In-depth practical guides

Normalise dates in a source dossier

Separate calendar dates, local times and instants before sorting records or converting to UTC.

Updated :

Document monitoring desk with a notebook, clock and resources to consult.

AI-generated illustration

Do these timestamps describe the same instant?

Publication, effective and access dates have different roles. Retaining role and precision prevents a misleading chronology; normalisation does not certify the producer’s clock.

  1. Keep original text

    Record field, language, convention and role. Keep 2026-10 as a month; do not invent a day or hour.

  2. Identify the offset

    Z or +02:00 locates an instant. Local time needs a zone and date-specific rules; a country or your computer’s setting is insufficient.

  3. Preserve ambiguity

    03/04/2026 needs the producer’s convention. Clock changes can repeat or skip local times; record uncertainty instead of silently choosing.

  4. Retain precision

    Keep original, normalised value, zone, offset and rule version. Sort determined instants separately from calendar dates.

How many instants match a local time?

A local clock reading does not always identify a single instant. Explore a documented zone’s rules without silently shifting a missing time or choosing one of two occurrences.

Explore a local time

Format: YYYY-MM-DDTHH:MM, minute precision, years 2000–2026. Example: 2026-10-25T02:30. No seconds or offset in this field.

Load a teaching case

Read the supplied clock time

Repeated time: multiple possible instants

Retain the candidates and seek the source’s offset or other evidence to choose between them. The tool does not select the first occurrence by default.

Local date and time
2026-10-25T02:30
Zone documented by the source
Europe/Paris
Compatible instants
2

Each candidate is shown at the start of the supplied minute (:00) as a reference point. This does not establish that the source recorded those seconds.

Candidate 1

Time with offset
UTC instant

Candidate 2

Time with offset
UTC instant
Show the relevant offset change
Transition in UTC
Immediately before
At the transition

Version of time-zone rules used : 2026.6

The tool works to minute precision for the listed zones and years 2000–2026. It uses the rules installed on the server. Verify that the original record documents the selected zone.

References for interpreting the result

From observation to decision

Two displayed dates, one chronology to check

Record A reads 2026-10-15T00:30:00+02:00. Record B reads 2026-10-14T23:00:00Z. The calendar dates are misleading: B is later than A. These timestamps are fictional and their offsets are explicitly supplied.

Fictional learning case: counts and dates illustrate the method and do not describe a real assignment.

  1. 22:30ZA, 14 October
  2. 23:00ZB, 14 October
  3. 30 minB after A
Case reading grid
ObservationUseful checkNext action
A: 2026-10-15T00:30:00+02:00UTC = local time − offset: subtracting 2 hours gives 2026-10-14T22:30:00Z.Keep both the original value and this normalized instant, even though the day changes.
B: 2026-10-14T23:00:00ZThe Z suffix denotes UTC. Compare 23:00Z with 22:30Z on the same day.Place A before B; the gap is 30 minutes.
C: 2026-10-15T00:30:00, no zoneNo offset or zone rule is supplied. The reader’s local time cannot determine this instant.Leave UTC undetermined and seek the producer’s documentation.
D: 2026-10-14, date onlyA calendar date supplies neither a time nor a zone. Adding 00:00Z would invent precision.Keep the calendar date separate from instants; do not place D in this precise order.
Your decision — Is it possible to conclude that D precedes A?

Explained answer

No. Without a time and zone, D does not define an instant that supports this comparison. Report A before B, 30 minutes apart, and state the limits of C and D separately.

Conversion rule: subtract the offset from local time to obtain UTC. RFC 3339 — Date and Time on the Internet: Timestamps, §4.2

Situations and decisions

Typical situations for preparing a check. They do not describe completed assignments or actual observations.

Same instant

Calculation example: 2026-01-15T10:00:00+02:00 equals 2026-01-15T08:00:00Z.

Date without time

2026-01-15 is a calendar date. Midnight UTC would add information the source never supplied.

Two dates, two events

Illustrative case: a report published on 10 June describes the situation on 31 March. A publication timeline uses June; an analysis of the situation uses March. Keep observation period, publication and access dates separate rather than selecting the latest date for every column.

Worked exercise

Two calendar dates for one instant

Fictional example: the values and documents teach the method and do not describe a real organisation.

Compare the two timestamps and the date-only record without inventing a missing time.

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

Starting information
ItemValueSupplied precision
A2026-01-02T00:30:00+02:00Time and offset
B2026-01-01T22:30:00ZUTC time
C2026-01-02Day only

Your checks

  1. Convert both complete timestamps to UTC.
  2. Compare instants, then calendar dates.
  3. Keep the third value at its original precision.
Read the answer

Answer and decision

Both timestamps identify the same instant: 22:30 UTC on 1 January. The local 2 January is not an extra event. The date-only value cannot be assigned midnight UTC or ordered precisely before or after that instant.

If the case changes

If aggregation uses local days, retain the original offset and explain that rule before grouping.

Record to retain

Keep raw text, precision, known offset, UTC value and aggregation rule.

