Record types

Types with a single data constructor can be defined in the following style, in which case they are called record types.

type Song = {title : Text, artist : Text, year : Nat}

A type like this has field names assigned to each argument of the type constructor, meaning it can be inspected or modified like this:

track13 = Song "Yesterday" "The Beatles" 1965
Song.title track13
⧨
"Yesterday"

use Text ++
track13 = Song "Yesterday" "The Beatles" 1965
Song.title.modify (yesterday -> yesterday ++ "(Remastered)") track13
  |> Song.year.set 2009
⧨
Song "Yesterday(Remastered)" "The Beatles" 2009

Compare this to a regular type declaration with unnamed arguments:

type Point = Point Nat Nat

The arguments have no name, so they are only identified positionally.

The effect of creating a record type is to generate some accessor methods, to help get, set, and modify each field.

• Song.title : Song -> Text
• Song.title.modify : (Text ->{g} Text) -> Song ->{g} Song
• Song.title.set : Text -> Song -> Song

There's currently no special syntax for creating, pattern matching, or decomposing records. That works the same as for regular data types:

decomposeSong =
  song = Song "Help!" "The Beatles" 1965
  (Song help beatles year) = song
  (help, beatles)
decomposeSong
⧨
("Help!", "The Beatles")