AdjustDateTimeStyleFormat: Don't adjust format when no conflicting options are present?

[anba]

AdjustDateTimeStyleFormat determines a new format even when no conflicting options are present. This leads to the following results:

js> console.log(new Date(0).toLocaleString("ja", {dateStyle: "full"}))                                                                           
1970年1月1日木曜日
js> console.log(new Temporal.PlainDate(1970, 1, 1).toLocaleString("ja", {dateStyle: "full"})) 
1970/1/1木曜日

The polyfill already appears to use the original format, because it returns:

> console.log(new Temporal.PlainDate(1970, 1, 1).toLocaleString("ja", {dateStyle: "full"}))
1970年1月1日木曜日

[ptomato]

Hmm, this shouldn't be happening according to my reading of the spec — do you know what's the difference between the two formats in the above code example? Is there something extra that should be in allowedOptions but isn't?

[anba]

AdjustDateTimeStyleFormat calls either BasicFormatMatcher or BestFitFormatMatcher, which compute the format using the data in %Intl.DateTimeFormat%.[[LocaleData]].[[<locale>]].[[formats]]. Whereas DateTimeStyleFormat uses the format data from %Intl.DateTimeFormat%.[[LocaleData]].[[<locale>]].[[styles]]. This distinction is based on how CLDR structures the data (<dateFormats> and <timeFormats> for the [[styles]] record and <availableFormats> for the [[formats]] record.)

Because there are two different data sources ([[styles]] and [[formats]]), calling AdjustDateTimeStyleFormat with a format record from DateTimeStyleFormat can lead to different outputs, as shown in the example.

---

There's a <datetimeSkeleton> element in CLDR to map styles to available formats, so we could change ECMA-402 to define [[styles]] in terms of the formats in [[formats]]. That way it's guaranteed that every format from DateTimeStyleFormat can also be returned from {Basic,BestFit}FormatMatcher, which ensures we get consistent formatting. Unfortunately ICU4C doesn't yet support returning <datetimeSkeleton> and CLDR data also needs some fixing first, see https://unicode-org.atlassian.net/browse/CLDR-14993.