Need to be more specific about "according to ISO-8601" in ToISOWeekOfYear #1641

FrankYFTang · 2021-07-14T09:32:53Z

Three of the AOs in Temporal has the language "according to ISO-8601"
https://tc39.es/proposal-temporal/#sec-temporal-toisodayofweek
https://tc39.es/proposal-temporal/#sec-temporal-toisodayofyear
https://tc39.es/proposal-temporal/#sec-temporal-toisoweekofyear

12.1.33 ToISODayOfWeek ( year, month, day )
1. Assert: year is an integer.
2. Assert: month is an integer.
3. Assert: day is an integer.
4. Let date be the date given by year, month, and day.
5. Return date's day of the week according to ISO-8601.
NOTE
Monday is 1 and Sunday is 7.
12.1.34 ToISODayOfYear ( year, month, day )
1. Assert: year is an integer.
2. Assert: month is an integer.
3. Assert: day is an integer.
4. Let date be the date given by year, month, and day.
5. Return date's ordinal date in the year according to ISO-8601.
12.1.35 ToISOWeekOfYear ( year, month, day )
1. Assert: year is an integer.
2. Assert: month is an integer.
3. Assert: day is an integer.
4. Let date be the date given by year, month, and day.
5. Return date's week number according to ISO-8601.

Please add link to ISO-8601 or add text to exaplain how that work.

and which revision of ISO-8601 are we talking about?
My understanding is ISO standard are not freely available but need to be purchase. IF we do not explain how does it work in these AO then it is hard to implement it or write test against it. RFC or other standard is much easier to reference to. By looking at these three AOs above, I have no clue how to implement them.

@justingrant @ptomato @Ms2ger @ryzokuken

The text was updated successfully, but these errors were encountered:

ryzokuken · 2021-07-14T09:34:31Z

The spec includes a grammar that should be followed, which is basically a profile of the ISO-8601 grammar in ecmarkup.

FrankYFTang · 2021-07-14T14:05:30Z

The spec includes a grammar that should be followed, which is basically a profile of the ISO-8601 grammar in ecmarkup.

sure, that work for all the Parse* AOs and great, but how would the ISO-8601 grammar to tell me how to
"Return date's day of the week according to ISO-8601." with an integer of year, month and day?
or
"Return date's ordinal date in the year according to ISO-8601." with an integer of year, month and day?
or
"Return date's week number according to ISO-8601." with an integer of year, month and day?

What I am asking is for Temporal to include similar level of text as ISO-8601 grammar from ISO8601 to answer how to implement these 3 AOs
ToISODayOfWeek / ToISODayOfYear / ToISOWeekOfYear

I do not believe the grammar can be used to answer that. IF it could, please show me one as example. Thanks

FrankYFTang · 2021-07-17T07:17:43Z

For example, from the spec itself, I don't think we can easily figure out the answer for the following and justify the value for the expected test, right?

let cal = new Temporal.Calendar("iso8601");
assertEquals(53, cal.weekOfYear(new Temporal.PlainDate(1977, 01, 01)));
assertEquals(53, cal.weekOfYear(new Temporal.PlainDate(1977, 01, 02)));

assertEquals(52, cal.weekOfYear(new Temporal.PlainDate(1977, 12, 31)));
assertEquals(52, cal.weekOfYear(new Temporal.PlainDate(1978, 01, 01)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1978, 01, 02)));

assertEquals(52, cal.weekOfYear(new Temporal.PlainDate(1978, 12, 31)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1979, 01, 01)));

assertEquals(52, cal.weekOfYear(new Temporal.PlainDate(1979, 12, 30)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1979, 12, 31)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1980, 01, 01)));

assertEquals(52, cal.weekOfYear(new Temporal.PlainDate(1980, 12, 28)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1980, 12, 29)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1980, 12, 30)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1980, 12, 31)));
assertEquals(1, cal.weekOfYear(new Temporal.PlainDate(1981, 01, 01)));

assertEquals(53, cal.weekOfYear(new Temporal.PlainDate(1981, 12, 31)));
assertEquals(53, cal.weekOfYear(new Temporal.PlainDate(1982, 01, 01)));
assertEquals(53, cal.weekOfYear(new Temporal.PlainDate(1982, 01, 02)));
assertEquals(53, cal.weekOfYear(new Temporal.PlainDate(1982, 01, 03)));

right?

FrankYFTang · 2021-07-19T21:05:28Z

for example, with the current spec text, it is very hard to review the above two PRs for test262 and reach conclusion that the tests were written correctly.

ptomato · 2021-09-02T19:26:24Z

Meeting, Sept. 2: We determined that adding a reference to the appropriate section in ISO 8601 should be sufficient.

