changes.txt

Changes for 2015 December 10:

   * 'astcheck' has a command-line option to set a path to
the data.  (The insistence that the asteroid elements had to
be in the same path as everything else was stupid.)  Also
did some more error checking to evade compiler warnings.

   * 'afuncs.h' has macros for hours_per_day,  minutes_per_hour,
seconds_per_minute,  and quantities inferred from that.  Numbers
such as 1440 (minutes per day) and 86400 (seconds/day) are not
exactly "magic numbers",  but they really should be defined...
many files are changed to use the newly defined constants.

   * 'astfuncs.cpp', 'classel.cpp': asinh() and atanh() in
there are now only added in for older MSVC,  which (rather
weirdly) didn't have inverse hyperbolic trig.

   * 'classel.cpp' had issues if z = vz = 0,  i.e.,  the input
state vector has _exactly_ zero inclination.  In that situation,
Omega and omega are undefined;  you have to make a somewhat
arbitrary choice,  which the code now does.

   * 'calendar.cpp' can now produce a year-on-a-page calendar,
suitable only for dollhouses.

   * 'colors.cpp' had some weird issues... I don't think it ever
handled the more recent loneos.pho file correctly.  It now does.

   * 'cospar.cpp' now returns,  for the earth,  a rotation
matrix that includes IAU 1980 nutation.  That required a new
setup_precession_with_nutation( ) function (see 'precess.cpp').

   * 'cospar.txt':  see the note in the file for 2015 Nov 17.
The Jupiter System III rotation rate wasn't quite right,  and
the rotation terms for Mercury were botched.

Changes for 2015 June 5:

   * Several files:  snprintf() is (rather bizarrely) not
supported in MSVC or OpenWATCOM.  So quite a few files now
have code that essentially "works around" the absence of a
working snprintf().  Not a big deal,  but then again,
supporting snprintf() in the standard manner wouldn't be a
big deal for MS or OW.

Changes for 2015 Apr 15:

   * 'rocks.cpp':  revised the test routine and investigated
why Triton doesn't work.  Flipping the signs for p and q got
it working,  at least to the point of generating values within
about 300-500 km of Horizons,  'triton.cpp',  and ephemerides
generated from observations in Find_Orb.

Changes for 2014 December 20:

   * full_ctime(), split_time(), and get_time_from_string()
all now have long double variants known as full_ctimel(),
split_timel(), and get_time_from_stringl() .  They are
identical to the originals,  except in the use of long doubles
and in that times are uniformly given/provided relative to
J2000 = JD 2451545.0.  Use of long doubles provides quite
a bit of added precision;  having times be relative to J2000
provides still more,  for dates within a millennium or so
of J2000.  The original,  garden-variety double functions
simply call the long double variants.

   * Several changes to fix compiler warnings generated
when using -Wextra and -pedantic with g++.

   * The leap second for the end of June 2015 is included.
Rather embarrassingly,  the code had extrapolated a leap
second to occur at the end of December 2014.  (It was a
very marginal case;  deciding when the leap second would be
inserted was probably close to a coin-flip for the Time
Lords in Paris.)

   * Partly as a result of the above,  I revised 'utc_test'
to do testing for when leap seconds will occur over a given
set of years.

Changes for 2014 March 13:

   * 'calendar.cpp', 'classel.cpp': added error checking and
assert()s.

   * 'cospar.cpp', 'precess.cpp':  to the latter,  I added a
new 'setup_precession_with_nutation( )' function.  This is
intended to create a precession matrix that adds in the effects
of nutation,  and is used in 'cospar' for determining the
earth's real orientation.  (Previously,  you just got the
mean orientation,  with nutation ignored.)  However,  I don't
think I have it right.  I added this because I had trouble
with radar observations.  (The problems turned out to be due
to something else.  Still,  someday,  the real orientation --
complete with nutations,  polar motion,  and offsets from
modelled nutation/precession/polar motion and observed --
really should be computed.)

   * 'delta_t.cpp':  added a new observed Delta-T for 2014,
and somewhat improved the syntax so that adding new values to
the table would be more straightforward.

   * 'jevent.cpp':  fixed a bug,  plus removed a limitation
on the maximum table size.

   * 'linlunar.mak':  we now have One Make File to Rule Them
All,  suited for almost all systems.  You can run :

make -f linlunar.mak MSWIN=1    (for MinGW on MS Windows)
make -f linlunar.mak CLANG=1    (for OS/X and BSD lacking gcc)
make -f linlunar.mak            (for Linux with gcc)

   Also,  the last two support 'clean' as a target.  Note
that we still have separate make files for MSVC and Watcom.

Changes for 2013 October 17:

   * 'calendar.cpp': improved the date file handling so you
could specify multiple dates for an event (in case it recurs
or spans more than one day).

   * 'astfuncs.cpp': Jonathan Myers pointed out some problems
with the kepler( ) function when handed low-eccentricity cases
with mean anomalies outside the -180 to 180 degree range.
I fixed this,  made a modest speed improvement for medium
eccentricity orbits,  and switched a Boolean value to actually
be of type bool.

   * 'get_time.cpp':  dates before 1582 October 4 should
have been parsed as being in the Julian calendar,  and those
after in Gregorian.  Instead,  all of 1582 October was parsed
as Julian.  This is fixed.

Changes for 2013 September 1:

   * 'cospar.txt': fixed rotation models for Pluto and its
satellites.  Because Pluto is now a "dwarf planet",  rather
than a "planet",  the north and south poles for it and its
satellites have been swapped.  $%*!&@ IAU pole definitions...

