I have been trying to figure out how to convert a timestamp to a date but with the trailing decimals at the end, so for example: Timestamp - C50204EC EC42EE92 is equivalent to S
Using the example from the explanation:
Timestamp - C50204EC EC42EE92 is equivalent to Sep 27, 2004 03:18:04.922896299 UTC.
Instant epoch = OffsetDateTime.of(1900, 1, 1, 0, 0, 0, 0, ZoneOffset.UTC).toInstant();
BigInteger timeStamp = new BigInteger("C50204ECEC42EE92", 16);
// To get the whole part and the fraction right, divide by 2^32
double secondsSince1900 = timeStamp.doubleValue() / 0x1_0000_0000L;
// Convert seconds to nanos by multiplying by 1 000 000 000
Instant converted = epoch.plusNanos(Math.round(secondsSince1900 * 1_000_000_000L));
System.out.println(converted);
Output is:
2004-09-27T03:18:04.922896384Z
It’s off by 85 nanoseconds. Likely better floating-point arithmetic can do even better. Edit: A little loss of precision is unavoidable since the original time stamp has a resolution of 2^-32 seconds, which is more than 4 times as fine as the nanosecond (10^-9 second) resolution of Instant
.
The Calendar
class that you were trying to use was always poorly designed and is now long outdated. Instead I do as Amongalen suggested in a comment, I am using java.time, the modern Java date and time API. Edit: For comparison Calendar
has millisecond resolution, so would at best give you a substabtial loss of precision.
I couldn’t let the 85 nanoseconds be. Here’s a version that preserves precision as far as possible and gives the expected result:
BigDecimal timeStamp = new BigDecimal(new BigInteger("C50204ECEC42EE92", 16));
// To get the whole part and the fraction right, divide by 2^32
BigDecimal bit32 = new BigDecimal(0x1_0000_0000L);
BigDecimal secondsSince1900 = timeStamp.divide(bit32);
// Convert seconds to nanos by multiplying by 1 000 000 000; round to long
long nanosSince1900 = secondsSince1900.multiply(new BigDecimal(TimeUnit.SECONDS.toNanos(1)))
.setScale(0, RoundingMode.HALF_UP)
.longValueExact();
Instant converted = epoch.plusNanos(nanosSince1900);
2004-09-27T03:18:04.922896300Z
1 nano too much? This is because I used half-up rounding in the call to setScale
. If instead I truncate (using RoundingMode.FLOOR
), I get the exact result from the explanation. So my version doesn’t lose more precision than theirs.
Oracle tutorial: Date Time explaining how to use java.time.