Number Format Separator View as Markdown

Introduction

Long numerical literals can be a source of poor readability in code. Take the following examples:

// what number is this?
197823459;
 
// which number is greater?
97802345932 > 97802349532;
 
// are these numbers equal?
9803458239 === 9803457239;

These are difficult to read and difficult reason about. To ameliorate this issue, this RFC proposes the introduction of a digit separator in numerical literals. This will enable the examples above to be rewritten as:

// what number is this?
197_823_459;
 
// which number is greater?
97_802_345_932 > 97_802_349_532;
 
// are these numbers equal?
9_803_458_239 === 9_803_457_239;

Proposal

This RFC will add support for using the underscore character as a separator in PHP's numerical literals. The separator will be supported by all numerical types and notations in PHP.

Example:

1_000_000; // versus 1000000
3.141_592; // versus 3.141592
0x02_56_12; // versus 0x025612
0b0010_1101; // versus 0b00101101
0267_3432; // versus 02673432
1_123.456_7e2 // versus 1123.4567e2

The underscores will be stripped out during the lexing stage, and so the runtime will not be affected in any way. For example:

var_dump(1_000_000); // int(1000000)

Chosen syntax

The digit separator is used to mark boundaries between digits - it is not used to separate digits from other characters. The following syntax choices are therefore based on this.

Disallow leading underscores

Leading underscores will not enhance readability and will conflict with constant naming conventions.

_100; // a valid constant in PHP

Disallow trailing underscores

Trailing underscores will not enhance readability - if anything, they will decrease it.

100_<=>1_000_+_CONST_NAME; // Parse error: syntax error, unexpected '_' (T_STRING)...

Disallow adjacent underscores

Allowing for two or more underscores to be placed together will provide no further readability benefits.

1__000___000_000; // Parse error: syntax error, unexpected '__000___000_000' (T_STRING)...

Enable underscores between digits only

Underscores are not allowed around the period for floats, around the 0x for hexadecimal notation, around the 0b for binary notation, or around the e for scientific notation. This is because readability will be negatively impacted, and it doesn't really serve the purpose of a “digit separator.”

100_.0; // Parse error:  syntax error, unexpected '_' (T_STRING) in...
100._01; // Parse error:  syntax error, unexpected '_01' (T_STRING) in...
0x_0123; // Parse error:  syntax error, unexpected 'x_0123' (T_STRING) in...
0b_0101; // Parse error:  syntax error, unexpected 'b_0101' (T_STRING) in...
1_e2; // Parse error: syntax error, unexpected '_e2' (T_STRING)...
1e_2; // Parse error: syntax error, unexpected 'e_2' (T_STRING)...

Enable for arbitrary grouping of digits

Underscores may be freely interspersed between arbitrary groups of digits, enabling for developers to group the digits as they see fit. One such argument for relaxing the interspersing of underscores is that not all countries group digits in sets of three [1].

1_234_567; // grouped in sets of 3
123_4567; // grouped in sets of 4 (Japan, China)
1_00_00_00_000; // grouped according to India's numbering system
 
0x11_22_33_44_55_66; // a number to be used as bytes, grouped by bytes
0x1122_3344_5566; // a number to be used as 16-bit data, grouped by word

Why the underscore character?

The underscore:

• Is easy to type on the majority of keyboard layouts
• Does not conflict with PHP's grammar
• Acts as a clear and unambiguous delineator between digits (unlike the comma or period)

It has also been widely adopted as a digit separator in a number of other languages, including:

• Ada
• D
• Java
• Julia
• Perl
• C#
• Ruby
• Elixir

Few other languages have deviated from using the underscore to separate digits. One notable exception is C++, where it could not use an underscore because of conflicts with user-defined literals (specifically in a hexadecimal context). Because PHP does not have such user-defined literals, there are no technical problems with using the underscore as a digit separator. This proposal therefore seeks to follow suite with the other languages.

Why no support for stringy numerics?

This RFC does not include stringy numerics because of the BC breakage involved. It would mean changing the coercion rules for strings to integers, which may potentially have wide-ranging impacts for PHP programs. Also, support for stringy numerics can be quite easily emulated in userland code.

If formatting stringy numerical literals is desired, then support for these can be added in the next major version of PHP.

Backward Incompatible Changes

There are no BC breaks with this feature.

Proposed PHP Version(s)

PHP 7.1

RFC Impact

To SAPIs

No impact.

To Existing Extensions

No impact.

To Opcache

No impact.

New Constants

No impact.

php.ini Defaults

No impact.

Open Issues

None so far.

Future Scope

Support for stringy numerics could be added in the next major version.

Vote

A simple yes/no voting option on whether to support a digit separator in PHP. A 2/3 majority is required.

Vote: Include a Digit Separator into PHP — Yes: 20 · No: 18 (closed)

Voting starts on January 13th and ends on January 20th.

Patches and Tests

PR: https://github.com/php/php-src/pull/1699

Implementation

After the project is implemented, this section should contain

1. the version(s) it was merged to
2. a link to the git commit(s)
3. a link to the PHP manual entry for the feature

References

Current discussion: https://marc.info/?l=php-internals&m=145149644624888&w=2

Previous discussion on separators for numerical literals: https://marc.info/?l=php-internals&m=142431171323037&w=2

[1]: http://www.statisticalconsultants.co.nz/blog/how-the-world-separates-its-digits.html

Rejected Features

Keep this updated with features that were discussed on the mail lists.