---
title: "Streams and data pipelines"
chapter: "08"
---

# Streams and data pipelines

A stream describes a pipeline over data. It does not store elements.

## Pipeline shape

1. Source: collection, array, file lines, generator.
2. Intermediate operations: `filter`, `map`, `flatMap`, `sorted`, `distinct`.
3. Terminal operation: `toList`, `collect`, `reduce`, `count`, `findFirst`.

Intermediate operations are lazy. Nothing runs until a terminal operation
needs results.

## Stateless behavior

Stream functions should be non-interfering and normally stateless. Mutating an
external list in `forEach` creates hidden coupling and breaks parallel safety.
Prefer collectors or immutable transformations.

## Mapping and flattening

`map` turns each element into one result. `flatMap` turns each element into
zero or more results and flattens them. `mapMulti` can emit multiple results
without creating an intermediate stream per input.

## Reduction and collectors

A reduction combines elements with an identity and associative accumulator.
For parallel execution, the operation must behave associatively. Collectors
build lists, maps, groups, partitions, summaries, and custom containers.
`toMap` needs a merge rule when keys may repeat.

## Parallel streams

Parallel does not mean faster. Work uses a common pool by default and depends
on data size, splitting, CPU cost, order, blocking, and contention. Avoid
parallel streams for blocking I/O and measure CPU pipelines before adopting.

## Reuse and side effects

A stream is single-use. Terminal operation closes the pipeline. Streams backed
by I/O should be closed with try-with-resources.

## Feynman check

A stream is a conveyor belt. Filters remove parcels, maps relabel them,
flatMap opens boxes into several parcels, and a collector packs the result.