ConceptDifficulty: Advanced1 min

Collectors: Ending a Stream in a List, a Set or a Map

Decide the shape you want out, and the collector is usually already written.

collect is the terminal operation (the last step, which produces a result and makes the stream run) that builds a collection from a stream, and the Collectors class supplies ready-made recipes, called collectors, for lists, sets, maps and strings.

Three pipelines ending in three shapes:

java

List<String> refs   = invoices.stream().map(Invoice::reference).toList();
Set<String> unique  = invoices.stream().map(Invoice::customer).collect(Collectors.toSet());
Map<String, Long> byCustomer =
        invoices.stream().collect(Collectors.groupingBy(Invoice::customer, Collectors.counting()));

Line 1 needs no collector at all, because toList() is a method on the stream itself. Lines 2 to 4 hand collect a recipe, and the recipe decides both the kind of collection and what happens to duplicates.

The collectors worth knowing by name
Collector	Produces
`toList()` on the stream	An unmodifiable list (adding to it throws), in encounter order
`Collectors.toSet()`	A set, with duplicates gone and no promised order
`Collectors.toMap(key, value)`	A map, throwing if two elements produce one key
`Collectors.groupingBy(key)`	A map from key to a list of the elements with it
`Collectors.joining(", ")`	One string, with a separator

groupingBy is the one that replaces the most code. A second collector inside it says what to do with each group, so counting, summing and collecting to a set are all the same call with a different second argument.

So the end of a pipeline is a choice about shape rather than about work. Decide what you want out, and the collector is usually already written.

← → move through sections and cards