问题
how expensive is 'new'? I mean, should I aim at reusing the same object or if the object is 'out of scope' it's the same as emptying it?
example, say a method creates a list:
List<Integer> list = new ArrayList<Integer>();
at the end of the method the list is no longer in use - does it mean that there's no memory allocated to it anymore or does it mean that there's a null pointer to it (since it was 'created').
Alternately, I can send a 'list' to the method and empty it at the end of the method with: list.removeAll(list); will that make any difference from memory point of view?
Thanks!
回答1:
its an array list, so creating a new object means allocating a slab of memory and zeroing it, plus any bookkeeping overhead. Clearing the list means zeroing the memory. This view would lead you to believe that clearing an existing object is faster. But, it's likely that the JVM is optimized to make memory allocations fast, so probably none of this matters. So just write clear, readable code, and don't worry about it. This is java after all, not c.
回答2:
at the end of the method the list is no longer in use - does it mean that there's no memory allocated to it anymore or does it mean that there's a null pointer to it (since it was 'created').
Means there are no references to it and object is eligible for GC.
Alternately, I can send a 'list' to the method and empty it at the end of the method with: list.removeAll(list); will that make any difference from memory point of view?
It's tradeoff between time/space. Removing elements from list is time consuming, even though you don't need to create new objects.
With the latest JVMs GC collection capabilities, it is ok to create new object WHEN REQUIRED (but avoiding object creation in loop is best). Longer references to an object sometimes make that object NOT eligible for GC and may cause memory leak if not handled properly.
回答3:
I don't know much about memory footprints in java, but I think emptying a List to reuse it, is not such a good idea because of the performance impact of emptying the List. And I think it is also in an OO perspective not a good idea, because you should have one object with just one purpose.
At the end of a method the object is indeed out of scope. But that doesn't mean it is garbage collected or even eligible for garbage collection, because others might still reference that List. So basically: if there are no objects references to that List then it might be elegible for garbage collection, but if it will be garbage collected it still unsure, if the List is still stored in the Young Generation space it can either be in the Eden space or Tenured space. Eden space is where objects are first allocated, when garbage collection happens and the object is still alive it will be moved to survivor space. If it still survives past that it will move on to the Tenured space, where I believe not much garbage collection happens. But all this depends how long an object lives, who refers to this object and where it is allocated
回答4:
how expensive is 'new'?
It definitely incurs some overhead. But it depends on how complex the object is. If you are creating an object with just few primitives, not that expensive. But if you are creating objects inside objects, may be collections of objects, if your constructor is reading some properties file to initialize object's member variables, EXPENSIVE!
But to be frank, if we need to create a new object, we have create it, there is no alternative. And if we don't need to and if we are still creating that is kind of bad programming.
at the end of the method the list is no longer in use - does it mean that there's no memory allocated to it anymore or does it mean that there's a null pointer to it (since it was 'created').
Once the object does not have any reference to it, it becomes out of scope, and it becomes eligible for garbage collection. Hence even if it has some memory allocated, it will be reclaimed by the GC at some later point, whenever it runs, we need not worry about it. (And we cannot guarantee when will GC run).
Emptying the collection at the end, I don't think will make things any better, because the same thing will happen to all the individual objects in the collection, as what happens to the collection itself. They will become eligible for GC.
回答5:
For small lists, it is probably a bit cheaper cheaper to clear() the list.
For the asymptotic case of really large lists in a really large heap, it boils down to whether the GC can zero a large chunk of memory faster than the for loop in clear() can. And I think it probably can.
However, my advice would be to ignore this unless you have convincing evidence (from profiling) that you have a high turn-over of ArrayList objects. (It is a bad idea to optimize based solely on your intuition.)
回答6:
It depends on how costly the object is, both in terms of initialization required and how large it's memory footprint is. It also depends heavily on the kind of application (what else does the application spend time on).
For your example with the ArrayList, its already very hard to give a definite answer - depending on how many entries there are in the list, clear() can be very expensive or very cheap, while a new ArrayList has almost constant cost.
The general rule of thumb is: Don't bother with reusing objects until you have measured that you have a performance problem, and then be very sure that creating the objects is the cause of that problem. Most likely there are more rewarding optimization opportunities in your application. A profiler will help identify the places where you spend the most time. Focus on those and better algoryhtms.
来源：https://stackoverflow.com/questions/9314176/whats-more-efficient-to-empty-an-object-or-create-a-new-one