")){
if((openActivityMap[packageName]?.get("lastAppliedAt") - currentTime) > 3600){
openActivityMap[packageName]?.put("isAlreadyApplied", false)
}
}
else{
openActivityMap[packageName]?.put("isAlreadyApplied", false)
}
}
}
I'm a bit late to the party, but I'd like to point out another solution here.
As I commented on the OP, heterogeneous maps with fixed string keys like this are usually better expressed with classes in Kotlin. For instance, in your case, the class for your main map's values could be the following:
data class PackageInfo(
var isAlreadyApplied: Boolean,
var lastAppliedAt: Long,
)
(you could obviously add more properties if need be)
This would save you all the casts on the final values.
Another point I'd like to make is that if you access the value for a key anyway, you don't need to check up front the existence of the key with containsKey. Maps return null for keys that are not associated with any value (this is why you need to check for null after getting the value).
The compiler cannot see the correlation between containsKey and the subsequent get or [] access. However, it's smart enough to understand a null check if you simply get the value first and then check for null.
This always applies unless you want to tell the difference between keys that aren't in the map and keys that are in the map but associated null values (which is quite rare).
All in all, I would write something like that:
fun main() {
val openActivityMap = mutableMapOf<String, PackageInfo>()
val packageName = "amazon"
val currentTime = 23454321234
val packageInfo = openActivityMap[packageName]
if (packageInfo != null) { // the key was found and the value is smart cast to non-null in the next block
if (packageInfo.isAlreadyApplied) {
if ((packageInfo.lastAppliedAt - currentTime) > 3600) {
packageInfo.isAlreadyApplied = false
}
} else {
packageInfo.isAlreadyApplied = false
}
}
}
data class PackageInfo(
var isAlreadyApplied: Boolean,
var lastAppliedAt: Long,
)
I would recommend writing tests first and working in small increments, but this should fix your compilation issues:
fun main() {
val openActivityMap = mutableMapOf<String, MutableMap<String, Any>>()
val packageName = "amazon"
val currentTime = 23454321234
if(openActivityMap.containsKey(packageName)){
if(openActivityMap[packageName]?.get("isAlreadyApplied") as Boolean){
if((openActivityMap[packageName]?.get("lastAppliedAt") as Long - currentTime) > 3600){
openActivityMap[packageName]?.put("isAlreadyApplied", false)
}
}
else {
openActivityMap[packageName]?.put("isAlreadyApplied", false)
}
}
}
EDIT: Also I prefer to avoid nullable variables and mutable objects in general, but I suppose there's an exception to every rule.
Couldn't you just declare your Map<String, Any> to return a Boolean instead of Any? So,
val openActivityMap = mutableMapOf<String, MutableMap<String, Boolean>>()
It looks like you're trying to use your second Map to store both Booleans and Ints, which is complicating the logic. You'll need to typecast if you decide to approach it without Typing.
There's a problem with the 2 statement below
if(openActivityMap[packageName]?.get("isAlreadyApplied"))
if((openActivityMap[packageName]?.get("lastAppliedAt") - currentTime) > 3600)
As we all know, an IF statement requires a boolean value for it's param. The types of both statement are unknown at compilation time as they are of a Generic type, Any. As such,
openActivityMap[packageName]?.get("isAlreadyApplied") could be a null or of type Any (Not Boolean).
openActivityMap[packageName]?.get("lastAppliedAt") could be a null or of type Any (an Int was expected here for computation).
This would throw compilation errors as the compiler does not know the types to go with. What could be done is to cast to it's proper types.
Solution
openActivityMap[packageName]?.get("isAlreadyApplied") as Boolean ?: false
((openActivityMap[packageName]?.get("lastAppliedAt") as Int ?: 0) - currentTime)
Giving a default value if it's null.
maybe you can try something like this
if (openActivityMap.containsKey(packageName)) {
val packageMap = openActivityMap[packageName]!!
val applyRequired = (packageMap["lastAppliedAt"] as Long - currentTime) > 3600
packageMap["isAlreadyApplied"] = packageMap.containsKey("isAlreadyApplied") && !applyRequired
}
btw. do you really want to have lastAppliedAt to be in te future? otherewise it will never be > 3600
In Kotlin I can:
val (specificMembers, regularMembers) = members.partition {it is SpecificMember}
However to my knowledge I can not do something like:
val (specificMembers as List<SpecificMember>, regularMembers) = members.partition {it is SpecificMember}
My question would be - is there's an idiomatic way to partition iterable by class and typecast it those partitioned parts if needed.
If you require that functionality more often, you may just reimplement the actual partition according to your needs, e.g.:
inline fun <reified U : T, T> Iterable<T>.partitionByType(): Pair<List<U>, List<T>> {
val first = ArrayList<U>()
val second = ArrayList<T>()
for (element in this) {
if (element is U) first.add(element)
else second.add(element)
}
return Pair(first, second)
}
with a usage similar as to follows:
val (specificMembers, regularMembers) = members.partitionByType<SpecificMember, Member>()
// where specificMembers : List<SpecificMember>
// and regularMembers : List<Member> for this example
Note that this way you can also set the second type to a more generic one. I leave that up to you whether this makes sense. At least this way an unchecked cast isn't necessary.
