See also python/cpython#113804. Potentially, we could reuse 'x'/'X' format codes, but I think it's a good idea to be compatible with gmpy2 (and C) here.

@javierelpianista ?

I

See also python/cpython#113804. Potentially, we could reuse 'x'/'X' format codes, but I think it's a good idea to be compatible with gmpy2 (and C) here.

@javierelpianista ?

Yes, I think it's better to use 'a' and 'A'.

I'm not a specialist in hex numbers but gmpy2 gives a different output than the one in test_format.py:

>>> import gmpy2
>>> gmpy2.get_context().precision = 5300
>>> print(f"{gmpy2.mpfr('1.234567890123456789'):.20a}") # This one is the same as gmpy2
0x1.3c0ca428c59fb71a4194p+0
>>> print(f"{gmpy2.mpfr('1.234567890123456789'):.0a}") # This one is different
0xap-3

whereas, according to the tests you wrote, the latter one should return "0x1p+0". Similar for the other two tests.

print(f"{gmpy2.mpfr('1.234567890123456789'):.0a}") # This one is different

It's not. The mpmath always use 1-normalization for normal numbers, while in the MPFR - the digit before decimal dot should be just non-zero. This is something they wish to fix, but... See https://sympa.inria.fr/sympa/arc/mpfr/2021-05/msg00002.html