FrankYFTang · 2021-09-02T22:12:13Z

Please be specific which PART of ISO 8601 and which section in that standard. I have hard time to locate them.
Notice ISO 8601 Part 1 is
"ISO 8601-1:2019 Date and time — Representations for information interchange — Part 1: Basic rules"
https://www.iso.org/standard/70907.html

and there are also
"ISO 8601-1:2019/DAMD 1
Date and time — Representations for information interchange — Part 1: Basic rules — Amendment 1"
https://www.iso.org/standard/81801.html

and there are Part 2 too
"ISO 8601-2:2019 Date and time — Representations for information interchange — Part 2: Extensions"
https://www.iso.org/standard/70908.html

and there are also previous version
"ISO 8601:2004 Data elements and interchange formats — Information interchange — Representation of dates and times"
https://www.iso.org/standard/40874.html

and
"ISO 8601:2000 Data elements and interchange formats — Information interchange — Representation of dates and times"
https://www.iso.org/standard/26780.html

and
"ISO 8601:1988 Data elements and interchange formats — Information interchange — Representation of dates and times"
https://www.iso.org/standard/15903.html

I assume we are referring to Part 1 of the 2019 version? Is that correc?
https://www.iso.org/obp/ui/#iso:std:iso:8601:-1:ed-1:v1:en

FrankYFTang · 2022-03-31T19:25:25Z

I really think at minimum you need to specify
ToISOWeekOfYear
https://tc39.es/proposal-temporal/#sec-temporal-toisoweekofyear
currently the text is

1. [Assert](https://tc39.es/ecma262/#assert): year is an [integer](https://tc39.es/ecma262/#integer).
2. [Assert](https://tc39.es/ecma262/#assert): month is an [integer](https://tc39.es/ecma262/#integer).
3. [Assert](https://tc39.es/ecma262/#assert): day is an [integer](https://tc39.es/ecma262/#integer).
4. Let date be the date given by year, month, and day.
5. Return date's week number according to ISO-8601 as an [integer](https://tc39.es/ecma262/#integer).

FrankYFTang · 2022-03-31T23:58:56Z

Notice my check in to v8 to implement
12.4.15 Temporal.Calendar.prototype.weekOfYear ( temporalDateLike )
is now blocked by this in
https://chromium-review.googlesource.com/c/v8/v8/+/3531567

Reviewer does not feel comfortable to approve my check in for an under specified specification text.

@sffc @ryzokuken

FrankYFTang · 2022-04-01T00:10:52Z

I read ISO 8601-1 , it does not specified how to convert from "year, month, and day" to week. It only define WHAT is "date's week number" without defining how to calculate it from year, month and day. We need the spec text to define HOW it is calculated here.

ECMA-262 already has a DayWithinYear equation. We had used some handwavy text in ToISODayOfYear about "the ordinal date in the year according to ISO-8601" but this change uses the logic already present in the spec. See: #1641

ECMA-262 already has a WeekDay equation. We had used some handwavy text in ToISODayOfWeek about "the day of week according to ISO-8601" but this change uses the logic already present in the spec. See: #1641

ECMA-262 already has a DayWithinYear equation. We had used some handwavy text in ToISODayOfYear about "the ordinal date in the year according to ISO-8601" but this change uses the logic already present in the spec. See: #1641

ECMA-262 already has a WeekDay equation. We had used some handwavy text in ToISODayOfWeek about "the day of week according to ISO-8601" but this change uses the logic already present in the spec. See: #1641

ptomato · 2022-05-06T01:44:01Z

@FrankYFTang #2178 should improve the situation for dayOfWeek and dayOfYear. I am still looking for a good reference for weekOfYear.

ECMA-262 already has a DayWithinYear equation. We had used some handwavy text in ToISODayOfYear about "the ordinal date in the year according to ISO-8601" but this change uses the logic already present in the spec. See: #1641

ECMA-262 already has a WeekDay equation. We had used some handwavy text in ToISODayOfWeek about "the day of week according to ISO-8601" but this change uses the logic already present in the spec. See: #1641

FrankYFTang · 2022-05-06T15:07:41Z

Just FYI, I am very closed to land all of my (non 402 version) of Temporal.Calendar code (the last CL got checked in but reverted last night for a simple issue, I will reland it today) EXCEPT this issue inside ToISOWeekOfYear blacking the review of Temporal.Calendar.prototype.weekOfYear. No need to rush since I still have a lot of other task need to complete the landing of other classes anyway. I will work on the (only support UTC version) of Temporal.TimeZone next. For anyone who like to see the progress, I post the tracking of the landing and all related CLs on https://docs.google.com/document/d/1UmZYIbe1Za4FtT8JvxLHx2yFtHqyuJhrjeo99D_IVk8