Changes for 2013 August 15:

   * 'classel.cpp': fixed a bug introduced in some situations
by the 2013 July 12 improvements.

   * 'calendar.cpp':  separated recurring annual events (now
in 'dates.txt') from one-time events (now in 'date####.txt',
where #### is the four-digit year).  Also allowed one to
specify events such as 'the last Thursday in November' and
so forth;  see 'dates.txt' for examples.

Changes for 2013 July 30:

   * 'cospar.txt', 'cospar.cpp',  'cosptest.cpp':  revised
to allow for orientations of numbered asteroids.  So a
"planet index" of -1 specifies (1) Ceres,  -4 specifies
(4) Vesta,  etc.

Changes for 2013 July 12:

   * 'comets.h': added a 'gm' parameter.  This means that
'calc_classical_elements' no longer needs GM passed in.

   * 'calendar.cpp',  'phases.cpp':  made 64-bit safe (both
broke in certain places when run on my 64-bit Ubuntu box.)

   * 'classel.cpp':  made changes for nearly-parabolic orbits
to get around a nasty loss-of-precision bug.  As it happens,  the
solution is almost identical to that used in 'astfuncs.cpp' for
the Kepler equation for near-parabolic orbits;  I really need to
combine the code and share the solution.  Also switched to use
of the 'gm' member of the ELEMENTS structure,  added some
asserts for debugging,  and did some code cleanup.

   * 'integrat.cpp', 'mpcorb.cpp', 'relativi.cpp':  made sure
GM was properly set in the ELEMENTS structure.

   * 'showelem.cpp':  switched to use of GM from the 'gm' member
of the ELEMENTS structure.  Also fixed a problem with rendering
of negative semimajor axes.

Changes for 2013 June 8:

   * 'astfuncs.cpp': ensured that the angular momentum was
computed in derive_quantities().

   * 'classel.cpp': added some asserts to investigate misbehavior
with near-parabolic orbits,  and added a test to avoid a possible
divide-by-zero bug.

   * 'rckin.cpp': made so it would run on Linux.

   * 'rocks.cpp': updated Pan & Daphnis with SAT357eq.

   * 'showelem.cpp': fixed a bug with "huge" semimajor axes,  and
caused it to use the perihelion distance within the ELEMENTS
structure instead of re-computing it (sometimes erroneously or
with poor precision,  when the orbit is parabolic).

   * 'cospar.txt':  fixed Mercury and Titan in accordance with
the 2009 IAU resolutions.

Changes for 2013 April 7:

   * 'showelem.cpp':  can now show barycentric elements for
'central object' index = -1.

   * 'get_time.cpp':  Jari Suomela pointed out that FITS input such
as '2009-10-01T03:14:15.9' was interpreted as 10 January if you
had selected day/month order.  FITS input should (and now is)
always assumed to be in month/day order.

   * 'big_vsop.cpp':  added a lot of documentation.

Changes for 2013 January 28:

   * Got all test routines (even some pretty irrelevant ones) to compile
under OpenWATCOM,  MinGW,  and MSVC.  (Previously,  some would compile
in one place but not another;  and a few were compiled in separate
batch files.)  This required a lot of changes to 'make' files,  plus
some minor changes in source code where different compilers balked
and/or gave cosmetic warnings.

   * 'refract4.cpp':  Noticed some problems when computing integrated
refractions for observations at greater than 90 degree zenith angle
(i.e.,  observed_alt < 0.) or made above the troposphere (height above
ground greater than 11 km).  The algorithm wasn't really meant to handle
such situations.  It still doesn't,  but now returns more gracefully.
Also added plenty of comments,  and broke out the test routine into a
new 'test_ref.cpp' program.

   * 'dist.cpp':  You can now get a reverse azimuth with the Vicenty code.
I also fixed up the root-finding algorithm to converge a little faster
and with less risk of error,  and set a variable OpenWATCOM thought
could be uninitialized (it couldn't be,  but OpenWATCOM wasn't quite
bright enough to realize why.)

   * showelem.cpp:  Fixed some improper handling of SHOWELEM_OMIT_PQ_MASK
(didn't always actually remove the PQ mask when set).

   * 'phases.cpp', 'jevent.cpp':  both of these had #include <malloc.h>,
now unnecessary on all systems and missing on some.  Also,  both had PI
#defined to insufficient precision.

   * 'calendar.cpp',  'gust86.cpp':  each of these relied on the ability
to initialize arrays with non-constant data.  OpenWATCOM doesn't like that.
Fortunately,  the ability to do this wasn't key to anything I was doing.

   * 'testprec.cpp': revised to have default from/to dates to test if
none are provided on the command line.

   * 'utc_test.cpp': added a header line and output of DUT1 = UT1 - UTC,
and added comments.

Changes for 2012 December 21:

   * 'cospar.cpp':  some speed improvements.  Also changes to some
parameters in 'cospar.txt' to bring it up to date;  look at notes
at the bottom of that file for details.

Changes for 2012 November 21:

   * 'vislimit.cpp':  added some comments and changed some variable names
to be more descriptive (for extinction coeffs from Rayleigh,  aerosols,
ozone,  and water).

Changes for 2012 March 20:

   * 'delta_t.cpp':  Revised the handling for Delta-T before the year
1620 to follow polynomial fits from the _Five Millennium Catalogue of
Solar Eclipses_ by Fred Espenak and Jean Meeus,  pages 12-13.  In doing
so,  I extended the ability to specify one's own Delta-T formulae to
include polynomials beyond the quadratic level.

