Monday, 4 December 2017

Java Optionals For More Expressive Code

Any of us who has programmed in a language that permits null references will have experienced what happens when you try to dereference one. Whether it results in a segfault or a NullPointerException, it’s always a bug. Tony Hoare described it as his billion-dollar mistake. The problem typically occurs when a function returns a null reference to a client that was unanticipated by the developer of the client. Say, in code like this:

User user = userRepository.find("Alice");

An astute programmer will straight away inquire what happens when no user matching “Alice” is found, but nothing in the find() method’s signature tells you what to expect. A typical Java solution to this in the past would have been to make the method throw a checked exception, maybe a UserNotFoundException. That would certainly communicate to the client programmer that this eventuality might happen, but it would do nothing to enhance the expressivity of their code. Catching exceptions results in code that hinders comprehension. In any case, checked exceptions fell out of favour and people tend no longer to write code that throws them.

Many programmers will resort instead to throwing an unchecked exception or returning a null reference. Both are as bad as each other, and for the same reasons: neither of them inform the programmer to expect this eventuality, and both of them will cause a runtime failure if not handled correctly. Java 8 introduced the Optional type to deal with this problem. Whenever you write a method that may or may not return a value, you should make the method return an Optional of whatever type you wish to return. So in our example above, find would return a value of type Optional<User>. The client code now needs to perform additional steps to test the presence of and then get the value:

Optional<User> userOpt = userRepository.find("Alice");
if (userOpt.isPresent()) {
    User user = userOpt.get();  

What’s more, if the code calls get() unguardedly then their IDE will probably warn them about it.

Lambdas Make Things Better

This solution is already a lot better, but there is more to Optional than that: if you stick to handling optionals this way, you are missing out on some opportunities to make your code more expressive.
The code snippet above was adapted from my own implementation of the “social networking” exercise that Codurance use to test job candidates. My actual code is more like:

Optional<User> userOpt = userRepository.find(subject);
if (userOpt.isPresent()) {
    User user = userOpt.get();

Optional has an ifPresent() method that allows us to provide a Consumer that will be called if the optional is present. The consumer’s argument will be the object being wrapped by the optional. This allows us to rewrite the code like this:

userRepository.find(subject).ifPresent(user -> printAllMessagesPostedToUser(user));

Indeed we can go one step further and replace the lambda with a method reference:


I think this communicates the programmer’s intent (in this case mine) much more clearly than the if statement.

Maddeningly, there is no ifNotPresent() counterpart and, even if there were, ifPresent is a void method so they would not be chainable anyway. Java 9 goes some way towards fixing this with its ifPresentOrElse(Consumer<T>, Runnable) method, but it is still not ideal.

Substituting Default Values

On the subject of when the optional value is not present, what can we do? Forgetting complaints about missing features, ifPresent() is only suitable for commands with side effects. If we were implementing a query then we may want to substitute a default value for an empty optional, for example:

if (optionalValue.isPresent()) {
    return optionalValue.get();
return defaultValue;

This can be very easily accomplished with Optional.orElse():

return optionalValue.orElse(defaultValue);

This also provides a convenient way of de-nulling values when you have to call a method that may return null and is not under your control. We have all written code similar to this before:

value = methodThatMayReturnNull();
if (value == null) {
    value = defaultValue;

You can use Optional.ofNullable() to refactor that code, because it returns Optional.empty() if the value is null:

value = Optional.ofNullable(methodThatMayReturnNull()).orElse(defaultValue);

I think this reads a bit better than using ObjectUtils.defaultIfNull to do the same thing. However, there is a caveat. You must not use Optional.orElse() to call a method that has side effects. For example, elsewhere in my social networking exercise I have code that searches for a user and returns it when found, otherwise it creates a new user:

Optional<User> userOpt = userRepository.find(recipient);
if (userOpt.isPresent()) {
    return userOpt.get();
return createUser();

You might assume that you can rewrite this code like this:

return userRepository.find(recipient).orElse(createUser());

You mustn’t do this, because createUser() will always be called whether the optional is present or not! This is almost certainly not what you want: at best you will make an unnecessary method call and, if the method has side effects, it may introduce a bug. Instead you should call Optional.orElseGet() and give it a Supplier that supplies the default value:

return userRepository.find(recipient).orElseGet(() -> createUser());

Now createUser() will only be called when the optional is not present, which is the behaviour I want. Once again we can replace the lambda with a method reference:

return userRepository.find(recipient).orElseGet(this::createUser);

Throwing Exceptions

It may be that, for you, it is an error condition when the optional is not present and you want to throw an exception. You can do this by calling Optional.orElseThrow() and passing it a Supplier that creates the exception:

return userRepository.find(recipient)
        .orElseThrow(() -> new RuntimeException("User " + recipient + " not found"));

Mapping Optional Values

Optional also has some methods that allow you to do operations similar to those on streams. For example, in another exercise I had some code that was structurally similar to this:

Optional<Amount> creditAmountOpt = transaction.getCreditAmount();
Optional<Amount> debitAmountOpt = transaction.getDebitAmount();

String formattedDepositAmount = creditAmountOpt.isPresent() ?
        formatAmount(creditAmountOpt.get()) : " ";

String formattedWithdrawalAmount = debitAmountOpt.isPresent() ?
        formatAmount(debitAmountOpt.get()) : " ";

return String.format(" %s| %s|", formattedDepositAmount, formattedWithdrawalAmount);

The context of this code was a class that prints a bank statement line: my Transaction class knew whether it was a deposit or a withdrawal but I did not want the statement line printer to know. So I had the Transaction interface return optional values for debit and credit amount: the statement line printer would format each value if it was present and substitute a blank space if not.

To avoid the conditional operators, we can use the method. This is very similar to the map method on the Stream API. It accepts a Function and calls it when the optional is present. It passes the wrapped value as the function argument, and wraps the return value in another Optional. So in this case it maps an Optional<Amount> to an Optional<String>. This allows us to rewrite the code like this:

return String.format(" %s| %s|",
        transaction.getDepositAmount().map(this::formatAmount).orElse(" "),
        transaction.getWithdrawalAmount().map(this::formatAmount).orElse(" "));

You might wonder what happens if you map a function that returns another optional – i.e. Function<T, Optional<U>> – in this case you end up with a result of type Optional<Optional<U>> which probably isn’t what you want. Again, similar to a stream, you can use flatMap() instead which will return an Optional<U> value.

The similarities with streams extend to Optional.filter() which evaluates the provided predicate if the optional value is present, and when the predicate evaluates to false then it will return an empty optional. It would be wise to avoid getting too cute though, without care you may end up with code that is difficult to understand. Optionals are best used to refactor code that is straightforward but long-winded into code that is straightforward and more concise.