Switch without break

Question

I have some switch statement as shown below. Notice there is no break. Findbugs is reporting error on the second case statement only. The error is : Switch statement found where one case falls through to the next case.

switch(x) {

    case 0:
        // code

    case 1:
        // code

    case 2:
        // code
}

Solution 1

Findbugs is flagging up that falling through from one case to the next is generally not a good idea if there's any code in the first one (although sometimes it can be used to good effect). So when it sees the second case and no break, it reports the error.

So for instance:

switch (foo) {
    case 0:
        doSomething();
    case 1:
        doSomethingElse();
    default:
        doSomeOtherThing();
}

This is perfectly valid Java, but it probably doesn't do what the author intended: If foo is 0, all three of the functions doSomething, doSomethingElse, and doSomeOtherThing run (in that order). If foo is 1, only doSomethingElse and doSomeOtherThing run. If foo is any other value, only doSomeOtherThing runs.

In contrast:

switch (foo) {
    case 0:
        doSomething();
        break;
    case 1:
        doSomethingElse();
        break;
    default:
        doSomeOtherThing();
        break;
}

Here, only one of the functions will run, depending on the value of foo.

Since it's a common coding error to forget the break, tools like Findbugs flag it up for you.

There's a common use-case where you have multiple case statements in a row with no intervening code:

switch (foo) {
    case 0:
    case 1:
        doSomething();
        break;
    case 2:
        doSomethingElse();
        break;
    default:
        doSomeOtherThing();
        break;
}

There, we want to call doSomething if foo is 0 or 1. Most tools won't flag this up as a possible coding error, because there's no code in the case 0 prior to the case 1 and this is a fairly common pattern.

Solution 2

I wrote these as comments but then it's not visible. I'm turning them into an answer. This is actually an extension to T.J.Crowder's answer.

You can find the related rule that causes Findbugs to report an error here.

You can prevent Findbugs to report this kind of errors by creating an xml file with the following content, say filter.xml and running the tool with -exclude filter.xml option. See filters on Findbugs.

<FindBugsFilter>
  <Match>
    <Bug category="PERFORMANCE" />
  </Match>
</FindBugsFilter>

Solution 3

Switch fall-throughs fall under the Findbugs category of "dodgy code". I think it only flags the first occurrence of a fall-through in a switch statement to cut down on the number of error messages.

Solution 4

Without the break they will fall though into each other so if x == 0 you'll go through all the code in every case statement block. Findbugs could either be wrong about the error, or it could be an error condition without the break, i.e. something in case 0 causes something in case 1 to break.

Without the exact code and error I can't really help further. Is the lack of breaks deliberate?

Solution 5

Usually bug analysis tools don't like fallthrough in code because in most cases, the user has just forgotten to write the break.

I don't know if there is a way to specifically disable the warning with FindBugs but Checkstyle tool recognize special comments such as /* fallthrough */ to assume that the user really wants the following code to be executed. Putting this kind of comment also improves readability. http://checkstyle.sourceforge.net/config_coding.html#FallThrough

Java code convention also mentions the use of fallthrough comment. http://www.oracle.com/technetwork/java/javase/documentation/codeconventions-142311.html

View more solutions

Comments

[T.J. Crowder]

On Stack Overflow, it's actually perfectly acceptable to edit someone else's answer to add useful information (such as the above). Go for it!

[melihcelik]

@T.J.Crowder unfortunately my edits were rejected, I had to undelete my answer. Thanks for the suggestion anyway.

[kcpr]

Is break really necessary at the end?

[T.J. Crowder]

@kcpr: Strictly speaking, you don't need a break; on the last case of a switch (default in the above). But I prefer consistency. It's purely a matter of style, the same bytecode is produced either way. (This isn't, of course, true if you leave off any other break;.)

[Kevin J. Chase]

I suspect the actual rule was the more general SF_SWITCH_FALLTHROUGH. SF_DEAD_STORE_DUE_TO_SWITCH_FALLTHROUGH means, "not only was there a fall-through, but it clobbered something you had tried to save in a previous case". That's much more likely to be a bug, in my opinion, so I would be much more hesitant to disable the DEAD_STORE rule than the FALLTHROUGH rule.

[Tasgall]

This is a very good description of the common issue and how switch works, but doesn't really answer the OP's question - though, strictly speaking, the OP didn't actually present a question...

[T.J. Crowder]

@Tasgall: How does it not answer the question? (Such as it is. :-) )

[Tasgall]

T.J.Crowder: The question - well, he didn't technically ask a question, but his situation and wording matches the question I was searching for at least :P - was how to make the findbugs tool not complain on an intentional fall through case. Your writeup on switch and why findbugs would flag it is very good and thorough, but doesn't show how to silence the warning if it was intentional. @L2M's answer below (adding "// fallthrough" or some variant where the "break;" would otherwise be) fixed it for me. Could you append that solution to your answer?

[T.J. Crowder]

@Tasgall: L2M's answer is interesting, I always include a "fallthrough" comment as a matter of style. Are you saying having the comment there makes FindBugs not report the problem?

[Emanuel Graf]

Why is this logical for java? Executing a case which statement is false just because a previous case was true?

[T.J. Crowder]

@EmanuelGraf: It's just how switch statements traditionally work. Java inherited this from C. And note that if you have two case statements without a break, you're explicitly saying "Do this for either of the above," so it's not that you're executing code for a case that's false. It's that you're executing code for one of the two (or more) cases that you want it for. (C# modifies switch to only allow stacked case labels if there's no code between them.)

[somejhereandthere]

Just to add that you can avoid the break with other statements like return doSomething(), but it supposes you are inside a function body and you don't want to proceed with the code over the switch-case statement. So actually a specific context.