machisuji / monocle

Lens library in Scala

Build

import sbt._
resolvers += Resolver.sonatypeRepo("releases")
resolvers += Resolver.sonatypeRepo("snapshots")

val scalaVersion   = "2.11.2" // or "2.10.4"
val libraryVersion = "0.5.1"  // or "1.0.0-SNAPSHOT"

libraryDependencies ++= Seq(
  "com.github.julien-truffaut"  %%  "monocle-core"    % libraryVersion,
  "com.github.julien-truffaut"  %%  "monocle-generic" % libraryVersion,
  "com.github.julien-truffaut"  %%  "monocle-macro"   % libraryVersion,         // since 0.4.0
  "com.github.julien-truffaut"  %%  "monocle-law"     % libraryVersion % "test" // since 0.4.0
)

Motivation

Monocle is a Lens library, or more generally an Optics library where Optics gather the concepts of Lens, Traversal, Optional, Prism and Iso. Monocle is strongly inspired by Haskell Lens.

What does it mean?

Optics are a set of purely functional abstractions to manipulate (get, set, modify) immutable objects. Optics compose between each other and particularly shine with nested objects.

Why do I need this?

Scala already provides getters and setters for case classes but modifying nested object is verbose which makes code difficult to understand and reason about. Let's have a look at some examples:

case class Street(name: String, ...)     // ... means it contains other fields
case class Address(street: Street, ...)
case class Company(address: Address, ...)
case class Employee(company: Company, ...)

Let's say we have an employee and we need to set the first character of his company street name address in upper case. Here is how we could write it in vanilla Scala:

val employee: Employee = ...

employee.copy(
  company = employee.company.copy(
    address = employee.company.address.copy(
      street = employee.company.address.street.copy(
        name = employee.company.address.street.name.capitalize // luckily capitalize exists
      }
    )
  )
)

As you can see copy is not convenient to update nested objects as we need to repeat at each level the full path to reach it. Let's see what could we do with Monocle:

val _name   : SimpleLens[Street  , String]  = ...  // we'll see later how to build Lens
val _street : SimpleLens[Address , Street]  = ...
val _address: SimpleLens[Company , Address] = ...
val _company: SimpleLens[Employee, Company] = ...

(_company composeLens _address composeLens _street composeLens _name).modify(employee, _.capitalize)

import monocle.syntax._ // to use optics as operator 

employee applyLens   _company
         composeLens _address
         composeLens _street
         composeLens _name
         modify (_.capitalize)
         
// or with some syntax sugar
employee |-> _company |-> _address |-> _street |-> _name modify (_.capitalize)

ComposeLens takes two Lens, one from A to B and another from B to C and creates a third Lens from A to C. Therefore, after composing _company, _address, _street and _name, we obtain a Lens from Employee to String (the street name).

More abstractions

In the above example, we used capitalize to upper case the first letter of a String. It works but it would be clearer if we could use Lens to zoom into the first character of a String. However, we cannot write such a Lens because a Lens defines how to focus from an object S into a mandatory object A and in our case, the first character of a String is optional as a String might be empty. For this we need a sort of partial Lens, in Monocle it is called Optional.

import monocle.syntax._
import monocle.function.HeadOption._ // to use headOption (a polymorphic optic)
import monocle.std.string._          // to get String instance for HeadOption


employee applyLens   _company
         composeLens _address
         composeLens _street
         composeLens _name
         composeOptional headOption
         modify toUpper
         
// or with some syntax sugar         
employee |-> _company |-> _address |-> _street |-> _name |-? headOption modify toUpper

Similarly to composeLens, composeOptional takes two Optional, one from A to B and another from B to C and creates a third Optional from A to C. All Lens can be seen as Optional where the optional element to zoom to is always present, hence composing an Optional and a Lens always produces an Optional (see class diagram for full inheritance relation between Optics).

For more examples, see the example module.

