In digital format, what should we do with grouping separators in hours, minutes and seconds field when the value is > 60 or > 999?

[FrankYFTang]

What should be the expected output of the following?

const df = new Intl.DurmationFormat("en", {style: "digital"})
df.format({hours: 300, minutes: 60, seconds: 120})

should it output "300:60:120" ? but how does it make sense?

How about

df.format({hours: 20, minutes: 3, seconds: 1234567})

"20:03:1,234,567" ?

or "20:03:1234567" ?

Currently, the spec will ask the implementation to output "20:03:1,234,567 "

[FrankYFTang]

@ben-allen @ryzokuken @sffc

[sffc]

The spec doesn't do balancing except in the narrow case of fractional second units in numeric mode. Users should eventually use Temporal.Duration to perform their balancing. So I believe this is mostly WAI. Except perhaps for this part:

"20:03:1,234,567 "

I think perhaps we should suppress grouping separators in digital format because otherwise it looks too confusing.

[FrankYFTang]

suppress grouping separators in digital format

Yea, that is the main reason I file this bug. But which part should we suppress grouping separators in digital format? All? or while unit is hours, minutes and seconds?

consider the following (I included the outcome of the current pec. Notice we need to consider what should do with days or larger units)

const df = new Intl.DurationFormat("en", {style: "digital"})
df.format({hours: 1234567, minutes: 20, seconds: 45}) // => "1,234,567:20:45"
df.format({hours: 12, minutes: 1234567, seconds: 20}) // => "12:1,234,567:20"
df.format({hours: 12, minutes: 34, seconds: 1234567}) // => "12:34:1,234,567"
df.format({hours: 12, minutes: 34, seconds: 56, milliseconds: 1234567}) //  => "12:34:1,290.567"
df.format({days: 1234567, hours: 3, minutes: 20, seconds: 45}). // => "1,234,567 days, 3:20:45"

what should be the use grouping suppressed output we should aim to change to?

[FrankYFTang]

@anba

[anba]

I think perhaps we should suppress grouping separators in digital format because otherwise it looks too confusing.

That suggestion sounds reasonable to me.

Using the example inputs from https://github.com/tc39/proposal-intl-duration-format/issues/192#issuecomment-2052598197:

Duration	Current	Proposed
{hours: 1234567, minutes: 20, seconds: 45}	1,234,567:20:45	1234567:20:45
{hours: 12, minutes: 1234567, seconds: 20}	12:1,234,567:20	12:1234567:20
{hours: 12, minutes: 34, seconds: 1234567}	12:34:1,234,567	12:34:1234567
{hours: 12, minutes: 34, seconds: 56, milliseconds: 1234567}	12:34:1,290.567	12:34:1290.567
{days: 1234567, hours: 3, minutes: 20, seconds: 45}	1,234,567 days, 3:20:45	1,234,567 days, 3:20:45

[FrankYFTang]

I support @anba 's proposal

[sffc]

TG2 notes: https://github.com/tc39/ecma402/blob/main/meetings/notes-2024-04-25.md#grouping-separators-in-durationformat

[FrankYFTang]

The PR implemented the TG2 agreed change is in https://github.com/tc39/proposal-intl-duration-format/pull/198 and got merged. I think we need someone to bring that PR to TG1 to got consensus for the change inside Stage 3.

[sffc]

TG2 discussion: https://github.com/tc39/ecma402/blob/main/meetings/notes-2024-08-22.md#durationformat-merge-issue