Wondering about using kotest to test Arrow Raise/E...

# arrow

[Chris Lee]

Wondering about using kotest to test Arrow Raise/Either w/ context receivers; prior to using context receivers we had an Either return value, and could use .shouldBeRight() and .shouldBeLeft. Now we need extra work to establish a context to execute the code being tested.

I’ve currently been using a pattern like this:

test("UUID converter works") {
            either {
                val uuid = UUID.randomUUID()
                converter.convert(ResolvedNodes.string(uuid.toString()), type, conversionContext())
                    .shouldBe(uuid)
            }.shouldBeRight()

            either {
                converter.convert(ResolvedNodes.string(""), type, conversionContext())
            }.shouldBeLeft()
                .shouldBeInstanceOf<ConversionError.TypeConversionError>()
                .message()
                .shouldBe(
                    "Error converting '' into 'java.util.UUID': IllegalArgumentException: Invalid UUID string: "
                )
        }

…which can be cleaned up a bit with a few helper shims:

test("UUID converter works") {
            shouldBeRight {
                val uuid = UUID.randomUUID()
                converter.convert(ResolvedNodes.string(uuid.toString()), type, conversionContext())
                    .shouldBe(uuid)
            }

            shouldBeLeft {
                converter.convert(ResolvedNodes.string(""), type, conversionContext())
            }.shouldBeInstanceOf<ConversionError.TypeConversionError>()
                .message()
                .shouldBe(
                    "Error converting '' into 'java.util.UUID': IllegalArgumentException: Invalid UUID string: "
                )
        }

        public inline fun <Error, A> shouldBeRight(@BuilderInference block: Raise<Error>.() -> A): A = either(block).shouldBeRight()
        public inline fun <Error, A> shouldBeLeft(@BuilderInference block: Raise<Error>.() -> A): Error = either(block).shouldBeLeft()

It seems like there are more opportunities to further improve this; ideally the semantics would be:
a) a way to execute code in a context w/o having to establish that context explicitly;
b) semantics similar to exceptions where the happy-path requires no extra code, anything raised fails the test
c) shouldRaise<ErrorType> similar to shouldThrow<ExceptionType> to handle the error path

Perhaps parts of this exist already and I haven’t yet tracked them down.

[Javier]

I haven’t checked the PR deeply yet, but I think @Youssef Shoaib [MOD] added some samples with Kotest in the arrow-context repo.