codesetconversion.java

来自「JAVA 所有包」· Java 代码 · 共 690 行 · 第 1/2 页

/*
 * @(#)CodeSetConversion.java	1.20 05/11/17
 *
 * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
 * SUN PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
 */
package com.sun.corba.se.impl.encoding;

import java.util.Map;
import java.util.HashMap;
import java.nio.ByteBuffer;
import java.nio.CharBuffer;
import java.nio.charset.Charset;
import java.nio.charset.CharsetEncoder;
import java.nio.charset.CharsetDecoder;
import java.nio.charset.CharacterCodingException;
import java.nio.charset.IllegalCharsetNameException;
import java.nio.charset.MalformedInputException;
import java.nio.charset.UnsupportedCharsetException;
import java.nio.charset.UnmappableCharacterException;
import com.sun.corba.se.impl.logging.ORBUtilSystemException;
import com.sun.corba.se.impl.logging.OMGSystemException;
import com.sun.corba.se.spi.logging.CORBALogDomains;

/**
 * Collection of classes, interfaces, and factory methods for
 * CORBA code set conversion.
 *
 * This is mainly used to shield other code from the sun.io
 * converters which might change, as well as provide some basic
 * translation from conversion to CORBA error exceptions.  Some
 * extra work is required here to facilitate the way CORBA
 * says it uses UTF-16 as of the 00-11-03 spec.
 *
 * REVISIT - Since the nio.Charset and nio.Charset.Encoder/Decoder
 *           use NIO ByteBuffer and NIO CharBuffer, the interaction
 *           and interface between this class and the CDR streams
 *           should be looked at more closely for optimizations to
 *           avoid unnecessary copying of data between char[] &
 *           CharBuffer and byte[] & ByteBuffer, especially
 *           DirectByteBuffers.
 *
 */
public class CodeSetConversion
{
    /**
     * Abstraction for char to byte conversion.
     *
     * Must be used in the proper sequence:
     *
     * 1)  convert
     * 2)  Optional getNumBytes and/or getAlignment (if necessary)
     * 3)  getBytes (see warning)
     */
    public abstract static class CTBConverter
    {
        // Perform the conversion of the provided char or String,
        // allowing the caller to query for more information
        // before writing.
        public abstract void convert(char chToConvert);
        public abstract void convert(String strToConvert);

        // How many bytes resulted from the conversion?
        public abstract int getNumBytes();

        // What's the maximum number of bytes per character?
        public abstract float getMaxBytesPerChar();

        public abstract boolean isFixedWidthEncoding();

        // What byte boundary should the stream align to before
        // calling writeBytes?  For instance, a fixed width
        // encoding with 2 bytes per char in a stream which
        // doesn't encapsulate the char's bytes should align
        // on a 2 byte boundary.  (Ex:  UTF16 in GIOP1.1)
        //
        // Note: This has no effect on the converted bytes.  It
        // is just information available to the caller.
        public abstract int getAlignment();

        // Get the resulting bytes.  Warning:  You must use getNumBytes()
        // to determine the end of the data in the byte array instead
        // of array.length!  The array may be used internally, so don't
        // save references.
        public abstract byte[] getBytes();
    }
    
    /**
     * Abstraction for byte to char conversion.
     */
    public abstract static class BTCConverter
    {
        // In GIOP 1.1, interoperability can only be achieved with
        // fixed width encodings like UTF-16.  This is because wstrings
        // specified how many code points follow rather than specifying
        // the length in octets.
        public abstract boolean isFixedWidthEncoding();
        public abstract int getFixedCharWidth();

        // Called after getChars to determine the true size of the
        // converted array.
        public abstract int getNumChars();

        // Perform the conversion using length bytes from the given
        // input stream.  Warning:  You must use getNumChars() to
        // determine the correct length of the resulting array.
        // The same array may be used internally over multiple
        // calls.
        public abstract char[] getChars(byte[] bytes, int offset, int length);
    }

    /**
     * Implementation of CTBConverter which uses a nio.Charset.CharsetEncoder
     * to do the real work.  Handles translation of exceptions to the 
     * appropriate CORBA versions.
     */
    private class JavaCTBConverter extends CTBConverter
    {
	private ORBUtilSystemException wrapper = ORBUtilSystemException.get(
	    CORBALogDomains.RPC_ENCODING ) ;

	private OMGSystemException omgWrapper = OMGSystemException.get(
	    CORBALogDomains.RPC_ENCODING ) ;

        // nio.Charset.CharsetEncoder actually does the work here
        // have to use it directly rather than through String's interface
        // because we want to know when errors occur during the conversion.
        private CharsetEncoder ctb;

        // Proper alignment for this type of converter.  For instance,
        // ASCII has alignment of 1 (1 byte per char) but UTF16 has
        // alignment of 2 (2 bytes per char)
        private int alignment;

        // Char buffer to hold the input. 
        private char[] chars = null;

        // How many bytes are generated from the conversion?
        private int numBytes = 0;

        // How many characters were converted (temporary variable
        // for cross method communication)
        private int numChars = 0;

        // ByteBuffer holding the converted input.  This is necessary
        // since we have to do calculations that require the conversion
        // before writing the array to the stream.
        private ByteBuffer buffer;

        // What code set are we using?
        private OSFCodeSetRegistry.Entry codeset;

        public JavaCTBConverter(OSFCodeSetRegistry.Entry codeset,
                                int alignmentForEncoding) {

            try {
                ctb = cache.getCharToByteConverter(codeset.getName());
                if (ctb == null) {
                    Charset tmpCharset = Charset.forName(codeset.getName());
                    ctb = tmpCharset.newEncoder();
                    cache.setConverter(codeset.getName(), ctb);
                }
            } catch(IllegalCharsetNameException icne) {

