} is fully defined as a specific chronology.In that case, the assumptions of that chronology are known at development time and specified in the code. When the chronology is defined in the generic type parameter as ? or otherwise unknown at development time, the rest of the discussion below applies.
To emphasize the point, declaring a method signature, field or variable as this interface type can initially seem like the sensible way to globalize an application, however it is usually the wrong approach. As such, it should be considered an application-wide architectural decision to choose to use this interface as opposed to {@code LocalDate}.
Architectural issues to consider
These are some of the points that must be considered before using this interface throughout an application. 1) Applications using this interface, as opposed to using just {@code LocalDate}, face a significantly higher probability of bugs. This is because the calendar system in use is not known at development time. A key cause of bugs is where the developer applies assumptions from their day-to-day knowledge of the ISO calendar system to code that is intended to deal with any arbitrary calendar system. The section below outlines how those assumptions can cause problems The primary mechanism for reducing this increased risk of bugs is a strong code review process. This should also be considered a extra cost in maintenance for the lifetime of the code.
2) This interface does not enforce immutability of implementations. While the implementation notes indicate that all implementations must be immutable there is nothing in the code or type system to enforce this. Any method declared to accept a {@code ChronoLocalDate} could therefore be passed a poorly ormaliciously written mutable implementation.
3) Applications using this interface must consider the impact of eras. {@code LocalDate} shields users from the concept of eras, by ensuring that {@code getYear()}returns the proleptic year. That decision ensures that developers can think of {@code LocalDate} instances as consisting of three fields - year, month-of-year and day-of-month.By contrast, users of this interface must think of dates as consisting of four fields - era, year-of-era, month-of-year and day-of-month. The extra era field is frequently forgotten, yet it is of vital importance to dates in an arbitrary calendar system. For example, in the Japanese calendar system, the era represents the reign of an Emperor. Whenever one reign ends and another starts, the year-of-era is reset to one.
4) The only agreed international standard for passing a date between two systems is the ISO-8601 standard which requires the ISO calendar system. Using this interface throughout the application will inevitably lead to the requirement to pass the date across a network or component boundary, requiring an application specific protocol or format.
5) Long term persistence, such as a database, will almost always only accept dates in the ISO-8601 calendar system (or the related Julian-Gregorian). Passing around dates in other calendar systems increases the complications of interacting with persistence.
6) Most of the time, passing a {@code ChronoLocalDate} throughout an applicationis unnecessary, as discussed in the last section below.
False assumptions causing bugs in multi-calendar system code
As indicated above, there are many issues to consider when try to use and manipulate a date in an arbitrary calendar system. These are some of the key issues. Code that queries the day-of-month and assumes that the value will never be more than 31 is invalid. Some calendar systems have more than 31 days in some months.
Code that adds 12 months to a date and assumes that a year has been added is invalid. Some calendar systems have a different number of months, such as 13 in the Coptic or Ethiopic.
Code that adds one month to a date and assumes that the month-of-year value will increase by one or wrap to the next year is invalid. Some calendar systems have a variable number of months in a year, such as the Hebrew.
Code that adds one month, then adds a second one month and assumes that the day-of-month will remain close to its original value is invalid. Some calendar systems have a large difference between the length of the longest month and the length of the shortest month. For example, the Coptic or Ethiopic have 12 months of 30 days and 1 month of 5 days.
Code that adds seven days and assumes that a week has been added is invalid. Some calendar systems have weeks of other than seven days, such as the French Revolutionary.
Code that assumes that because the year of {@code date1} is greater than the year of {@code date2}then {@code date1} is after {@code date2} is invalid. This is invalid for all calendar systemswhen referring to the year-of-era, and especially untrue of the Japanese calendar system where the year-of-era restarts with the reign of every new Emperor.
Code that treats month-of-year one and day-of-month one as the start of the year is invalid. Not all calendar systems start the year when the month value is one.
In general, manipulating a date, and even querying a date, is wide open to bugs when the calendar system is unknown at development time. This is why it is essential that code using this interface is subjected to additional code reviews. It is also why an architectural decision to avoid this interface type is usually the correct one.
Using LocalDate instead
The primary alternative to using this interface throughout your application is as follows. - Declare all method signatures referring to dates in terms of {@code LocalDate}.
- Either store the chronology (calendar system) in the user profile or lookup the chronology from the user locale
- Convert the ISO {@code LocalDate} to and from the user's preferred calendar system duringprinting and parsing
This approach treats the problem of globalized calendar systems as a localization issue and confines it to the UI layer. This approach is in keeping with other localization issues in the java platform.
As discussed above, performing calculations on a date where the rules of the calendar system are pluggable requires skill and is not recommended. Fortunately, the need to perform calculations on a date in an arbitrary calendar system is extremely rare. For example, it is highly unlikely that the business rules of a library book rental scheme will allow rentals to be for one month, where meaning of the month is dependent on the user's preferred calendar system.
A key use case for calculations on a date in an arbitrary calendar system is producing a month-by-month calendar for display and user interaction. Again, this is a UI issue, and use of this interface solely within a few methods of the UI layer may be justified.
In any other part of the system, where a date must be manipulated in a calendar system other than ISO, the use case will generally specify the calendar system to use. For example, an application may need to calculate the next Islamic or Hebrew holiday which may require manipulating the date. This kind of use case can be handled as follows:
- start from the ISO {@code LocalDate} being passed to the method
- convert the date to the alternate calendar system, which for this use case is known rather than arbitrary
- perform the calculation
- convert back to {@code LocalDate}
Developers writing low-level frameworks or libraries should also avoid this interface. Instead, one of the two general purpose access interfaces should be used. Use {@link TemporalAccessor} if read-only access is required, or use {@link Temporal}if read-write access is required.
Specification for implementors
This interface must be implemented with care to ensure other classes operate correctly. All implementations that can be instantiated must be final, immutable and thread-safe. Subclasses should be Serializable wherever possible. Additional calendar systems may be added to the system. See {@link Chronology} for more details.
In JDK 8, this is an interface with default methods. Since there are no default methods in JDK 7, an abstract class is used.