.RKT File Format – The #lang family

How a language name is found

A language name after #lang is a restricted module path: the ASCII letters and digits, _, -, + and / (not at the start or end). The reader turns it into a module path in two steps, trying the first and then the second: a reader submodule of the main module of the language ((submod name reader)), or the module name/lang/reader. For scribble/base this is scribble/base/lang/reader. The module found must export read and read-syntax, which read the rest of the file and return a module form. Because the language must be in a collection to be found, it has to be installed (raco pkg install, or a link made with raco link), except through the meta-languages below. The reader module may also export get-info, which tells tools (DrRacket, the viewer in principle) how to colour the language, which buttons to show and how to configure the run-time.

Racket tools expect a module to start with #lang; a file that does not is read as a single expression sequence by load and the REPL.

The languages

#lang htdp/bsl
; A Posn is a structure: (make-posn Number Number)
(define-struct posn (x y))
(define (distance-to-0 p)
  (sqrt (+ (sqr (posn-x p)) (sqr (posn-y p)))))
(check-expect (distance-to-0 (make-posn 3 4)) 5)

#lang datalog
parent(john, douglas).
parent(douglas, ada).
ancestor(X, Y) :- parent(X, Y).
ancestor(X, Y) :- parent(X, Z), ancestor(Z, Y).
ancestor(john, W)?

#lang rhombus
// keyword and optional arguments: a keyword starts with ~
fun six(_, ~plus: amt = 0):
  6 + amt
six("anything", ~plus: 7)

fun add1(x) :: Int:
  match x
  | _ :: Int : x + 1
  | ~else: x

cond
| 1 == 2: "also oops"
| ~else: "sensible"

Datalog, in detail

The Datalog reader's lexical rules are written in its documentation: % starts a comment that runs to the end of the line; a variable is a sequence of Unicode letters, digits and underscores that begins with an upper-case letter; an identifier is any run of printing characters that contains none of ( ) ` ' = : . ~ ? " % or a space and does not begin with a Latin capital letter (a hyphen is allowed); a string is in double quotes with the escapes \", \\ and a backslash-newline; a number is 0 or an optional - and digits with no leading zero; true and false are the booleans. A literal is a predicate symbol with an optional parenthesised list of comma-separated terms. A clause is a head literal with an optional body after :-; it is an assertion when it ends ., a retraction with ~, and a query with ?. a = b and a != b are built-in literals, and (name). imports an external predicate.

Rhombus, in detail

Rhombus builds on shrubbery notation, whose token rules are written down: comments are // to the end of the line and /* */ that nest; numbers may hold _ between digits and may be decimal, floating point (with e), hexadecimal 0x, octal 0o, binary 0b or a fraction, with #inf, #neginf and #nan for the special values; #true, #false and #void are literals; strings and byte strings (#"…") have Racket's escapes but no raw newline; an identifier is Unicode alphanumerics and _; a keyword is ~ and an identifier; and an operator is a run of symbolic characters. Grouping is by indentation, : for a block and | for alternatives. A #{…} escape lets Racket S-expressions in. The words def, fun, let, class, match, if, cond, for, import and the like are ordinary bindings of the language and not lexical.

Writing a language

1. A module language is a module that exports the bindings the language gives its programs, including #%module-begin, #%app, #%datum, #%top and #%top-interaction when it changes them. (provide (all-from-out racket/base) (rename-out [my-app #%app])) is the standard shape. #lang s-exp "mylang.rkt" can then use it with no further setup.
2. A reader is a module (usually mylang/lang/reader.rkt) that exports read and read-syntax. syntax/module-reader does most of the work: (module reader syntax/module-reader mylang) makes a reader that reads S-expressions and puts them in a module whose language is mylang.
3. A collection and a package make the name available: the language mylang is in the collection mylang, installed with raco pkg install or raco link.
4. get-info makes tools treat the language well: 'color-lexer for syntax colouring, 'drracket:toolbar-buttons, 'configure-runtime for setting up printing and reading when the module runs.

Not described

Known, not yet covered

• Reference: the list of languages is the one that the Racket Guide gives (typed, lazy, frtime, scribble, r5rs, r6rs and the teaching languages named as a group) plus Datalog and Rhombus. The names of the individual teaching and course languages (htdp/bsl and relatives, plai and relatives) and their forms are from memory: the HtDP and PLAI manuals were not downloaded. Other well-known languages made with #lang (Pollen, Rosette, Brag, Racklog, Algol 60, Hackett, Pie and others) are not covered.
• Reference: the Rhombus syntax is described from the token-parsing chapter of the shrubbery notation and a demonstration file; the Rhombus reference and guide were not read, so the keyword list, the forms of the language and the rules of grouping are not described in full.
• Reference: the full protocol of get-info and read-language, the configure-runtime submodule, module-language-info and the way a language is loaded by racket -l were not described.
• Viewer: only the S-expression languages, Scribble and at-exp, Datalog and Rhombus have their own readings. A file in any other #lang is read as S-expressions, which is wrong for languages with another syntax (for example, Pollen markup). The Rhombus colouring marks comments, strings, numbers, keywords ~key, literals and a short list of words (def fun let class interface match if cond when unless for import and others, from memory), and treats a name followed by ( as a function; indentation is not used. The Datalog colouring has no handling of the (name). import form beyond ordinary punctuation.

See also

• File structure – how a file starts
• Macros – the macros that languages are made of
• Scheme: implementations and dialects – Racket among the Schemes