Decoded Epoch type often has wrong timescale

[ljbade]

The hifitime Epoch struct supports different timescales but the decoder often initialises them with the wrong types.

For reference here is how each type of RINEX file should set the timescale:

• clocks -> UTC
• meteo -> UTC
• observation
  • use the value from the TIME OF FIRST OBS header line
  • mandatory header for RINEX 3+, but in RINEX 2 it is only requires for mixed GPS/GLONASS files
  • so if not present default to GPS for pure GPS file, and GLO for pure GLONASS file
• navigation (EOP/EPH/STO/ION) records
  • for RINEX 3+ the value is whatever constellation the satellite is in, so for each record look at the satellite ID, and use that
  • e.g. if the satellite ID is G01 use GPS timescale, E01 use Galileo etc
  • for RINEX 2 the value is the constellation type of the file (since RINEX 2 uses a different file for GPS or GLONASS)

[ljbade]

While working on this I also found a bug in hifitime relates to the GNSS time scales - nyx-space/hifitime#209 - so be sure to the dependency once thats fixes

Also there isn't support for IRNSS/QZSS yet in hifitime - nyx-space/hifitime#210 - so might need to fall back to something else, for QZSS I would use GPS in the meantime. Seems like the decoder doesn't support IRNSS so not an immediate issue there.

[gwbres]

Hello @ljbade,

I am not surprized by incorrect behaviors, due to the many possible cases, that you thoroughly described.
If you feel like correcting these initial conditions, feel free to submit a PR I'll be happy to merge it

[ljbade]

OK, I will work on a small PR based on the changes I made.

[ljbade]

Started on a PR in https://github.com/ljbade/rinex/tree/time-scale

I still need to check that it handles the different cases and add tests then I'll make a PR.

[ljbade]

I have got stuck fixing up the epoch formatting functions to match the parse functions. I can't figure out how to get the Gregorian format from a hifitime::Epoch in some given time scale. There is only a function for returning the Gregorian values in UTC or TAI, and the Epoch::compute_gregorian function is private.

Basically I want to ensure the output RINEX formatted observation times are in the correct time scale for the type of file (e.g. GPS, Galileo, etc) to match the rules for parsing observation times.

[gwbres]

@ljbade,

I have got stuck fixing up the epoch formatting functions to match the parse functions

I'm sorry I won't be able to provide direct help to your branch until I get back from the week end, I can't access a build machine until then

I can't figure out how to get the Gregorian format from a hifitime::Epoch in some given time scale

From the Hifitime API, I only see these methods to answer your requirement

let epoch = Epoch::default();
let new_epoch = epoch.in_time_scale(TimeScale::GST); // let's you convert into a timescale
// you also have this method, but it does not output an Epoch directly
let  iso8601_formatted_str = epoch.to_gregorian_str(TimeScale::GST);

Basically I want to ensure the output RINEX formatted observation times are in the correct time scale for the type of file (e.g. GPS, Galileo, etc) to match the rules for parsing observation times.

In the end, the best thing to do is to enhance the rinex/tests/parsing.rs::test_parser method to have it test the truth table you gave in this topic, for every type of RINEX file

[gwbres]

I initiated support for QZSST. GLONASST and IRNST will follow, it's easier to have them introduced one at a time

[ljbade]

I did find the in_time_scale and the to_gregorian_str string function but that returns it in the wrong format. I will look to see if there is a generic strftime type function that might work.

Otherwise I might make a feature request on the hifitime page.

[gwbres]

I did find the in_time_scale and the to_gregorian_str string function but that returns it in the wrong format. I will look to see if there is a generic strftime type function that might work.

Otherwise I might make a feature request on the hifitime page.

Another option would be

let epoch = Epoch::default();
let duration = epoch.to_duration();
// you also have this, to translate into another timescale at the same time
let duration = epoch.to_duration_time_scale(TimeScale::GST);
// and then use decompose
let (_sign, y, m, d, hh, mm, ss, nanos) = duration.decompose();

[ChristopherRabotin]

@ljbade yes, hifitime supports generic formatting like strftime. You can find examples here: https://github.com/nyx-space/hifitime/blob/master/tests/efmt.rs . The documentation on this feature is a bit hard to find (it even took me a few wrong clicks and I knew what to look for): https://docs.rs/hifitime/latest/hifitime/efmt/format/struct.Format.html .

If you wish to format an epoch in Gregorian, the simplest should be format!("{my_epoch:?}"): that will print your epoch in the current time scale in Gregorian format (cf. https://docs.rs/hifitime/latest/src/hifitime/epoch.rs.html#2872-2891 ).

Version 4 will provide better support: nyx-space/hifitime#231 .