I tried more distant objects like Jupiter ... Neptune.
This revealed that at increasing distances, the convergence
threshold in inverse_terra needed to increased also.
So now I use 1 AU as a baseline, and scale up linearly
for more distant objects.
Asking the latitude and longitude directly beneath
the Sun causes inverse_terra not to converge, because the
convergence increment `W` never got below 1.48e-8, but the
convergence limit was 1.0e-8. I increased the limit to 2.0e-8
in all programming language versions.
I'm hoping that is a big enough tolerance for all cases now,
but I will do more testing to see if further fixes are required
for even more distant bodies than the Sun.
Replace the abstract class with a parameter of function type.
This allows the documentation to fully explain how to use
`CorrectLightTravel` without having to look at the code.
Added EclipticGeoMoon as output to the temp/*_check.txt files as 'm' lines.
This ensures that all the languages calculate nearly identical values.
Optimized EclipticGeoMoon a little more by eliminating a redundant
call to mean_obliq.
While trying to convert ecliptic coordinates from mean
equinox of date to true equinox of date, I ran into excessive
overhead from the IAU2000B nutation model. The fact that it
uses 77 trigonometric terms made the calculations a lot slower.
https://apps.dtic.mil/sti/pdfs/AD1112517.pdf
Page 4 in the above document mentions a shorter series
“NOD version 2” that has 13 terms instead of 77 as used in IAU2000B.
I had not noticed NOD2 before, because it appears only in
the FORTRAN version of NOVAS 3.x, not the C version.
After reading the FORTRAN code, I realized NOD2 is the same
as IAU2000B, only it keeps the first 13 of 77 terms.
The terms are already arranged in descending order of
significance, so it is easy to truncate the series.
Based on this discovery, I realized I could achieve all of
the required accuracy needed for Astronomy Engine by
keeping only the first 5 terms of the nutation series.
This tremendously speeds up nutation calculations while
sacrificing only a couple of arcseconds of accuracy.
It also makes the minified JavaScript code smaller:
Before: 119500 bytes.
After: 116653 bytes.
So that's what I did here. Most of the work was updating
unit tests for accepting slightly different calculation
results.
The nutation formula change did trigger detection of a
lurking bug in the inverse_terra functions, which convert
a geocentric vector into latitude, longitude, and elevation
(i.e. an Observer object). The Newton's Method loop in
this function was not always converging, resulting in
an infinite loop. I fixed that by increasing the
convergence threshold and throwing an exception
if the loop iterates more than 10 times.
I also fixed a couple of bugs in the `demotest` scripts.
Modified the code generator and source templates to allow
using fewer than 77 terms from the IAU2000B nutation model.
The number of terms causes calculations to be far slower than
I would like, plus most of these terms provide far less than
one arcsecond of difference in the output.
I want to experiment with truncated versions of the series.
https://apps.dtic.mil/sti/pdfs/AD1112517.pdf
Page 4 in the above document references a shorter series
"NOD version 2" that has only 13 terms instead of 77 as used in IAU2000B.
I found that code and confirmed the 13 terms are the first 13
consecutive terms from the original 77.
This commit does not change the number of terms calculated;
it only enables doing so by changing a single #define in
generate/codegen.c.
Once I change the nutation model, I'm sure there will be multiple
tweaks to unit tests to get everything working again.
I had a copy-n-paste typo in the `dec` parameters
for all of the DefineStar functions. Fixed it.
The TypeScript version of HelioState did not handle
user-defined stars. Added support there.
Because we instantly know the heliocentric
distance of a user-defined star, there is no
need to convert it into a vector and then take
the length of the vector.
All of the HelioDistance functions now return
the distance directly, as an optimization.
Also, I decided it didn't make sense to have a
default definition for user-defined stars.
If the caller doesn't define a star, it should
be treated as an invalid body.
DefineStar now requires passing in the heliocentric
distance of the star expressed in light-years.
That way, I can directly support returning vectors
to a star from HelioVector, GeoVector, etc.
SearchRiseSet and SearchHourAngle now work with user-defined stars.
I'm starting to implement the ability to define
up to 8 distinct points in the sky as "stars"
that will be allowed as a `body` parameter to
some Astronomy Engine functions, to be determined.
Made sure all the altitude search functions
verify that the geographic latitude and target altitude
are valid numbers in the range [-90, +90].
Reworked the C version of the code to be clearer:
eliminated goofy ALTDIFF macro, split out max
altitude derivative into its own function MaxAltitudeSlope,
just like the other language implementations do.
Minor rewording of comments in MaxAltitudeSlope functions.
Python InvalidBodyError now includes the invalid body
in the diagnostic message.
The following Python functions now support searching
in forward or reverse chronological order:
SearchRiseSet
SearchAltitude
SearchHourAngle
Made some minor performance improvements to the
other implementations: return sooner if we
go past time window.
The quadratic interpolator used by `Search` was returning
an unused output: `x`. Search does not need this dimensionless
value; it only cares about the time solution `t` and the slope
of the function at `t`. Removed `x` from the return value
to make slightly smaller/faster code.
