Bindings

A binding gives a value a name. const means the name and everything reachable through it never changes; var means it can be changed through this name. There is no third form, and no type in the language has a mutable twin.

Example#

const answer = 42
var counter = 0
counter = counter + 1

var list = [1, 2]
list.append 3
const fixed = list

print "{answer} {counter} {list} {fixed}"

Syntax#

const <pattern> [: <Type>] = <expression>
var   <pattern> [: <Type>] = <expression>

A pattern destructures here too, so const (quotient, remainder) = divide(7, 2) and const Point(x, y) = p are bindings that split their value into several names at once. See Cases and match for the pattern vocabulary a binding shares with match. A module's own top-level const is the one exception: it binds a single name, because a name another file imports has to be one thing (see Top-level code).

Rules#

1. A binding is always initialized. var x and var x: Int are both compile errors. No type in the language has an implied default value, so there is nothing a bare declaration could mean.

   Kompiliert nicht

   var later: Int
   // error: A binding needs a value

2. const is deep from the perspective of the binding. Through a const binding you can neither reassign, nor assign a field, nor call a method that is a var fn. var means "changeable through this path".

   Kompiliert nicht

   const fixed = [1, 2]
   fixed.append 3
   // error: `append` needs a `var`

3. The binding decides about the value, not the type. There is no List and MutableList, no Point and MutablePoint, and no read-only view of anything. A parameter without var is the immutable form.

4. Assigning, passing and capturing is a copy. Two bindings never refer to the same value, so a change through one is invisible through the other. The exception is a shared type, which has an identity.

   type Point {
     var x: Int
     var y: Int
   }
   
   var first = Point 1, 2
   const second = first
   first.x = 99
   print "{first} {second}"
   // prints Point(x: 99, y: 2) Point(x: 1, y: 2)

   That prints Point(x: 99, y: 2) Point(x: 1, y: 2).

5. A type annotation is optional, and a literal adapts to it. Without one, an integer literal is Int64 and a decimal literal is Float64. With one, the literal takes the expected type, and a literal that does not fit it is a compile error at the literal.

   const ratio: Float = 1
   const byte: UInt8 = 0xFF
   const million = 1_000_000

   Kompiliert nicht

   const small: Int8 = 300
   // error: `300` does not fit into `Int8`

6. An empty collection literal needs a type from context. [] and [:] carry no element type of their own.

   const empty: List<String> = []
   const table: Map<String, Int> = [:]
   print "{empty} {table}"

7. A name can be shadowed in a nested scope and never redeclared in the same one. The parameters of a function and the top level of its body are one scope, so a const there may not take a parameter's name. A nested block and a closure are scopes of their own. There is no silent shadowing anywhere in the language.

8. A change that cannot have an effect is a compile error. A var that is changed and never read afterwards, and the discarded result of a method that only reads its receiver, are both errors: with value semantics they are always mistakes rather than defensive lines. Discard on purpose with const _ = ....

9. An expression statement has the type Void or Never, unless the call has a var receiver or a var argument. That is the rule behind rule 8: parser.bump() changes something and stays a statement, while Email.tryFrom(text) and 1 + 2 are values that go nowhere.

10. Assignment is a statement, not an expression. a = b = c does not parse, and if x = y is not a comparison that was mistyped - it does not parse either.

What this is not#

const is not final, and it is not readonly. In Java, C# and TypeScript those words freeze the binding and leave the object writable. Here the value is frozen too:

var writable = [1, 2]
writable.append 3
print writable

const frozen = [1, 2]
frozen.append 3
// error: `append` needs a `var`

var is not let mut, because there is nothing to borrow. A var binding owns its value. Handing it to a function that may change it is a var parameter, and that reference exists only for the duration of the call.

Taking an element out of a collection is not a reference to it. This is the copy trap. The first program below prints 1, the second prints 1 0: the copy changed and the list did not. Rule 8 makes a change that is never read a compile error. The second program reads the copy in its last line, so there is nothing to report: a copy that is changed and then read is a legal program that does something else than was meant.

type Counter {
  var count: Int = 0

  var fn increment() {
    count = count + 1
  }
}

var counters = [Counter(), Counter()]
counters[0].increment()
print counters[0].count
// prints 1

type Counter {
  var count: Int = 0

  var fn increment() {
    count = count + 1
  }
}

var counters = [Counter(), Counter()]
var first = counters[0]
first.increment()
print "{first.count} {counters[0].count}"

A const binding is not a compile-time constant. Its initializer runs at runtime like any other expression. The one place where a constant has to be evaluable at compile time is a top-level const of a module, which is a different rule.

Related#

• Declaring a type - fields, var fields and what is generated.
• Mutation and var paths - what has to be var from the binding down.
• Values and bindings - the same material as a learning step.
• Why values instead of references - the argument behind rule 4.
• Coming from Rust - if let mut and &mut are what you reach for.