Documentation reference : IETF — RFC 3339, section 4.2 · Consulted on

Record to retain

  • Original text, field and date role.
  • Precision, explicit offset or identified zone.
  • Normalised value, convention and rule version.
  • Ambiguities, partial dates and refused conversions.
  • Ordering before and after conversion, explaining any change of calendar day.

Official references

Related source profiles

Related protocols

Checks before reaching a conclusion

  • Is each date’s role explicit?
  • Is the offset sourced?
  • Is original precision retained?
  • Are ambiguous values kept out of instant sorting?

Frequently asked questions

Is it possible to use the computer’s time zone?

Only when evidence establishes that it is the zone of the event. The reader’s setting does not establish the source’s local time.

What does -00:00 mean?

RFC 3339 uses it for an unknown local offset when UTC time is known. Preserve that status instead of treating it as evidence of a local zone.

Can timestamps be sorted directly as text?

Only with compatible formats and offsets. Strings with different offsets can be misordered lexically. Convert determined instants to a common representation and keep partial or unresolved local dates separate.

Is it possible to draw the format from one line?

A line can give an index without establishing the agreement of the entire file. Examining multiple cases and documentation allows you to identify mixed formats or values that are impossible to interpret.

TOOLS / REGISTERS

Information to record

Keep original, normalised value, zone, offset and rule version. Sort determined instants separately from calendar dates.

  • Access date and time zone
  • Exact URL
  • Exact column name
  • Observed passage / value
  • Event, publication and observation dates
  • Logical type and format
  • Dataset / series code
  • Version / revision
  • Method / definition
  • Observed status
  • Limit / uncertainty
  • Decision / next action
  • Review owner
  • Next review

DOCUMENTED REFERENCE

Europe/Paris: two instants, the same displayed time

On 26 October 2025, two UTC instants one hour apart both map to 02:30 in Europe/Paris with different offsets. This table applies IANA tzdb 2025b rules to four real instants; it is not a fictional case.

2025-10-26 · UTCEurope/Paris00:30Z02:30 +02:0001:30Z02:30 +01:00
Europe/Paris: two instants, the same displayed time
UTC instantEurope/Paris timeInterpretation
2025-10-26T00:30:00Z02:30 +02:00Before the autumn change: 02:30 local, two hours ahead of UTC.
2025-10-26T01:30:00Z02:30 +01:00After the change: 02:30 again, one hour ahead. The instants are distinct.
2025-03-30T00:30:00Z01:30 +01:00On 30 March 2025, before the spring jump: 01:30 local.
2025-03-30T01:30:00Z03:30 +02:00One UTC hour later: 03:30 local. Local 02:30 does not exist in this zone on that date.

One civil day, several possible durations

Six cases frozen in IANA tzdb 2025b, calculated between consecutive local midnights. Paris gives 23 h or 25 h at the selected changes; Lord Howe gives 23.5 h or 24.5 h. This is not a schedule of future changes.

Local calculation without transmitting inputs. Durations use days of 86,400 seconds without adding leap seconds; they do not measure clock drift.

Example calculated; initial data remain available without JavaScript.

Starting local midnight
2025-03-30T00:00:00+01:00
Next local midnight
2025-03-31T00:00:00+02:00
UTC start
2025-03-29T23:00:00Z
UTC end
2025-03-30T22:00:00Z
Duration · minutes
1380
Duration · hours
23
0 h24 h25 h23 h
The bar shows calculated duration; the vertical mark shows 24 hours. Fixed scale from 0 to 25 hours.

Rules version : IANA tzdb 2025b.

Six days calculated from frozen rules
IANA zoneCivil dayStarting local midnightNext local midnightUTC startUTC endDuration · hours
Europe/Paris2025-03-302025-03-30T00:00:00+01:002025-03-31T00:00:00+02:002025-03-29T23:00:00Z2025-03-30T22:00:00Z23
Europe/Paris2025-10-262025-10-26T00:00:00+02:002025-10-27T00:00:00+01:002025-10-25T22:00:00Z2025-10-26T23:00:00Z25
Australia/Lord_Howe2025-04-062025-04-06T00:00:00+11:002025-04-07T00:00:00+10:302025-04-05T13:00:00Z2025-04-06T13:30:00Z24.5
Australia/Lord_Howe2025-10-052025-10-05T00:00:00+10:302025-10-06T00:00:00+11:002025-10-04T13:30:00Z2025-10-05T13:00:00Z23.5
Europe/Paris2025-06-012025-06-01T00:00:00+02:002025-06-02T00:00:00+02:002025-05-31T22:00:00Z2025-06-01T22:00:00Z24
Etc/UTC2025-03-302025-03-30T00:00:00+00:002025-03-31T00:00:00+00:002025-03-30T00:00:00Z2025-03-31T00:00:00Z24

Implication for the record

“26/10/2025 02:30, Paris” cannot choose between the two instants. Keep original string, zone, supplied offset and resolution decision. Civil-time rules may change: record the version rather than fixing one year-round offset.

Are two traces showing 02:30 simultaneous?

