Hi. Since upgrading to 1.1.3 I have the following ...

# announcements

[kschlesselmann]

Hi. Since upgrading to 1.1.3 I have the following issue: I have this value object

@Embeddable
data class RouterId(val routerId: String) {

    init {
        val octets = routerId.split(":")

        if (octets.size != 6) {
            throw IllegalArgumentException("$routerId does not consist of 6 octets")
        }

        for (octet in octets) {
            Integer.parseInt(octet, 16)
        }
    }

    override fun toString(): String {
        return routerId
    }
}

which now fails with

java.lang.IllegalArgumentException: Parameter specified as non-null is null: method kotlin.text.StringsKt__StringsKt.split, parameter $receiver
	at kotlin.text.StringsKt__StringsKt.split(Strings.kt) ~[kotlin-stdlib-1.1.3.jar:1.1.3]
	at kotlin.text.StringsKt__StringsKt.split$default(Strings.kt:1153) ~[kotlin-stdlib-1.1.3.jar:1.1.3]
	at domain.model.aircraft.RouterId.<init>(RouterId.kt:9) ~[main/:na]
        ....

during creation with JPA. In my build.gradle I use

plugins {
    id "org.jetbrains.kotlin.jvm" version "1.1.3"
    id "org.jetbrains.kotlin.plugin.noarg" version "1.1.3"
    id "org.jetbrains.kotlin.plugin.jpa" version "1.1.3"

    id "org.springframework.boot" version "1.5.4.RELEASE"
}

any suggestions? The code runs fine using 1.1.2.

[Czar]

kschlesselmann: I believe, it should not actually run in 1.1.2, probably 1.1.3 fixes the bug. the no-arg plugin generates a no-arg constructor, and you have init {} which works with a constructor parameter. JPA will call no-arg constructor, then your init will run and will find a routerId == null.

[kschlesselmann]

Yreah, I thought something like that. The code really works in 1.1.2. So what would you suggest in this case?

[lovis]

you could just make routerId optional or provide a default value like 0:0:0:0:0:0

[kschlesselmann]

@lovis But this would defeat the purpose of a proper value object

[lovis]

I kind of agree, but data classes are not value objects. Your problem is that your framework needs a constructor that you don’t want, so that’s imho the best workaround.

[lovis]

or you move the creation logic to a companion object. (aka factory method) whoops that won’t help here. (but it would be better anyway because now you actually throw in the constructor, which is imho bad)

[lovis]

but what what will break if you use a default routerId? you could even add a method isValid or something that checks for that particular one

[lovis]

the more I think about it I really don’t see how it defeats the purpose of a value object

[kschlesselmann]

Why is this data class no value object?

[lovis]

because the jvm does not support value objects (aka structs). so data classes are still reference objects, just with an equals implementation that makes it behave like a value object would.
however, I’m being pedantic here, sorry 😉

[lovis]

The important part is that a “value object” should
* be immutable ☑️
* have no identity (i.e equals depends on the fields, not on the object identity) ☑️
* be created in a valid state ☑️
and using 0:0:0:0:0:0 seems like a valid state for the object itself

[kschlesselmann]

So kotlin wonÄt complain if the routerId is still a val which gets modified by JPA later on?

[kschlesselmann]

I'm quite aware that those are not real value objects because the JVM doesn't support that 😉

[lovis]

like i said i was being pedantic. it didn’t help you so i’m sorry about that 🙂
it shouldn’t complain. but I don’t know how your framework works, for gson, which uses reflection it does work like that

[kschlesselmann]

I'm using hibernate here. It was so nice to just use the plugin and have unwanted constructors generated … damn JPA 😛