The alternative is also shown by Simon with the let-usage. You can also directly cast the result of partition (without let and another Pair) to whatever fits, e.g.:
val (specificMembers, regularMembers) = members.partition {it is SpecificMember} as Pair<List<SpecificMember>, List<Member>>
The partition function will return a Pair<List<T>, List<T>> with T being the generic type of your Iterable. You can transform the partitioned values again using e.g. let:
val (specificMembers, regularMembers) = lists
.partition { it is SpecificMember }
.let { Pair(it.first as List<SpecificMember>, it.second) }
I’m trying to use Kotlin’s when block to look up an element in different maps. After confirming the element exists, the code subsequently does not smart-cast the resulting lookup in the map to not null.
Below is a minimum working example: is it possible to rework it such that !! is not needed?
fun main(args: Array<String>) {
val string = "abc"
val map1 = mapOf('a' to 5)
val map2 = mapOf('b' to 4)
when (val char = string.firstOrNull()) {
null -> println("Nothing to find")
in map1 -> println("Found in map1: ${map1[char]!!+1}")
in map2 -> println("Found in map2: ${map2[char]!!-1}")
else -> println("Unrecognised character $char")
}
}
Unfortunately, in Kotlin, functions can't have contracts of the form "if f returns true, then g doesn't return null." Hence, the compiler doesn't use information about definitely successful contains calls.
The workaround with !! is OK in this case because you can be sure that get returns not null. Implementation of complex patterns in when (KT-186) would cover this use case by allowing declaring a variable inside when clauses and providing static guarantees that it's not null.
In Kotlin documentation I found the following example:
for ((index, value) in array.withIndex()) {
println("the element at $index is $value")
}
Is it possible (and how) to do the similar with 2D matrix:
for ((i, j, value) in matrix2D.withIndex()) {
// but iterate iver double index: i - row, j - column
if (otherMatrix2D[i, j] > value) doSomething()
}
How to make support this functionality in Kotlin class?
While the solutions proposed by miensol and hotkey are correct it would be the least efficient way to iterate a matrix. For instance, the solution of hotkey makes M * N allocations of Cell<T> plus M allocations of List<Cell<T>> and IntRange plus one allocation of List<List<Cell<T>>> and IntRange. Moreover lists resize when new cells are added so even more allocations happen. That's too much allocations for just iterating a matrix.
Iteration using an inline function
I would recommend you to implement a very similar and very effective at the same time extension function that will be similar to Array<T>.forEachIndexed. This solution doesn't do any allocations at all and as efficient as writing nested for cycles.
inline fun <T> Matrix<T>.forEachIndexed(callback: (Int, Int, T) -> Unit) {
for (i in 0..cols - 1) {
for (j in 0..rows - 1) {
callback(i, j, this[i, j])
}
}
}
You can call this function in the following way:
matrix.forEachIndexed { i, j, value ->
if (otherMatrix[i, j] > value) doSomething()
}
Iteration using a destructive declaration
If you want to use a traditional for-loop with destructive declaration for some reason there exist a way more efficient but hacky solution. It uses a sequence instead of allocating multiple lists and creates only a single instance of Cell, but the Cell itself is mutable.
data class Cell<T>(var i: Int, var j: Int, var value: T)
fun <T> Matrix<T>.withIndex(): Sequence<Cell<T>> {
val cell = Cell(0, 0, this[0, 0])
return generateSequence(cell) { cell ->
cell.j += 1
if (cell.j >= rows) {
cell.j = 0
cell.i += 1
if (cell.i >= cols) {
return#generateSequence null
}
}
cell.value = this[cell.i, cell.j]
cell
}
}
And you can use this function to iterate a matrix in a for-loop:
for ((i, j, item) in matrix.withIndex()) {
if (otherMatrix[i, j] > value) doSomething()
}
This solution is lightly less efficient than the first one and not so robust because of a mutable Cell, so I would really recommend you to use the first one.
These two language features are used for implementing the behaviour that you want:
For-loops can be used with any class that has a method that provides an iterator.
for (item in myItems) { ... }
This code will compile if myItems has function iterator() returning something with functions hasNext(): Boolean and next().
Usually it is an Iterable<SomeType> implementation (some collection), but you can add iterator() method to an existing class as an extension, and you will be able to use that class in for-loops as well.
For destructuring declaration, the item type should have componentN() functions.
val (x, y, z) = item
Here the compiler expects item to have component1(), component2() and component3() functions. You can also use data classes, they have these functions generated.
Destructuring in for-loop works in a similar way: the type that the iterator's next() returns must have componentN() functions.
