Dotted identifiers and λ<arg>.code syntax

7.7

top ← prev up next →

Dotted identifiers and λ<arg>.code syntax

Georges Dupéron <georges.duperon@gmail.com>

This #lang language overrides #%module-begin and #%top-interaction from racket/base, and splits identifiers which contain dots, following these rules:

A single dot splits the identifier, and the dot is replaced with #%dot-separator. If an identifier is split by one or more non-consecutive dots, all the resulting identifiers, including the occurrences #%dot-separator are placed in a syntax list, starting with #%dotted-id, so that a.b.c gets transformed into (#%dotted-id a #%dot-separator b #%dot-separator c).
A leading dot (which is not followed by another dot) is allowed, and is replaced with #%dot-separator, like dots occurring in the middle of the identifier.
A dot immediately preceded or followed by an ellipsis … can be omitted, so that a.….b, a….b, a.…b and a…b are all translated to (#%dotted-id a #%dot-separator … #%dot-separator b).
Two or more dots do not split the identifier, but one of the dots is removed (i.e. it escapes the other dots).
If an identifier ends with a dot, a single trailing dot is removed and the identifier is otherwise left intact (i.e. the trailing dot escapes the whole identifier).
Identifiers consisting only of dots are left unchanged, as well as the following: ..+, ...+, ..*, ...*, …, …+, …* and ::....

Furthermore the syntax λarg₁.arg₂.….argₙ. (expr …) is recognised as a shorthand for (λ (arg₁ arg₂ … argₙ) (expr …)), so that λx. (+ x 2) is roughly translated to (λ (x) (+ x 2)). If the var part is left empty, then it defaults to %1, %2 and so on. The number of parameters is determined from the syntactical contents of the function’s body, before performing macro-expansion. The term λ. (+ %1 %2) is therefore roughly translated to (λ (%1 %2) (+ %1 %2)). The variable named % can be used as a shorthand for %1, so that λ. (+ % 10) is therefore roughly translated to (λ (%) (+ % 10)).

Since this substitution is performed on the whole program, before macro-expansion, these notations are performed regardless of the context in which an expression occurs. For example, the quoted term 'a.b will also get translated to '(#%dotted-id a #%dot-separator b). In this way, the #%module-begin from dotlambda works a bit like if it were a reader extension.

Warning: There probably are some issues with hygiene, especially in mixed contexts (e.g. literate programs, or typed/racket programs with untyped code at phase 1). I will think about these issues and adjust the behaviour in future versions. Future versions may therefore not be 100% backward-compatible with the current version, but the general syntax of dotted identifiers should hopefully not change much.

syntax parameter
(#%dotted-id ids-and-separators …)

The default implementation currently translates a.b.c.d to (d (c (b a))), and .a.b.c to (λ (x) (c (b (a x)))).

This behaviour can be altered using syntax-parameterize. I don’t think syntax parameters can be modified globally for the whole containing file like parameters can (via (param new-value)), so the exact mechanism used to customise the behaviour of #%dotted-id may change in the future.

syntax
#%dot-separator

Indicates the presence of a (possibly implicit) dot. The original string (usually "." or the empty string "" for an implicit dot before or after an ellipsis) is normally stored in the 'dotted-original-chars syntax property of the occurrence of the #%dot-separator identifier.

syntax
(#%module-begin . body)

Overridden form of #%module-begin and racket/base

syntax
(#%top-interaction . expression)

Overridden form of #%top-interaction from racket/base

1 Typed version of dotlambda

#lang typed/dotlambda

package: dotlambda

Like #lang dotlambda, but overrides #%module-begin and #%top-interaction from typed/racket/base, instead.

syntax
(#%module-begin . body)

Overridden form of #%module-begin and racket/base

syntax
(#%top-interaction . expression)

Overridden form of #%top-interaction from racket/base

top ← prev up next →