codesetconversion.java

来自「java jdk 1.4的源码」· Java 代码 · 共 675 行 · 第 1/2 页

                return buffer;

            } catch (MalformedInputException mie) {
                // There were illegal Unicode char pairs
                throw new DATA_CONVERSION(mie.getMessage(),
                                          MinorCodes.BAD_UNICODE_PAIR,
                                          CompletionStatus.COMPLETED_NO);
            } catch (UnknownCharacterException uce) {
                // A character doesn't map to the desired code set.
                // CORBA formal 00-11-03.
                throw new DATA_CONVERSION(uce.getMessage(),
                                          MinorCodes.CHAR_NOT_IN_CODESET,
                                          CompletionStatus.COMPLETED_NO);
            } catch (ConversionBufferFullException cbfe) {
                // If this happens, then the CharToByteConverter was lying
                // about the maximum chars per byte.
                throw new INTERNAL(cbfe.getMessage(),
                                   MinorCodes.BTC_CONVERTER_FAILURE,
                                   CompletionStatus.COMPLETED_NO);
            }
        }

        /**
         * Utility method to find a ByteToCharConverter in the
         * cache or create a new one if necessary.  Throws an
         * INTERNAL if the code set is unknown.
         */
        protected ByteToCharConverter getConverter(String javaCodeSetName) {
         
            try {
                ByteToCharConverter result
                    = cache.getByteToCharConverter(javaCodeSetName);

                if (result == null) {
                    result = ByteToCharConverter.getConverter(javaCodeSetName);
                    cache.setConverter(javaCodeSetName, result);

                    // When substitution mode is false, the converter
                    // will throw exceptions when it encounters illegal
                    // encodings.
                    result.setSubstitutionMode(false);
                }

                return result;

            } catch(UnsupportedEncodingException uee) {
                // This can only happen if one of our entries has
                // an invalid name.
                throw new INTERNAL("Invalid converter name: " + javaCodeSetName);
            }
        }
    }

    /**
     * Special converter for UTF16 since it's required to optionally
     * support a byte order marker while the internal Java converters
     * either require it or require that it isn't there.
     *
     * The solution is to check for the byte order marker, and if we
     * need to do something differently, switch internal converters.
     */
    private class UTF16BTCConverter extends JavaBTCConverter
    {
        private boolean defaultToLittleEndian;
        private boolean converterUsesBOM = true;

        private static final char UTF16_BE_MARKER = (char) 0xfeff;
        private static final char UTF16_LE_MARKER = (char) 0xfffe;

        // When there isn't a byte order marker, used the byte
        // order specified.
        public UTF16BTCConverter(boolean defaultToLittleEndian) {
            super(OSFCodeSetRegistry.UTF_16);

            this.defaultToLittleEndian = defaultToLittleEndian;
        }

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

            if (hasUTF16ByteOrderMarker(bytes, offset, numBytes)) {
                if (!converterUsesBOM)
                    switchToConverter(OSFCodeSetRegistry.UTF_16);

                converterUsesBOM = true;

                return super.getChars(bytes, offset, numBytes);
            } else {
                if (converterUsesBOM) {
                    if (defaultToLittleEndian)
                        switchToConverter(OSFCodeSetRegistry.UTF_16LE);
                    else
                        switchToConverter(OSFCodeSetRegistry.UTF_16BE);

                    converterUsesBOM = false;
                }

                return super.getChars(bytes, offset, numBytes);
            }
        }

        /**
         * Utility method for determining if a UTF-16 byte order marker is present.
         */
        private boolean hasUTF16ByteOrderMarker(byte[] array, int offset, int length) {
            // If there aren't enough bytes to represent the marker and data,
            // return false.
            if (length >= 4) {

                int b1 = array[offset] & 0x00FF;
                int b2 = array[offset + 1] & 0x00FF;

                char marker = (char)((b1 << 8) | (b2 << 0));
                
                return (marker == UTF16_BE_MARKER || marker == UTF16_LE_MARKER);
            } else
                return false;
        }

        /**
         * The current solution for dealing with UTF-16 in CORBA
         * is that if our sun.io converter requires byte order markers,
         * and then we see a CORBA wstring/wchar without them, we 
         * switch to the sun.io converter that doesn't require them.
         */
        private void switchToConverter(OSFCodeSetRegistry.Entry newCodeSet) {

            // Use the getConverter method from our superclass.
            btc = getConverter(newCodeSet.getName());
        }
    }

    /**
     * CTB converter factory for single byte or variable length encodings.
     */
    public CTBConverter getCTBConverter(OSFCodeSetRegistry.Entry codeset) {
        int alignment = (!codeset.isFixedWidth() ?
                         1 :
                         codeset.getMaxBytesPerChar());
            
        return new JavaCTBConverter(codeset, alignment);
    }

    /**
     * CTB converter factory for multibyte (mainly fixed) encodings.
     *
     * Because of the awkwardness with byte order markers and the possibility of 
     * using UCS-2, you must specify both the endianness of the stream as well as 
     * whether or not to use byte order markers if applicable.  UCS-2 has no byte 
     * order markers.  UTF-16 has optional markers.
     *
     * If you select useByteOrderMarkers, there is no guarantee that the encoding
     * will use the endianness specified.
     *
     */
    public CTBConverter getCTBConverter(OSFCodeSetRegistry.Entry codeset,
                                        boolean littleEndian,
                                        boolean useByteOrderMarkers) {

        // UCS2 doesn't have byte order markers, and we're encoding it
        // as UTF-16 since UCS2 isn't available in all Java platforms.
        // They should be identical with only minor differences in
        // negative cases.
        if (codeset == OSFCodeSetRegistry.UCS_2)
            return new UTF16CTBConverter(littleEndian);

