6.1 KiB
effectful
Note: this is a pre-release.
An easy to use, performant extensible effects library with seamless integration with the existing Haskell ecosystem.
Main features:
-
Very fast (benchmark results with GHC 8.8.4: countdown, filesize, comparison with eff).
-
Easy to use API (no boilerplate code and dealing with arcane types).
-
Correct semantics in presence of runtime exceptions (no more lost or discarded state).
-
Seamless integration with the existing ecosystem (
exceptions,monad-control,unliftio-core,resourcetetc.). -
Effects can be defined for either
-
static dispatch (as fast as it gets, single interpretation) or
-
dynamic dispatch (a bit slower, multiple interpretations),
depending on your needs.
-
Motivation
Do we really need yet another library for handling effects? There's freer-simple, fused-effects, polysemy, eff and probably a few more.
Unfortunately, of all of them only eff is a promising proposition because of
reasonable performance characteristics (see the talk "Effects for Less" linked
below for more information) and potential for good interoperability with the
existing ecosystem.
The second point is arguably the most important, because it allows focusing on things that matter instead of reinventing all kinds of wheels and is crucial for adoption of the library.
However, eff uses delimited continuations underneath, which:
-
Are not yet supported by GHC (though the proposal for including support for them has been accepted).
-
Are quite hard to understand.
-
Make the library "too powerful" in a sense as it faces a few issues with no clear path towards their resolution.
On the other hand, if support for continuations is excluded from the handler
monad, then the ability to define effects with non-linear control flow (such as
NonDet) is lost. Arguably it's a small price to pay for predictability,
because such specialized effects are needed rarely and locally, at which point a
dedicated, well established solution such as
conduit,
list-t or
logict can be used.
This is where effectful comes in. The Eff monad it uses is essentially a
ReaderT over IO on steroids, allowing us to dynamically extend its
environment with data types that represent effects or their handlers.
Because this concept is so simple:
-
It's reasonably easy to understand what is going on under the hood.
-
The
Effmonad being a reader allows for seamless interoperability with ubiquitous classes such asMonadBaseControlandMonadUnliftIOas well as support for handling runtime exceptions without worrying about lost or discarded state (see the talk "Monad Transformer State" linked below for more information).
What is more:
-
The
Effmonad is concrete, so GHC has many possibilities for optimization, which results in a very fast code at a default optimization level. There is no need to mark every functionINLINEor enable additional optimization passes, it just works. -
If an advanced effect with non-linear control flow is needed, you can always stick a transformer that implements it on top of
Effin a local context.
In conclusion, effectful aims to reduce duplication and bring back performance
to "boring" transformer stacks, most of which are a dozen of newtype'd StateT
or ReaderT transformers, each with a few associated operations (usually tied
to a type class), not to replace monad transformers altogether.
Usage
The effect system and its effects are split among several libraries:
-
The
effectful-corelibrary contains the main machinery of the effect system itself and a few basic effects. It aims for a small dependency footprint and provides the building blocks for more advanced effects. -
TBD
effectful-resource,effectful-process, ... -
Finally, the
effectfullibrary which comes with 'batteries included'. It is build on top of the other libraries and re-exports the functionality of those.
Example
A Filesystem effect with two handlers, one that runs in IO and another that
uses an in-memory virtual file system can be found
here.
Resources
Resources that inspired the rise of this library and had a lot of impact on its design.
Talks:
-
Effects for Less by Alexis King.
-
Monad Transformer State by Michael Snoyman.
Blog posts:
-
ReaderT design pattern by Michael Snoyman.
-
Exceptions Best Practices by Michael Snoyman.