sffc Stage 2.7 Review

[sffc]

All minor things. Good for Stage 2.7.

General:

• Did you consider making IntlMV have the following shape: a Record or one of the enum values (if a Record, [[Value]] is allowed to be negative zero), instead of a Record where the [[Value]] field can me a MV or one of the enum values, and the stringDigits are ignored? (not sure if it is better)

GetNumberFormatPattern:

• Description seems to be missing some <ins>

PartitionNumberPattern and elsewhere:

• It's a little weird to take an argument of type "x (a mathematical value or an Intl mathematical value)" because "Intl mathematical value" is a superset of "mathematical value". Just wrap the mathematical value into the new Record.

ToIntlMathematicalValue:

• You revert stringDigits to 0 when the number is constrained to infinity but not when it is constrained to 0. Bug?

@eemeli

[eemeli]

All minor things. Good for Stage 2.7.

🎉

Did you consider making IntlMV have the following shape: a Record or one of the enum values (if a Record, [[Value]] is allowed to be negative zero), instead of a Record where the [[Value]] field can me a MV or one of the enum values, and the stringDigits are ignored? (not sure if it is better)

I went with the current shape because that has sentinel values in only one place, rather than two separate places.

Description seems to be missing some <ins>

Oops, will fix.

PartitionNumberPattern and elsewhere:
It's a little weird to take an argument of type "x (a mathematical value or an Intl mathematical value)" because "Intl mathematical value" is a superset of "mathematical value". Just wrap the mathematical value into the new Record.

That shape of argument is used by PartitionNumberPattern and FormatNumeric, which are used by other parts of ECMA-402, where that wrapping would need to be done:

• 11.5.5 FormatDateTimePattern
• 13.5.9 FormatNumericHours
• 13.5.10 FormatNumericMinutes
• 13.5.11 FormatNumericSeconds
• 13.5.15 PartitionDurationFormatPattern
• 18.5.2 PartitionRelativeTimePattern
• 20.3.1 BigInt.prototype.toLocaleString

All of the above are currently untouched by this change, and my sense was that having a short "If x is an Intl mathematical value, then ... Else, ..." block in PartitionNumberPattern would be more readable than adding wrappers in all of the above.

ToIntlMathematicalValue:
You revert stringDigits to 0 when the number is constrained to infinity but not when it is constrained to 0. Bug?

No, this is intentional, and needed to account for nf.format('0.00'), i.e. cases where the value 0 itself has trailing zeros. The stringDigits value is not consulted for infinite values, so for those cases it doesn't really matter what value is assigned to it.

[sffc]

ToIntlMathematicalValue:
You revert stringDigits to 0 when the number is constrained to infinity but not when it is constrained to 0. Bug?

No, this is intentional, and needed to account for nf.format('0.00'), i.e. cases where the value 0 itself has trailing zeros. The stringDigits value is not consulted for infinite values, so for those cases it doesn't really matter what value is assigned to it.

How about nf.format("1.2345e-1000"), which constrains to 0. Should it have 5 sigfigs or zero sigfigs?

(I lack permission to reopen the issue)

[ljharb]

1.2345e-1000 has 1005 significant figures, doesn't it?