<stdfloat>

Is this because float16_t lives in std?

No. formatter only works for struct/class/union/enum/array.

What is needed to format/print a std::complexstd::float16_t? Is this something I can add a formatter for in my own code?

Yes. Just add a new formatter specialization.

Fixing this requires an abi imcompatible change in fmt library. This will not available until next major release.

You can cast to f32 at now, or use a struct wrapper and implement formatter for it.

What needs to change - is it type_mapper?

Casting at the call site isn't really feasible throughout my code base, sadly.

What is needed to format/print a std::complexstd::float16_t? Is this something I can add a formatter for in my own code?

Normally you should only add formatters for code you control, so a supported way to do this is via a wrapper type in your namespace.

I'm happy to help with a PR to fix this properly but need some pointers on where to start. ABI isn't an issue for me. While it seems I'm the first person to try formatting a float16_t I can't imagine I'll be the last.

The FP implementation will soon be replaced with https://github.com/vitaut/zmij so the first step would be to add float16_t support there.

I’d like to work on this issue. I’ve looked into the formatting gap for std::float16_t/std::float32_t and std::complex<...>, and it appears to be related to missing formatter support rather than a bug in existing formatting behavior. If this is acceptable, I’d be happy to investigate and prepare a patch/PR for it. Please let me know if you’d like me to take this on.

By all means, I have completely failed to get round to trying more myself.

@shreyas-s1403 the FP formatter is about to be replaced so I wouldn't recommend building anything on top of existing facilities, see earlier discussion.