I have the following code which I am trying to use for two purposes:
1) Call an API and get result as a POJO
2) Sanitize this object (POJO) before I display it in the UI
private fun getWinbackDataItems(rewardPurpose: String) /*Single<WinbackBaseItem>*/ {
val x = repository.getRewardsList(rewardPurpose)
.filter {
it.result?.rewards != null
}.map { winback ->
winback.result?.rewards?.asSequence()?.filter { rewardsItem ->
rewardsItem?.id != null && rewardsItem.title != null
}?.toList()?.take(3)?.map {
WinbackListItem(it?.id, it?.title!!, false)
}?.toList()
}
}
The point of contention for me is the line below:
itemListSanitized.add(WinbackListItem(it.id, it.title, false))
At this point I assume the filter has removed all nulls from the original list but to my amazement I find that I have to null check on it and all its content while adding them to the new list.
What do I miss here, pardon my naivety as I have just begun reactive
I take it that you are working not against executing code but against your IDE's warning messages or just the ability for this code to compile. What you're probably running up against is that earlier checks for null won't necessarily allow the compiler to assume non-null values later on, because in the meantime, other code in a different thread could have run and changed the values.
So when you create a WinbackListItem, you can safely assume that certain items are not null, and yet the compiler can't be sure of this, because it can't know what else is going on in your process space. So the compiler requires that you tell it not to worry about null values (!!) or that you check the values again. This is just the way Kotlin works. It's often a PITA, but it's just how it is.
I played with the posted code just to be sure I knew what I was talking about. Here is code that I was able to run:
private fun getWinbackDataItems(rewardPurpose: String) /*Single<WinbackBaseItem>*/ {
val x = repository.getRewardsList(rewardPurpose)
.filter {
it.result?.rewards != null
}.map { winback ->
winback.result?.rewards?.asSequence().filter { rewardsItem ->
rewardsItem.id != null && rewardsItem.title != null
}.toList().take(3).map {
println(it.id)
println(it.title)
WinbackListItem(it.id!!, it.title!!, false)
}.toList()
}.count()
}
I created some very simple classes and objects to satisfy this code and let it run. Note that I took out some unnecessary '?' null checks. I played with input values until I was convinced that it.id and it.title can never be null when the WinbackListItem constructor is called. And yet, the two !! on its parameters, or something else making sure they are not null, are required given this definition of WinbackListItem that won't accept null parameter values:
class WinbackListItem(val id: Int, val title: String, val huh: Boolean)
Related
I have some code which looks like this, where param is of a data class type:
val options = if (param.language == null) {
param.copy(language = default())
} else {
param
}
Now, however, the language object has been moved into a hierarchy of nullable objects, so the check must look like this:
if (param.subObj?.nextObj?.language == null) { ... }
How do I use the copy idiom in this case?
One way to do this is:
val newParam = when {
param.subObj == null -> param.copy(subObj = SubObj(nextObj = NextObj(language = Language())))
param.subObj.nextObj == null -> param.copy(subObj = param.subObj.copy(nextObj = NextObj(language = Language())))
param.subObj.nextObj.language == null -> param.copy(subObj = param.subObj.copy(nextObj = param.subObj.nextObj.copy(language = Language())))
else -> param
}
I agree that this doesn't look very clean but this seems to be the only way to me, because at each step you need to check if the current property is null or not. If it is null, you need to use the default instance otherwise you need to make a copy.
Could you do something like this?
// you could create a DefaultCopyable interface if you like
data class SubObj(val prop1: Double? = null, val nextObj: NextObj? = null) {
fun copyWithDefaults() =
copy(prop1 = prop1 ?: 1.0, nextObj = nextObj?.copyWithDefaults() ?: NextObj())
}
data class NextObj(val name: String? = null) {
fun copyWithDefaults() = copy(name = name ?: "Hi")
}
I think you need a special function because you're not using the standard copy functionality exactly, you need some custom logic to define defaults for each class. But by putting that function in each of your classes, they all know how to copy themselves, and each copy function that works with other types can just call their default-copy functions.
