Undocumented handling of thousands separator for floats and differences wrt Decimal

Bug report

Bug description:

In docs we have:

The ',' option signals the use of a comma for a thousands separator for floating-point presentation types and for integer presentation type 'd'. For other presentation types, this option is an error. For a locale aware separator, use the 'n' integer presentation type instead.

The '_' option signals the use of an underscore for a thousands separator for floating-point presentation types and for integer presentation type 'd'. For integer presentation types 'b', 'o', 'x', and 'X', underscores will be inserted every 4 digits. For other presentation types, specifying this option is an error.

But this doesn't specify how separators are inserted for customized alignment, consider:

>>> format(0.0, '>020_f')
'0000000000000.000000'
>>> format(0.0, '<020_f')
'0.000000000000000000'
>>> format(0.0, '=020_f')
'0_000_000_000.000000'
>>> format(0.0, '^020_f')
'0000000.000000000000'

Probably it's obvious, that separators might be inserted only in case of 0-padding and = alignment, though maybe this should be mentioned explicitly.

Note also that neither from above supported by Decimal/Fraction, e.g.:

>>> format(Decimal(0.0), '=020_f')
Traceback (most recent call last):
  File "<python-input-13>", line 1, in <module>
    format(Decimal(0.0), '=020_f')
    ~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: invalid format string
>>> format(Fraction(0.0), '=020_f')
'            0.000000'
>>> format(float(Fraction(0.0)), '=020_f')
'0_000_000_000.000000'

BTW, the rationale for Fraction's behavior (see https://github.com/python/cpython/pull/100161) is "no special-casing of the particular '0=' fill-character/alignment combination", which is rather weak, IMO.

Either this should be fixed (preferred) or documented.

CPython versions tested on:

CPython main branch

Operating systems tested on:

No response

Linked PRs

• gh-131030
• gh-131067
• gh-132155
• gh-136241
• gh-136242
• gh-136649

首先阅读 format specification mini-language 文档，并复现 issue 中的 float、Decimal 和 Fraction 示例。查看链接的 PR，了解已经在进行的工作。完成的标准是以一致的方式解决分隔符和对齐行为，或记录观察到的规则。