Name: jl125535 Date: 02/25/2002


FULL PRODUCT VERSION :
java version "1.4.0-rc"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.0-rc-b91)
Java HotSpot(TM) Client VM (build 1.4.0-rc-b91, mixed mode)

A DESCRIPTION OF THE PROBLEM :
There are many places where code like this

   String x = someString + anotherString;

occurs in the JDK class source.

As the compiler converts the expression to use a
StringBuffer and StringBuffer initially has an internal
buffer length of 16, if the resulting String is over 16
characters long aditional new char[] memory allocations and
garbage collection is required within StringBuffer.

   StringBuffer xx = new StringBuffer(someString.length() +
anotherString.length();
   xx.append(someString).append(anotherString);
   String x = xx.toString();


A 15-20% improvement in speed can be expected;

-------------------------------------------------------------------------------
Here is a list of locations that use direct String concatenation.

I've ignored direct String concatenation in Exception throw statements
and some debugging code, since this code is called rarely. For those
cases, speed is less important than the greater readability and
slight reduction in text size.


java.util.jar.Attributes.writeMain()

            out.writeBytes(vername+": "+version+"\r\n");

   is probably better as

            out.writeBytes(vername);
            out.writeBytes(": ");
            out.writeBytes(version);
            out.writeBytes("\r\n");



java.util.jar.Attributes.read()

   value should probably be a StringBuffer, but this is complicated by
   using the deprecated String constructor. Should probably use
       public String(byte bytes[], int offset, int length, String
charsetName)
   string constructor.

java.util.jar.Manifest.read(InputStream is)

   name may be better as a StringBuffer, but this is complicated by
   using the deprecated String constructor. Should probably use
       public String(byte bytes[], int offset, int length, String
charsetName)
   string constructor.


java.util.logging.ConsoleHandler.configure()

   4 occurances of cname +".level", or similar; as this
   is a one off configuration then it is probably not
   worth converting.

java.util.logging.ErrorManager.error()

        String text = "java.util.logging.ErrorManager: " + code;
        if (msg != null) {
            text = text + ": " + msg;
        }
        System.err.println(text);

   should be

        StringBuffer text = new
StringBuffer(1024).append("java.util.logging.ErrorManager:
").append(code);
        if (msg != null) {
            text.append(": ").append(msg);
        }
        System.err.println(text.toString());

   a size of 1024 should take care of all but the very longest msg
Strings.


java.util.logging.FileHandler.configure()

   8 occurances of cname +".level", or similar; as this
   is a one off configuration then it is probably not
   worth converting.

java.util.logging.FileHandler.generate()

   The String word is appended in multiple places, converting
   it to a StringBuffer would probably save space and would be
   quicker.

java.util.logging.LogManager.addLogger()

   A couple of occurances of appending ".level" to a String
   probably not worth converting.

java.util.logging.LogManager.readConfiguration()

                System.err.println("" + ex);

   should this not be

                System.err.println(ex.toString());


java.util.logging.LogManager.AccessController.doPrivileged()

   One occurance of String "+", but its probably not worth
   converting to use StringBuffer.


                System.err.println("" + ex);

   should this not be

                System.err.println(ex.toString());

            
java.util.logging.MemoryHandler.configure()

   6 occurances of cname +".level", or similar; as this
   is a one off configuration then it is probably not
   worth converting.


java.util.logging.MemoryHandler.MemoryHandler()

                System.err.println("" + ex);

   should this not be

                System.err.println(ex.toString());


java.util.logging.SocketHandler.configure()

   6 occurances of cname +".level", or similar; as this
   is a one off configuration then it is probably not
   worth converting.



java.util.logging.StreamHandler.configure()

   6 occurances of cname +".level", or similar; as this
   is a one off configuration then it is probably not
   worth converting.