Example implementation (not pretending to be best at performance, see below):
Class with destructuring support:
class Cell<T>(val i: Int, val j: Int, val item: T) {
operator fun component1() = i
operator fun component2() = j
operator fun component3() = item
}
Or using data class:
data class Cell<T>(val i: Int, val j: Int, val item: T)
Function that returns List<Cell<T>> (written as an extension, but can also be a member function):
fun <T> Matrix<T>.withIndex() =
(0 .. height - 1).flatMap { i ->
(0 .. width - 1). map { j ->
Cell(i, j, this[i, j])
}
}
The usage:
for ((i, j, item) in matrix2d.withIndex()) { ... }
UPD Solution offered by Michael actually performs better (run this test, the difference is about 2x to 3x), so it's more suitable for performance critical code.
The following method:
data class Matrix2DValue<T>(val x: Int, val y: Int, val value: T)
fun withIndex(): Iterable<Matrix2DValue<T>> {
//build the list of values
}
Would allow you to write for as:
for ((x, y, value) in matrix2d.withIndex()) {
println("value: $value, x: $x, y: $y")
}
Bear in mind though that the order in which you declare data class properties defines the values of (x, y, value) - as opposed to for variable names. You can find more information about destructuring in the Kotlin documentation.
If I create an array, then fill it, Kotlin believes that there may be nulls in the array, and forces me to account for this
val strings = arrayOfNulls<String>(10000)
strings.fill("hello")
val upper = strings.map { it!!.toUpperCase() } // requires it!!
val lower = upper.map { it.toLowerCase() } // doesn't require !!
Creating a filled array doesn't have this problem
val strings = Array(10000, {"string"})
val upper = strings.map { it.toUpperCase() } // doesn't require !!
How can I tell the compiler that the result of strings.fill("hello") is an array of NonNull?
A rule of thumb: if in doubts, specify the types explicitly (there is a special refactoring for that):
val strings1: Array<String?> = arrayOfNulls<String>(10000)
val strings2: Array<String> = Array(10000, {"string"})
So you see that strings1 contains nullable items, while strings2 does not. That and only that determines how to work with these arrays:
// You can simply use nullability in you code:
strings2[0] = strings1[0]?.toUpperCase ?: "KOTLIN"
//Or you can ALWAYS cast the type, if you are confident:
val casted = strings1 as Array<String>
//But to be sure I'd transform the items of the array:
val asserted = strings1.map{it!!}
val defaults = strings1.map{it ?: "DEFAULT"}
Why the filled array works fine
The filled array infers the type of the array during the call from the lambda used as the second argument:
val strings = Array(10000, {"string"})
produces Array<String>
val strings = Array(10000, { it -> if (it % 2 == 0) "string" else null })
produces Array<String?>
Therefore changing the declaration to the left of the = that doesn't match the lambda does not do anything to help. If there is a conflict, there is an error.
How to make the arrayOfNulls work
For the arrayOfNulls problem, they type you specify to the call arrayOfNulls<String> is used in the function signature as generic type T and the function arrayOfNulls returns Array<T?> which means nullable. Nothing in your code changes that type. The fill method only sets values into the existing array.
To convert this nullable-element array to non-nullable-element list, use:
val nullableStrings = arrayOfNulls<String>(10000).apply { fill("hello") }
val strings = nullableStrings.filterNotNull()
val upper = strings.map { it.toUpperCase() } // no !! needed
Which is fine because your map call converts to a list anyway, so why not convert beforehand. Now depending on the size of the array this could be performant or not, the copy might be fast if in CPU cache. If it is large and no performant, you can make this lazy:
val nullableStrings = arrayOfNulls<String>(10000).apply { fill("hello") }
val strings = nullableStrings.asSequence().filterNotNull()
val upper = strings.map { it.toUpperCase() } // no !! needed
Or you can stay with arrays by doing a copy, but really this makes no sense because you undo it with the map:
val nullableStrings = arrayOfNulls<String>(10000).apply { fill("hello") }
val strings: Array<String> = Array(nullableStrings.size, { idx -> nullableStrings[idx]!! })
Arrays really are not that common in Java or Kotlin code (JetBrains studied the statistics) unless the code is doing really low level optimization. It could be better to use lists.
Given that you might end up with lists anyway, maybe start there too and give up the array.
val nullableStrings = listOf("a","b",null,"c",null,"d")
val strings = nullableStrings.filterNotNull()
But, if you can't stop the quest to use arrays, and really must cast one without a copy...
You can always write a function that does two things: First, check that all values are not null, and if so then return the array that is cast as not null. This is a bit hacky, but is safe only because the difference is nullability.
First, create an extension function on Array<T?>:
fun <T: Any> Array<T?>.asNotNull(): Array<T> {
if (this.any { it == null }) {
throw IllegalStateException("Cannot cast an array that contains null")
}
#Suppress("CAST_NEVER_SUCCEEDS")
return this as Array<T>
}
Then use this function new function to do the conversion (element checked as not null cast):
val nullableStrings = arrayOfNulls<String>(10000).apply { fill("hello") }
val strings = nullableStrings.asNotNull() // magic!
val upperStrings = strings.map { it.toUpperCase() } // no error
But I feel dirty even talking about this last option.
There is no way to tell this to the compiler. The type of the variable is determined when it is declared. In this case, the variable is declared as an array that can contain nulls.
The fill() method does not declare a new variable, it only modifies the contents of an existing one, so it cannot cause the variable type to change.