Editorial: Fewer GetOffsetNanosecondsFor calls during InterpretISODateTimeOffset (prefer/reject)

[arshaw]

It's possible to avoid the GetOffsetNanosecondsFor call. For each candidate instant, you can derive its offset by comparing it to the UTC-zoned year/month/day/etc. Applicable to offset:prefer/reject only.

Modified (and passing) test262:
tc39/test262@main...fullcalendar:test262:temporal-fewer-calls-offset-prefer-reject

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.60%. Comparing base (02a6d14) to head (177e3e1).
Report is 146 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2789   +/-   ##
=======================================
  Coverage   96.60%   96.60%           
=======================================
  Files          23       23           
  Lines       12312    12324   +12     
  Branches     2272     2272           
=======================================
+ Hits        11894    11906   +12     
  Misses        356      356           
  Partials       62       62           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ptomato]

My personal opinion on this: at this point in the process I don't want to bring further optimizations of user calls to TC39. We already had a normative change last year to fix the most egregious ones. I'm happy to add optimizations in cases where we already requested and approved a normative change due to an outright bug, like #2758 or #2760, especially if you are taking the initiative and I don't have to come up with them myself! But I'm strongly in favour of no new normative changes, unless there is really a bug that gives actual wrong results.

It's certainly possible to squeeze more out of this, we can potentially keep auditing user code calls indefinitely, but it effectively pushes out the ship date in browsers by 2 months every time we do it. Custom calendars are a niche use case, in the normal string-calendar-ID case browsers can optimize all of these calls away because they're not observable, so to me it's just not worth it.

@arshaw Are you joining the Temporal meeting tomorrow? We can talk about this and hear from others.

[arshaw]

@ptomato, could you please give me info about meeting time/invite? I'm not really in the loop.

[ptomato]

@arshaw My bad, I thought you had a standing invitation. The time and video link are the same as last time you attended. I believe they are also on the TC39 public calendar.

[arshaw]

thanks @ptomato , I'll be in the meeting, but 10 mins late (just got your email).

See this comment for an aggregation of all reduced-user-calls I found:
fullcalendar/temporal-polyfill#3 (comment)

[ptomato]

After #2826 we should rewrite this to be an editorial PR. Marking as draft for now.

[ptomato]

Rebased. This is now an editorial change and doesn't require any test262 changes. I added the corresponding spec text.