Fast, safe and composable streaming abstractions

Streaming abstractions that combine, transform and reduce large amounts of sequential data efficiently, in constant space and without leaking resources.

Streaming uses composable stream producers (sources), consumers (sinks) and transformers (flows).

The following features are provided:

• Constant memory usage: large or infinite streams can be computed in constant and small space. Buffering of the input is possible when needed.
• Excellent performance: all models were designed with performance at the core. See https://github.com/rizo/streams-bench for detailed comparison with other libraries.
• Resource safety: resources in effectful streaming pipelines are allocated lazily and released as early as possible. Resources are guaranteed to be terminated even when streams rise exceptions.
• Flexibility: both push-based and pull-based models are implemented to allow efficient zipping, concatenation and other streaming operations.
• Streaming notation: build streams and sinks using a convenient comprehension and applicative notations.