• This repository has been archived on 03/May/2019
  • Stars
    star
    184
  • Rank 209,187 (Top 5 %)
  • Language
    Scala
  • License
    Other
  • Created almost 10 years ago
  • Updated almost 10 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Functional type classes for Scala

The Scala Structures Library

Gitter

The Scala Structures library is a zero-dependency Scala library that defines commonly used type classes for functional programming.

There are two modules, core and laws. The former provides the API of the library and the latter provides Discipline based laws for each of the type classes. The laws project is intended to be used in test scope, to verify that instances of the type classes are law abiding.

FAQ

Is this a fork of Scalaz?

Yes and no. This is a reimplementation of the core type classes which borrows heavily from the ideas in Scalaz. It differs in a number of significant ways though:

  • Structures defines type classes (and a minimal set of types needed for the implementation of those type classes), whereas Scalaz encompasses much more functionality, including data types (e.g., NonEmptyList, Validation, \/), replacements for standard library types (e.g., Order), and monad transformers (e.g., Reader / Kleisli). Structures has a much smaller focus than Scalaz.
  • Structures attempts to fit in with the Scala standard library when possible, even when the standard library provides unsafe functionality. This is contrary to Scalaz, which often provides safe alternatives to standard library constructs. For instance, Maybe is equivalent to Option but omits the unsafe get operation, and Order is equivalent to scala.math.Ordering, but defines the result of a comparison to be an enumerated type rather than an integer. There is great value in using safe alternatives, but providing those alternatives is not in scope of Structures.
  • Structures defines a small subset of the type classes defined by Scalaz. For example, Structures does not provide abstractions for modeling categories, arrows, or adjoint functors.
  • Structures does not provide special syntax support for working with implicit type class instances of odd shapes. Scalaz provides the Unapply type for dealing with odd shapes.

In summary, Structures models a very small subset of what is provided by Scalaz, and does so in a different way, with a different goal.

Why is minimizing API footprint desirable? Can't I just ignore stuff I don't use?

One very real cost of a large API is that there is more stuff to change over time. These changes tend to be organized in to versioned releases, with some claim of binary compatibility or incompatibility. Tooling may enforce the binary compatibility claim. However, tooling typically operates on versioned releases rather than individual artifacts like types and methods. As a result, even if you use the stablest part of a large library, you have to ensure the version of the library you use is binary compatible with the version used by all other libraries in your application, or otherwise risk linkage errors resulting from no tool checked binary compatibility.

Structures aims to be a small, stable library, with binary incompatible releases occurring at the same frequency as major Scala releases.

Will this force library authors to choose between Scalaz and Structures?

Structures is an experiment. It is focused on minimizing the number of type classes and integrating them closely with Scala language features and the standard library. It is entirely possible that this experiment fails in the sense that it no longer should exist as a library. In which case, the overall effort would still be valuable by demonstrating why a smaller subset of type classes is not useful.

OK, but what about cats?

The Typelevel cats project is similar in many ways to Structures. However, cats is more ambitious in what it is attempting. Eventually, cats will grow in to a number of modules that together have similar scope to Scalaz. The focus of Structures is to provide a minimal subset of common FP type classes. As an example of how this goal results in differences, cats will likely include something like Scalaz's Unapply syntax, whereas that is out of scope for Structures.

In short, we believe that there is room for multiple libraries in this space.

Acknowledgements

This project is influenced by a number of other projects. Notably:

This project is also influenced by a number of individuals, who will be added here with their agreement.