   I also implemented a new scheme for "future" UTC.  The problems here
are that one doesn't know in advance when leap seconds will be inserted
(you can only make an educated guess,  one that lines up with your guess
about "future" plain old UT) and that in a few centuries,  we'll be adding
leap seconds two or more at a time.  I leave the second issue to our
dozen-greats-grandchildren;  the approach taken for the first issue is
described in 'delta_t.cpp'.

Changes for 2012 March 22:

   * 'classel.cpp':  Realized that for certain hyperbolic orbits near
periapsis,  the mean anomaly would be poorly defined and sometimes
become NaN (resulting in the time of periapsis also being NaN).
Similar logic is now used for both hyperbolic and parabolic cases.

   * 'ps_1996.cpp':  New program that shows basics of computing
geocentric lunar/planetary positions in J2000 coordinates.  Handling
of topocentric correction,  precession,  nutation,  and aberration
will come later,  perhaps.

Changes for 4 February 2012:

   * 'delta_t.cpp':  Cleaned up the UTC-UT code to include fewer cryptic
numbers,  and added a 'utc_test.cpp' routine to make sure I'd not
broken anything.

Changes for 23 January 2012:

   * 'refract4.cpp':  Got an e-mail mentioning that this function could
go into an infinite loop.  I did a little digging,  and found out that
it could indeed do so at temperatures below -85 C.  This is mostly
a theoretical issue (the coldest temperature ever measured on earth was
-89.2 C,  at the Russian research station at Lake Vostok in Antarctica),
but was easy enough to fix.  The test program for 'refract4' also has a
few new command line options and spits out error messages.

   * Revised 'delta_t.cpp' to include the new leap second for July 2012,
and made the leap second table slightly more readable and easy to modify.

Changes for 20 January 2012:

   * Revised 'get_time.cpp' to handle time input with 'AM',  'A.M.',
'PM', 'P.M.',  or lowercase versions of these.  For the last two of
these,  twelve hours are added (i.e.,  it does what a human would expect
it to do.)  This may cause ambiguity someday if I try to add the Jewish
calendar,  for which the standard abbreviation is AM (Anno Mundi,
"Year of the World").  But I'll burn that bridge when I get to it.

Changes for 6 January 2012:

   * Revised 'astephem.cpp' so that it could be thoroughly controlled
via command line,  with object selected,  starting time of the ephemeris,
number of steps,  and step size given.  If they are not given,  it
goes ahead and asks for these details through the console,  as before.
Also revised it so that unnumbered objects could be selected.

   * Everything that compiles in the MinGW version now also compiles
in Linux (except for 'relativi').  g++ was somewhat pickier than MinGW
in some areas.  For example,  ignoring the return values of fread()
and fgets() resulted in warnings;  one also is nagged about putting
braces and parentheses in certain places.  None of this affects the
function of the code (and a few of the suggestions were actually
good ideas),  but it did result in a lot of small changes.

Changes for 21 Nov 2011:

   * 'get_time.cpp' now has broader and more logical capabilities
for handling lunar phases.  A time entry such as

21 Nov 2011 +4w 3q -2d

   will be interpreted,  logically enough,  as "start from 21 Nov
2011;  add four weeks;  find the nearest last quarter;  back up
two days from that point."  Spaces may be omitted:

21n2011+4w3q-2d

   Phases 'nm', '1q', 'fm',  and '3q' are handled.

Changes for 23 Sep 2011:

   * 'precess.cpp' was revised to cache the precession matrix
(same as a previous version did,  but using the new scheme in which
precession is always done in ecliptic coordinates).  'prectest.cpp'
is a test program I used to verify that the precession code was
doing what it was supposed to do.

   * 'obliquit.cpp' caches values to improve speed.  The formula
used is quite accurate for years -8000 to +12000,  but diverges
badly outside that range.  Rather than return "garbage" values
wherein the obliquity rapidly reaches +/-90 degrees,  I capped
the obliquity at the values reached at the ends of the valid range.

   * 'obliqui2.cpp' contains a couple of alternative obliquity
formulae.  'oblitest.cpp' tests them out.

Changes for 13 Aug 2011:

   * 'classel' had some issues with circular and very near-circular
orbits.  You could get "not a number" and "infinite" results in
these cases,  due to mathematical singularities.  I rewrote much
of the function to handle this.

   Three vector cross-products are computed in that code,  which
really called for a separate vector_cross_product() function,
now to be found in miscell.cpp.

Changes for 29 Jul 2011:

   * 'lunar2.cpp' relied on structures being packed to four-byte
boundaries and on 'long' being a 32-bit type.  That broke on a
64-bit compile on an Ubuntu box.

   * As part of some contract work,  I broke up 'riseset3.cpp'
so that it contained only the core functionality,  putting the
test routine into 'tables.cpp'.  I added a new 'lun_tran.cpp'
with functions for computing the time when the moon transits,
as well as the 'anti-transit' time (time when the moon is at
its lowest altitude).

Changes for 24 May 2011:

   * Prompted by Jean-Claude Pelle,  I made some major changes to
the COSPAR routines (code in 'cospar.cpp',  data in 'cospar.txt').
These are mostly described in 'cospar.txt'.  Note that there is now
a 'cosptest.cpp' unit test program,  which I used to verify that
nothing got broken in the course of these improvements.  (Except
for the fact that more objects now have COSPAR data,  users should
see no change.)

