std.format: %b, %o, %x and %X on negative integers gives wrong result

[dlangBugzillaToGithub]

bugzilla (@berni44) reported this on 2020-01-26T13:29:58Z

Transfered from https://issues.dlang.org/show_bug.cgi?id=20535

CC List

• bugzilla (@WalterBright)
• qs.il.paperinik

Description

Currently format casts the value to it's unsigned version and prints that. This is inconsistent with %d. Additionally, if this behaviour is wished, it can easily be accomplished by the user casting himself before formatting.

void main()
{
    import std.format : format;

    assert(format("%b",-17) == "-10001"); // fails
    assert(format("%o",-17) == "-21"); // fails
    assert(format("%x",-17) == "-11"); // fails
    assert(format("%X",-17) == "-11"); // fails
}

[dlangBugzillaToGithub]

qs.il.paperinik commented on 2020-12-03T18:54:29Z

Reading https://dlang.org/phobos/std_format.html#formattedWrite section 'b','d','o','x','X':
> [...] If the argument is a signed type and the FormatChar is d it is converted to a signed string of characters, otherwise it is treated as unsigned. [...]

So, clearly, the intended result for negative numbers is not a minus sign and magnitude, but two's complement.

[dlangBugzillaToGithub]

bugzilla (@WalterBright) commented on 2021-03-01T19:55:38Z

Even, if the current specs state it, this is inconsistent behavior and while it's easy for a user to get the current behavior with a changed implementation it is not possible to get the wished for behavior with the current implementation. So I'm reopening this, but changing to enhancement.