Skip to content

Lilaisms

Code concepts and functions that are specific to, and idiomatic to, the lila codebase.

How lila uses scala

lila makes little use of object orientation and particularly inheritance, but:

  • Immutability is required everywhere. Notable exceptions can be made:
  • in akka actors (because race conditions are no longer a concern)
  • in performance sensitive functions. In this case, the mutability must be contained in the function scope. The function must expose an immutable signature.
  • in atomic structures like java's AtomicReference or ConcurrentHashMap.
  • Strong typing is preferred. For instance, FiniteDuration is better than Int, and case class Name(value: String) is better than String. The more the value is used in the code, and the more useful it is to type it correctly. This rule is actually loosely respected in lila, but that's the direction we're going to.

Weird code you will only see in lila

If the function you're looking for is not here, try the scala doc or the cats doc.

Disclaimer

Some of these lilaisms are driven by a compulsive tendency to play code golf. However, they also present benefits in terms of type safety. In any case, contributors are never expected to use lilaisms. You can make use of them, but a pull request will likely not be rejected because it isn't idiomatic to lila. However, immutability and type safety remain requirements (with discussable exceptions).

Debugging (actually useful)

pp

pp means print & pass, it's an inline println. It prints the value and returns it.

"foo".pp                   // prints "foo" to stdout, and returns "foo"

player.pp.make(move.pp).pp // behaves like player.make(move), but prints
                           // player, move, and the result of player.make(move)

"foo".pp("context")        // prints "context: foo" to stdout, and returns "foo"

thenPp

Like pp but applies to futures, and prints their result.

fetchUser(id).thenPp       // returns Future[Option[User]] and prints the Option[User] when available

Type aliases

For code golf purposes, common types have been aliased.

Fu[A]         // Future[A]
Funit         // Future[Unit]

Future shortcuts

def fuccess[A](a: A) = Future successful a
def fufail[A <: Throwable, B](a: A): Fu[B] = Future failed a
def fufail[A](a: String): Fu[A] = fufail(base.LilaException(a))
val funit = fuccess(())

Int functions

Reminder: In scala, a.b(c) == a b c. For instance, 1.+(2) == 1 + 2.

// I think it reads better
a atMost b                  // a min b
a atLeast b                 // a max b

Option functions

42.some                     // Option(42) // or Some(42): Option[Int]
none[Int]                   // None: Option[Int]

maybeInt | 0                // maybeInt.getOrElse(0) // but with extra type safety
maybeInt ifTrue boolean     // maybeInt.filter(_ => boolean)
maybeInt ifFalse boolean    // maybeInt.filter(_ => !boolean)
maybeInt has value          // maybeInt contains value // but with extra type safety

~maybeInt                   // maybeInt | 0 // where 0 is provided by the Zero[Int] typeclass instance
~maybeString                // maybeString | "" // thanks to Zero[String] typeclass instance
~maybeList                  // maybeList | Nil // thanks to Zero[List[_]] typeclass instance

maybeInt so f               // maybeInt.fold(0)(f)
                            // e.g. 42.some so (_ + 1) == 43
                            //      none[Int] so (_ + 1) == 0
maybeString so f            // maybeString.fold("")(f)

// etc, with every type having a Zero instance.

Boolean functions

boolean option 42           // if boolean then 42.some else None
                            // or 42.some ifTrue boolean

boolean so 42               // if boolean then 42 else 0
boolean so "foo"            // if boolean then "foo" else ""
boolean so List(1, 2)       // if boolean then List(1, 2) else List.empty[Int]
boolean so 42.some          // if boolean then 42.some else none
boolean so fuccess(42)      // if boolean then fuccess(42) else fuccess(0)
boolean so asyncComputation // if boolean then asyncComputation else funit

// etc, with every type having a Zero instance.

Future functions

val fu: Future[Int]
val f: Int => Boolean

fu dmap f                   // fu map f // but runs on the same thread (perf tweak)
fu dforeach f               // fu foreach f // but runs on the same thread (perf tweak)

fu.void                     // fu.map(_ => ()) // discards the result, returns Funit
fu inject "foo"             // fu.map(_ => "foo") // replaces the result

Sequencing futures and effects

fu >> otherFu               // fu.flatMap(_ => otherFu) // sequence without using first result
fu >>- effect               // fu andThen { case _ => effect } // run a side effect after completion

fu1 >> fu2 >>- effect1 >>- effect2 // sequences f1 and f2, then runs effect1 then effect2

val futures: List[Future[Int]]

futures.parallel: Future[List[Int]] // executes all futures and return one with a list of values

Coding standards

Only mentioning what's not already rewritten by scalafmt.

Calling methods

obj.method(arg)  // yes, best
obj method arg   // yes if cute and not confusing
obj.method {     // yes
  argBlock
}
obj.method (arg) // no
obj method (arg) // no
obj method(arg) // no

obj.method(arg1, arg2)  // yes
obj.method (arg1, arg2) // no
obj method (arg1, arg2) // no
obj method(arg1, arg2) // no

obj.methodWithoutArg // yes
obj methodWithoutArg // no