The problem there though is:
fun main() {
val thing = SubObj(3.0)
val newThing = thing.copyWithDefaults()
println("$thing\n$newThing")
}
> SubObj(prop1=3.0, nextObj=null)
> SubObj(prop1=3.0, nextObj=NextObj(name=null))
Because nextObj was null in SubObj, it has to create one instead of copying it. But the real default value for name is null - it doesn't know how to instantiate one with the other defaults, that's an internal detail of NextObj. You could always call NextObj().copyWithDefaults() but that starts to look like a code smell to me - why isn't the default value for the parameter the actual default value you want? (There are probably good reasons, but it might mean there's a better way to architect what you're up to)
I have a list of objects with an optional id as String and I want to make a map out of it.
I want to have the keys of my map as non nullable: so something like this:
data class Foo(
val id: String? = null
val someStuff: String? = null,
)
val foo = listOf(Foo("id1"), Foo())
val bar = foo.filterNot { it.id == null }.associateBy { it.id }
Here bar type is Map<String?, Foo> but not Map<String, Foo>
My workaround is to add a non null asserted call: !!, but it doesn't seem clean.
Is there an easy and safe way to do this?
This looks like something that contracts could help with, but currently a contract expression can't access properties of the class in use.
As a workaround, you could define a 2nd class that has a non-null id, like so
data class Foo(
val id: String? = null,
val someStuff: String? = null
)
data class Foo2(
val id: String,
val someStuff: String? = null
)
val foo = listOf(Foo("id1"), Foo())
val bar = foo
.mapNotNull { if (it.id != null) Foo2(it.id, it.someStuff) else null }
.associateBy { it.id }
There's a six-year-old open feature request for Map.filterNotNullKeys() and a four-year old open feature request for Map.associateByNotNull().
In my opinion, the associateBy { it.id!! } would be cleanest for readability. But you could do it like this:
val bar = foo.mapNotNull { it.id?.run { it.id to it } }.toMap()
As for your actual question, that logic is way too many steps for the compiler to infer. Your last function call to associateBy sees a nullable, so it infers a nullable. For the compiler to figure this out, it would have to step back and see that the List that you call associateBy on happens to have filtered out certain objects in a way that happens to ensure that a certain nullable property won't be null within this specific list, and it's the same property that you are associating with. Now imagine it has to do this for every call to any generic function, and the various lambdas involved could potentially have multiple lines of code. Compile times would skyrocket.
I have a List of objects.
importerResponse.applications is empty (size=0)
This is my code:
val isDeployed = importerResponse.applications
.flatMap(Application::instances)
.map(Instance::state)
.all { state -> DEPLOYED == state }
isDeployed is true in this case. How can this be? I want it to resolve into false if applications is empty.
Why would you want that? All the elements in the collection satisfy your predicate.
You can check the documentation:
https://kotlinlang.org/api/latest/jvm/stdlib/kotlin.sequences/all.html
If you want you can explicitly check for the collection being empty.
This should give you what you want:
val isDeployed = importerResponse.applications
.flatMap(Application::instances)
.map(Instance::state)
.count { state -> DEPLOYED == state } > 0```
The all method might be looking for any element that doesn't meet the condition, since you don't have any, it defaults to true.
You can achieve what you want by doing something similar to this:
val isDeployed = importerResponse.applications
.flatMap(Application::instances)
.map(Instance::state)
.let { it.size() > 0 && it.all { state -> DEPLOYED == state } }
Note that let allows you to reuse the same expression without recalculating it twice.
It can be a little bit confusing, why "any" returns "false" on empty collections, but "all" return true, because "all" seems to be more limiting than "any" (based on human language).
But if you ask as one example "if all persons in a room are male", than that's still true, if the room is empty. 0 out of 0 persons are ALL.