Enhanced the JavaScript function Astronomy.SearchMoonPhase
to allow searching forward in time when the `limitDays`
argument is positive, or backward in time when `limitDays`
is negative.
Added unit test "moon_reverse" to verify this new feature.
Fixed 3 of the language implementations where I forgot
to make GeoVector return the observation time, not the
backdated time. This is important to preserve existing
behavior.
Finished coding the Python version of the gravity simulator.
No unit tests have been written yet.
Cleaned up documentation in the other languages.
Made some functions static that did not need to be members.
The JavaScript version of the gravity simulator is
now working. I had one bug in the acceleration formula.
Finished the unit tests and also streamlined them a little.
Implemented half of the GravitySimulator unit tests.
Fixed a bug in the GravitySimulator constructor.
Did some refactoring of test.js:
- Wrote Debug() function as shortcut for conditional output.
- Created generic JPL Horizons state file reader.
It makes more sense to report Jupiter's moons with
individually named structure fields rather than an array.
It reduces the overall code and documentation size,
and outside of unit testing, there are few cases
where iterating over an array of moons is more
lucid than using the names of the moons.
This is a breaking change, but hopefully very few
developers are using this function yet.
Fixing the breakage is very simple.
The existing lunar libration functions in the
other languages (C, C#, Python, JavaScript) were
calculating the Moon's ecliptic latitude and longitude
in radians, not degrees as intended. They have been fixed.
Implemented the libration function for Kotlin.
For years before 1582 or years after 3668, the Seasons functions
were unable to find many equinoxes and/or solstices.
The problem was that over time, the Earth's axis precesses
enough that the calendar dates of these events drifts outside
the fixed search ranges I had provided for them.
I expanded the search ranges so all season changes can be found
for a much wider range of years, as verified by unit tests:
C/C++: -2000..9999
C#: 1..9999
JavaScript: -2000..9999
Python: 1..9999
Kotlin: 1..9999
Note: C#, Python, and Kotlin currently do not allow
years values below +1. In fact, I discovered we were not
noticing when an invalid year was passed into the Kotlin code.
I updated that code to throw an exception when the year does
not match what was expected. It is disturbing that the
GregorianCalendar class silently ignores invalid years!
Constricted the search tolerance from 1 second to 0.01
seconds for the seasons search, to ensure more consistent
behavior.
Fixed a bug in the Kotlin search() function's
quadratic interpolation that was causing the convergence
to be slower than it should have been.
There was already an internal function for calculating
Greenwich Apparent Sidereal Time (GAST). By request,
I have exposed this function for outside users.
Added a minimal unit test to verify the function is
callable and returns the correct result for one case.
This function is already exhaustively tested by unit
tests that verify other functions that already called
this function when it was internal, so minimal testing
is sufficient in this case.
Added the following new functions to all 4 languages:
MassProduct: find the GM product for all Solar System bodies.
LagrangePoint: calculate L1..L5 state vectors for a pair of bodies.
LagrangePointFast: calculate L1..L5 state vectors given
state vectors and GM products of a pair of bodies.
In languages that support it, using hypot(x,y) is a little
easier to read than sqrt(x*x + y*y). Some documentation
(e.g. the man page for the C function) leads me to believe
hypot might also be better behaved than sqrt in some cases.
The JavaScript Math.hypot() is especially nice because it works
for any number of dimensions, so I can use it in 2D and 3D cases.
C only allows 2D usage, as does Python 3.7. Python 3.8 added
support for any number of dimensions, but I don't want to break
compatibility with Python 3.7 just yet. Therefore, in C and Python,
I am only using hypot for 2D cases.
C# does not appear to have any kind of hypot function,
so no changes were made to the C# code.
Thanks to https://github.com/ebraminio for this suggestion.
See this discussion:
https://github.com/cosinekitty/astronomy/issues/150
For the case of calculating a map, where each pixel
on the map represents a different location on the Earth,
it is more efficient to factor out expensive calculation
of sidereal times, assuming the entire map represents
some phenomenon at a single moment in time.
For example, to determine whether the Moon is visible
at different places on the Earth, the following
functions can be calculated across thousands of
different (lat, lon) geographic coordinates around
the world:
ObserverVector
Rotation_EQD_HOR
Before iterating over the map pixels, a program
can call GeoMoon, then convert EQJ coordinates to EQD.
Then by passing the same time value in a loop to
ObserverVector and Rotation_EQD_HOR, the program
can calculate a vector from the observer to the Moon
in EQD coordinates, then convert EQD to HOR.
The z-coordinate of the horizontal coordinates
determines whether the Moon is above or below the
observer's horizon at that point on the Earth.
This calculation pattern performed redundant
sidereal time calculations for each pixel on the map.
I changed the code for all 4 languages to cache
sidereal time so that it only needs to be calculated
once.
In the C version of Astronomy Engine, this resulted
in a speedup factor of about 2.3 in the above use case.
(See the function MapPerformanceTest in generate/ctest.c.)
The code generator was creating slightly different numeric
values for the Pluto state tables and the Jupiter rotation matrix.
I decreased the output precision by one decimal digit.
This should allow the code generator to produce identical
source code on both Linux and macOS.
