Maybe<T>
is a monadic wrapper similar to java.util.Optional
, but with a different intention. By leveraging Optional<T>
benefits, it provides a functional API that safely allows to work with operations that throw checked (and unchecked) exceptions.
The motivation of Maybe<T>
is to help developers avoid imperative try/catch blocks while promoting safe exception handling which lives by functional programming principles.
- Type-safe differentiation between resolving a value vs. runnning effects.
- Rich and intuitive API based on
java.util.Optional
. - Full interoperability with
java.util.Optional
. - Includes a safe
Either<L, R>
type where only one side can be present at a time. - Method reference friendly - The API provides methods with overloads that makes it easier to use method reference syntax.
As of v3.3.0
, this library is compatible with JDK11+ by using Multi-Release JARs. However, using at least JDK17 to enjoy the new Java enhancements and features is highly recommended.
For example, the JDK17+ version of Either<L, R>
uses a combination of sealed classes and records classes, effectively making it an algebraic data type in Java. It means that Either<L, R>
is an interface no other class can implement. The only implementations are the Left
and Right
records, which live within the interface. In short, it's a composite type created by combining other types.
If you need a JDK8-compatible version of
Maybe<T>
, you can use v2 instead. However, much like Java 8, v2 has reached its end-of-life, so it will not get any more features, patches, or updates.
- Due to changes on GitHub policies (and by consequence on Maven), it's no longer allowed to use
com.github
as a valid group ID prefix. To honor that and maintain consistency, as of v3, the artifact ID was renamed toio.github.joselion.maybe
. If you want to use a version before v3, you can still find it usingcom.github.joselion:maybe
artifact. - A
SolveHandler
can no longer be empty. It either has the solved value or an error. - The method
SolveHandler#filter
was removed to avoid the posibility of an inconsitent empty handler. - The
WrapperException
type was removed. Errors now propagate downstream with the API. - The method
EffectHandler#toMaybe
was removed as it didn't make sense for effects. - All
*Checked.java
functions were renamed toThrowing*.java
- For example,
FunctionChecked<T, R, E>
was renamed toThrowingFunction<T, R, E>
- For example,
Maybe is available in Maven Central. You can checkout the latest version with the badge above.
Gradle
implementation('io.github.joselion:maybe:x.y.z')
Maven
<dependency>
<groupId>io.github.joselion</groupId>
<artifactId>maybe</artifactId>
<version>x.y.z</version>
</dependency>
We'd use Maybe<T>
for 3 different cases:
- Solve: When we need to obtain a value from a throwing operation.
- Effects: When we need to run effects that may throw exception(s), so no value is returned.
- Closeables: When we need to use a closeable resources on another operation (as in
try-with-resource
blocks)
We can create simple instances of Maybe using Maybe.of(value)
or Maybe.empty()
so we can chain throwing operations to it that will create the handlers. We also provide the convenience static methods .from(ThrowingSupplier)
and .from(ThrowingRunnable)
to create handlers directly from lambda expressions. Given the built-in lambda expression do not allow checked exception, we provide a few basic functional interfaces like ThrowingFunction<T, R, E>
, that are just like the built-in ones, but with a throws E
declaration. You can find them all in the util packages of the library.
Once a solver operation runs we'll get a SolveHandler instance. This is the API that lets you handle the possible exception and produce a final value, or chain more operations to it.
final Path path = Paths.get("foo.txt");
final List<String> fooLines = Maybe.from(() -> Files.readAllLines(path))
.doOnError(error -> log.error("Fail to read the file", error)) // where `error` has type IOException
.orElse(List.of());
// or we could use method reference
final List<String> fooLines = Maybe.of(path)
.solve(Files::readAllLines)
.doOnError(error -> log.error("Fail to read the file", error)) // where `error` has type IOException
.orElseGet(List::of); // the else value is lazy now
The method .readAllLines(..)
on example above reads from a file, which may throw a IOException
. With the solver API we can run an effect if the exception was thrown. The we use .orElse(..)
to safely unwrap the resulting value or another one in case of failure.
When an effect operation runs we'll get a EffectHandler instences. Likewise, this is the API to handle any possinble exception the effect may throw. This handler is very similar to the SolveHandler, but given an effect will never solve a value, it does not have any of the methods related to manipulating or unwrapping the value.
Maybe
.from(() -> {
final String to = ...
final String from = ...
final String message = ...
MailService.send(message, to, from);
})
.doOnError(error -> { // the `error` has type `MessagingException`
MailService.report(error.getMessage());
});
In the example above the .send(..)
methods may throw a MessagingException
. With the effect API we handle the error running another effect, i.e. reporting the error to another service.
Maybe also offers a way to work with AutoCloseable resources in a similar way the try-with-resource
statement does, but with a more functional approach. We do this by creating a CloseableHandler instance from an autoclosable value, which will hold on to the value to close it at the end. The resource API lets you solve values or run effects using a closable resource, so we can ultimately handle the throwing operation with either the SolveHandler or the EffectHandler.
