What is the difference of using TemporalAmount or TemporalUnit in Java 8?

Question

I write some piece of code in Java 8 which use time arithmetic. I realize that I can implement in differentways. Lets look at simple code below. Of course it is the same result but I confused which way is mostly applied or most efficient to make arithmetic operations in Java 8 ?

LocalTime time = LocalTime.now();
// 1st way
LocalTime plusOp = time.plus(Duration.ofMinutes(10L));
// 2nd way
LocalTime plusOp2 = time.plus(10L, ChronoUnit.MINUTES);
System.out.println(plusOp);
System.out.println(plusOp2);
// 3. way simply
 time.plusMinutes(10L);

Thanks in advance.

Answer 1

Duration can only handle fixed-length periods, such as "hours", "minutes", "seconds", "days" (where it assumes exactly 24 hours per day). You can't use "months" with Duration, because a month varies in length.

Period - the other common TemporalAmount implementation - represents years, months and days separately.

Personally I would recommend:

• When you know the unit beforehand, use the appropriate plusXxx method, e.g. time.plusMinutes(10). That's about as easy to read as it gets.
• When you're trying to represent "logical" calendar amounts, use Period
• When you're trying to represent "fixed length" amounts, use Duration

Here's an example of where Period and Duration can differ:

import java.time.*;

public class Test {
    public static void main(String[] args) {
        ZoneId zone = ZoneId.of("Europe/London");
        // At 2015-03-29T01:00:00Z, Europe/London goes from UTC+0 to UTC+1
        LocalDate transitionDate = LocalDate.of(2015, 3, 29);
        ZonedDateTime start = ZonedDateTime.of(transitionDate, LocalTime.MIDNIGHT, zone);
        ZonedDateTime endWithDuration = start.plus(Duration.ofDays(1));
        ZonedDateTime endWithPeriod = start.plus(Period.ofDays(1));
        System.out.println(endWithDuration); // 2015-03-30T01:00+01:00[Europe/London]
        System.out.println(endWithPeriod);   // 2015-03-30T00:00+01:00[Europe/London]
    }
}

I wouldn't worry about the efficiency until you really need to - at which point you should have a benchmark so you can test different options.