Difference in hexadecimal floating-point (`a`) format between `fmt` and `std` isn't documented

[statementreply]

https://godbolt.org/z/TaK4s5YTM

When formatting floating-point numbers to hexadecimal (a) format, fmt includes the 0x prefix (similar to printf %a), but std doesn't (specified in terms of to_chars). Perhaps std should be fixed.

[soumik15630m]

while (print_xdigits > 0 && xdigits[print_xdigits] == '0') --print_xdigits;

  buf.push_back('0');
  buf.push_back(specs.upper() ? 'X' : 'x');
  buf.push_back(xdigits[0]);

it's a design choice most probably

[vitaut]

Good catch, thanks! Let me investigate why std::to_chars doesn't print 0x like printf.

[soumik15630m]

Happy to help

[vitaut]

I think it's an unintentional change in std::format caused by switching to std::to_chars in one of the revisions of P0645. The reference implementation of std::format always printed 0x like pretty much every other facility and I missed the fact that std::to_chars does something different. The <charconv> paper doesn't mention it so it's unclear why it went with this "innovation". @statementreply, could you by any chance submit an LWG issue?

[vitaut]

LWG issue: https://cplusplus.github.io/LWG/issue4515