Bump date-fns to ^3.3.1

[renovate]

This PR contains the following updates:

Package	Change	Age	Adoption	Passing	Confidence
date-fns	^2.30.0 -> ^3.3.1

---

Release Notes

date-fns/date-fns (date-fns)

v3.3.1

Compare Source

Kudos to @​kossnocorp and @​fturmel for working on the release.

Fixed

• Fixed DST issue in getOverlappingDaysInIntervals, resulting in an inconsistent number of days returned for intervals starting and ending in different DST periods.

• Fixed functions incorrectly using trunc instead of round. The bug was introduced in v3.3.0. The affected functions: differenceInCalendarDays, differenceInCalendarISOWeeks, differenceInCalendarWeeks, getISOWeek, getWeek, and getISOWeeksInYear.

v3.3.0

Compare Source

On this release worked @​kossnocorp, @​TheKvikk, @​fturmel and @​ckcherry23.

Fixed

• Fixed the bug in getOverlappingDaysInIntervals caused by incorrect sorting of interval components that led to 0 for timestamps of different lengths.

• Fixed bugs when working with negative numbers caused by using Math.floor (-1.1 → -2) instead of Math.trunc (-1.1 → -1). Most of the conversion functions (i.e., hoursToMinutes) were affected when passing some negative fractional input. Also, some other functions that could be possibly affected by unfortunate timezone/date combinations were fixed.

  The functions that were affected: format, parse, getUnixTime, daysToWeeks, hoursToMilliseconds, hoursToMinutes, hoursToSeconds, milliseconds, minutesToMilliseconds, millisecondsToMinutes, monthsToYears, millisecondsToHours, millisecondsToSeconds, minutesToHours, minutesToSeconds, yearsToQuarters, yearsToMonths, yearsToDays, weeksToDays, secondsToMinutes, secondsToHours, quartersToYears, quartersToMonths and monthsToQuarters.

• Fixed the Czech locale's formatDistance to include 1 in formatDistance.

• Fixed differenceInSeconds and other functions relying on rounding options that can produce a negative 0.

• Added a preprocessor to the locales API, enabling fixing a long-standing bug in the French locale. (#​1391)

• Added missing yearsToDays to the FP submodule.

• Made functions using rounding methods always return 0 instead of -0.

Added

• Added format alias formatDate with corresponding FormatDateOptions interface.

v3.2.0

Compare Source

This release is brought to you by @​kossnocorp, @​fturmel, @​grossbart, @​MelvinVermeer, and @​jcarstairs-scottlogic.

Fixed

• Fixed types compatability with Lodash's flow and fp-ts's pipe. (#​3641)

• Fixed inconsistent behavior of roundToNearestMinutes.

Added

• Added exports of format, lightFormat, and parse internals that enable 3rd-parties to consume those.

v3.1.0

Compare Source

This release is brought to you by @​kossnocorp, @​makstyle119 and @​dmgawel.

Fixed

• Fixed the plural form of weeks in Swedish.

Added

• Added yearsToDays function.

• Added warning about using protected tokens like Y or D without passing a corresponding option. See #​2950.

v3.0.6

Compare Source

On this release worked @​imwh0im, @​jamcry and @​tyrw.

Fixed

• Fixed bug in areIntervalsOverlapping caused by incorrect sorting (#​3614)

v3.0.5

Compare Source

This release is brought to you by @​goku4199.

Fixed

• Fixed internal toDate not processing string arguments properly

v3.0.4

Compare Source

This release is brought to you by @​kossnocorp.

Fixed

• Fixed isWithinInterval bug caused by incorrectly sorting dates (#​3623).

v3.0.3

Compare Source

Fixed

• Rolled back pointing ESM types to the same d.ts files. Instead now it copies the content to avoid the Masquerading as CJS problem reported by "Are the types wrong?".

v3.0.2

Compare Source

Fixed

• Fixed yet another issue caused by ESM types by pointing to the same d.ts files.

• Added package.json to exports to provide access to tooling.

• Fixed TypeScript 5.4 build break by using the latest type names.

v3.0.1

Compare Source

Fixed

• Fixed an error in certain environments caused by d.mts files exporting only types.

v3.0.0

Compare Source

Changed

• BREAKING: date-fns is now a dual-package with the support of both ESM and CommonJS. The files exports are now explicitly in the package.json. The ESM files now have .mjs extension.

• BREAKING: The package now has a flat structure, meaning functions are now named node_modules/date-fns/add.mjs, locales are node_modules/date-fns/locale/enUS.mjs, etc.

• BREAKING: Now all file content’s exported via named exports instead of export default, which will require change direct imports i.e. const addDays = require(‘date-fns/addDays’) to const { addDays } = require(‘date-fns/addDays’).

• BREAKING: TypeScript types are now completely rewritten, check out the d.ts files for more information.

• BREAKING: constants now is not exported via the index, so to import one use import { daysInYear } from "date-fns/constants";. It improves compatibility with setups that modularize imports like Next.js.

• BREAKING: Functions now don’t check the number of passed arguments, delegating this task to type checkers. The functions are now slimmer because of this.

• BREAKING The arguments are not explicitly converted to the target types. Instead, they are passed as is, delegating this task to type checkers.

