Java short circuit evaluation

If perfectAgent is genuinely null, that code won't throw an exception (at least assuming there aren't weird threading things going on, changing it from non-null to null half way through the expression). I would be utterly shocked if you could produce a short but complete program demonstrating it doing so.

So yes, your intuition is right - this shouldn't be a problem. Look elsewhere for the cause. I strongly suspect that perfectAgent isn't actually null, and that you're running into any of the other situations in that code which could cause an exception.

I suggest you try to extract that bit of code out into a short but complete example - if you can do so, I'll eat my metaphorical hat; if not, you'll hopefully find the problem while you attempt the extraction.

What makes you think that perfectAgent really is null? Try inserting this code before it:

if (perfectAgent == null)
{
    System.out.println("Yup, it's null");
}

Another very, very slim possibility is that you've run into a JIT bug - but I highly doubt it.

Java does have short circuit evaluation. Perhaps entry is null and so entry.getKey() is causing the NullPointerException. Another possibility is that getAddress() either returns null or has a NullPointerException happening inside somewhere (if it's more complicated than a simple return statement).

EDIT: I see your edit where you claim this:

More to the point, it is impossible to execute perfectAgent.getAddress() ...

But what if perfectAgent.getAddress() is successfully executed and returns null? See what I mean...

You ensure that perfectAgent is not null, so one or more of perfectAgent.getAddress() or entry or entry.getKey() must be null. Or getAddress() or getKey() are hitting an NPE in their implementation.

To debug this sort of thing, look first at the stack trace to pin down the location. This would tell you if it's happening in getAddress() or in getKey() or in the pasted code snippet that calls them. Next, if it's in this snippet, add some code before the if to test which is null. You can use good old System.err.println() or assertions. (If you use assertions, be sure to enable them with the java command's -enableassertions flag.)

Update: So my interpretation turned out to be wrong ... the problem presented two contradictory facts (there was an NPE on this line and yet the short-circuit should have happened) and I automatically assumed the first fact was true and the second false when in fact it was a different problem entirely due to turning off the auto-build in Eclipse. Duh! In debugging something "impossible" it helps to be radically skeptical.

There are three references other than perfectAgent that could be null:

• perfectAgent.getAddress()
• entry
• entry.getKey()

Break up the statement or run it in a debugger.

Big mistery. I copied your line of code and tested with perfectAgent == null, entry == null, entry.getKey() == null and combinations of those: No NPE in my test bed (Java 1.6).

Whatever annoying bug it is, I doubt that it has something to do with short circuit evaluation. If it's this line causing NPE, than, as far as I can say, perfectAgent is not null. Good luck and - show us the bug once you've catched it :)

Try formatting your code like this:

if( 
  (perfectAgent != null) 
  && (
      perfectAgent.getAddress()
      .equals(
       entry.getKey()
      )
     ) 
  ) {

It should give you a better stack trace line entry.