FrankYFTang · 2022-05-06T19:57:39Z

Here are some reference, but not from ISO8601
https://en.wikipedia.org/wiki/ISO_week_date#Weeks_per_year
my implementation in v8 https://chromium-review.googlesource.com/c/v8/v8/+/3531567/4/src/objects/js-temporal-objects.cc
ICU implementation https://github.com/unicode-org/icu/blob/main/icu4c/source/i18n/gregocal.cpp#L870

FrankYFTang · 2022-05-06T20:01:50Z

ICU4J's implementatino of WoY
https://github.com/unicode-org/icu/blob/289d9703a03494ef0d9dfec122169b22bd9fc84f/icu4j/main/classes/core/src/com/ibm/icu/util/Calendar.java#L5147

ptomato · 2022-05-20T01:41:41Z

Note to myself, here is what HTML says about week numbers: https://html.spec.whatwg.org/#weeks

FrankYFTang · 2022-06-14T23:25:10Z

@pdunkel - are you going to work on better specifying https://tc39.es/proposal-temporal/#sec-temporal-toisoweekofyear ?

ptomato · 2022-06-21T15:20:19Z

I'm intending to do it, but there are a lot of PRs that I want to review before the July agenda deadline first. If someone else wants to work on it, that would be fine with me.

FrankYFTang · 2022-07-28T16:58:06Z

I landed most of the non-intl implementation to v8 now except since/until of PlainDateTime and ZoneDateTime (see https://chromium-review.googlesource.com/c/v8/v8/+/3750098 ) and the implementation of this (see https://chromium-review.googlesource.com/c/v8/v8/+/3531567 ) Please raise the priority to be higher. If you think the algorithm I used in https://chromium-review.googlesource.com/c/v8/v8/+/3531567/4/src/objects/js-temporal-objects.cc is reasonable and conforming to ISO8601, we may convert that to spec text. Also how does the polyfill implement this? Can we convert that to the spec text?

FrankYFTang · 2022-08-16T01:03:14Z

Polyfill currently has the implementation of ToISOWeekOfYear in
https://github.com/tc39/proposal-temporal/blob/main/polyfill/lib/ecmascript.mjs#L2387
as

  WeekOfYear: (year, month, day) => {
    let doy = ES.DayOfYear(year, month, day);
    let dow = ES.DayOfWeek(year, month, day) || 7;
    let doj = ES.DayOfWeek(year, 1, 1);

    const week = MathFloor((doy - dow + 10) / 7);

    if (week < 1) {
      if (doj === 5 || (doj === 6 && ES.LeapYear(year - 1))) {
        return 53;
      } else {
        return 52;
      }
    }
    if (week === 53) {
      if ((ES.LeapYear(year) ? 366 : 365) - doy < 4 - dow) {
        return 1;
      }
    }

    return week;
  },

Can we just turn this into algorithm to for the spec text?

ptomato · 2022-08-16T17:20:25Z

I don't know the provenance of this code, and can't guarantee that it's bug-free, so I'm not sure it's the best basis for a spec algorithm. If possible, I would like to avoid writing our own algorithm in the spec, and refer to one from a reliable source instead.

Here's an overview of the research I've done so far, if you'd like to work on this. I think the most comprehensive explanation is given in https://webspace.science.uu.nl/~gent0113/calendar/isocalendar.htm which is basically an algorithm with a 400-year-long lookup table to verify it with. This seems like the most promising source to refer to and/or build a spec algorithm from.

I also found https://myweb.ecu.edu/mccartyr/ISOwdALG.txt and https://howardhinnant.github.io/date_algorithms.html and then this Perl module https://metacpan.org/pod/Date::WeekOfYear.

Ideally a book or a published peer-reviewed paper would be best, but it's been so long since I searched academia that I'm not sure I even remember where to start. I wonder if it would be acceptable to the editors to refer to the The Mathematics of ISO 8601 page, or maybe one of the sources they refer to.

sffc · 2022-08-16T17:46:15Z

Does the spec contain the logic required to convert an ISO DateTime to a timestamp (seconds since epoch)?

If yes, then we've already opened the can of works of putting parts of the ISO 8601 arithmetic logic into ECMA-262, and I think we should likely add other pieces of it that Temporal requires, such as the week-of-year math.

ptomato · 2022-08-16T17:51:24Z

We have MakeDay which does not contain a description of the logic for converting an ISO calendar date to a timestamp, but on the other hand we have DaysInYear which does contain a full description of how to tell whether an ISO year is a leap year.

pipobscure · 2022-08-16T21:16:02Z

I cobbled this together from the ISO spec. So with minor adjustments it’s a good base.

