Module chrono::naive::time [−] [src]

ISO 8601 time without timezone.

Leap Second Handling

Since 1960s, the manmade atomic clock has been so accurate that it is much more accurate than Earth's own motion. It became desirable to define the civil time in terms of the atomic clock, but that risks the desynchronization of the civil time from Earth. To account for this, the designers of the Coordinated Universal Time (UTC) made that the UTC should be kept within 0.9 seconds of the observed Earth-bound time. When the mean solar day is longer than the ideal (86,400 seconds), the error slowly accumulates and it is necessary to add a leap second to slow the UTC down a bit. (We may also remove a second to speed the UTC up a bit, but it never happened.) The leap second, if any, follows 23:59:59 of June 30 or December 31 in the UTC.

Fast forward to the 21st century, we have seen 26 leap seconds from January 1972 to December 2015. Yes, 26 seconds. Probably you can read this paragraph within 26 seconds. But those 26 seconds, and possibly more in the future, are never predictable, and whether to add a leap second or not is known only before 6 months. Internet-based clocks (via NTP) do account for known leap seconds, but the system API normally doesn't (and often can't, with no network connection) and there is no reliable way to retrieve leap second information.

Chrono does not try to accurately implement leap seconds; it is impossible. Rather, it allows for leap seconds but behaves as if there are no other leap seconds. Various operations will ignore any possible leap second(s) except when any of the operands were actually leap seconds.

If you cannot tolerate this behavior, you must use a separate TimeZone for the International Atomic Time (TAI). TAI is like UTC but has no leap seconds, and thus slightly differs from UTC. Chrono 0.3 does not provide such implementation, but it is planned for 0.4.

Representing Leap Seconds

The leap second is indicated via fractional seconds more than 1 second. This makes possible to treat a leap second as the prior non-leap second if you don't care about sub-second accuracy. You should use the proper formatting to get the raw leap second.

All methods accepting fractional seconds will accept such values.

use chrono::{NaiveDate, NaiveTime, UTC, TimeZone};

let t = NaiveTime::from_hms_milli(8, 59, 59, 1_000);

let dt1 = NaiveDate::from_ymd(2015, 7, 1).and_hms_micro(8, 59, 59, 1_000_000);

let dt2 = UTC.ymd(2015, 6, 30).and_hms_nano(23, 59, 59, 1_000_000_000);

Note that the leap second can happen anytime given an appropriate time zone; 2015-07-01 01:23:60 would be a proper leap second if UTC+01:24 had existed. Practically speaking, though, by the time of the first leap second on 1972-06-30, every time zone offset around the world has standardized to the 5-minute alignment.

Date And Time Arithmetics

As a concrete example, let's assume that 03:00:60 and 04:00:60 are leap seconds. (In reality, of course, leap seconds are separated by at least 6 months.)

Time + Duration:

  • 03:00:00 + 1s = 03:00:01.
  • 03:00:59 + 60s = 03:02:00.
  • 03:00:59 + 1s = 03:01:00.
  • 03:00:60 + 1s = 03:01:00. Note that the sum is identical to the previous.
  • 03:00:60 + 60s = 03:01:59.
  • 03:00:60 + 61s = 03:02:00.
  • 03:00:60.1 + 0.8s = 03:00:60.9.

Time - Duration:

  • 03:00:00 - 1s = 02:59:59.
  • 03:01:00 - 1s = 03:00:59.
  • 03:01:00 - 60s = 03:00:00.
  • 03:00:60 - 60s = 03:00:00. Note that the result is identical to the previous.
  • 03:00:60.7 - 0.4s = 03:00:60.3.
  • 03:00:60.7 - 0.9s = 03:00:59.8.

Time - Time:

  • 04:00:00 - 03:00:00 = 3600s.
  • 03:01:00 - 03:00:00 = 60s.
  • 03:00:60 - 03:00:00 = 60s. Note that the difference is identical to the previous.
  • 03:00:60.6 - 03:00:59.4 = 1.2s.
  • 03:01:00 - 03:00:59.8 = 0.2s.
  • 03:01:00 - 03:00:60.5 = 0.5s. Note that the difference is larger than the previous, even though the leap second clearly follows the previous whole second.
  • 04:00:60.9 - 03:00:60.1 = (04:00:60.9 - 04:00:00) + (04:00:00 - 03:01:00) + (03:01:00 - 03:00:60.1) = 60.9s + 3540s + 0.9s = 3601.8s.

In general,

  • Time + Duration unconditionally equals to Duration + Time.

  • Time - Duration unconditionally equals to Time + (-Duration).

  • Time1 - Time2 unconditionally equals to -(Time2 - Time1).

  • Associativity does not generally hold, because (Time + Duration1) - Duration2 no longer equals to Time + (Duration1 - Duration2) for two positive durations.

    • As a special case, (Time + Duration) - Duration also does not equal to Time.

    • If you can assume that all durations have the same sign, however, then the associativity holds: (Time + Duration1) + Duration2 equals to Time + (Duration1 + Duration2) for two positive durations.

Reading And Writing Leap Seconds

The "typical" leap seconds on the minute boundary are correctly handled both in the formatting and parsing. The leap second in the human-readable representation will be represented as the second part being 60, as required by ISO 8601.

use chrono::{UTC, TimeZone};

let dt = UTC.ymd(2015, 6, 30).and_hms_milli(23, 59, 59, 1_000);
assert_eq!(format!("{:?}", dt), "2015-06-30T23:59:60Z");

There are hypothetical leap seconds not on the minute boundary nevertheless supported by Chrono. They are allowed for the sake of completeness and consistency; there were several "exotic" time zone offsets with fractional minutes prior to UTC after all. For such cases the human-readable representation is ambiguous and would be read back to the next non-leap second.

use chrono::{DateTime, UTC, TimeZone};

let dt = UTC.ymd(2015, 6, 30).and_hms_milli(23, 56, 4, 1_000);
assert_eq!(format!("{:?}", dt), "2015-06-30T23:56:05Z");

let dt = UTC.ymd(2015, 6, 30).and_hms(23, 56, 5);
assert_eq!(format!("{:?}", dt), "2015-06-30T23:56:05Z");
assert_eq!(DateTime::parse_from_rfc3339("2015-06-30T23:56:05Z").unwrap(), dt);

Since Chrono alone cannot determine any existence of leap seconds, there is absolutely no guarantee that the leap second read has actually happened.

Structs

NaiveTime

ISO 8601 time without timezone. Allows for the nanosecond precision and optional leap second representation.