Exception Handling for Frame connection information (CK coverage)

It is well known that when no CK information is available for a reference frame, SPICE throws an exception. This is perfectly fine. Nevertheless a tool like this should be able to provide to the user different options on how to handle these exceptions given that, especially for measured CKs, CK files have gaps and if present this will make these operational CK files un-usable.

My proposal is to notify GAPS in CK data and whenever possible, in plots, show the existing data and leave the CK gaps as gaps in the plot.

In addition the user should also be able to activate or deactivate this behavior in order to be able to catch CK coverage gaps as exceptions if needed.

Excellent suggestion!

It is already possible to compute the time windows on which a given ID is valid in a the Pool. It should be possible to re-use the SpicePool._ck_cov() method to extract CK gaps as well.

Do you recommend to catch SPICE exception early on, when the user define the trajectory (if any computed time is in a ck-gap w.r.t instrument ID) or when the error is thrown by SPICE for an invalid computation?

I would recommend to catch it when the error is thrown given that gaps are the norm in SPICE Kernel data sets. And in fact it would be useful that a similar method to _ck_cov() could provide the coverage from all loaded kernels, instead of having to specify one CK.

OK, there is already a logger in place, so it should be possible to catch the SPICE error and display an explicit warning to the user instead, and fill the gaps with NaN values.

It should also be possible to add Trajectory.gaps(**kwargs) method to return of a collection of Patch that could be used to grey-out these areas in timeline plots, similar to this:

(I need to explore how matplotlib Axes.axvspan() function manages to display vertical span rectangle across the Axes independently of Axes.set_ylim().)

---

You already have similar kernel coverage checker for pck and spk as well. Currently they are used as substitute values when the user is lazy and don't provide a start/stop in the trajectory temporal window definition.

(Note: it's also used in Flyby definition as an initial guess to locate CA, but it might be removed in the future because it generates errors in some cases…).

In practice, it should be possible to either:

1. Re-use the same methods to add a new SpicePool.gaps(*refs) similar to SpicePool.windows(*refs) or/and SpicePool.coverage(*refs), that contains only gaps (if they exists).
2. Change SpicePool.windows(*refs) or/and SpicePool.windows(*refs) to include gaps in the response directly.

I think 1 is better solution but I'm not yet 100% sure…

marked this issue as related to #72 (closed)

mentioned in merge request !56 (merged)

closed with merge request !56 (merged)

With !56 (merged), coverage gaps intervals can be retrieved by reference for all the kernels loaded into the pool (it includes CK, PCK and SPK kernels):

>>> SpicePool.gaps('JUICE_SPACECRAFT_PLAN', fmt='TDB')
[
    ['2023-11-21 01:29:11.998 TDB', '2024-10-31 02:29:45.998 TDB'],
    ['2024-08-21 08:19:33.998 TDB', '2026-09-27 23:49:02.998 TDB'],
    ['2024-11-04 08:34:55.998 TDB', '2027-11-20 19:56:45.998 TDB'],
    ...,
    ['2035-10-01 10:55:35.998 TDB', '2035-10-02 02:50:52.998 TDB'],
    ['2035-10-02 10:50:52.998 TDB', '2035-10-03 02:46:15.998 TDB'],
    ['2035-10-03 10:46:15.998 TDB', '2035-10-04 02:41:58.998 TDB']
]

These values are based on SpicePool.windows().

Multiple references is also supported (the resulting list will merge all the gaps in a single list). The gaps intervals are always sorted by start time value.

Theses gaps can also be retrieved at the TourConfig level:

gaps = tour.gaps('JUICE_SPACECRAFT_PLAN', fmt='TDB')

	t_start	t_end
0	2023-11-21T01:28:02.816	2024-10-31T02:28:36.816
1	2024-08-21T08:18:24.816	2026-09-27T23:47:53.816
2	2024-11-04T08:33:46.816	2027-11-20T19:55:36.816
3	2026-09-28T23:47:53.816	2029-01-17T06:19:53.816
...	...	...

In this case, the result is a EventsList object. This allows the user the use theses gaps to filter any Trajectory object:

>>> gap = gaps[0]
>>> mask_traj = traj ^ gap  # `traj ^ gaps` also works
<MaskedSpacecraftTrajectory> Observer: JUICE | Target: JUPITER
 - First UTC start time: 2023-06-01T00:00:00.000
 - Last UTC stop time: 2025-01-01T00:00:00.000
 - Nb of pts: 236 (+345 masked)
 - Nb of segments: 2

Then:

_, ax = plt.subplots(figsize=(12, 4))

ax.plot(traj.utc, traj.dist, ':', color='0.7', lw=1)
ax.plot(traj.utc, mask_traj.dist, color='C0')

ax.axvspan(gap.start, gap.stop, color='0.9', alpha=.5)

ax.set_ylabel('Distance to Jupiter (km)')
ax.set_title(f'Mask trajectory in {gap}');

---

This does not full implement the SPICE exception handling but it allow the user to explicitly discard the points in the coverage gaps. Catching the SPICE exception is difficult because it is not obvious how to determine what is the best reference that need to taking into account in the gaps selection. At the moment, there is no obvious path on how to implement this but any related MR is welcome .