https://kotlinlang.org
Join Slack
Is there a way to use Optics to work with KotlinX Serialization's JsonObject hierarchy? i'd like to ...
# arrow
d
Is there a way to use Optics to work with KotlinX Serialization's JsonObject hierarchy? i'd like to pinpoint on certain parts of a big, very nested json and just modify it...
🙌 1
s
Yes, I've build a specialised DSL based on Optics for it. https://github.com/nomisRev/kotlinx-serialization-jsonpath
It offers special functions to more easily access deeper parts of a very nested 
JsonElement
.
d
It's production-ready?
s
Yes, of course. What would you consider not-production ready?
d
Well, it's not 1.0.0 😊... in some libraries that's considered experimental or WIP...
s
Ah, I guess I could/should release it as 1.0.0 😅
The API is stable, and tested.
d
I hope you'll reference all these great libraries in the new site 😃... I keep on thinking that something's not possible/not implemented just to discover that you've already written a library for it 😅.
s
We have a page for this already 😉 I am pushing today again to have it released in preview, but it'll probably be for in 2 weeks. I have some much needed PTO next week 😂
d
Minor detail... my JsonObject keys contains a 
.
... is there a way to escape them?
Is there a reference for the JsonPath select syntax you're using there?
Ok, I think I got it JsonPath["..."]["..."]...
s
I'm not sure I understood your question
d
Copy code
{
  "some.key": {
    "other-key": [
      "value"
    ]
  }
}
Copy code
JsonPath.select("some.key.other-key") // would look for { "some": { "key": { "other-key" ...
Another thing... to modify 
"value"
I'd have to map my set of strings to `JsonElement`s I guess... I wouldn't be able to create a function that allows me to set the list from a collection of strings directly?
s
You can by calling 
.string
or 
.int
on the DSL
It currently it's possible to escape 
.
, but that is something we could quite easily add. We can parameterise splitting on a parameterised 
Char
rather than hardcoding 
.
Otherwise using 
[][]
can currently be used as a workaround. Feel free to open a issue, and I'll try to get to is asap ☺️ A PR fixing it is of course also very welcomed. The codebase is quite small, so it's less intimidating to approach than Arrow itself.
Would give a good excuse to bump to 1.0.0 I guess 😄
d
I'm still not sure how to set the list "value" is in:
Copy code
val listRef = JsonPath["some.key"]["other-key"].every.string

listRef.set(jsonElem, listOf("other-value")) // this doesn't work
s
Copy code
JsonPath["some.key"]["other-key"].every.string
  .modify(jsonElem) { original: String -> "other-value" }
d
I want to overwrite the whole list
s
JsonPath["some.key"]["other-key"].jsonArray()
would give you 
Optional<JsonElement, List<JsonElement>>
but then you'd want to turn 
List<JsonElement>
into 
List<String>
🤔
Not sure how to do that conveniently
Copy code
JsonPath["some.key"]["other-key"].jsonArray().modify(jsonElem) {
  listOf(JsString("other-value"))
}
this definitely works
Not sure if it'd be easy to make that better, since JsonArray can contain JsElement of different types
d
Yeah, so in my case, it doesn't...
s
But the type system doesn't know that 🧌
d
It seems like Intellij doesn't like your setup for some reason... There's no code navigation, it doesn't seem to catch that those are source sets...
s
That's really weird.. I am having the same issue now but it has the same layout as all my other projects 😕
d
I go to project structure, mark them as source sets... they get highlighted (with red lines under arrow deps).. I try to resync with the gradle file, it goes back to not being source sets...
Maybe Gradle 8 changed something for KMM projects?
I wanted to give a PR a shot... but with my little knowledge of optics (I was hoping to learn more about optics along the way...), and no code navigation... I guess I'll wait until there's a solution for this...
s
Might be related to missing Kotlin sources after all: 
Could not find kotest-common.klib (io.kotest:kotest-common-iosx64:5.5.5).
I currently don't have time to investigate this
d
No rush 😊, just reporting...
@simon.vergauwen Can regular optics be combined with that libraries optics?
Copy code
@JvmInline
value class Bar(val json: JsonObject) {
  companion object {
     val path = JsonPath....
  }
}

@optics data class Foo(val bar: Bar)