Not necessarily. In Paris on 26 October 2025, 02:30 +02:00 and 02:30 +01:00 are 3,600 seconds apart. Compare resolved instants while preserving originals. Without an offset, retain ambiguity rather than choosing arbitrarily.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

Timestamps: check the calendar before comparing strings

Different spellings can represent the same instant. Conversely, alphabetical order can contradict time order when offsets or fractions differ. The comparator retains the input text and calculates within an explicit profile without asking the browser to infer a time zone.

02:30 +02:0000:30 Z02:30 +01:00A → 00:30 Z · B → 01:30 Z · B − A = 3600 s
Timestamps: check the calendar before comparing strings
ReferenceValue or formWhat it identifies
Gregorian2000 ✓ · 1900 / 2100 ✕The day must exist in its month and year. In the Gregorian calendar 2000 is a leap year while 1900 and 2100 are not; a regular expression alone does not check this rule.
UTClocal − offsetThe numeric offset is local time minus UTC. Subtract +02:00 and add the absolute value of −04:00; an offset does not supply an IANA zone name.
RFC 9557 §2Z / +00:00 / −00:00Under the RFC 9557 update, Z can express a known UTC instant without knowing the original local offset. +00:00 retains its preferred-UTC-reference meaning; −00:00 remains a distinct notation to preserve.
fraction1…9 digits in this readerRFC 3339 fractions have one or more digits. This reader calculates at most nine exactly; extra digits are reported outside scope without silent rounding.
orderingtext order ≠ instant orderText sorting requires, among other conditions, the same offset written identically and the same number of fractional digits. The reader reports text order separately from instant order.

Compare two timestamps

Format YYYY-MM-DDTHH:MM:SS[.fraction]Z or ±HH:MM; t and z accepted. No added spaces or inferred zone. Years and UTC result 0001–9999, 0–9 fractional digits, at most 256 characters per field. 24:00 is not accepted in this profile.

Local calculation without transmitting inputs. Durations use days of 86,400 seconds without adding leap seconds; they do not measure clock drift.

Example calculated; initial data remain available without JavaScript.

A · UTC representation
2025-10-26T00:30:00Z
B · UTC representation
2025-10-26T01:30:00Z
Difference B − A · model seconds
3600
Text order A / B
A > B
Instant order A / B
A < B
Fraction digits A / B
0 / 0
Offset written in A
+02:00
Offset written in B
+01:00
Initial example data
{
    "first": "2025-10-26T02:30:00+02:00",
    "second": "2025-10-26T02:30:00+01:00"
}

Implication for the record

This profile uses years and a UTC result from 0001 to 9999, a proleptic Gregorian calendar and calculation days of 86,400 seconds. It does not correct clocks or count intervening leap seconds. Second 60 is outside scope: validity requires the relevant announcements, not automatic conversion. IXDTF zone suffixes are not parsed.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

Partial dates: preserve the source’s precision

A publication may give only a year or month. Filling it with January 1 invents precision absent from the document. Compare compatible periods before ordering two events.

Partial dates: preserve the source’s precision
ReferenceValue or formWhat it identifies
2024YYYYYear known, month and day unknown. Keep “2024” in the citation.
2024-02YYYY-MMMonth known; day unknown. February 2024 covers 29 calendar days.
2024-02-29YYYY-MM-DDDay known; this implies neither a time nor a time zone.

Compare without inventing a day

Enter YYYY, YYYY-MM or YYYY-MM-DD (years 0001 to 9999). A year or month is represented by its full civil calendar period solely to reason about precision.

The periods overlap: the available precision does not establish their order.

Date A
Precision: year
2024-01-01 → 2024-12-31
Date B
Precision: month
2024-02-01 → 2024-02-29
Bands are proportional to the number of covered calendar days. They represent precision, not event duration.

The displayed bounds are calculated, not retrieved publication dates. This tool does not handle times, time zones, seasons, historical calendars or the full EDTF format.

Calculated teaching examples, not observations of real documents.

What the precision lets you conclude
Date ADate BSupported conclusion
20242024-02The periods overlap: the available precision does not establish their order.
2024-022024-02-29The periods overlap: the available precision does not establish their order.
2024-022024-03A precedes B at the supplied precision.
2024-032024-02-29A follows B at the supplied precision.
2024-02-292024-02-29Same civil calendar period: this establishes neither that two events are identical nor their order within a day.

Implication for the record

The displayed bounds are calculated, not retrieved publication dates. This tool does not handle times, time zones, seasons, historical calendars or the full EDTF format.

Primary documentation

Documentation consulted on .

DEEPER READING / PRACTICE

Normalise dates while preserving their meaning

A corpus blends dates of event, publication and consultation. A useful standardization retains the original value and describes the transformation applied.

Updated on .

Priority checks

Identify the role of the date, format, accuracy and known zone. Preserve ambiguous cases and partial dates instead of giving them an inexistent accuracy.

Interpretation pitfalls

One day and one month can be reversed according to the convention used. Converting a partial date into a full date introduces information that the source does not provide.

Working record to retain

The date table contains original, standardized value, role, accuracy and rule. Uncertain conversions are marked for subsequent verification.