Numeric Literal Syntax

DEP #:

11

Type:

Standards Track

Author:

Carl Gay

Status:

Final

Created:

11-Feb-2019 (Thomas Edison’s birthday)

Last-Modified:

13-Feb-2019

Post-History:

04-Mar-2019

Target-Version:

Open Dylan 2019.1

Abstract

This DEP proposes to change the literal syntax for Dylan integers and floating point numbers by allowing the underscore character to be used between any two digits.

Motivation

This change is designed to aid readability of large integers and large or high precision floats.

Consider

define constant $nanos-per-second = 1000000000;

vs

define constant $nanos-per-second = 1_000_000_000;

This should eliminate the need to count groups of three digits to be certain you’re reading it correctly.

Proposal

The Dylan lexical grammar is changed as follows:

  • For clarity, a new character class is introduced with a single character, underscore (_):

    underscore:
    _
  • underscore opt is added to the second alternative of the decimal-integer, binary-integer, octal-integer, and hex-integer productions:

    binary-integer:
    binary-digit
    binary-integer underscore opt binary-digit
    octal-integer:
    octal-digit
    octal-integer underscore opt octal-digit
    decimal-integer:
    decimal-digit
    decimal-integer underscore opt decimal-digit
    hex-integer:
    hex-digit
    hex-integer underscore opt hex-digit

The proposed change in floating point numeric syntax follows because the floating-point production references the decimal-integer production.

Examples of valid numeric literals include:

1_000_000
-4_000_000
2.000_002
3_000.000_123
4.0e1_000
#xdead_beef
#b1111_0000

and also:

1_2_3_4_5_6_7  // valid but not recommended

Examples of uses of underscore that are not allowed include:

1__2
_123
-123_
1_.23
1._23
1.23_
#x_feed
#o777_

Backward Compatibility

Dylan code that uses the new syntax is not compatible with older versions of Dylan compilers. Any code that uses the new syntax creates a dependency on the Open Dylan version in which this feature is released. In particular, this syntax should not be used in compiler code (or supporting libraries) until it is supported by the oldest bootstrapping compiler.