hifitime 3

Precise date and time handling in Rust built on top of a tuple of two integers allowing representation of durations and epochs with the exactly one nanosecond precision for 32,768 years before 01 January 1900 and 32,767 years after that reference epoch, all that in only 80 bits!

The Epoch used is TAI Epoch of 01 Jan 1900 at midnight, but that should not matter in day-to-day use of this library.

Features

• Leap seconds (as announced by the IETF on a yearly basis)
• Julian dates and Modified Julian dates
• Clock drift via oscillator stability for simulation of time measuring hardware (via the simulation feature)
• UTC representation with ISO8601 formatting (and parsing in that format #45)
• High fidelity Ephemeris Time / Dynamic Barycentric Time (TDB) computations from ESA's Navipedia (caveat: up to 10ms difference with SPICE near 01 Jan 2000)
• Trivial support of time arithmetic (e.g. TimeUnit::Hour * 2 + TimeUnit::Second * 3)
• Supports ranges of Epochs and TimeSeries (linspace of Epochs and Durations)
• Support for custom representations of time (e.g. NASA GMAT Modified Julian Date)
• Trivial support of other time representations, such as TDT (cf #44)

Almost all examples are validated with external references, as detailed on a test-by-test basis.

Validation example

Validation is done using NASA's SPICE toolkit, and specifically the spiceypy Python wrapper.

The most challenging validation is the definition of Ephemeris Time, which is very nearly the same as the Dynamic Barycentric Time (TDB). These calculations in hifitime are from ESA's Navipedia.

Hifitime uses a fixed offset for the computation of Ephemeris Time, as is recommended in Navipedia. For TDB however, the offset is based on the centuries since J2000 TT and therefore time varying. I believe that SPICE uses TDB for all dates after J2000 TT. Hence, in the following validation, we will be comparing the SPICE ET with the Hifitime TDB.

The following examples are executed as part of the standard test suite (cf. the function called spice_et_tdb).

Note: the differences shown here are likely due to a combination of SPICE using a different formulation for the calculation (using the constants in the SPICE kernels) and computing everything on a 64-bit floating point value. By design, a 64-bit floating point value has approximation errors. Hifitime performs all calculations on integers, which do not suffer from rounding errors.

Case 1

In SPICE, we chose to convert the UTC date 2012-02-07 11:22:33 UTC into Ephemeris Time. SPICE responds with 381885819.18493587. Initializing the same UTC date in hifitime and requesting the TDB leads to 381885819.18493646, which is an error of 596.05 nanoseconds.

Case 2

In SPICE, we chose to convert the UTC date 2002-02-07 00:00:00.000 UTC into Ephemeris Time. SPICE responds with 66312064.18493876. Initializing the same UTC date in hifitime and requesting the TDB leads to a difference 633.29 nanoseconds.

Case 3

This tests that we can correctly compute TDB time which will have a negative number of days because the UTC input is prior to J2000 TT. In SPICE, we chose to convert the UTC date 1996-02-07 11:22:33 UTC into Ephemeris Time. SPICE responds with -123035784.81506048. Initializing the same UTC date in hifitime and requesting the TDB leads to a difference 640.74 nanoseconds.

Case 4

In SPICE, we chose to convert the UTC date 2015-02-07 00:00:00.000 UTC into Ephemeris Time. SPICE responds with 476580220.1849411. Initializing the same UTC date in hifitime and requesting the TDB leads to a difference 655.65 nanoseconds.

Case 5

In SPICE, we chose to convert the TDB Julian Date in days 2452312.500372511 into Ephemeris Time, and initialize a Hifitime Epoch with that result (66312032.18493909). We then convert that epoch back into days of Julian Date TDB and Julian Date ET, which lead a difference below machine precision for the TDB computation and 0.46 nanoseconds for the ET computation on a 64-bit floating point (f64/double).

Notes

Please report and bugs by clicking here.

Leap second support

Each time computing library may decide when the extra leap second exists as explained in the IETF leap second reference. To ease computation, hifitime decides that second is the 60th of a UTC date, if such exists. Note that this second exists at a different time than defined on NASA HEASARC. That tool is used for validation of Julian dates. As an example of how this is handled, check the Julian day computations for 2015-06-30 23:59:59, 2015-06-30 23:59:60 and 2015-07-01 00:00:00.

Ephemeris Time vs Dynamic Barycentric Time (TDB)

ET and TDB should now be identical. However, hifitime uses the European Space Agency's definition of TDB, detailed here. It seems that SPICE uses the older definition which has a fixed offset from TDT of 0.000935 seconds. This difference is more prominent around the TDB epoch of 01 January 2000.

Changelog

3.1.0

• Add #![no_std] support
• Add to_parts to Duration to extract the centuries and nanoseconds of a duration
• Allow building an Epoch from its duration and parts in TAI system
• Add pure nanosecond (u64) constructor and getter for GPST since GPS based clocks will count in nanoseconds

Possibly breaking change

• Errors::ParseError no longer contains a String but an enum ParsingErrors instead. This is considered possibly breaking because it would only break code in the cases where a datetime parsing or unit parsing was caught and handled (uncommon). Moreover, the output is still Display-able.

3.0.0

• Backend rewritten from TwoFloat to a struct of the centuries in i16 and nanoseconds in u64. Thanks to @pwnorbitals for proposing the idea in #107 and writing the proof of concept. This leads to at least a 2x speed up in most calculations, cf. this comment.
• Fix GPS epoch, and addition of a helper functions in Epoch by @cjordan

2.2.3

• More deterministic as_jde_tdb_days() in Epoch. In version 2.2.1, the ephemeris time and TDB days were identical down to machine precision. After a number of validation cases in the rotation equations of the IAU Earth to Earth Mean Equator J2000 frame, the new formulation was shown to lead to less rounding errors when requesting the days. These rounding errors prevented otherwise trivial test cases. However, it adds an error of 40.2 nanoseconds when initializing an Epoch with the days in ET and requesting the TDB days.

Note: this was originally published as 2.2.2 but I'd forgotten to update one of the tests with the 40.2 ns error.