Historically, the fundamental time unit—the second—was derived from Earth’s rotation around its axis. There are 24 hours or 24 x 60 x 60 = 86400 seconds in a full revolution, so it seems just a question of astronomical measurements to precisely define a second. Unfortunately, Earth wobbles slightly, and a more precise definition was needed. In 1967, a new precise definition of a second, matching the historical definition, was derived from an intrinsic property of atoms of caesium-133. Since then, a network of atomic clocks keeps the official time.
Ever so often, the official time keepers synchronize the absolute time with the rotation of Earth. At first, the official seconds were slightly adjusted, but starting in 1972, “leap seconds” were occasionally inserted. (In theory, a second might need to be removed once in a while, but that has not yet happened.) There is talk of changing the system again. Clearly, leap seconds are a pain, and many computer systems instead use “smoothing” where time is artificially slowed down or sped up just before the leap second, keeping 86,400 seconds per day. This works because the local time on a computer isn’t all that precise, and computers are used to synchronizing themselves with an external time service.
The Java Date and Time API specification requires that Java uses a time scale that
- Has 86,400 seconds per day
- Exactly matches the official time at noon each day
- Closely matches it elsewhere, in a precisely defined way
That gives Java the flexibility to adjust to future changes in the official time.
In Java, an Instant represents a point on the time line. An origin, called the epoch, is arbitrarily set at midnight of January 1, 1970 at the prime meridian that passes through the Greenwich Royal Observatory in London. This is the same convention used in the UNIX/POSIX time. Starting from that origin, time is measured in 86,400 seconds per day, forward and backward, to nanosecond precision. The Instant values go back as far as a billion years (instant.MIN). That’s not quite enough to express the age of the universe (around 13.5 billion years) but it should be enough for all practical purposes. After all, a billion years ago, the earth was covered in ice and populated by microscopic ancestors of today’s plants and animals. The largest value, Instant.MAX, is December 31 of the year 1,000,000,000.
The static method call Instant.now() gives the current instant. You can compare two instants with the equals and compareTo methods in the usual way, so you can use instants as timestamps.
To find out the difference between two instants, use the static method Duration.between. For example, here is how you can measure the running time of an algorithm:
Instant start = Instant.now(); runAlgorithm();
Instant end = Instant.now();
Duration timeElapsed = Duration.between(start, end);
long millis = timeElapsed.toMillis();
A Duration is the amount of time between two instants. You can get the length of a Duration in conventional units by calling toNanos, toMillis, toSeconds, toMinutes, toHours, or toDays.
If you need nanosecond precision, be aware of overflow. A long value can hold almost 300 years of nanoseconds. If your durations are shorter than that, simply convert them to nanoseconds. You can use longer durations—a Duration object stores the number of seconds in a long, and the number of nanoseconds in an additional int. The Duration API has a number of methods, shown at the end of this section, for carrying out arithmetic.
For example, if you want to check whether an algorithm is at least ten times faster than another, you can compute
Duration timeElapsed2 = Duration.between(start2, end2);
This is just to show the syntax. Since the algorithms aren’t going to run for hundreds of years, you can simply use
boolean overTenTimesFaster = timeElapsed.toNanos() * 10 < timeElapsed2.toNanos();
In the example program in Listing 6.1, you can see how to use the Instant and Duration classes for timing two algorithms.
Source: Horstmann Cay S. (2019), Core Java. Volume II – Advanced Features, Pearson; 11th edition.