Align the grammar of the `Decimal` string constructor with `float`'s

[KommuSoft]

The grammar does not cover the case with underscores between the digits of a Decimal. The PEP-515 introduced underscores for integers, floating-point numbers and decimals. The documentation for integers and floating-point numbers are updated to cover underscores, but the grammar documentation for the Decimal does not.

Linked PRs

• gh-128185: Align the grammar of the Decimal string constructor with float's #128315
• gh-128185: align Decimal docs with spec (case irrelevant for nan/inf's) #128323
• [3.13] gh-128185: Align Decimal docs with spec (case irrelevant for nan/inf) (GH-128323) #132875

[picnixz]

The grammar does not cover the case with underscores between the digits of a Decimal

Whtat do you mean digits of a decimal? a Decimal object is constructed from either a float or from a string. Are you talking about the string constructor? AFAIK Decimal("123_456") is allowed.

Do you mean it's not specified on the docs? namely that you can have _ in the string form?

[KommuSoft]

Yes, but the documentation on that seems outdated, since the grammar mentioned in the docs does not mention the underscore style.

[picnixz]

Well.. it says (emphasis mine)

it should conform to the decimal numeric string syntax after leading and trailing whitespace characters, as well as underscores throughout, are removed:

So technically, it's fine? we don't want to overcharge the grammar I think.

[KommuSoft]

Well the grammar for the float(..) includes the underscore, and except for signaling NaNs, it is identical for Decimals, so now it is inconsistent. I think the documentation should either mention the underscores, and do not add it to the grammar, or include it in the grammar, but not present it in two different ways.

[picnixz]

I see. I think it would be fine if we remove the mention of underscores in the text and add it to the grammar. Concerning leading and trailing whitespaces, I think this should be outside the grammar. WDYT?

[skirpichev]

it is identical for Decimals

It's not.

>>> Decimal('1__.__2')
Decimal('1.2')
>>> float('1__.__2')
Traceback (most recent call last):
  File "<python-input-3>", line 1, in <module>
    float('1__.__2')
    ~~~~~^^^^^^^^^^^
ValueError: could not convert string to float: '1__.__2'

That's known, see e.g. #88433

now it is inconsistent

It's ok. So far we have an important module in the stdlib (called "this").

I think the documentation should either mention the underscores, and do not add it to the grammar

That's what it does: "If value is a string, it should conform to the decimal numeric string syntax after leading and trailing whitespace characters, as well as underscores throughout, are removed: [grammar follows]".

[picnixz]

That's what it does: "If value is a string, it should conform to the decimal numeric string syntax after leading and trailing whitespace characters, as well as underscores throughout, are removed: [grammar follows]".

What I meant is to remove "as well as underscores throughout" and put the _ in the grammar definition. For floats, we have:

but for decimals, we have

so we could update decimal-part to include _ as for the floats. That would at least align the two grammars a bit more.