为什么我们需要一个接一个的案件陈述?

为什么编译器不在开关中的每个代码块后自动放置 break 语句?是因为历史原因吗?您希望何时执行多个代码块?

75993 次浏览

Not having an automatic break added by the compiler makes it possible to use a switch/case to test for conditions like 1 <= a <= 3 by removing the break statement from 1 and 2.

switch(a) {
case 1: //I'm between 1 and 3
case 2: //I'm between 1 and 3
case 3: //I'm between 1 and 3
break;
}

Sometimes it is helpful to have multiple cases associated with the same code block, such as

case 'A':
case 'B':
case 'C':
doSomething();
break;


case 'D':
case 'E':
doSomethingElse();
break;

etc. Just an example.

In my experience, usually it is bad style to "fall through" and have multiple blocks of code execute for one case, but there may be uses for it in some situations.

So you do not have to repeat code if you need several cases to do the same thing:

case THIS:
case THAT:
{
code;
break;
}

Or you can do things like :

case THIS:
{
do this;
}
case THAT:
{
do that;
}

In a cascade fashion.

Really bug/confusion prone, if you ask me.

Exactly, because with some clever placement you can execute blocks in cascade.

You can do all sorts of interesting things with case fall-through.

For example, lets say you want to do a particular action for all cases, but in a certain case you want to do that action plus something else. Using a switch statement with fall-through would make it quite easy.

switch (someValue)
{
case extendedActionValue:
// do extended action here, falls through to normal action
case normalActionValue:
case otherNormalActionValue:
// do normal action here
break;
}

Of course, it is easy to forget the break statement at the end of a case and cause unexpected behavior. Good compilers will warn you when you omit the break statement.

because there are situations where you want to flow through the first block for example to avoid writing the same code in multiple blocks but still be able to divide them for mroe control. There are also a ton of other reasons.

Java comes from C and that is the syntax from C.

There are times where you want multiple case statements to just have one execution path. Below is a sample that will tell you how many days in a month.

class SwitchDemo2 {
public static void main(String[] args) {


int month = 2;
int year = 2000;
int numDays = 0;


switch (month) {
case 1:
case 3:
case 5:
case 7:
case 8:
case 10:
case 12:
numDays = 31;
break;
case 4:
case 6:
case 9:
case 11:
numDays = 30;
break;
case 2:
if ( ((year % 4 == 0) && !(year % 100 == 0))
|| (year % 400 == 0) )
numDays = 29;
else
numDays = 28;
break;
default:
System.out.println("Invalid month.");
break;
}
System.out.println("Number of Days = " + numDays);
}
}

I think it is a mistake. As a language construct it is just as easy to have break as the default and instead have a fallthrough keyword. Most of the code I have written and read has a break after every case.

Java is derived from C, whose heritage includes a technique known as Duff's Device . It's an optimization that relies on the fact that control falls through from one case to the next, in the absence of a break; statement. By the time C was standardized, there was plenty of code like that "in the wild", and it would have been counterproductive to change the language to break such constructions.

Why doesn't the compiler automatically put break statements after each code block in the switch?

Leaving aside the good desire to be able to use the identical block for several cases (which could be special-cased)...

Is it for historical reasons? When would you want multiple code blocks to execute?

It's mainly for compatibility with C, and is arguably an ancient hack from the days of old when goto keywords roamed the earth. It does enable some amazing things, of course, such as Duff's Device, but whether that's a point in its favor or against is… argumentative at best.

Historically, it's because the case was essentially defining a label, also known as the target point of a goto call. The switch statement and its associated cases really just represent a multiway branch with multiple potential entry points into a stream of code.

All that said, it has been noted a nearly infinite number of times that break is almost always the default behavior that you'd rather have at the end of every case.

As far as the historical record goes, Tony Hoare invented the case statement in the 1960s, during the "structured programming" revolution. Tony's case statement supported multiple labels per case and automatic exit with no stinking break statements. The requirement for an explicit break was something that came out of the BCPL/B/C line. Dennis Ritchie writes (in ACM HOPL-II):

For example, the endcase that escapes from a BCPL switchon statement was not present in the language when we learned it in the 1960s, and so the overloading of the break keyword to escape from the B and C switch statement owes to divergent evolution rather than conscious change.

I haven't been able to find any historical writings about BCPL, but Ritchie's comment suggests that the break was more or less a historical accident. BCPL later fixed the problem, but perhaps Ritchie and Thompson were too busy inventing Unix to be bothered with such a detail :-)

It is an old question but actually I ran into using the case without break statement today. Not using break is actually very useful when you need to combine different functions in sequence.

e.g. using http response codes to authenticate user with time token

server response code 401 - token is outdated -> regenerate token and log user in.
server response code 200 - token is OK -> log user in.

in case statements:

case 404:
case 500:
{
Log.v("Server responses","Unable to respond due to server error");
break;
}
case 401:
{
//regenerate token
}
case 200:
{
// log in user
break;
}

Using this you do not need to call log in user function for 401 response because when the token is regenerated, the runtime jumps into the case 200.

