Before making these changes, I had the following discrepancies
between the calculations made by the different programming
language implementations of Astronomy Engine:
C vs C#: 5.55112e-17, worst line number = 6
C vs JS: 2.78533e-12, worst line number = 196936
C vs PY: 1.52767e-12, worst line number = 159834
Now the results are:
Diffing calculations: C vs C#
ctest(Diff): Maximum numeric difference = 5.55112e-17, worst line number = 5
Diffing calculations: C vs JS
ctest(Diff): Maximum numeric difference = 1.02318e-12, worst line number = 133677
Diffing calculations: C vs PY
ctest(Diff): Maximum numeric difference = 5.68434e-14, worst line number = 49066
Diffing calculations: JS vs PY
ctest(Diff): Maximum numeric difference = 1.02318e-12, worst line number = 133677
Here is how I did this:
1. Use new constants HOUR2RAD, RAD2HOUR that directly convert between radians and sidereal hours.
This reduces tiny roundoff errors in the conversions.
2. In VSOP longitude calculations, keep clamping the angular sum to
the range [-2pi, +2pi], to prevent it from accumulating thousands
of radians. This reduces the accumulated error in the final result
before it is fed into trig functions.
The remaining discrepancies are largely because of an "azimuth amplification" effect:
When converting equatorial coordinates to horizontal coordinates, an object near
the zenith (or nadir) has an azimuth that is highly sensitive to the input
equatorial coordinates. A tiny change in right ascension (RA) can cause a much
larger change in azimuth.
I tracked down the RA discrepancy, and it is due to a different behavior
of the atan2 function in C and JavaScript. There are cases where the least
significant decimal digit is off by 1, as if due to a difference of opinion
about rounding policy.
My best thought is to go back and have a more nuanced diffcalc that
applies less strict tests for azimuth values than the other calculated values.
It seems like every other computed quantity is less sensitive, because solar
system bodies tend to stay away from "poles" of other angular coordinate
systems: their ecliptic latitudes and equatorial declinations are usually
reasonably close to zero. Therefore, right ascensions and ecliptic longitudes
are usually insensitive to changes in the cartesian coordinates they
are calculated from.
Now there are constants for the mean radii of Jupiter's
four major moons available in the C, C#, Python, and JavaScript
versions of Astronomy Engine.
Clarified that these are all mean radii.
Fixed some lingering "//" comments in the C code
(I want to keep ANSI C code as portable as possible.)
Now callers can create time objects from either UT (UT1/UTC civil time)
or ephemeris/dynamical Terrestrial Time (TT). The new TT functions
numerically solve to find the UT that produces the given TT based
on the Delta-T value at that UT. This is always a very fast
numerical convergence, because TT and UT are almost perfectly
linear over brief time windows.
I increased the error tolerance slightly for the Jupiter moons model.
This shrank the model tables significantly, giving me some more
breathing room to stay under 100K download size.
I don't like how close I am to my 100K target size, now
that I'm calculating Jupiter's moons.
Simplified the spin() function so its minified code is smaller.
I will look for other things I can shrink too.
I am starting the process of implementing calculation
of Jupiter's four largest moons: Io, Europa, Ganymede, Callisto.
This commit just contains constant declarations for the
equatorial, polar, and volumetric mean radii of Jupiter.
The positions of the moons will be related to the center
of Jupiter and be expressed in Jupiter equatorial radius units,
so I felt it would be good to give users a way to convert to
kilometers, which can in turn be converted to AU.
Also made time parameters to rotation matrix functions be of
type FlexibleDateTime, and internally convert them to AstroTime.
This should be the policy of all exposed functions in the
JavaScript version of Astronomy Engine.
Exposed KM_PER_AU to outside callers.
Use a private enumerated type to select which direction
the precession and nutation is to be done:
- from date to J2000
- from J2000 to date
Normalize the order of parameters to be consistent
between precession() and nutation(), and across languages.
Pass in AstroTime instead of a pair of floating point TT
values (one of which had to be 0).
Added TypeScript version of ObserverVector(),
but it has not yet been documented or tested.
Astronomy Engine used to use USNO historical and predictive tables,
along with linear interpolation, to calculate Delta-T values.
The problem with the USNO tables is, they did not work well outside
a few centuries around present day.
Later I replaced with Espenak & Meeus piecewise polynomials
that work over a much larger time span (thousands of years).
I just discovered there were still comments in the code referring
to the USNO models. I updated the ones I could find to reflect
the current truth about how the code works today.
This is technically a breaking change, but only for clients
that use the cartesian coordinates in an ecliptic coordinate
return type. Before now, the coordinates were just separate
floating-point members ex, ey, ez. Now they are a standard
vector type.
The purpose is to allow seamless interfacing with vector
rotation functions, and to be consistent with the equatorial
coordinate types.
Instead of declaring all the "body" parameters in the
TypeScript/JavaScript code to be strings, I created a
string-valued enumerated type called Body.
The same string values can still be passed in from JavaScript
code, or callers can use syntax like Astronomy.Body.Moon.
This improves the type checking inside the TypeScript source,
plus it adds better documentation for each of the parameters.
In the generated Markdown documentation, the user can click
on the Body type and see all the supported bodies.
The other three supported languages (C, C#, Python)
already use enumerated types for bodies, so this
brings the JavaScript version more in sync with them.
Now that equatorial coordinates include both angles
and cartesian coordinates, there is no need for the
VectorFromEquator function. It has been removed
from all four supported languages.
The expression "VectorFromEquator(equ, time)" can be
replaced with "equ.vec" in any calling code.
This caused me to discover I had forgotten to finish
making the necessary changes to astronomy.ts for saving
the cartesian vector inside the EquatorialCoordinates class.
I also realized I had made a mistake in the documentation
for the y-coordinate of the vector: it is the June solstice;
there is no such thing as a September solstice!
Also fixed some mistakes in demo tests: if something failed,
I was printing out the wrong filename (camera.c instead of camera.cs).
I'm about to start working on adding a new output
from the Horizon functions. It was a good time to better
document the ideas behind these calculations, before
adding anything new. These are internal comments only
and do not affect generated documentation.
While I was in there, I noticed extra code that was
checking for impossible return values from atan2().
I eliminated these.
In the TypeScript/JavaScript code, the functions MakeObserver and MakeSpherical
are no longer needed, because the classes Observer and Spherical are now exported,
along with their constructors. I deleted those functions and reworked callers
to use the equivalent constructors instead.
Also fixed a few breakages in the html/browser examples that crept in recently.
Although it looks less pretty in the generated Markdown,
I think it makes more sense to consistently use @brief in
the jsdoc comments for functions and classes, so that the generated
documentation.json might be more useful some day.
Added @brief comments for places it was missing.
Fixed obsolete remarks in the documentation for AstroTime:
this type, and its constructor, are both exported now.
Fixed broken link to EclipseEvent function.