Changes for 2 Mar 2011:

   * 'jsats' has been changed from the E2x3 theory of Lieske to the
later E5 theory.  I'd not realized it,  but the first edition of
Meeus' _Astronomical Algorithms_ used E2x3.  E5 is almost identical,
except that the coefficients have changed.  So the switch was not
especially difficult (though somewhat tedious).

   However,  comparisons using JPL _Horizons_ ephemerides make it look
as if the improvement isn't all that great; errors are still about 100
km over 1950-2020.  Which is usually more than sufficient.  However, if
the errors are at that level, we're adding in a lot of terms that are
essentially meaningless.  At some point,  I may #ifdef out a lot of
small terms.

   Errors grow outside the 1950-2020 range,  especially for Callisto
and Ganymede.  These show some quadratic drift along-orbit;  presumably,
one is (in the Horizons model) accelerating whilst the other decelerates.
(The E5 theory has no quadratic terms in mean motion;  I'm somewhat
inclined to add them for Callisto and Ganymede.)

   * 'htc20b':  new code to compute ephemerides for Helene, Telesto
and Calypso (small satellites of Saturn).  I haven't actually used
this code anywhere,  though the test code at the end indicates that
at least I'm getting the same results as the original version did.

   * 'precess2':  new code to compute precession matrices,  using
the formulae for _ecliptic_ precession from Meeus' _Astronomical
Algorithms_.  Previously,  I used the equatorial precession formulae.
But looking at the coefficients,  I was struck by the much smaller
magnitude of the cubic terms for the ecliptic formulae.  I realized
that this made sense,  since precession is almost entirely a rotation
around the ecliptic plane;  expressing it in equatorial coordinates
is not really a good way of doing things.  And indeed,  if you use
the ecliptic algorithm for precession,  the results are stable out
to many more millennia than they are for the equatorial formula.

   'precess2' is not currently used anywhere.  I hope to incorporate
it in the next version of this code,  in place of the current 'precess'.
(Which I eventually did;  see above changes for 23 Sep 2011.)

   * 'get_time' now allows entries such as '1q',  'fm-3', 'NM+4',
or '3q-1' to get times of the nearest first quarter phase,  or the
time three days before full moon,  or four days after new moon,
or the day before the third-quarter phase.  In each case,  the
event nearest the currently-chosen time will be selected,  so that
the time will change by (at most) half a lunation.

   * 'easter' now contains test code for determining the frequency at
which Easter will occur for a given date,  the years when it will occur
for a given date,  and a table of Easter dates for a span of years.

Changes for 10 Jan 2011:

   * In several files,  corrected 'millenia' to 'millennia'.

Changes for 8 Jan 2011:

   * 'jsats':  Prompted by an e-mail from Thomas Meyer,  I realized
that the Galilean satellite code had errors wherein the mean longitude
was sometimes used in places where the true longitude should have been
used.  This led to errors of as much as 1/200 of a Jovian radius,  or
about 1/4 of a satellite radius. This has been fixed.

   In the process of fixing this,  I realized that the code was poorly
coded in some ways,  and I rewrote a lot of it.  (There is a new
'jsattest.cpp' program which I used to verify that my changes in the
code didn't cause changes in ephemerides... except for my fix of the
above bug,  which _did_ cause changes.  Beneficial changes,  however.)

   Also,  note that the code is now GPL'ed.

Changes for 6 Jan 2011:

   * 'astcheck' now checks possible matches to ensure that they have
similar motion to the observed object.  This strongly suppresses
spurious matches,  enabling the program to search out to a degree
(instead of just a quarter degree) and to mag 22 (instead of mag 21).
I also revised the error/help message to be more informative.

Changes for 4 Jan 2011:

   * Prompted by a conversation with Yaakov Zik,  I added some
comments to 'delta_t.cpp'.  It is functionally unchanged,  just
commented better.  Also revised 'lunar.mng' (make file for the
MinGW compiler) so it wouldn't needlessly recompile certain files
(I'd messed up the original makefile.)

Changes for 3 Jan 2011:

   * Prompted by an e-mail from Thomas Meyer,  I added a lot
of comments in 'jsats.cpp' and fixed up 'jevent.cpp' to work
somewhat more rationally.  The former was functionally unchanged
(just commented better).  The latter now provides times in UT
instead of TD,  and truncates to minutes instead of rounding.

Changes for 15 Dec 2010:

   * I wanted to be able to output dates without spaces in
'full_ctime( )',  so I added a FULL_CTIME_NO_SPACES macro
in date.h and added a few corresponding lines to 'miscell.cpp'.
This was done to support some things going on in 'astcheck.cpp',
a new program that reads 80-column MPC astrometry data and looks
for corresponding objects.  'astcheck' is now included with this
code,  since it's a decent demonstration of some of the things
in this library.

Change for 1 Oct 2010:

   * Realized that 'get_time' handled decimal year input
relative to J2000 = JD 2451545,  i.e.,  1.5 January 2000.
Integer years were handled relative to 1 January,  which
makes more sense.  So I redefined J2000 to be JD 2451544.5.
Now,  entering a value such as 1999.99999 gives a value
shortly before 1 January 2000,  instead of just before midday
on that day.

Changes for 19 Sep 2010:

   * Added observed Delta-T data for January 2010 to the array of
Delta-T values in 'delta_t.cpp'.  Also added three projected leap
seconds for 2013,  2016,  and 2019,  to keep UTC-UT1 close to zero
over the next decade.  (Those projected leap seconds will almost
certainly prove to be wrong.)