Lens Creation

There are four ways to create SimpleLens, each with their pro and cons:

1. The first method is the most verbose but it fully documents the type of the SimpleLens created. Remark, this is the only constructor that can also be used for Lens (i.e. with 4 type parameters):

   val _company = SimpleLens[Employee, Company](_.company, (e, c) => e.copy(company = c))

2. The next solution is slightly shorter as it does not require to define the second type parameter of SimpleLens (it is inferred from the type of the first argument):

   val _company = SimpleLens[Employee](_.company)((e, c) => e.copy(company = c))

3. Now we start using the heavy artillery ... Macro. This method can only be used on case classes and since macros are experimental in scala, there is no guarantee that future version of Monocle will support it (even though we will do our best):

    import monocle.Macro._ // require monocle-macro dependency
   
    val _company = mkLens[Employee, Company]("company") // company is checked at compiled time to be a valid accessor

   Note: this macro is deprecated in 0.5.1

4. An alternative Macro syntax uses a dedicated object to capture the class, and a simple closure to define the field. This syntax is more IDE-friendly.

   val lenser = Lenser[Employee]
   
   val _company = lenser(_.company) 

   A Lenser can be in-lined or re-used to avoid specifying the class type parameter.

5. Finally, the boiler plate free solution with macro annotation (which are probably the most experimental part of macros). Adding @Lenses annotation on case class will generate SimpleLens for every single accessor of the case class. These generated SimpleLens are in the companion object of the case class (even if there is no companion object declared). Nevertheless, this solution has several disadvantages:

   1. users need to add the macro paradise plugin to their project.
   2. IDE have a poor support for Macro annotation, therefore it is likely your IDE will not know that the generated SimpleLens exist (but it will compile). If you want a better IDE support, please vote on the following issue.
   3. this solution can only be applied when you control the case classes since you need to annotate them. This means that you cannot use this technique for classes defined in another project.

   @Lenses
   case class Employee(company: Company, name: String, ...)
   
   // generates Employee.company: SimpleLens[Employee, Company]
   // and       Employee.name   : SimpleLens[Employee, String]

Polymorphic Optics and Instance Location Policy

A polymorphic optic is an optic that is applicable to different types. For example, headOption is an Optional from some type S to its optional first element of type A. In order to use headOption (or any polymorphic optics), you need to:

1. import the polymorphic optic in your scope via import monocle.function.headoption._ or import monocle.function._
2. have the required instance of the type class HeadOption in your scope, e.g. if you want to use headOption from a List[Int], you need an instance of HeadOption[List[Int], Int]. This instance can be either provided by you or by Monocle.

Monocle defines polymorphic optic instances in the following packages:

1. monocle.std for standard Scala library and Scalaz classes, e.g. List, Vector, Map, IList, OneAnd
2. monocle.generic for Shapeless classes, e.g. HList, CoProduct

An example shows how to use Monocle imports.

Overview

Sub Projects

Core contains the main library concepts: Lens, Traversal, Prism, Iso, Getter and Setter. Core only depends on scalaz for type classes.

Law defines Iso, Lens, Prism, Setter and Traversal laws using scalacheck.

Macro defines a macro to reduce Lens creation boiler plate.

Generic is an experiment to provide highly generalised Lens and Iso using HList from shapeless. Generic focus is on neat abstraction but that may come at additional runtime or compile time cost.

Example shows how other sub projects can be used.

Contributor Handbook

We are happy to have as many people as possible contributing to Monocle. Therefore, we made this small workflow to simplify the process:

1. Select or create an issue (issues tagged with label "padawan-friendly" are designed for Scala novice)
2. Comment on the issue letting everyone knows that you are working on it.
3. Fork Monocle
4. Work on your fork until you are satisfied (label your commits with issue number)
5. Submit a pull request
6. We will review your pull request and merge it back to master

If you have any questions, we have irc channel on freenode #scala-monocle and a mailing group

Thank you for you contribution!

Contributors

Julien Truffaut - @JulienTruffaut
Ross Huggett - [email protected] / @rosshuggett
Ilan Godik - [email protected] / NightRa
Adam Warski - @adamwarski
Dale Wijnand - @dwijnand