foo.copy { Foo.Bar.path set ... }
s
(Foo.bar compose Bar.path) set ...
Not sure if 
+
is still available as alias for 
compose
.
d
compose is red
None of the following functions can be called with the arguments supplied. compose(Fold<in CustomRestrictions, out TypeVariable(C)>) where C = TypeVariable(C) for fun <C> compose(other: Fold<in CustomRestrictions, out C>): Fold<RestrictionAction, C> defined in arrow.optics.POptional compose(PEvery<in CustomRestrictions, out CustomRestrictions, out TypeVariable(C), in TypeVariable(D)>) where C = TypeVariable(C), D = TypeVariable(D) for fun <C, D> compose(other: PEvery<in CustomRestrictions, out CustomRestrictions, out C, in D>): PEvery<RestrictionAction, RestrictionAction, C, D> defined in arrow.optics.POptional compose(POptional<in CustomRestrictions, out CustomRestrictions, out TypeVariable(C), in TypeVariable(D)>) where C = TypeVariable(C), D = TypeVariable(D) for infix fun <C, D> compose(other: POptional<in CustomRestrictions, out CustomRestrictions, out C, in D>): POptional<RestrictionAction, RestrictionAction, C, D> defined in arrow.optics.POptional compose(POptionalGetter<in CustomRestrictions, RestrictionAction, out TypeVariable(C)>) where C = TypeVariable(C) for fun <C> compose(other: POptionalGetter<in CustomRestrictions, RestrictionAction, out C>): POptionalGetter<RestrictionAction, RestrictionAction, C> defined in arrow.optics.POptional compose(PSetter<in CustomRestrictions, out CustomRestrictions, out TypeVariable(C), in TypeVariable(D)>) where C = TypeVariable(C), D = TypeVariable(D) for fun <C, D> compose(other: PSetter<in CustomRestrictions, out CustomRestrictions, out C, in D>): PSetter<RestrictionAction, RestrictionAction, C, D> defined in arrow.optics.POptional compose(PTraversal<in CustomRestrictions, out CustomRestrictions, out TypeVariable(C), in TypeVariable(D)>) where C = TypeVariable(C), D = TypeVariable(D) for fun <C, D> compose(other: PTraversal<in CustomRestrictions, out CustomRestrictions, out C, in D>): PTraversal<RestrictionAction, RestrictionAction, C, D> defined in arrow.optics.POptional
s
What type is 
path
?
d
Optional<JsonElement, List<JsonElement>>
Ohh... the value class.
It should have been Optional<Bar, List<JsonElement>>
But how do I adapt that? value classes can't have optics.
And it's a bit overkill.
s
You could write an 
Iso
for it and put it in between
d
?
s
Copy code
@JvmInline
value class Bar(val json: JsonElement) {
  companion object {
     val path: Optional<Bar, List<JsonElement>> =
       json compose JsonPath....

     val json: Iso<Bar, JsonElement> = Iso(
       get = { it.json },
       reverseGet = { Bar(it) }
     )
  }
}

@optics data class Foo(val bar: Bar)

foo.copy { (Foo.Bar compose Bar.path) set ... }
d
Thanks!
So I could also make such an Iso for my previous List<JsonElement> to List<String>, no?
s
It would be in violation of the laws which is a set of tests that the Optics type should adhere in order for them to work correctly. You could create an unsafe 
Iso
which makes a type-unsafe assumption that 
List<JsonElement> == List<String>
.
TL;DR it will requires an unsafe cast, which violates the type system but yes anything is possible in code 😄
d
Setting has no problems with types... only reading. But I guess it's all one big package deal in optics?
If there would be a way to implement only reverseGet and just use the setter part of optics for such cases.
Or alternatively, even possibly have some kind of Either to return for the get case which would just return Left if the cast wasn't successful
But I guess that would make things much more complex
s
That is possible with 
Prism
but I'm not sure on the top of my head how to do it for 
List<JsonElement> -> List<String>
. We have it for 
JsonElement -> String
.
d
Ok, I just learned what Iso is... but what's Prism?
s
d
Ok... so a Prism allows to return an Either on 
get
... so there's no such concept for each element in a collection?
s
Nope, but I though you didn't want that? Otherwise you can just use 
path.every.string
You can still compose 
.every
after you go from 
List<JsonElement>
to 
List<String>
though, and it should be possible to write a Prism for that
d
I need to set the whole list (overwrite the current contents)... every does that?
Something like val baz = Optional<Bar, List<String>>
that allows 
baz.set(bar, listOf("one", "two"))
s
Ok... so a Prism allows to return an Either on 
get
... so there's no such concept for each element in a collection?
Then I don't understand your question
d
Well, I don't know Prism's return type so I just expressed it as Optional...
But whatever gives that set(...)
s
Both Optional and Prism have 
set
d
I guess this should do the trick:
Copy code
private val prism: Prism<List<JsonElement>, List<String>> = Prism(
    getOption = { orig -> 
            orig.mapNotNull { it.jsonPrimitive.contentOrNull }
                .takeIf { orig.size == it.size }?.some() ?: none() 
        },
    reverseGet = { it.map { value -> JsonPrimitive(value) } }
)
Maybe I'm mixing up Optional here... but there IS getOption on Prism... 🤔, Maybe I'm not doing this correctly? Or maybe the better practice would be the Either alternative...?
s
is something not working?
d
No, everything's fine 😃, just wondering if I'm doing things "right".
s
Yes, you're doing it right ☺️ You can think of 
Prism
as 
when/is
and 
Optional
as 
?.
. The new website is planned to launch end of this month btw 😉
d
You can think of 
Prism
as 
when/is
and 
Optional
as 
?.
.
Not sure I understand that -- they both have the same getOption...?
Couldn't I have just used Optional instead of Prism in the same way?
Or would that be abusing Optional...
s
Yes, but Optics always exists out of 2 operations. 
Prism
doesn't have 
set
but 
reverseGet
which is required to go from 
List<String>
back to 
List<JsonElement>
.
d
Just a bit curious... why does the 
@optics
annotation generate inline vals? Wouldn't it be more efficient to have regular vals...?
s
It's not possible to generate regular val with backing field as an extension on 
Companion
. In order for use to do that we need stable compiler plugins, if we get those than we can apply more optimisations.
9 Views
Open in Slack
PreviousNext
Powered by

kotlinlang

A modern programming language that makes developers happier.

Powered by