Changes for 13 May 2010:

   * Martin Ettl kindly passed on some results from 'cppcheck'
(http://sourceforge.net/projects/cppcheck/), which revealed
the following issues:

[phases.cpp:295]: (Error) Resource leak: log_file
[phases.cpp:295]: (Error) Resource leak: data_file
[de_plan.cpp:299]: (Error) Resource leak: ifile
[persian.cpp:216]: (Error) Memory leak: jds
[mpcorb.cpp:29]: (Error) Uninitialized variable: rval
[rckin.cpp:75]: (Error) Resource leak: ifile
[elp82dat.cpp:383]: (Error) Deallocating a deallocated pointer: ifile
[colors.cpp:275]: (Error) Resource leak: ifile

   As it happens,  none of these are particularly worrisome.  In
certain processing programs,  I've relied on the fact that files
are closed and memory freed when the program terminates.  All of
these issues,  except for the one in 'elp82dat',  are of that sort.
I am reasonably certain that the 'elp82dat' reflects a problem in
cppcheck.

   Still,  I'm impressed with the fact that 'cppcheck' was able to
find such issues,  and would like to be able to use it in the
future.  So I've fixed all of the above issues (except the elp82dat
one),  in hopes of getting a clean bill of health from cppcheck in
the future.

Changes for 28 Mar 2010:

   * showelem.cpp:  All calendar conversions were being done with the
Gregorian calendar only.  I revised it so that the Julian calendar is
used before October 1582.  (Basically just meant using
CALENDAR_JULIAN_GREGORIAN rather than the default.)

Changes for 2 Feb 2010:

   * 'date.h':  added some macros demonstrating how one can set the
precision and format for date output.

Changes for 15 Feb 2010:

   * 'chinese.dat':  Following a question from Mark Sims,  I realized
that certain years in the data file for the Chinese calendar were set
incorrectly.  Specifically,  any year in which the preceding year had
an '11i' or '12i' intercalary month was stored with garbage data.  This
occurred in 68 years over the 5990-year range of the original
'chinese.dat'.  This was due to a small bug in the processing programs.
I've fixed this,  and generated a new 'chinese.dat' covering the years
-3000 to +7000 Gregorian (the previous one covered -3000 to +2990).

   Be advised that,  since the Chinese calendar is an observational one,
the extrapolations into the past and future get to be iffy as uncertainty
in Delta-T grows.  If the time of New Moon is very close to midnight
Chinese time (zone UT+8 hours),  a small change in Delta-T may cause a
month to begin a day earlier or later.  The first such case is the
lunar month corresponding to a New Moon at 28 Sep 2057 16:00:17 UT,
just after midnight in the UT+8 zone.  But some sources (including the
Hong Kong observatory) have this occurring just _before_ midnight.  The
uncertainty in Delta-T for 2057 is such that either of us could be right.
As one goes further into the future (or past),  uncertainty in Delta-T
grows,  and so does uncertainty about the calendar.

   Similar issues plague the Jalali calendar and the French Republican one
(sort of;  note that it's unclear if the Republican calendar was intended
to be algorithmic or observational).

Changes for 29 Jan 2010:

   * 'triton.cpp'/'marsuran.cpp':  Realized that 'marsuran' was essentially
no longer in use,  and in some cases shouldn't be used (the Uranian and
Martian satellite solutions were abysmal).  Ideally,  I'd just drop this
file completely.  But it also contains an algorithm for Triton's position,
which I need to keep because the 'rocks' solution is wrong (see comment
below for 'rocks.cpp'.)

   So I removed most of 'marsuran.cpp' and renamed it to 'triton.cpp'.
The renaming seemed wise,  given that all Martian and Uranian satellites
had been completely removed!

   * 'rocks.cpp':  realized that the data for Triton must be wrong somewhere;
the orbit given in this file (from JPL's NEP050 solution) results in
prograde motion.  I've not puzzled it all out yet,  but it's definitely
wrong.  Added a suitable comment to warn about this.

   Also,  Uranians are all based on URA091 now (instead of,  mostly,
URA086X.)

   * 'get_time.cpp':  Removed the 'scan_for_time' variable,  which was
occasionally set,  but never actually used anywhere.

   * 'cospar.cpp':  Realized that errors were mis-handled.  There's
code to extract the planet orientation and spin rate from 'cospar.txt',
then save it for the next call... but the error code was _not_ saved
for the next call.  Some of my code was therefore thinking it was getting
a perfectly decent orientation matrix,  even though it wasn't.

   Also,  added orientation data for the Plutonian moons Hydra and Nix.
No 'official' model exists for these,  but as described in 'cospar.txt',
I was able to make a reasonable rotation model for them.

Changes for 11 Jan 2010:

   * 'showelem.cpp':  now supports more digits of precision,  enough to
support the full range of 64-bit floats.  This isn't apt to be very useful
to anyone;  I added it to simplify some debugging I was doing in Find_Orb.

Changes for 24 Nov 2009:

   * Fixed some loss-of-precision problems in 'classel.cpp'.  Specifically,
I noticed that the argument of perihelion was computed using arcsin,
so that if sin(arg_per) was near +/-1,  you'd lose precision.  In fact,
one could get a result slightly outside the -1 <= sin(arg_per) <= 1 range.
Also did a lot of cleanup.