I am kinda new using kotlin and I was wondering if I can do something like this.
I have a list with objects of type Person,
Person has properties like name, id but can be null.
So I made something like this:
return persons.filter {
it.name != null && it.id != null
}.map {
it.id to it.name
}.toMap()
I personally dont see the error but the compiler keeps telling me I should return a map of not nulls.
Is there any way I can do it using filter and map using lambdas functions?
Use mapNotNull to combine filter and map:
persons.mapNotNull {
val id = it.id
val name = it.name
if (id != null && name != null) Pair(id, name) else null
}.toMap()
Pulling id and name into local variables should make sure they'll get inferred as not-null in Pair(id, name) but may not be necessary.
The reason your approach doesn't work is that persons.filter { ... } just returns List<Person>, there's no way to say "list of Persons with non-null name and id" or to represent it internally in the compiler.
btw, you can even get rid of if in mapNotNull:
persons.mapNotNull {
val id = it.id ?: return#mapNotNull null
val name = it.name ?: return#mapNotNull null
id to name
}.toMap()
May be you can simply change
.map {
it.id to it.name
}
into
.map {
it.id!! to it.name!!
}
The suffix !! operator converts String? into String, throwing exception if the value with type String? is null, which in our case can't be true, due to the previously applied filter.
Use of !! should be limited to cases where you take responsibility of explicitly asserting that value can't be null: you're saying to the compiler that values are String even if the type is String? - and should be imho used with caution.
Compiler can't infer the type domain restriction from String? to String from the predicate passed to filter, but you can, so I think !! usage can be a valuable approach...
I'm trying to convert my nested for loop to asSequence in Kotlin. Here, my goal is to get and update the value of all my object array from another object array with the same key.
nested for loop:
val myFields = getMyFields()
val otherFields = getOtherFields()
for (myField in myFields) { // loop tru the my fields
for (otherField in otherFields) { // find the same fields
if (myField.key == otherField.key) { // if the same, update the value
val updatedMyField = myField.copy(value = otherValue.value)
myFields[myFields.indexOf(myField)] = updatedMyField // update my field value
break
}
}
}
What I've tried:
val updatedMyFields = getMyFields().asSequence()
.map { myField ->
getOtherFields().asSequence()
.map { otherField ->
if (myField.key == otherField.key) {
return#map otherField.value
} else {
return#map ""
}
}
.filter { it?.isNotEmpty() == true }
.first()?.map { myField.copy(value = it.toString()) }
}
.toList()
but this does not compile as it will return List<List<MyField>>.
I'm just looking for something much cleaner for this.
As comments suggest, this would probably be much more efficient with a Map.
(More precisely, a map solution would take time proportional to the sum of the list lengths, while the nested for loop takes time proportional to their product — which gets bigger much faster.)
Here's one way of doing that:
val otherFields = getOtherFields().associate{ it.key to it.value }
val myFields = getMyFields().map {
val otherValue = otherFields[it.key]
if (otherValue != null) it.copy(value = otherValue) else it
}
The first line creates a Map from the ‘other fields’ keys to their values. The rest then uses it to create a new list from ‘my fields’, substituting the values from the ‘other fields’ where present.
I've had to make assumptions about the types &c, since the code in the question is incomplete, but this should do the same. Obviously, you can change how it merges the values by amending the it.copy().
There are likely to be even simpler and more efficient ways, depending on the surrounding code. If you expanded it into a Minimal, Complete, and Verifiable Example — in particular, one that illustrates how you already use a Map, as per your comment — we might be able to suggest something better.
Why do you want to use asSequence() ? You can go for something like that:
val myFields = getMyFields()
val otherFields = getOtherFields()
myFields.forEach{firstField ->
otherFields.forEach{secondField ->
if (firstField.key == secondField.key) {
myFields[myFields.indexOf(firstField)] = secondField.value
}
}
}
This will do the same job than your nested for loop and it's easier to read, to understand and so to maintain than your nested asSequence().