Terminal Operations in Stream Processing

A stream pipeline usually ends with a terminal operation. Its job is to turn a whole stream of values into an answer. Asking for the total of a list of numbers or the largest of a set of values are common examples of this kind of question. Writing a separate terminal operation for every possible question would be impractical, so the Stream API offers a more general tool called reduce, which can express many of these computations.

A second tool, collect, answers a different kind of question. Instead of producing one value, it gathers the elements of the stream into a new collection. Both tools end a pipeline. This article explains how each one works and how to tell them apart by looking at what each one returns.

Think of reduce as a rule for combining two values at a time. You give the stream a rule that takes two inputs and returns one output. The stream applies that rule repeatedly, merging values until only one result is left. The best way to see this is to follow the running value step by step. Before you read the diagram, try to predict what the running value will be after each element is combined.

Each arrow in the diagram is one application of the combining rule. The value on the left of each arrow is the accumulated value so far, and the new element is the input from the stream. The final node is the only value that remains once every element has been merged. This is why reduce is said to produce one result from many values.

The source points out a useful gap in the built-in terminal operations. The Stream API has no ready-made terminal operation that computes the product of the values in a stream. You can still get that result, because reduce lets you supply the combining rule yourself. The source's own illustration is the product of all the non-zero elements in an array of double values. That case shows a custom rule filling a missing built-in.

The rule you give reduce must be a BinaryOperator, and the source adds a condition on it: the binary operator should be associative. An operator is associative when the way its applications are grouped does not change the result. The source states the consequence directly: the order in which the operator is applied does not matter. Addition is the standard example, since adding three numbers gives the same total whether you combine the first two or the last two first.

The diagram shows the same values 10, 5, and 2 with two different groupings. Grouping the first two values first gives 5, and then subtracting 2 gives 3. Grouping the last two values first gives 3, and then subtracting that from 10 gives 7. Subtraction is not associative, so the grouping changes the answer. Addition does not have this problem, which is why a sum is safe to compute with reduce regardless of how the stream is split up.

The collect operation has a different purpose. The source describes it as accumulating the input elements into a new Collection. Where reduce returns a single combined value, collect returns a collection that holds the elements that reached it. The usual form names the collector as Collectors.toList(), so the elements are gathered into a List.

The boundary sits at the collect step. Upstream of it, values are still flowing through the stream one at a time. At the collect step, the elements that reached it are accumulated into a new collection object, and that object is what the program keeps after the pipeline finishes. The stream itself is consumed by this process, and the List is the durable result.

This is the practical skill the source points toward. When a final answer is wrong, the cause is usually one of the stages that produced it. For a collect result, the filter is the first place to look. For a reduce result, check both the stage that selects values and the reduce stage that combines them. Tracing each stage's input and output shows which one diverged from what you expected.

Both reduce and collect end a stream pipeline, and the difference is what they return. Reduce combines the stream's values with a BinaryOperator until one result remains. Collect accumulates the elements into a new Collection, such as a List built with Collectors.toList().