You can makes easily to separate other type of number, month, count.
This is better then if in this case;

public static void spanishNumbers(String span){


span = span.toLowerCase().replace(" ", "");
switch (span){
case "1":
case "jan":  System.out.println("uno"); break;
case "2":
case "feb":  System.out.println("dos"); break;
case "3":
case "mar":  System.out.println("tres"); break;
case "4":
case "apr":  System.out.println("cuatro"); break;
case "5":
case "may":  System.out.println("cinco"); break;
case "6":
case "jun":  System.out.println("seis"); break;
case "7":
case "jul":  System.out.println("seite"); break;
case "8":
case "aug":  System.out.println("ocho"); break;
case "9":
case "sep":  System.out.println("nueve"); break;
case "10":
case "oct": System.out.println("diez"); break;
}
}

I am now working on project where I am in need of break in my switch statement otherwise the code won't work. Bear with me and I will give you a good example of why you need break in your switch statement.

Imagine you have three states, one that waits for the user to enter a number, the second to calculate it and the third to print the sum.

In that case you have:

  1. State1 - Wait for user to enter a number
  2. State2 - Print the sum
  3. state3 - Calculate the sum

Looking at the states, you would want the order of exaction to start on state1, then state3 and finally state2. Otherwise we will only print users input without calculating the sum. Just to clarify it again, we wait for the user to enter a value, then calculate the sum and prints the sum.

Here is an example code:

while(1){
switch(state){
case state1:
// Wait for user input code
state = state3; // Jump to state3
break;
case state2:
//Print the sum code
state = state3; // Jump to state3;
case state3:
// Calculate the sum code
state = wait; // Jump to state1
break;
}
}

If we don't use break, it will execute in this order, state1, state2 and state3. But using break, we avoid this scenario, and can order in the right procedure which is to begin with state1, then state3 and last but not least state2.

As people said before, it is to allow fall-through and it is not a mistake, it is a feature. If too many break statements annoy you, you can easily get rid of them by using return statements instead. This is actually a good practice, because your methods should be as small as possible (for the sake of readability and maintainability), so a switch statement is already big enough for a method, hence, a good method should not contain anything else, this is an example:

public class SwitchTester{
private static final Log log = LogFactory.getLog(SwitchTester.class);
public static void main(String[] args){
log.info(monthsOfTheSeason(Season.WINTER));
log.info(monthsOfTheSeason(Season.SPRING));
log.info(monthsOfTheSeason(Season.SUMMER));
log.info(monthsOfTheSeason(Season.AUTUMN));
}


enum Season{WINTER, SPRING, SUMMER, AUTUMN};


static String monthsOfTheSeason(Season season){
switch(season){
case WINTER:
return "Dec, Jan, Feb";
case SPRING:
return "Mar, Apr, May";
case SUMMER:
return "Jun, Jul, Aug";
case AUTUMN:
return "Sep, Oct, Nov";
default:
//actually a NullPointerException will be thrown before reaching this
throw new IllegalArgumentException("Season must not be null");
}
}
}

The execution prints:

12:37:25.760 [main] INFO lang.SwitchTester - Dec, Jan, Feb
12:37:25.762 [main] INFO lang.SwitchTester - Mar, Apr, May
12:37:25.762 [main] INFO lang.SwitchTester - Jun, Jul, Aug
12:37:25.762 [main] INFO lang.SwitchTester - Sep, Oct, Nov

as expected.

The break after switch cases is used to avoid the fallthrough in the switch statements. Though interestingly this now can be achieved through the newly formed switch labels as implemented via JEP-325.

With these changes, the break with every switch case can be avoided as demonstrated further :-

public class SwitchExpressionsNoFallThrough {


public static void main(String[] args) {
Scanner scanner = new Scanner(System.in);
int value = scanner.nextInt();
/*
* Before JEP-325
*/
switch (value) {
case 1:
System.out.println("one");
case 2:
System.out.println("two");
default:
System.out.println("many");
}


/*
* After JEP-325
*/
switch (value) {
case 1 ->System.out.println("one");
case 2 ->System.out.println("two");
default ->System.out.println("many");
}
}
}

On executing the above code with JDK-12, the comparative output could be seen as

//input
1
// output from the implementation before JEP-325
one
two
many
// output from the implementation after JEP-325
one

and

//input
2
// output from the implementation before JEP-325
two
many
// output from the implementation after JEP-325
two

and of course the thing unchanged

// input
3
many // default case match
many // branches to 'default' as well