Docs / GenType / Supported Types
Edit

Supported Types

Some types and values in ReScript do not map directly to JavaScript and need to be converted whenever a value crosses the boundary. This document gives an overview on how genType's convertion works on different types.

Int

ReScript values e.g. 1, 2, 3 are unchanged. So they are exported to JS values of type number.

Float

ReScript values e.g. 1.0, 2.0, 3.0 are unchanged. So they are exported to JS values of type number.

String

ReScript values e.g. "a", "b", "c" are unchanged. So they are exported to JS values of type string.

Optionals

ReScript values of type e.g. option(int), such as None, Some(0), Some(1), Some(2), are exported to JS values null, undefined, 0, 1, 2. The JS values are unboxed, and null/undefined are conflated. So the option type is exported to JS type null or undefined or number.

Nullables

ReScript values of type e.g. Js.Nullable.t(int), such as Js.Nullable.null, Js.Nullable.undefined, Js.Nullable.return(0), Js.Nullable.return(1), Js.Nullable.return(2), are exported to JS values null, undefined, 0, 1, 2. The JS values are identical: there is no conversion unless the argument type needs conversion.

Records

ReScript record values of type e.g. {x:int} such as {x:0}, {x:1}, {x:2}, are exported to JS object values {x:0}, {x:1}, {x:2}. This requires a change of runtime representation from arrays to objects. So they are exported to JS values of type {x:number}.

Since records are immutable by default, their fields will be exported to readonly property types in Flow/TS. Mutable fields are specified in ReScript by e.g. {mutable mutableField: string}.

The @genType.as annotation can be used to change the name of a field on the JS side of things. So e.g. {[@genType.as "y"] x:int} is exported as JS type {y:int}.

If one field of the ReScript record has option type, this is exported to an optional JS field. So for example ReScript type {x: option(int)} is exported as JS type {x?: number}.

Objects

ReScript object values of type e.g. {. "x":int} such as {"x": 0}, {"x": 1}, {"x": 2}, are exported as identical JS object values {x:0}, {x:1}, {x:2}. This requires no conversion. So they are exported to JS values of type {x:number}. A conversion is required only when the type of some field requires conversions.

Since objects are immutable by default, their fields will be exported to readonly property types in Flow/TS. Mutable fields are specified in ReScript by e.g. { @set "mutableField": string }.

It is possible to mix object and option types, so for example the ReScript type {. "x":int, "y":option(string)} exports to JS type {x:number, ?y: string}, requires no conversion, and allows option pattern matching on the ReScript side.

Object field names follow ReScript's mangling convention (so e.g. _type in ReScript represents type in JS):

Remove trailing "__" if present. Otherwise remove leading "_" when followed by an uppercase letter, or keyword.

Tuples

ReScript tuple values of type e.g. (int, string) are exported as identical JS values of type [number, string]. This requires no conversion, unless one of types of the tuple items does. While the type of ReScript tuples is immutable, there's currently no mature enforcement in TS/Flow, so they're currenty exported to mutable tuples.

Variants

Ordinary variants (with capitalized cases, e.g. | A | B(int)) and polymorphic variants (with a backtick, e.g. | `A | `B(int)) are represented in the same way, so there's no difference from the point of view of JavaScript. Polymorphic variants don't have to be capitalized.

Variants can have an unboxed, or a boxed representation. The unboxed representation is used when there is at most one case with a payload, and that payload has object type; otherwise, a boxed representation is used. Object types are arrays, objects, records and tuples.

Variants without payloads are essentially sequences of identifiers. E.g. type [@genType] type days = | Monday | Tuesday. The corresponding JS representation is "Monday", "Tuesday". Similarly, polymorphic variant type [@genType] type days = [ | `Monday | `Tuesday ] has the same JS representation.

When at most one variant case has a payload, and if the payload is of object type, e.g. [ | Unnamed | Named({. "name": string, "surname": string}) ] then the representation is unboxed: JS values are e.g. "Unnamed" and {name: "hello", surname: "world"}. Similarly for polymorphic variants. Note that this unboxed representation does not use the label "Named" of the variant case with payload, because that value is distinguished from the other payload-less cases by its type: an object.

