Building Streams – Streams

16.4 Building Streams

A stream must have a data source. In this section we will explore how streams can be created from various data sources: collections, arrays, specified values, generator functions, strings, and I/O channels, among others.

Aspects to Consider When Creating Streams

When creating a stream from a data source, certain aspects to consider include whether the stream is:

• Sequential or parallel
• Ordered or unordered
• Finite or infinite
• Object or numeric

Sequential or Parallel Stream

A sequential stream is one whose elements are processed sequentially (as in a for loop) when the stream pipeline is executed by a single thread. Figure 16.1 illustrates the execution of a sequential stream, where the stream pipeline is executed by a single thread.

A parallel stream is split into multiple substreams that are processed in parallel by multiple instances of the stream pipeline being executed by multiple threads, and their intermediate results combined to create the final result. Parallel streams are discussed in detail later (p. 1009).

The different ways to create a stream on a data source that are illustrated in this section result in a sequential stream. A parallel stream can only be created directly on a collection by invoking the Collection.parallelStream() method (p. 897).

The sequential or parallel mode of an existing stream can be modified by calling the BaseStream.sequential() and BaseStream.parallel() intermediate operations, respectively (p. 933). A stream is executed sequentially or in parallel depending on the execution mode of the stream on which the terminal operation is initiated.

Ordered or Unordered Stream

The encounter order of a stream refers to the way in which a stream makes its elements available for processing to an operation in a pipeline. For such data sources as a list, the encounter order of the initial stream is the same as the order of the elements in the list, whereas a stream created with a set of values does not have an encounter order, as the elements of a set are considered to be unordered.

The encounter order of a stream may be changed by an intermediate operation. For example, the sorted() operation may impose an encounter order on an unordered stream (p. 929), and the unordered() operation may designate a stream unordered (p. 932). Also, some terminal operations might choose to ignore the encounter order; an example is the forEach() operation (p. 948).

For ordered sequential streams, an identical result is produced when identical stream pipelines are executed on an identical data source—that is, the execution is deterministic. This guarantee does not hold for unordered sequential streams, as the results produced might be different.

Processing of unordered parallel streams may have better performance than for ordered parallel streams in identical stream pipelines when the ordering constraint is removed, as maintaining the order might carry a performance penalty.