String vs. StringBuffer

[iago]

kp once said this, and I decided to see how much difference there actually is between a String and StringBuffer.  My test code is this:

Code Select Expand

       public static void main(String args[])
       {
               String b = new String();
               StringBuffer c = new StringBuffer(50001);

               Timer a = new Timer();
               for(int i = 0; i < 50000; i++)
               {
                       b = b + "a";
               }
               System.out.println(a.toString());

               b = null;

               a.reset();
               for(int i = 0; i < 50000; i++)
               {
                       c.append(a);
               }
               b = c.toString();
               System.out.println(a.toString());

       }


The output is this:

Code Select Expand

It has been 28373ms (28s) since the timer was created.
It has been 977ms (0s) since the timer was created.

Bottom line: StringBuffer can make a HUGE difference.

[Banana fanna fo fanna]

Strings are immutable, thats why.

[iago]

Strings are immutable, thats why.

I know why, but I wanted to see how much of a difference it made, just for the sake of curiousity.

[Grok]

I don't know Java, but from looking at your code it appears with b you forced it to grow char-by-char, while with c you preallocated its memory.  So it would seem natural to expect working with b to take longer.  Did I misinterpret?

[Yoni]

To add to Grok's comment, do Java Strings have a "reserve" function similar to C++ std::string::reserve, so you can pre-allocate 50001 bytes? I think the comparison would be more interesting this way, since then both benchmarks do the same thing in different ways, instead of benchmark #1 having to do more work as it is now.

[iago]

I forgot to mention this, but I also tried this without allocating a StringBuffer of a constant size, like this:
StringBuffer s = new StringBuffer();

And I tried doing this:
t = new Timer();
StringBuffer s = new StringBuffer(50001);

In both cases, it was ~150ms more.

The reason for the speed difference is what Storm said - when you use a String, it's making a new string in every loop rather than resuing the old one.  This takes up considerably more resources.

Kp once posted a good example somewhere, but I couldn't find it

[Grok]

And I tried doing this:
t = new Timer();
StringBuffer s = new StringBuffer(50001);

In both cases, it was ~150ms more.

150ms more is a huge difference from

It has been 28373ms (28s) since the timer was created.
It has been 977ms (0s) since the timer was created.

[iago]

And I tried doing this:
t = new Timer();
StringBuffer s = new StringBuffer(50001);

In both cases, it was ~150ms more.

150ms more is a huge difference from

It has been 28373ms (28s) since the timer was created.
It has been 977ms (0s) since the timer was created.

I can't copy/paste, since I'm running this test on a different computer, but here are my results without pre-allocating:
1119ms
1319ms
1077ms

And with pre-allocating within the timer block:
1093ms
1173ms
1088ms

And pre-allocating without the timer block:
1075ms
1088ms
1078ms

In any case, it's still considerably faster to use StringBuffer.  The slower results today probably indicate the server is busy with something.

[Tuberload]

StringBuffer versus String
What is the performance impact of the StringBuffer and String classes?
http://www.javaworld.com/javaworld/jw-03-2000/jw-0324-javaperf_p.html