Maybe.withResource(myResource)
.solve(res -> {
// Return something using `res`
});
Maybe.withResource(myResource)
.effect(res -> {
// do somthing with `res`
});
In many cases, the resource you need will also throw an exception when we obtain it. We encourage you to first handle the exception that obtaining the resource may throw, and then map the value to a CloseableHandler to handle the next operation. For this SolveHandler provides a .mapToResource(..)
method so you can map solved values to resources.
public Properties parsePropertiesFile(final String filePath) {
return Maybe.of(filePath)
.solve(FileInputStream::new)
.catchError(err -> /* Handle the error */)
.mapToResource(Function.identity())
.solve(inputStream -> {
final Properties props = new Properties();
props.load(inputStream);
return props;
})
.orElseGet(Properties::new);
}
We know the first solved value extends from
AutoCloseable
, but the compiler doesn't. We need to explicitly map the value withFunction.identity()
so the compiler can safely ensure that the resource can be closed.
Some operation may throw multiple type of exceptions. We can choose how to handle each one using one of the .catchError(..)
matcher overloads. This method can be chained one after another, meaning the first one to match the exception type will handle the error. However, the compiler cannot ensure exhaustive matching of the error types (for now), so we'll always need to handle a default case with a terminal operator.
Maybe.of(path)
.solve(Files::readAllLines) // throws IOException
.catchError(FileNotFoundException.class, err -> ...)
.catchError(FileSystemException.class, err -> ...)
.catchError(EOFException.class, err -> ...)
.orElse(err -> ...);
The API has overloads where parameters can be passed as lambda expressions. Even though the parameter types are not ambiguous, there's one particular case where a lambda expression may not be inferred as the expected type:
() -> {
throw new Exception("...");
}
arg -> {
throw new Exception("...");
}
If the lambda expression finishes throwing an exception, the compile will not be able to tell the difference between a type that returns a value and a type that doesn't. For example, the first lambda expression above can be either a ThrowingRunnable<Exception>
or a ThrowingSupplier<Object, Exception>
. Similarly, the second expression can either be a ThrowingConsumer<Object, Exception>
or a ThrowingFunction<Object, Object, Exception>
.
If you ever run into ambiguity issues in the Maybe API, you have 2 easy options to solve the problem:
- (Recommended) Explicity setting the generic type(s) of the method will deambiguate the overload you expect to be used.
Maybe.<IOException>from(() -> { // param is a `ThrowingRunnable<IOException>`
throw new IOException("...");
});
Maybe.<String, IOException>from(() -> { // param is a `ThrowingSupplier<String, IOException>`
throw new IOException("...");
});
- Casting the lambda expression to the specific
@FunctionalInterface
type you want to use.
Maybe.from((ThrowingRunnable<IOException>) () -> {
throw new IOException("...");
});
Maybe.from((ThrowingSupplier<String, IOException>) () -> {
throw new IOException("...");
});
An awesome extra of Maybe
, is that it provides a useful Either<L, R> type which guarantees that the only one of the sides (left L
or right R
) is present per instance. That is possible thanks to:
Either<L, R>
is a sealed interface. It cannot be implemented by any class nor anonimously instantiated in any way.- There only exist 2 implementations of
Either<L, R>
:Either.Left
andEither.Right
. In those implementations, only one field is used to store the instance value. - It's not possible to create an
Either<L, R>
instance of anull
value.
The Either<L, R>
makes a lot of sense when resolving values from throwing operations. At the end of the day, you can end up with either the solved value (Rigth
) or the thrown exception (Left
). You can convert from a SolveHandler<T, E>
to an Either<E, T>
usong the SolveHandler#toEither
terminal operator.
To use Either
on its own, use the factory methods to create an instance and the API to handle/unwrap the value:
public Either<String, Integer> fizzOrNumber(final int value) {
return value % 7 == 0
? Either.ofLeft("fizz")
: Either.ofRight(value);
}
public static void main (final String[] args) {
final var sum = IntStream.range(1, 25)
.boxed()
.map(this::fizzOrNumber)
.map(either ->
either
.onLeft(fizz -> log.info("Multiple of 7: {}", fizz))
.onRight(value -> log.info("Value: {}", value))
.unwrap(
fizz -> 0,
Function.identity()
)
)
.reduce(0, Integer::sum);
log.info("The sum of non-fizz values is: {}", sum);
}
Take a look at the documentation to see all the methods available in the Either<L, R>
API.
The API provides full interoperability with Java's Optional
. You can use Maybe.from(Optional)
overload to create an instance from an optional value, or you can use the terminal operator .toOptional()
to unwrap the value to an optional too. However, there's a change you might want to create a Maybe<T>
withing the Optional API or another library like Project Reactor, like from a .map(..)
method. To make this esier the API provides overloads to that create partial applications, and when fully applied return the specific handler.
So instead of having nested lambdas like this:
Optional.ofNullable(rawValue)
.map(str -> Maybe.from(() -> Base64.getDecoder().decode(str)))
.map(decoded -> decoded.catchError(...));
You can use the partial application overload and use method reference syntax:
Optional.ofNullable(rawValue)
.map(Maybe.partial(Base64.getDecoder()::decode))
.map(decoded -> decoded.catchError(...));
You can find more details of the API in the latest Javadocs.If you need to check the Javadocs of an older version you can also use the full URL as shown below. Just replace <x.y.z>
with the version you want to see:
https://javadoc.io/doc/io.github.joselion/maybe/<x.y.z>
Suggestions are always welcome! Please create an issue describing the request, feature, or bug. I'll try to look into it as soon as possible 🙂
Contributions are very welcome! To do so, please fork this repository and open a Pull Request to the main
branch.