                // This can only happen if one of our Entries has
                // an invalid name.
		throw wrapper.invalidCtbConverterName(icne,codeset.getName());
            } catch(UnsupportedCharsetException ucne) {

                // This can only happen if one of our Entries has
                // an unsupported name.
		throw wrapper.invalidCtbConverterName(ucne,codeset.getName());
            }

            this.codeset = codeset;
            alignment = alignmentForEncoding;
        }

        public final float getMaxBytesPerChar() {
            return ctb.maxBytesPerChar();
        }

        public void convert(char chToConvert) {
            if (chars == null)
                chars = new char[1];
            
            // The CharToByteConverter only takes a char[]
            chars[0] = chToConvert;
            numChars = 1;

            convertCharArray();
        }

        public void convert(String strToConvert) {
            // Try to save a memory allocation if possible.  Usual
            // space/time trade off.  If we could get the char[] out of
            // the String without copying, that would be great, but
            // it's forbidden since String is immutable.
            if (chars == null || chars.length < strToConvert.length())
                chars = new char[strToConvert.length()];

            numChars = strToConvert.length();
            
            strToConvert.getChars(0, numChars, chars, 0);

            convertCharArray();
        }
        
        public final int getNumBytes() {
            return numBytes;
        }
        
        public final int getAlignment() {
            return alignment;
        }

        public final boolean isFixedWidthEncoding() {
            return codeset.isFixedWidth();
        }

        public byte[] getBytes() {
            // Note that you can't use buffer.length since the buffer might
            // be larger than the actual number of converted bytes depending
            // on the encoding.
            return buffer.array();
        }

        private void convertCharArray() {
            try {
                
                // Possible optimization of directly converting into the CDR buffer.
                // However, that means the CDR code would have to reserve
                // a 4 byte string length ahead of time, and we'd need a
                // confusing partial conversion scheme for when we couldn't
                // fit everything in the buffer but needed to know the
                // converted length before proceeding due to fragmentation.
                // Then there's the issue of the chunking code.
                //
                // For right now, this is less messy and basic tests don't
                // show more than a 1 ms penalty worst case.  Less than a
                // factor of 2 increase.

                // Convert the characters
                buffer = ctb.encode(CharBuffer.wrap(chars,0,numChars));

                // ByteBuffer returned by the encoder will set its limit
                // to byte immediately after the last written byte.
                numBytes = buffer.limit();

            } catch (IllegalStateException ise) {
                // an encoding operation is already in progress
		throw wrapper.ctbConverterFailure( ise ) ;
            } catch (MalformedInputException mie) {
                // There were illegal Unicode char pairs
		throw wrapper.badUnicodePair( mie ) ;
            } catch (UnmappableCharacterException uce) {
                // A character doesn't map to the desired code set
                // CORBA formal 00-11-03.
		throw omgWrapper.charNotInCodeset( uce ) ;
            } catch (CharacterCodingException cce) {
                // If this happens, then some other encoding error occured
		throw wrapper.ctbConverterFailure( cce ) ;
            }
        }
    }

    /**
     * Special UTF16 converter which can either always write a BOM
     * or use a specified byte order without one.
     */
    private class UTF16CTBConverter extends JavaCTBConverter
    {
        // Using this constructor, we will always write a BOM
        public UTF16CTBConverter() {
            super(OSFCodeSetRegistry.UTF_16, 2);
        }

        // Using this constructor, we don't use a BOM and use the
        // byte order specified
        public UTF16CTBConverter(boolean littleEndian) {
            super(littleEndian ? 
                  OSFCodeSetRegistry.UTF_16LE : 
                  OSFCodeSetRegistry.UTF_16BE, 
                  2);
        }
    }

    /**
     * Implementation of BTCConverter which uses a sun.io.ByteToCharConverter
     * for the real work.  Handles translation of exceptions to the 
     * appropriate CORBA versions.
     */
    private class JavaBTCConverter extends BTCConverter
    {
	private ORBUtilSystemException wrapper = ORBUtilSystemException.get(
	    CORBALogDomains.RPC_ENCODING ) ;

	private OMGSystemException omgWrapper = OMGSystemException.get(
	    CORBALogDomains.RPC_ENCODING ) ;

        protected CharsetDecoder btc;
        private char[] buffer;
        private int resultingNumChars;
        private OSFCodeSetRegistry.Entry codeset;

        public JavaBTCConverter(OSFCodeSetRegistry.Entry codeset) {
            
            // Obtain a Decoder
            btc = this.getConverter(codeset.getName());

            this.codeset = codeset;
        }

        public final boolean isFixedWidthEncoding() {
            return codeset.isFixedWidth();
        }

        // Should only be called if isFixedWidthEncoding is true
        // IMPORTANT: This calls OSFCodeSetRegistry.Entry, not
        //            CharsetDecoder.maxCharsPerByte().
        public final int getFixedCharWidth() {
            return codeset.getMaxBytesPerChar();
        }

        public final int getNumChars() {
            return resultingNumChars;
        }

        public char[] getChars(byte[] bytes, int offset, int numBytes) {

            // Possible optimization of reading directly from the CDR
            // byte buffer.  The sun.io converter supposedly can handle
            // incremental conversions in which a char is broken across
            // two convert calls.
            //
            // Basic tests didn't show more than a 1 ms increase
            // worst case.  It's less than a factor of 2 increase.
            // Also makes the interface more difficult.


            try {

                ByteBuffer byteBuf = ByteBuffer.wrap(bytes, offset, numBytes);
                CharBuffer charBuf = btc.decode(byteBuf);