potoki

Simple streaming in IO

https://github.com/metrix-ai/potoki

Latest on Hackage:2.1.4.1

This package is not currently in any snapshots. If you're interested in using it, we recommend adding it to Stackage Nightly. Doing so will make builds more reliable, and allow stackage.org to host generated Haddocks.

MIT licensed by Nikita Volkov
Maintained by Metrix.AI Tech Team

This library provides a new simpler approach to the IO-streaming problem.

In difference to libraries like "pipes", "conduit", "streaming", this library is specialised to streaming in the IO monad, which greatly simplifies the abstractions that it provides. This simplification is motivated by the fact that the majority of streaming tasks are performed in IO anyway.

Also, unlike the mentioned libraries, "potoki" API doesn't treat streaming as a side operation in its abstractions, which allows it to express the composition of streams using the standard typeclass instances, thus simplifying the API even further.

Naturally, being simpler limits the application area of this library. Thus it is not capable of transforming custom context monads and etc. It is a tradeoff, but, as we expect, the user will rarely be affected by it.

Another benefit of being specialized to IO is the ability to optimize for performance better. It must however be mentioned that this is only theoretical and no benchmarks have yet been performed.

In some of the mentioned regards "potoki" is similar to the "io-streams" library. However, unlike that library it approaches composition with the standard typeclass instances. Also, in difference to "io-streams", "potoki" doesn't use exceptions for control-flow. In fact, "potoki" doesn't use exceptions whatsoever, instead it makes failures explicit, using the standard Either type.

"potoki" comes with automated resource-management (acquisition and clean-up), concurrency and buffering features.