Changes for 21 Oct 2009:

   * Made a lot of small changes to evade gcc and MinGW compiler warnings.
I also investigated an old optimization bug in 'de_plan.cpp',  but was
unable to replicate it.  For the nonce,  at least,  I'm leaving optimizations
fully enabled.

Changes for 20 Sep 2009:

   * Fixed a bug in 'date.cpp' wherein a 'month_data' buffer of 13
bytes was subjected to the line

   memset( month_data, 0, 14);

   I think this got by because the extra byte was blank anyway.  But
the latest gcc caught this bug and considered it to be 'stack smashing'.
In the investigation of this bug,  I cleaned up some of the date.cpp and
jd.cpp code: no changes in functionality,  but it's easier to read.

Changes for 9 Apr 2009:

   * Added some comments to 'integrat' explaining how the integration
scheme works.

Changes for 4 Apr 2009:

   * Added a 'make' file for MinGW (gcc for Windows).  Everything compiled
Just Fine,  except for an issue with the use of asinh() in astfuncs.cpp.
That source file now recognizes that asinh is only needed with Microsoft
Visual C/C++.

   * COM_FILE.CPP:  when extracting a periodic comet name,  preference
now goes to the permanent designation.  So if given P/Van Ness (213P)
(P/2005 R2),  the name extracted would be (213P).  This should help
to evade some problems with certain comets appearing twice.

Changes for 21 Mar 2009:

   * Various files:  This code does a fair bit of direct reading of structures
from files on disk.  To do that,  it relies heavily on sizeof(short) = 2
and sizeof( long) = 4.  This is,  of course,  stupid,  and means any
64-bit compiles will break.  Where structures are read from disk,  they
now use int16_t and int32_t,  instead of short and long.

Changes for 16 Mar 2009:

   * GET_TIME.CPP:  If the input date/time has a field with a decimal
point,  you can be pretty sure that that is the day of the month. The
code now makes use of this fact for puzzling out the formats.  (For
example,  both '9.75/4' and '4/9.75' would be interpreted as 9 April
18:00.)

Changes for 13 Feb 2009:

   * ROCKS.CPP:  Got data for Deimos and Phobos from Bob Jacobson at JPL.
Rearranged the rock elements slightly,  and they're now in numerical order.
Also wrote a little 'rckin' program to take the JPL format and convert it
into C code,  fixed a mistake of long standing in the numbering of
Uranus XXV = Perdita,  and enabled certain satellites that had been
disabled (simply because Guide wasn't using them).

   * BIG_VSOP.CPP:  Added a file 'big_vsop.txt',  which describes how this
function and the data are set up.

   * INTEGRAT.CPP:  Improvements to handling input dates in various formats,
fix to allow Pluto to be in the file.  See 'integrat.htm' for details.

   * GUST86.CPP:  Code to compute positions for five main satellites of
Uranus (Miranda = Uranus V, Ariel = Uranus I, Umbriel = Uranus II,
Titania = Uranus III,  Oberon = Uranus IV).  Thanks are due to Chris
Marriott,  the author of _SkyMap Pro_,  for the original version of this
code.  I've had this sitting on my hard drive for an embarrassingly long
time.  See 'uranus1.cpp' and 'uranus2.cpp' for usage demonstrations.

Changes for 23 Oct 2008:

   * ASTFUNCS.CPP:  the kepler( ) function has gone unchanged for about
a decade,  so I was surprised to find _two_ bugs in it.  For certain
hyperbolic,  very high mean anomaly cases,  it failed to exit a loop,
because it looked for fabs( err) < thresh .  In reality,  delta_curr
was becoming suitably small,  and that was the value that should have
been tested.

   Also,  for elliptical cases,  the mean anomaly really ought to be
kept in the range -pi < mean_anom < pi.  For M slightly outside that
range,  you lose a little precision and waste an iteration or two.
For M far outside that range (i.e.,  "many orbits completed"),  you
lose _lots_ of accuracy.

   This also caused me to see some possibilities for loss of precision
with nearly-parabolic orbits,  resulting in a new 'near_parabolic'
function that does essentially the same math,  but in a way that avoids
subtracting nearly-equal quantities.

   * DELTA_T.CPP:  added newly announced leap second for end of 2008.

   * ROCKS.CPP: Updated orbital elements for Pan and Daphnis,  from
SAT276 to SAT291.

   * GET_TIME.CPP: added better handling of offsets,  so one can have
'13:14-10m +3h' interpreted as 16:04,  for example.  Also,  days of the
week can be entered (which meant I had to add some day-of-week functions
to 'date.cpp').

Changes for 6 Apr 2008:

   * GET_TIME.CPP: function mostly extracted from Guide that parses
time strings, sort of an inverse of 'full_ctime'.  It attempts to handle
odd inputs 'logically' (i.e.,  "2007-10-17" will be understood even if
those three values are in an odd order,  because 17 is clearly a day
of the month and 2007 a year,  leaving 10 as the month).

   * GET_TEST.CPP runs assorted tests of the above function.

   * MISCELL.CPP: fixed a rather stupid series of bugs in 'full_ctime'
which caused instants a few microseconds shy of midnight to appear as
2:24:00 (i.e., one-tenth of a day).

   * DIST_PA.CPP now has a "reverse" function:  given a point and
distance and position angle,  it computes the resulting point.

   * VISLIMIT.CPP was set up so that the instant the moon went below
the horizon,  its contributed light went to zero.  I'm not sure what
the actual behavior should be,  but that was definitely wrong.  I
revised it to have a fairly sudden exponential decrease as it goes
below the horizon.

