How task dates survive export and parsing

Dataset Last tested 25 October 2026

A task date can move because one application treats it as a calendar day and another treats it as a moment in time. Preserve the intended meaning alongside the value, then compare the imported display with the original. In this synthetic comparison, the same date-only string produced different instants in JavaScript and Python.

What the comparison measures

The collection sends each string to JavaScript’s new Date() and Python’s datetime.fromisoformat(). The exact versions are Node.js 24.19.0 and CPython 3.12.14. These are deliberately selected cases, including ordinary dates, explicit offsets, daylight-saving transitions and invalid inputs. They are not a sample of transaction software or a score for either language.

The two APIs do not have the same starting model. JavaScript’s specified date-only form without an offset uses UTC, while a date-time form without an offset uses local time.1 Python can produce a value without time-zone information.2 Our Python runner attaches America/Los_Angeles with fold=0 to those naive values. That attachment is our declared policy, not a default that every Python application shares.

Open the canonical JSON, CSV or method. Each record retains acceptance, parsed components, precision, errors, local display and round-trip results for both parsers. Original observations and synthetic inputs are licensed CC BY 4.0; the original runners are MIT licensed.

All twenty inputs

The table shows UTC results exactly as recorded under that configuration. Z and +00:00 both denote UTC; “Rejected” means that this API did not produce a usable date. Acceptance does not establish that an input was a valid calendar date or suitable for a task.

InputNode UTC resultPython UTC resultObservation
2026-06-152026-06-15T00:00:00.000Z2026-06-15T07:00:00+00:00Date-only interpretation differs
2026-01-012026-01-01T00:00:00.000Z2026-01-01T08:00:00+00:00Date-only interpretation differs
2026-06-15T00:00:00Z2026-06-15T00:00:00.000Z2026-06-15T00:00:00+00:00Same explicit UTC instant
2026-06-15T17:00:00-07:002026-06-16T00:00:00.000Z2026-06-16T00:00:00+00:00Same offset-bearing instant
2026-06-15T17:00:002026-06-16T00:00:00.000Z2026-06-16T00:00:00+00:00Same under declared zone
2026-03-08T01:30:002026-03-08T09:30:00.000Z2026-03-08T09:30:00+00:00Before spring gap
2026-03-08T02:30:002026-03-08T10:30:00.000Z2026-03-08T10:30:00+00:00Same instant; different intermediate steps
2026-11-01T01:30:002026-11-01T08:30:00.000Z2026-11-01T08:30:00+00:00Earlier autumn occurrence
2026-11-01T01:30:00-07:002026-11-01T08:30:00.000Z2026-11-01T08:30:00+00:00Explicit earlier offset
2026-11-01T01:30:00-08:002026-11-01T09:30:00.000Z2026-11-01T09:30:00+00:00Explicit later offset
2024-02-292024-02-29T00:00:00.000Z2024-02-29T08:00:00+00:00Valid leap day; interpretation differs
2026-02-292026-03-01T00:00:00.000ZRejectedNode: March UTC, February local
2026-04-312026-05-01T00:00:00.000ZRejectedNode: May UTC, April local
2026-06-15T24:00:002026-06-16T07:00:00.000ZRejectedNode accepts next-day midnight
2026-06-15T12:00:00.123456Z2026-06-15T12:00:00.123Z2026-06-15T12:00:00.123456+00:00Node loses sub-millisecond digits
2026-06-15T12:00:00+05:302026-06-15T06:30:00.000Z2026-06-15T06:30:00+00:00Same explicit offset conversion
2026-W25-1Rejected2026-06-15T07:00:00+00:00Python accepts week date
20260615Rejected2026-06-15T07:00:00+00:00Python accepts basic date
06/15/20262026-06-15T07:00:00.000ZRejectedV8 accepts slash form
2026-13-01RejectedRejectedBoth reject invalid month

A date-only task is not automatically midnight

For 2026-06-15, Node produced midnight UTC. Its recorded display in America/Los_Angeles was 17:00 on 14 June. Python first parsed a naive midnight on 15 June; after our zone attachment it represented 07:00 UTC. Both paths accepted the text, but they answered different questions about what it meant.

If a task means “review on 15 June,” preserving a calendar date may be the desired design. If it means “join a call at a particular moment,” the value needs enough information to identify that moment. Record which kind of field the receiving application expects before changing the exported string. Adding a time or stripping an offset changes the information being supplied; it is not merely cosmetic formatting.

The explicit UTC input and the offset-bearing examples let you check this distinction. A local display on the previous day can still represent the same instant. Compare the intended calendar label for a date-only field and the UTC instant for a scheduled event, rather than using one equality check for both.

Parsing is not calendar validation

In this run, Node accepted 2026-02-29 and 2026-04-31. Their UTC results were 1 March and 1 May, but the declared America/Los_Angeles display showed 28 February at 16:00 and 30 April at 17:00. Those plausible-looking local dates were not the requested dates. Python rejected both inputs.

The out-of-range day rollover and acceptance of 06/15/2026 are observations of Node’s V8 engine in this run, not guarantees of the language standard or other engines. The specification permits implementation-specific parsing for nonstandard forms.1 Node also accepted 24:00:00 as the next midnight, while Python rejected the hour. A nonempty result alone did not validate the calendar input.

Keep a rejection visible. Do not replace it with today’s date, an empty completion date or an inferred deadline. For an export check, compare the supplied components with the resulting components and decide explicitly whether a change is allowed. The JSON includes the original strings and actual errors so another engineer can inspect that decision.

The clock-change cases need their policy attached

The spring example supplies 02:30 on 8 March 2026 in the declared zone. Node resolved it to a displayed 03:30. Python’s attachment initially retained 02:30 with the earlier offset; converting through UTC and back produced 03:30. Their recorded UTC instants match, but attaching Python’s zone did not validate that the original wall time existed.

For the autumn example, the naive 01:30 input selected the earlier occurrence under the declared policies. Python’s fold parameter distinguishes the earlier and later offsets during a repeated local interval.3 The two explicit-offset inputs in the table show different UTC instants. We did not run a separate naive fold=1 policy, so the dataset makes no claim about that unrun configuration.

Precision and round trips answer narrower questions

The six-digit fractional-second input retained microseconds in Python; Node retained milliseconds. A date can consequently look the same in a display while its stored precision differs. The recorded round trip serializes and reparses within the same API. It does not simulate a calendar upload or certify an application’s export.

Before moving a real checklist, make a small fictional copy containing a calendar-only task, a scheduled instant, an owner label and completion states. Compare each field after export and import, including missing values. This dataset tests dates only: owner and completion-state preservation still need that separate application check. The private checklist planner is an independent local tool, not an importer for transaction platforms.

Repeat the comparison when an input changes

The runner instructions give exact commands, original inputs and expected environment details. Node’s documented TZ option supplies the named zone for its command.4 Runtime, runner and time-zone-data hashes accompany the results; different builds or policies should be recorded as a new run. A licensed professional determines any contractual or statutory date before it becomes a task.

Not covered on this page

Sources

  1. ECMAScript Date.parse. Living draft labelled ECMAScript 2027; read 25 October 2026. ↩ ↩2

  2. CPython datetime documentation. Pinned v3.12.14 source; read 25 October 2026. ↩

  3. CPython zoneinfo documentation. Pinned v3.12.14 source; read 25 October 2026. ↩

  4. Node.js TZ option. Pinned 24.19.0 documentation; read 25 October 2026. ↩

Code on this page is original and MIT licensed.

How we test: methodology. Found a mistake? Corrections.