Improve JulianDateTime

[igr]

Let's discuss here all the issues and corner cases of the JulianDateTime

@cusher let's list everything we may fix with the JulianDateTime.

[cusher]

I came up a slightly obnoxious test case for precision. Quick calculation to start off with:

1 JD = 86400s
0.00000000001 JD = 864ns
0.00000000578 JD = 499392ns = 0.499392ms
0.00000000579 JD = 500256ns = 0.500256ms

So in theory if we add/subtract 0.00000000578 from any JulianDate that is a whole number of millis, it should not round away, while if we add/subtract 0.00000000579 it should round away.

And so:

	@Test
	void bigDecimalToLocalDateTimeDelta() {
		JulianDate jdt = new JulianDate(BigDecimal.valueOf(2457754.4));
		LocalDateTime ldt = LocalDateTime.of(2016, 12, 31, 21, 36, 0, 0);

		assertEquals(ldt, jdt.toLocalDateTime());
		assertEquals(ldt, jdt.sub(0.00000000578).toLocalDateTime());
		assertEquals(ldt, jdt.add(0.00000000578).toLocalDateTime());
		assertEquals(ldt.minus(1, ChronoUnit.MILLIS), jdt.sub(0.00000000579).toLocalDateTime());
		assertEquals(ldt.plus(1, ChronoUnit.MILLIS), jdt.add(0.00000000579).toLocalDateTime());
	}

That's the only other thing I can think of.

[igr]

I've added the following:

final int time_millisecond = (int) Math.round(d_millis - 0.0085);

in the method toLocalDateTime() and this makes this test green (other tests as well).

Do you think we can commit this @cusher ?

[cusher]

@igr This test should already pass, in fact I added an updated version of this check in testDeltaRounding. The change I made that got this test passing was removing the + 1.0e-10 in toLocalDateTime:

jodd-util/src/main/java/jodd/time/JulianDate.java

Line 317 in 695d909

frac = fraction + 0.5 - ka + 1.0e-10;

jodd-util/src/main/java/jodd/time/JulianDate.java

Line 319 in 833d872

frac = fraction + 0.5 - ka;

[igr]

Got it, I was solving the same issue in a different place! I just didn't pull 🤦‍♂️ (sorry, doing couple of things in the same time:)

We can release now I would say :)