   * JD.CPP shows Delta-T for the selected time.  That just provided
me with a convenient testing mechanism.  It also now uses the 'get_time.cpp'
function,  so the date on the command line can be more flexible (and
so I can test the 'get_time.cpp' functions.)

Changes for 2 Feb 2008:

   * ROCKS.CPP:  Updated most of the rocks and added some new ones.

   * COSPAR.CPP:  Revised so it gets data from 'cospar.txt'.  This
allows it to generate correct rotation data,  including the previously
neglected smaller terms,  for all planets and satellites,  including
many previously ignored as "insufficiently important".

   From time to time,  changes may be made to the COSPAR data in
'cospar.txt'.  (For example,  the model for Mercury was recently
improved,  and one matching the average rotation rate of the clouds
of Venus was added.)  Those changes will be described at the end of
that file,  in the comments section.

   The previous implementation is still provided,  as 'cospar2.cpp',
just for reference;  I don't expect to be making any use of it.

   * DELTA_T.CPP:  Updated the Delta-T lookup table.

   * EASTER.CPP:  23 Mar 2008 is Easter,  and a neighbor of mine asked
about when Easter would be so early again.  I revised the test routine
so you can find out in which years Easter will fall on a given day
(in addition to the previous,  more common "when will Easter be this
year").

   * SHOWELEM.CPP:  Epochs with precision greater than .01 day were
"shifted" to the nearest .01 day and the mean anomaly adjusted
accordingly.  This is fixed:  if you specify the "usual" sort of
epoch (i.e.,  midnight UT for some date),  the output will look as
it always did (truncated to one decimal place),  but otherwise,
decimal places will all get shown.

Changes for 3 July 2007:

   * JD.CPP:  Added code so that most cases where day/month/year
is entered on the command line,  instead of year/month/day,  will
be corrected.

Changes for 8 June 2007:

   * ROCKS.CPP:  Converted some variables to type 'const' to
help me see what the heck is going on.

   * INTEGRAT.CPP:  Made some small revisions because MPC altered
the 'mpcorb' format very slightly,  so that when the program looked
for the names 'Ceres',  'Pallas',  and 'Vesta',  it didn't find them.
Now it will,  whether the input is in the old or new format.

Changes for 23 April 2007:

   * JD.CPP:  Added an error message for cases where no date is
specified on the command line,  removed some unused macros and
variables,  and changed certain "variables" to be of type const.
(This is largely cosmetic:  if a particular "variable" is,  in
fact,  constant,  I like to make that fact obvious.)

   * ASTEPHEM.CPP:  This program now gets orbital elements from
'mpcorb.dat' (or 'mpcorbcr.dat'),  rather than from Guide-format
elements.  As before,  it uses those elements to generate
asteroid ephemerides.

   * COSPAR.CPP,  MISCELL.CPP:  compiling with gcc -Wall caused
a few warning messages to appear,  mostly for cases where things
defined as 'int' should have been 'long int' or vice versa.
Again,  cosmetic (unless one were on a system wherein 'int'
differs from 'long'.)

Mid-2006 changes:

   * ASTFUNCS.CPP:  A couple of lines in the 'kepler' function for
handling "hyperbolic, large-mean-anomaly cases".  Such cases caused
overflow problems (when _really_ hyperbolic and large-MA... but such
cases arose with objects in planetary encounters.) */

   * DIST_PA.CPP:  As comments in the source indicate,  there was a
roundoff problem with certain position angles where the RAs were
identical, or nearly so.  I essentially had to do a total rewrite,
abandoning the haversine-based approach.

   * SHOWELEM.CPP:  Distances are now shown in kilometers (not AU)
whenever they are less than 400000 km (roughly the distance from the
earth to the moon).

   * DELTA_T.CPP: Added leap second for 1 January 2006,  plus comments
on which leap second is which.

Changes for 29 March 2005:

   * SHOWELEM.CPP:  The 'decimal_day_to_dmy' function now takes a
'calendar' argument,  instead of assuming Gregorian.  That allows
use of the function by assorted calendar-dependent routines in Guide.
Most other uses will just pass in a 0 (Gregorian) value.

   * REFRACT4.CPP:  Some code for a 'reverse_integrated_refraction'
function,  which inverts the 'integrated_refraction' function.

Changes for 3 Feb 2005:

   * DATE.CPP:  Found and fixed a bug in the 'is_hebrew_leap_year' function.
   * SOLSEQN.CPP:  Added some error handling for cases where the 'big_vsop.bin'
file isn't found.
   * Generally:  Switched use of asin( ) to asine( ),  i.e.,  the version that
protects against domain errors.  I was getting such errors in Find_Orb,
and was trying to track them down.  It turned out to have nothing to do
with asin( ) calls in this library... but they're a good idea anyway,
just on general principle.

Changes for 16 Dec 2004:

   * DATE.CPP:  Changed a lot of variables to be of type 'const'.
This has no bearing on the actual operation of the code,  but it
does mean that a human reading the code can say:  "OK,  I know
these variables aren't going to change within the function."

Changes for 1 Dec 2004:

   * DELTA_T.CPP:  Previously,  there was an adjustment to Delta-T
within the code now marked by "#ifdef TRY_OMITTING",  for years before
1620.  The idea was that the formulae for Delta-T were originally
derived with an assumed lunar secular acceleration of 26 arcseconds
per century^2.  If you use a theory such as VSOP or ELP,  which
assume a secular acceleration of -23.8946 arcsec/cy^2,  then you
have to assume the earth's spin is,  in the long run,  decelerating
at a different rate... after all,  if the moon is receding at a different
rate,  it's soaking up the earth's angular momentum at a different rate,
too.

