Java String, StringBuilder and StringBuffer

I had a phone conversation with someone yesterday afternoon about Java's String and StringBuffer and how they work. After 24 hours had elapsed my thoughts on String still comes back to bug me - as if I missed out something.

The urge to scratch this itch finally made me fire up “vi” from my terminal. Yes I know what you are thinking, I'm no "vi" expert, but do I really need Eclipse to write 15 lines of code? :)

Here is the source:

1:  public class Str {  
2:      public void sayHelloWorld() {  
3:          String str1 = "hello " + "world ";  
4:          String str2 = "hello world again";  
5:      }  
6:      public void sayGoodbyeWorld() {  
7:          String str3 = "goodbye world " + 3;  
8:      }  
9:      public void sayYouSayMe() {  
10:          String str4 = "";  
11:          for (int i = 0; i < 3; i++) {  
12:              str4 += "say you say me ";  
13:          }  
14:      }  
15:  }  

Compiling the source with JDK 6 followed by
“javap -c Str" gives us:

1:  Compiled from "Str.java"  
2:  public class Str extends java.lang.Object{  
3:  public Str();  
4:   Code:  
5:    0:    aload_0  
6:    1:    invokespecial    #1; //Method java/lang/Object."<init>":()V  
7:    4:    return  
8:  public void sayHelloWorld();  
9:   Code:  
10:    0:    ldc    #2; //String hello world   
11:    2:    astore_1  
12:    3:    ldc    #3; //String hello world again  
13:    5:    astore_2  
14:    6:    return  
15:  public void sayGoodbyeWorld();  
16:   Code:  
17:    0:    ldc    #4; //String goodbye world 3  
18:    2:    astore_1  
19:    3:    return  
20:  public void sayYouSayMe();  
21:   Code:  
22:    0:    ldc    #5; //String   
23:    2:    astore_1  
24:    3:    iconst_0  
25:    4:    istore_2  
26:    5:    iload_2  
27:    6:    iconst_3  
28:    7:    if_icmpge    36  
29:    10:    new    #6; //class java/lang/StringBuilder  
30:    13:    dup  
31:    14:    invokespecial    #7; //Method java/lang/StringBuilder."<init>":()V  
32:    17:    aload_1  
33:    18:    invokevirtual    #8; //Method java/lang/StringBuilder.append:(Ljava/lang/String;)Ljava/lang/StringBuilder;  
34:    21:    ldc    #9; //String say you say me   
35:    23:    invokevirtual    #8; //Method java/lang/StringBuilder.append:(Ljava/lang/String;)Ljava/lang/StringBuilder;  
36:    26:    invokevirtual    #10; //Method java/lang/StringBuilder.toString:()Ljava/lang/String;  
37:    29:    astore_1  
38:    30:    iinc    2, 1  
39:    33:    goto    5  
40:    36:    return  
41:  }  

The disassembled JDK 6 code from above reveals quite a bit about this simple class.

• Line 2: the compiler automatically extends our class from java.lang.Object.

• Line 3-7: the compiler automatically inserts a default Str constructor that invokes java.lang.Object constructor

• Line 10: ldc #2 suggests that “hello world” string is stored as the 2^nd item from constant pool table, where “hello " and "world” is automagically concatenated for you by the compiler

• Line 12: ldc #3 suggests that “hello world again” string is stored as the 3^rd item from constant pool table

• Line 17: ldc #4 suggests that “goodbye world 3” is stored as the 4^th item from constant pool table, where the compiler again automagically concatenated integer “3” to "goodbye world "
  

• Line 22: ldc #5 suggests that our empty string is stored as the 5^th item from constant pool table

• Line 29-36: tells us that javac inserts StringBuilder to concatenate "say you say me " strings with "say you say me " stored as the 9^th item from constant pool table
  

• Line 36: StringBuilder object invokes the toString method and assigns it back to “str4”

Let's try compiling the same source with JDK 1.4 by running “javac -source 1.4 -target 1.4 Str.java” then followed by “javap -c Str” which gives us:

1:  Compiled from "Str.java"  
2:  public class Str extends java.lang.Object{  
3:  public Str();  
4:   Code:  
5:    0:    aload_0  
6:    1:    invokespecial    #1; //Method java/lang/Object."<init>":()V  
7:    4:    return  
8:  public void sayHelloWorld();  
9:   Code:  
10:    0:    ldc    #2; //String hello world   
11:    2:    astore_1  
12:    3:    ldc    #3; //String hello world again  
13:    5:    astore_2  
14:    6:    return  
15:  public void sayGoodbyeWorld();  
16:   Code:  
17:    0:    ldc    #4; //String goodbye world 3  
18:    2:    astore_1  
19:    3:    return  
20:  public void sayYouSayMe();  
21:   Code:  
22:    0:    ldc    #5; //String   
23:    2:    astore_1  
24:    3:    iconst_0  
25:    4:    istore_2  
26:    5:    iload_2  
27:    6:    iconst_3  
28:    7:    if_icmpge    36  
29:    10:    new    #6; //class java/lang/StringBuffer  
30:    13:    dup  
31:    14:    invokespecial    #7; //Method java/lang/StringBuffer."<init>":()V  
32:    17:    aload_1  
33:    18:    invokevirtual    #8; //Method java/lang/StringBuffer.append:(Ljava/lang/String;)Ljava/lang/StringBuffer;  
34:    21:    ldc    #9; //String say you say me   
35:    23:    invokevirtual    #8; //Method java/lang/StringBuffer.append:(Ljava/lang/String;)Ljava/lang/StringBuffer;  
36:    26:    invokevirtual    #10; //Method java/lang/StringBuffer.toString:()Ljava/lang/String;  
37:    29:    astore_1  
38:    30:    iinc    2, 1  
39:    33:    goto    5  
40:    36:    return  
41:  }  

The disassembled code for 1.4 and 1.6 looks the same except that 1.4 uses StringBuffer instead of StringBuilder as StringBuilder was introduced as part of JDK 5 and later. So it looks like a little bit of String optimisation is done on the Javac level not on the JVM level.