        // We can write UTF-16 with or without a byte order marker.
        if (codeset == OSFCodeSetRegistry.UTF_16) {
            if (useByteOrderMarkers)
                return new UTF16CTBConverter();
            else
                return new UTF16CTBConverter(littleEndian);
        }

        // Everything else uses the generic JavaCTBConverter.
        //
        // Variable width encodings are aligned on 1 byte boundaries.
        // A fixed width encoding with a max. of 4 bytes/char should
        // align on a 4 byte boundary.  Note that UTF-16 is a special
        // case because of the optional byte order marker, so it's
        // handled above.
        //
        // This doesn't matter for GIOP 1.2 wchars and wstrings
        // since the encoded bytes are treated as an encapsulation.
        int alignment = (!codeset.isFixedWidth() ?
                         1 :
                         codeset.getMaxBytesPerChar());
        
        return new JavaCTBConverter(codeset, alignment);
    }

    /**
     * BTCConverter factory for single byte or variable width encodings.
     */
    public BTCConverter getBTCConverter(OSFCodeSetRegistry.Entry codeset) {
        return new JavaBTCConverter(codeset);
    }

    /**
     * BTCConverter factory for fixed width multibyte encodings.
     */
    public BTCConverter getBTCConverter(OSFCodeSetRegistry.Entry codeset,
                                        boolean defaultToLittleEndian) {

        if (codeset == OSFCodeSetRegistry.UTF_16 ||
            codeset == OSFCodeSetRegistry.UCS_2) {

            return new UTF16BTCConverter(defaultToLittleEndian);
        } else {
            return new JavaBTCConverter(codeset);
        }
    }

    /** 
     * Follows the code set negotiation algorithm in CORBA formal 99-10-07 13.7.2.
     *
     * Returns the proper negotiated OSF character encoding number or
     * CodeSetConversion.FALLBACK_CODESET.
     */
    private int selectEncoding(CodeSetComponentInfo.CodeSetComponent client,
                               CodeSetComponentInfo.CodeSetComponent server) {

        // A "null" value for the server's nativeCodeSet means that
        // the server desired not to indicate one.  We'll take that
        // to mean that it wants the first thing in its conversion list.
        // If it's conversion list is empty, too, then use the fallback
        // codeset.
        int serverNative = server.nativeCodeSet;

        if (serverNative == 0) {
            if (server.conversionCodeSets.length > 0)
                serverNative = server.conversionCodeSets[0];
            else
                return CodeSetConversion.FALLBACK_CODESET;
        }

        if (client.nativeCodeSet == serverNative) {
            // Best case -- client and server don't have to convert
            return serverNative;
        }

        // Is this client capable of converting to the server's
        // native code set?
        for (int i = 0; i < client.conversionCodeSets.length; i++) {
            if (serverNative == client.conversionCodeSets[i]) {
                // The client will convert to the server's
                // native code set.
                return serverNative;
            }
        }

        // Is the server capable of converting to the client's
        // native code set?
        for (int i = 0; i < server.conversionCodeSets.length; i++) {
            if (client.nativeCodeSet == server.conversionCodeSets[i]) {
                // The server will convert to the client's
                // native code set.
                return client.nativeCodeSet;
            }
        }

        // See if there are any code sets that both the server and client
        // support (giving preference to the server).  The order
        // of conversion sets is from most to least desired.
        for (int i = 0; i < server.conversionCodeSets.length; i++) {
            for (int y = 0; y < client.conversionCodeSets.length; y++) {
                if (server.conversionCodeSets[i] == client.conversionCodeSets[y]) {
                    return server.conversionCodeSets[i];
                }
            }
        }

        // Before using the fallback codesets, the spec calls for a
        // compatibility check on the native code sets.  It doesn't make
        // sense because loss free communication is always possible with
        // UTF8 and UTF16, the fall back code sets.  It's also a lot
        // of work to implement.  In the case of incompatibility, the
        // spec says to throw a CODESET_INCOMPATIBLE exception.
        
        // Use the fallback
        return CodeSetConversion.FALLBACK_CODESET;
    }

    /**
     * Perform the code set negotiation algorithm and come up with
     * the two encodings to use.
     */
    public CodeSetComponentInfo.CodeSetContext negotiate(CodeSetComponentInfo client,
                                                         CodeSetComponentInfo server) {
        int charData
            = selectEncoding(client.getCharComponent(),
                             server.getCharComponent());

        if (charData == CodeSetConversion.FALLBACK_CODESET) {
            charData = OSFCodeSetRegistry.UTF_8.getNumber();
        }

        int wcharData
            = selectEncoding(client.getWCharComponent(),
                             server.getWCharComponent());

        if (wcharData == CodeSetConversion.FALLBACK_CODESET) {
            wcharData = OSFCodeSetRegistry.UTF_16.getNumber();
        }

        return new CodeSetComponentInfo.CodeSetContext(charData,
                                                       wcharData);
    }

    // No one should instantiate a CodeSetConversion but the singleton
    // instance method
    private CodeSetConversion() {}

    // initialize-on-demand holder
    private static class CodeSetConversionHolder {
	static final CodeSetConversion csc = new CodeSetConversion() ;
    }

    /**
     * CodeSetConversion is a singleton, and this is the access point.
     */
    public final static CodeSetConversion impl() {
	return CodeSetConversionHolder.csc ;
    }

    // Singleton instance
    private static CodeSetConversion implementation;

    // Number used internally to indicate the fallback code
    // set.
    private static final int FALLBACK_CODESET = 0;

    // Provides a thread local cache for the sun.io
    // converters.
    private CodeSetCache cache = new CodeSetCache();
}