• BREAKING: Functions that accept Interval arguments now do not throw an error if the start is before the end and handle it as a negative interval. If one of the properties in an Invalid Date, these functions also do not throw and handle them as invalid intervals.

  • areIntervalsOverlapping normalize intervals before comparison, so { start: a, end: b } is practically equivalent to { start: b, end: a }. When comparing intervals with one of the properties being Invalid Date, the function will return false unless the others are valid and equal, given the inclusive option is passed. Otherwise, and when even one of the intervals has both properties invalid, the function will always return false.

  • getOverlappingDaysInIntervals now normalizes intervals before comparison, so { start: a, end: b } is practically equivalent to { start: b, end: a }. If any of the intervals’ properties is an Invalid Date, the function will always return 0.

  • isWithinInterval now normalizes intervals before comparison, so { start: a, end: b } is practically equivalent to { start: b, end: a }. If any of the intervals’ properties is an Invalid Date, the function will always return false.

  • intervalToDuration now returns negative durations for negative intervals. If one or both of the interval properties are invalid, the function will return an empty object.

  • The eachXOfInterval functions (eachDayOfInterval, eachHourOfInterval, eachMinuteOfInterval, eachMonthOfInterval, eachWeekendOfInterval, eachWeekendOfMonth, eachWeekendOfYear, eachWeekOfInterval, eachYearOfInterval) now return a reversed array if the passed interval’s start is after the end. Invalid properties will result in an empty array. Functions that accept the step option now also allow negative, 0, and NaN values and return reversed results if the step is negative and an empty array otherwise.

• BREAKING: intervalToDuration now skips 0 values in the resulting duration, resulting in more compact objects with only relevant properties.

• BREAKING: roundToNearestMinutes now returns Invalid Date instead of throwing an error when nearestTo option is less than 1 or more than 30.

• BREAKING: IE is no longer supported.

• BREAKING: Now all functions use Math.trunc rounding method where rounding is required. The behavior is configurable on a per-function basis.

• BREAKING: Undocumented onlyNumeric option was removed from nn and sv locales. If you relied on it, please contact me.

• BREAKING: Flow is not supported anymore. If you relied on it, please contact me.

• BREAKING: The locales now use regular functions instead of the UTC version, which should not break any code unless you used locales directly.

Added

• All functions that accept date arguments now also accept strings.

• All functions now export options interfaces.

• Now functions allow passing custom Date extensions like UTCDate. They will detect and use the arguments constructor to generate the result of the same class.

• eachMonthOfInterval, eachQuarterOfInterval, eachWeekOfInterval, and eachYearOfInterval now accept the step option like most of the eachXOfInterval functions.

• A new interval function that validates interval, emulating the v2 interval functions behavior.

• differenceInX functions now accept options and allow setting up roundingMethod that configures how the result is rounded. Math.trunc is the default method.

---

Configuration

📅 Schedule: Branch creation - "on sunday before 6:00am" in timezone UTC, Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about these updates again.

---

• If you want to rebase/retry this PR, check this box

---

This PR has been generated by Mend Renovate. View repository job log here.

[mui-bot]

Deploy preview: https://deploy-preview-11508--material-ui-x.netlify.app/

Generated by 🚫 dangerJS against b64b3b4

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[LukasTy]

My plan of action in this regard would be:

1. Release the AdapterDateFnsV3 and let it mature by allowing people to adopt it and test its stability;
2. Keep it like this until we plan to release v7 stable;
3. Before v7 stable, if all is good with date-fns@v3, change the default date-fns package version in our repo to v3;
4. Invert ignores in the adapters to keep TS happy
5. Consider if AdapterDateFnsV3 should become the new AdapterDateFns and AdapterDateFns -> AdapterDateFnsV2 for the v7 stable.

WDYT @flaviendelangle? 🤔

[flaviendelangle]

It all make sense
Just for 5., I'd say that we should wait for the next major.
The goal being to provide the best experience to the larger chunk of our community.
Most of our users are using date-fns@2, so making the usage of our components with date-fns@2 as simple as possible and non breaking is the goal IMHO.

I would suggest we move AdapterDateFns code to a AdapterDateFnsV2 folder but still export an alias AdapterDateFns during v7.
That way people can migrate when they want, without a breaking change.
And then on v8 if the adoption of date-fns@3 is not a disaster, we decide either to remove AdapterDateFns and keep the two with a suffix, or to remove AdapterDateFnsV3 and to make AdapterDateFns use date-fns@3.

My assumption is that this migration will be quite long because it's breaking every single import people have to date-fns in their codebase, unlike the last Luxon major who was painless for probably 90% of our users.

[flaviendelangle]

Just to give some figures on the migration pace

[tylerlaprade]

I just updated my date-fns to v3, but now @mui/x-date-pickers is broken. Is there any way for me to resolve this before this PR lands?

[flaviendelangle]

Which version are you using?
Since v6.19.0 or v7.0.0-alpha.8 you have a new AdapterDateFnsV3 that is compatible with the new major of date-fns

This PR only focuses on which version we run on our own infra

[tylerlaprade]

Oh perfect, I was using AdapterDateFns without any version number, thanks!

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.