whats faster - test for identity vs. (cast + booleanValue())

[Tom Hughes]

anyone know what's faster ?
MyArray is an array of Object and the 4th element is always a Boolean.

  
if (MyArray[0] == Boolean.TRUE) //do summit

or

  
if (((Boolean) MyArray[0]).booleanValue()) //do summit

I'm writing a parser to performance is ery important.
cheers,
Tom

[Zakaria Haque]

In the following code you posted, identity check is not what you want, as a matter of fact, as long as you are creating the Boolean instance with a new operator, identity check with Boolean.TRUE will always return false. Rather than using,
if (MyArray[0] == Boolean.TRUE) //do summit
you should use
if (MyArray[0].equals(Boolean.TRUE)) //do summit
Among the following two expresions
MyArray[0].equals(Boolean.TRUE)
((Boolean) MyArray[0]).booleanValue()
It does not make much sense to think about speed, since the performance gain is almost absolutely insginificant.

[Jim Yingst]

If you're using SDK 1.4, then there's no reason to ever use new Boolean(boolean) anyway - use Boolean.valueOf(boolean) instead. This allows you to omit the pointless creation of many identical objects - and also allows you to get away with using == rather than equals(), if you like. Identity check with == will in fact be slightly faster than value check with equals(). But still, it's unlikely to be noticeable - and is prone to failure if you or another programmer accidentally use new Boolean() rather than Boolean.valueOf(). More significant is the elimination of unnecessary object creation, as already mentioned.

[Tom Hughes]

Thanks for your replies.
In regards to 'creating' my objects, I think what I was doing originally is faster than both of your suggestions. A typical line where I create my object array would be like this.

keyStack.push(new Object [] {qname, Boolean.FALSE, Boolean.TRUE, attributes});  

In this way I can check for identity and do not actually instatiate any objects. I figure Boolean.valueOf(boolean) does pretty much this, but seeing as in my case I always know at compile time what I am going to add to the Object array, it is quicker to do it my way.
I know the peformance difference is going to be minimal but in this case I am writing a parser, so for parsing large messages this code is going to be executed a huge amount of times, so I figured if there was a quicker way to do it, I may as well do it as the difference it makes to me coding the thing is negligible.
cheers,
Tom

[Tom Hughes]

Does anyone know the answer to this query ?
cheers,
Tom

[Jim Yingst]

You're right, using the literal values is faster than creating new ones. The valueOf() method also uses the literal values internally - if you're getting the value of a boolean variable (i.e. you don't know in advance whether it's true or false) then it makes perfect sense. But if, at the point you need a Boolean, you know exactly which one you want, then there's no reason to use Boolean.valueOf(true) instead of Boolean.TRUE.
As for your query - I answered it, didn't I?

Identity check with == will in fact be slightly faster than value check with equals().

Anyway, it should be easy to test this sort of thing yourself. If you can't detect a difference by testing, then it really doesn't matter.

[Tom Hughes]

my apologies, your right - THANKS A LOT !!
Tom

[Mr. C Lamont Gilbert]

Originally posted by Jim Yingst:
Identity check with == will in fact be slightly faster than value check with equals().

Why? a value check with equals() is nothing more than an identity check on the objects pointer.

[Jim Yingst]

No, for Boolean it's overridden so that new Boolean(true).equals(new Boolean(true)) returns true. So there can be a bit more overhead, as it's necessary to actually retrieve the Boolean object from the heap to find out what its internal boolean value is.