This is a tiny framework for modelling success/failure of operations in Kotlin. In short, it is a model in type of Result<V, E : Exception>.
Result<V, E: Exception> is to provide higher abstraction of operation that can be ended with result either success or failure. The is somewhat similar to Kotlin's nullable types (T?) (https://kotlinlang.org/docs/reference/null-safety.html).
Result.Success represents value in case of success, and Result.Failure represents error in case of failure which is upper bounded with Exception type.
repositories {
jcenter()
}
dependencies {
compile 'com.github.kittinunf.result:result:<latest-version>' //for jvm
compile 'com.github.kittinunf.result:result-coroutines:<latest-version>' //for kotlin's coroutine support
}
This model is highly inspired by "Railway Oriented Programming" concept.
Result allows one to express series of success/failure operations in Kotlin as;
Result.of(operation)
.flatMap { normalizedData(it) }
.map { createRequestFromData(it) }
.flatMap { database.updateFromRequest(it) }Work with Result is easy
//multi-declaration
val (value, error) = result
//get
val value: Int = result.get<Int>() ?: 0
val ex: Exception = result.get<Exception>()!!
//success
result.success {
}
//failure
result.failure {
}
//fold is there, if you want to handle both success and failure
result.fold({ value ->
//do something with value
}, { error ->
//do something with error
})Combine several results in a validation (without stopping at the first error)
val r1: Result<Int, Exception> = Result.of(1)
val r2: Result<Int, Exception> = Result.of{throw Exception("Not a number")}
val r3: Result<Int, Exception> = Result.of(3)
val r4: Result<Int, Exception> = Result.of{throw Exception("Division by zero")}
val validation = Validation(r1, r2, r3, r4)
validation.hasFailure //true
validation.failures.map{it.message} //[Not a number, Division by zero]Result is suitable whenever there is a need to represent an operation that has the possibility of failure. Error handling can be cumbersome to work with.
Result helps process the operations in a nice, functional way, while maintaining readability to your code.
Let's consider a need to read data from foo, and to perform some further validation
fun process(): String {
try {
val foo = File("/path/to/file/foo.txt").readText()
val isSuccessful = processData(foo)
if (!isSuccessful) {
return "Data is corrupted and cannot be processed"
}
} catch (e: Exception) {
//do something if error
Logger.log(ERROR, e.message())
}
}However, things start getting ugly when we have chain of operations being run sequentially, such as
fun process(): String {
try {
val foo = File("/path/to/file/foo.txt").readText()
val isSuccessful = normalizedData(foo)
if (!isSuccessful) {
return "Data cannot be processable"
}
val request = createRequestFromData(foo)
try {
val result = database.updateFromRequest(request)
if (!result) {
return "Record in DB is not found"
}
} catch (dbEx: DBException) {
return "DB error, cannot update"
}
} catch (e: Exception) {
//do something if error
Logger.log(ERROR, e.message())
}
}Ouch, it looks pretty bleak.
Let's see how Result can help us.
First, we break things down into a small set of model in Result.
- Read a file
val operation = { File("/path/to/file/foo.txt").readText() }
Result.of(operation) // Result<String, FileException>- Normalize a data
fun normalizedData(foo): Result<Boolean, NormalizedException> {
Result.of(foo.normalize())
}- Create a request from data
fun createRequestFromData(foo): Request {
return createRequest(foo)
}- Update DB with Request
fun database.updateFromRequest(request): Result<Boolean, DBException> {
val transaction = request.transaction
return Result.of(db.openTransaction {
val success = db.execute(transaction)
if (!success) {
throw DBException("Error")
}
return success
})
}The whole operation can be chained by the following;
Result.of(operation)
.flatMap { normalizedData(it) }
.map { createRequestFromData(it) }
.flatMap { database.updateFromRequest(it) }The creates a nice "happy path" of the whole chain, also handle error as appropriate. It looks better and cleaner, right?.
In some case, one wants to model an always successful operation. Result<V, NoException> is a good idea for that.
NoException is to indicate that there is no exception to throw. E.g.
// Add operation can never be failure
fun add(i: Int, j: Int) : Result<Int, NoException>Nice thing about modelling in this way is to be able to compose it with others "failable" operations in Result.
map and flatMap
map transforms Result with given transformation (V) -> U. As a result, we are able to transform V into a new V in the case where Result is Result.Success.
When Result is Result.Failure, error is re-wrapped into a new Result.
flatMap is similar to map, however it requires transformation in type of (V) -> Result<U, ...>.
mapError and flatMapError
mapError ((E) -> E2) and flatMapError ((E) -> Result<E2, ...>) are counterpart of map and flatMap. However, they are operate on Result.Failure. It is quite handy when one needs to do some transformation on given Exception into a custom type of Exception that suits ones' need.
These classes are an exact copy of the Result and Validation classes respectively. Use these classes if you are planning on using coroutines in your functions.
If interested, here are more articles that one might enjoy.
- http://fsharpforfunandprofit.com/posts/recipe-part2/
- https://speakerdeck.com/swlaschin/railway-oriented-programming-a-functional-approach-to-error-handling
- https://github.com/swlaschin/Railway-Oriented-Programming-Example
Credit to Scott Wlaschin
Result is brought to you by contributors.
Result is released under the MIT license.