Adjustment ES.DayOfWeek here conforms to the Date.prototype.getDay schema of having Sunday be 0 rather than the ISO schema where Sunday is 7. So for an algortithm spec that needs smoothing out.

The value of week can roughly be described as:

how many Thursdays have happened in that year?
if no Thursdays have happened the it’s the last week of the previous year

ljharb · 2022-08-16T21:17:40Z

would an alternative for the day stuff be "get day, plus 1, mod 7"?

pipobscure · 2022-08-16T21:52:46Z

would an alternative for the day stuff be "get day, plus 1, mod 7"?

No, because

getDay(Monday) = 1
1+1 mod 7 = 2

where the correct value for Monday is 1

unless I misunderstood what you mean

ptomato · 2022-08-23T18:07:28Z

Closed by #2378.

This was referenced Jul 19, 2021

Add tests for Temporal.Calendar.p*.weekOfYear tc39/test262#3060

Closed

Add tests for Temporal.Calendar.prototype.dayOf* tc39/test262#3051

Merged

ptomato added the spec-text Specification text involved label Sep 2, 2021

FrankYFTang changed the title ~~Need more info about "according to ISO-8601"~~ Need to be more specific about "according to ISO-8601" in ToISOWeekOfYear Mar 31, 2022

ptomato mentioned this issue May 6, 2022

Editorial: Use time value operations in date calculations #2178

Merged

ptomato added the editorial label May 10, 2022

FrankYFTang mentioned this issue Aug 16, 2022

Editorial: Spec out ToISOWeekOfYear w/ algorithm #2378

Merged

ptomato closed this as completed Aug 23, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Need to be more specific about "according to ISO-8601" in ToISOWeekOfYear #1641

Need to be more specific about "according to ISO-8601" in ToISOWeekOfYear #1641

FrankYFTang commented Jul 14, 2021 •

edited by justingrant

Loading

ryzokuken commented Jul 14, 2021

FrankYFTang commented Jul 14, 2021

FrankYFTang commented Jul 17, 2021

FrankYFTang commented Jul 19, 2021

ptomato commented Sep 2, 2021

FrankYFTang commented Sep 2, 2021

FrankYFTang commented Mar 31, 2022

FrankYFTang commented Mar 31, 2022

FrankYFTang commented Apr 1, 2022 •

edited

Loading

ptomato commented May 6, 2022

FrankYFTang commented May 6, 2022

FrankYFTang commented May 6, 2022

FrankYFTang commented May 6, 2022

ptomato commented May 20, 2022

FrankYFTang commented Jun 14, 2022

ptomato commented Jun 21, 2022

FrankYFTang commented Jul 28, 2022

FrankYFTang commented Aug 16, 2022

ptomato commented Aug 16, 2022

sffc commented Aug 16, 2022

ptomato commented Aug 16, 2022 •

edited

Loading

pipobscure commented Aug 16, 2022

ljharb commented Aug 16, 2022

pipobscure commented Aug 16, 2022

ptomato commented Aug 23, 2022

Need to be more specific about "according to ISO-8601" in ToISOWeekOfYear #1641

Need to be more specific about "according to ISO-8601" in ToISOWeekOfYear #1641

Comments

FrankYFTang commented Jul 14, 2021 • edited by justingrant Loading

ryzokuken commented Jul 14, 2021

FrankYFTang commented Jul 14, 2021

FrankYFTang commented Jul 17, 2021

FrankYFTang commented Jul 19, 2021

ptomato commented Sep 2, 2021

FrankYFTang commented Sep 2, 2021

FrankYFTang commented Mar 31, 2022

FrankYFTang commented Mar 31, 2022

FrankYFTang commented Apr 1, 2022 • edited Loading

ptomato commented May 6, 2022

FrankYFTang commented May 6, 2022

FrankYFTang commented May 6, 2022

FrankYFTang commented May 6, 2022

ptomato commented May 20, 2022

FrankYFTang commented Jun 14, 2022

ptomato commented Jun 21, 2022

FrankYFTang commented Jul 28, 2022

FrankYFTang commented Aug 16, 2022

ptomato commented Aug 16, 2022

sffc commented Aug 16, 2022

ptomato commented Aug 16, 2022 • edited Loading

pipobscure commented Aug 16, 2022

ljharb commented Aug 16, 2022

pipobscure commented Aug 16, 2022

ptomato commented Aug 23, 2022

FrankYFTang commented Jul 14, 2021 •

edited by justingrant

Loading

FrankYFTang commented Apr 1, 2022 •

edited

Loading

ptomato commented Aug 16, 2022 •

edited

Loading