If there is more than one case with payload, or if the single payload has not type object, a boxed representation is used. The boxed representation has shape {tag: "someTag", value: someValue}. For example, type | A | B(int) | C(string) has values such as "A" and {tag: "B", value: 42} and {tag: "C", value: "hello"}. Polymorhphic variants are treated similarly. Notice that payloads for polymorphic variants are always unary: `Pair(int,int) has a single payload of type (int,int). Instead, ordinary variants distinguish between unary Pair((int,int)) and binary Pair(int,int) payloads. All those cases are represented in JS as {tag: "Pair", value: [3, 4]}, and the conversion functions take care of the different ReScript representations.

The @genType.as annotation can be used to modify the name emitted for a variant case on the JS side. So e.g. | [@genType.as "Arenamed"] A exports ReScript value A to JS value "Arenamed". Boolean/integer/float constants can be expressed as | [@genType.as true] True and | [@genType.as 20] Twenty and | [@genType.as 0.5] Half. Similarly for polymorphic variants. The @genType.as annotation can also be used on variants with payloads to determine what appears in { tag: ... }.

For more examples, see Variants.res and VariantsWithPayload.res.

NOTE: When exporting/importing values that have polymorphic variant type, you have to use type annotations, and cannot rely on type inference. So instead of let monday = `Monday, use let monday : days = `Monday. The former does not work, as the type checker infers a type without annotations.

Arrays

Arrays with elements of ReScript type t are exported to JS arrays with elements of the corresponding JS type. If a conversion is required, a copy of the array is performed.

Immutable arrays are supported with the additional ReScript library ImmutableArray.res/.resi, which currently needs to be added to your project. The type ImmutableArray.t(+'a) is covariant, and is mapped to readonly array types in TS/Flow. As opposed to TS/Flow, ImmutableArray.t does not allow casting in either direction with normal arrays. Instead, a copy must be performed using fromArray and toArray.

Functions and Function Components

ReScript functions are exported as JS functions of the corresponding type. So for example a ReScript function foo : int => int is exported as a JS function from numbers to numbers.

If named arguments are present in the ReScript type, they are grouped and exported as JS objects. For example foo : (~x:int, ~y:int) => int is exported as a JS function from objects of type {x:number, y:number} to numbers.

In case of mixed named and unnamed arguments, consecutive named arguments form separate groups. So e.g. foo : (int, ~x:int, ~y:int, int, ~z:int) => int is exported to a JS function of type (number, {x:number, y:number}, number, {z:number}) => number.

Function components are exported and imported exactly like normal functions. For example:

RE
[@genType] [@react.component] let make = (~name) => React.string(name);

For renaming, named arguments follow ReScript's mangling convention:

Remove trailing "__" if present. Otherwise remove leading "_" when followed by an uppercase letter, or keyword.

For example:

RES
@genType let exampleFunction = (~_type) => "type: " ++ _type

Imported Types

It's possible to import an existing TS/Flow type as an opaque type in ReScript. For example,

RES
@genType.import("./SomeFlowTypes") type weekday

defines a type which maps to weekday in SomeFlowTypes.js. See for example Types.re and SomeFlowTypes.js.

Recursive Types

Recursive types which do not require a conversion are fully supported. If a recursive type requires a conversion, only a shallow conversion is performed, and a warning comment is included in the output. (The alternative would be to perform an expensive conversion down a data structure of arbitrary size). See for example Types.res.

First Class Modules

ReScript first class modules are converted from their array ReScript runtime representation to JS Object types. For example,

RES
module type MT = { let x: int let y: string } module M = { let y = "abc" let x = 42 } export firstClassModule: module(MT) = module(M)

is exported as a JS object of type

RES
{"x": number, "y": string}

Notice how the order of elements in the exported JS object is determined by the module type MT and not the module implementation M.

Polymorphic Types

If a ReScript type contains a type variable, the corresponding value is not converted. In other words, the conversion is the identity function. For example, a ReScript function of type {payload: 'a} => 'a must treat the value of the payload as a black box, as a consequence of parametric polymorphism. If a typed back-end is used, the ReScript type is converted to the corresponding generic type.

Exporting Values from Polymorphic Types with Hidden Type Variables

For cases when a value that contains a hidden type variable needs to be converted, a function can be used to produce the appropriate output:

Doesn't work

RES
@genType let none = None
JS
export const none: ?T1 = OptionBS.none; // Errors out as T1 is not defined

Works

RES
@genType let none = () => None
JS
const none = <T1>(a: T1): ?T1 => OptionBS.none;

Promises

Values of type Js.Promise.t(arg) are exported to JS promises of type Promise<argJS> where argJS is the JS type corresponding to arg. If a conversion for the argument is required, the conversion functions are chained via .then(promise => ...).