java.util.prefs.AbstractPreferences.AbstractPreferences()

            absolutePath = (parent==root ? "/" + name
                                         : parent.absolutePath() + "/"
+ name);

should probably be

            if (parent==root) {
               absolutePath = "/" + name;
            } else {
               String pap = parent.absolutePath();
               StringBuffer sb = new
StringBuffer(pap.length()+1+name.length());
               sb.append(pap).append("/")+append(name);
               absolutePath = sb.toString();
            }



java.util.prefs.AbstractPreferences.toString()

        return (this.isUserNode() ? "User" : "System") +
               " Preference Node: " + this.absolutePath();


   should probably be

        StringBuffer sb = new StringBuffer(1024); // should be long
enough
        this.isUserNode() ? sb.append("User") : sb.append("System");
        return sb.append(" Preference Node:
").append(this.absolutePath()).toString();



java.util.prefs.Preferences.nodeName()

        return "/" + packageName.replace('.', '/');

   should probably be

        StringBuffer sb = new StringBuffer(1+packageName.length());
        return sb.append("/").append(packageName.replace('.', '/'));


java.util.regex.Matcher.appendReplacement()

   groupString does not seem to be required.


        String s = replacement;
        StringBuffer result = null;

        // If no match, return error
        if (first < 0)
            throw new IllegalStateException("No match available");

        // Process substitution string to replace group references
with groups
        int cursor = 0;
        result = new StringBuffer();

   could be


        // If no match, return error
        if (first < 0)
            throw new IllegalStateException("No match available");

        // Process substitution string to replace group references
with groups
        int cursor = 0;
        String s = replacement;
        StringBuffer result = new StringBuffer();


java.util.regex.Matcher.appendTail()

        String tail = getSubSequence(lastAppendPosition,
getTextLength()).toString();
        sb.append(tail);

   could be:

        sb.append(getSubSequence(lastAppendPosition,
getTextLength()));



java.util.regex.Pattern.normalize()

        StringBuffer newPattern = new StringBuffer();

   could be:

        StringBuffer newPattern = new StringBuffer(patternLength*2);
// unlikely to more than double in size


   and

                newPattern.append("(?:" + ea + ")");

   could be:

                newPattern.append("(?:").append(ea).append(")");




java.util.regex.Pattern.normalizeCharClass()

            result = new String("(?:"+charClass.toString()+
                                eq.toString()+")");

   could be:

            StringBuffer sb = new
StringBuffer(4+charClass.getLength()+eq.getLength());
            sb.append("(?:").append(charClass).append(eq).append(")");
            result = sb.toString();



java.util.regex.Pattern.produceEquivalentAlternation()

        // Add combined permutations
        for(int x=0; x<perms.length; x++) {
            String next = base + perms[x];
            if (x>0)
                result.append("|"+next);
            next = composeOneStep(next);
            if (next != null)
                result.append("|"+produceEquivalentAlternation(next));
        }

could be:

        // Add combined permutations
        StringBuffer next = new StringBuffer(1024);
        for(int x=0; x<perms.length; x++) {
            next.setLength(0);
            next.append(base).append(perms[x]);
            if (x>0)
                result.append("|").append(next);
            next = composeOneStep(next);
            if (next != null)
                result.append("|").append(produceEquivalentAlternation(next)));
        }


    private StringBuffer composeOneStep(StringBuffer input) {
        String firstTwoChars = input.substring(0,2);
        String result = Normalizer.compose(firstTwoChars, false, 0);

        if (result.equals(firstTwoChars))
            return null;
        else {
            String remainder = input.substring(2);
            input.setLength(0);
            return input.append(result).append(remainder);
        }
    }

or even further reduced to:

        StringBuffer next = new StringBuffer(1024);
        for(int x=0; x<perms.length; x++) {
            next.setLength(0);
            next.append(base).append(perms[x]);
            if (x>0) {
                result.append("|").append(next);
            }
            String firstTwoChars = next.substring(0,2);
            String result = Normalizer.compose(firstTwoChars, false,
0);

            if (! result.equals(firstTwoChars) ) {
                String remainder = next.substring(2);
                next.setLength(0);
                next.append(result).append(remainder);
                result.append("|").append(produceEquivalentAlternation(next)));
            }
        }


   Nothing to do with StringBuffers, but there seems to be scope for
   factorizing out the cases of:

        boolean study(TreeInfo info) {
            info.minLength++;
            info.maxLength++;
            return next.study(info);
        }

   into a SingleCharNode extending Node and then subclasses of
   Node (SingleA, SingleU, etc.) could extend this.


(Review ID: 139191)
======================================================================