   However,  I'm now using DE-405 and DE-406 for just about everything.
I could have made sure that Delta-T was adjusted for DE-40*,  or the
other way around.  I chose the latter.  (Dealing with these assorted
time systems is a heck of a mess!)

   * ALT_AZ.CPP:  Some new code to add "supergalactic" transformations
as a parallel to the galactic ones.

Changes for 7 April 2004:

   * DATE.CPP:  the "modern Persian" calendar is now supported.  This
corresponds quite closely to the traditional Jalaali calendar,  but
instead of being a purely observational calendar,  683 leap years are
spread evenly over a 2820-year span.

Changes for 17 Dec 2003:

   * MISCELL.CPP:  the 'full_ctime' has been mostly rewritten to allow
handling of different calendars,  and more flexible formats and setting
of levels of precision.  Most of this was required by my Guide software,
to support a new feature wherein people could specify exactly how times
and dates are to be formatted (see
http://www.projectpluto.com/update8.htm#time_format .)  All this also
meant some changes to the FULL_CTIME macros in 'date.h'.

   * DATE.CPP:  the code for setting/storing month names used to copy
in the month names provided to set_month_name(),  which meant you
couldn't have excessively long names (more than six bytes plus a
trailing NUL). Now,  instead,  set_month_name() maintains a table of
pointers to the user-specified month names,  and provision is made
for a thirteenth month name (useful for the Hebrew and French
calendars).  All this was required for the same 'time_format' dialog
described in the URL given in the previous paragraph.

   * DELTA_T.CPP: brought the table of delta_t values up to date.
Values for late 2003 were off by about 1.2 seconds!

   * INTEGRAT.CPP: this code used to assume that the first,  second,
and fourth asteroids in an 'mpc_orb.dat' file were Ceres,  Pallas,  and
Vesta.  It then assumed that it could compute positions for these three
asteroids,  and compute perturbations caused by them.  That was a bad
move,  since there are 'mpc_orb'-formatted files that don't contain these
four objects.  The code now looks for the actual asteroid names "Ceres",
etc. instead.

   * NUTATION.CPP: fixed an erroneous coefficient reported by James
Miller and Mark Huss;  that repair required me to clarify the code a
bit. It's not quite as cryptic as previous versions were.

Changes for 12 Dec 2002:

   * Added REFRACT4.CPP,  code to compute a truly "accurate" refraction
value via numerical integration.  Compare to the code in REFRACT.CPP
(which is much simpler,  but not quite as precise.)

   * COLORS.CPP was fixed to read 'loneos.pho' properly (it was getting
confused by lines where the RA was given to .001-second precision).

   * The Republican calendar code in DATE.CPP was returning weird
values (and sometimes just locking up!) for dates prior to
JD 2007729.5 or after JD 2598322.5.  The values outside that range
are still not necessarily exact,  but should at least be close,
with more "graceful" degradation of accuracy.

   * REFRACT.CPP had an error,  where I thought the input value was
in degrees,  but it was really in arcminutes.

   * INTEGRAT.CPP and SHOWELEM.CPP had slightly "bad" planetary masses
(in particular,  that for Earth was off by about .5%!)

Changes for 6 June 2002:

   * There's now a LUNARDLL.MAK file,  for creating a 32-bit Windoze DLL.

   * The #defines for AU_PER_DAY and AU_IN_KM were moved into
'afuncs.h'.  I hadn't realized this,  but there were a few places
where both were defined with fewer digits than one might ideally like.
This affects a slew of files.

   * DATE.CPP:  the method for handling the French Revolutionary
(Republican) calendar was changed a bit.  You can still use the
"algorithmic" methods,  but there's some new code that allows you
to use the original ugly scheme in which the year begins on the
autumnal equinox,  so leap years occur at either four or five-year
intervals.  (#define's can switch you back to the other methods.)
The code looks much like that for the Persian Jalaali calendar,
and I used the same program,  PERSIAN.CPP (q.v.),  slightly modified,
to generate the lookup table.

   * DELTA_T.CPP now contains a table of values of Delta-T to .01
second,  not .1 second,  precision.  I made the size of the table a
#define,  since I keep increasing it a bit from time to time. There
are some new notes at the bottom of the file,  because at some point,
I want to include some code to give,  not just Delta-T = TDT - UT1,
but TDT - UTC.  For the latter,  I'll mostly just need a table of
leap seconds.  But none of this affects the code yet.

   * DIST_PA.CPP:  in reply to an inquiry as to how this function
works,  I added some comments and revised a bug in how the PA was done.

   * EASTER.CPP:  mostly cosmetic change of some values to type 'const'.

   * PERSIAN.CPP:  code added to do a similar analysis for the French
Revolutionary (Republican) calendar.  See DATE.CPP.  (I also switched
to use of the "full VSOP" as used in BIG_VSOP.CPP,  rather than the
"small VSOP" used previously.  That gave better accuracy,  but also
meant I had to make some minor changes to the make files.)

   * PRECESS.CPP:  Jordi Mas reported that the function
'setup_ecliptic_precession( )' was broken,  and has probably been
so since it was first written.  (I never actually had a situation
where I used it.)  It's fixed now.

   * SSATS.CPP:  After I got an inquiry about these formulae,  I added
some comments near the top,  and made some cosmetic changes such as
converting some values to be of type 'const'.