/*
 * Copyright (c) 1997, 2012, Oracle and/or its affiliates. All rights reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.  Oracle designates this
 * particular file as subject to the "Classpath" exception as provided
 * by Oracle in the LICENSE file that accompanied this code.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 */

/*
 * @(#)MimeUtility.java       1.45 03/03/10
 */


package com.sun.xml.internal.messaging.saaj.packaging.mime.internet;

import java.io.*;
import java.util.*;

import javax.activation.DataHandler;
import javax.activation.DataSource;

import com.sun.xml.internal.messaging.saaj.packaging.mime.MessagingException;
import com.sun.xml.internal.messaging.saaj.packaging.mime.util.*;
import com.sun.xml.internal.messaging.saaj.util.SAAJUtil;

/**
 * This is a utility class that provides various MIME related
 * functionality. <p>
 *
 * There are a set of methods to encode and decode MIME headers as
 * per RFC 2047. A brief description on handling such headers is
 * given below: <p>
 *
 * RFC 822 mail headers <strong>must</strong> contain only US-ASCII
 * characters. Headers that contain non US-ASCII characters must be
 * encoded so that they contain only US-ASCII characters. Basically,
 * this process involves using either BASE64 or QP to encode certain
 * characters. RFC 2047 describes this in detail. <p>
 *
 * In Java, Strings contain (16 bit) Unicode characters. ASCII is a
 * subset of Unicode (and occupies the range 0 - 127). A String
 * that contains only ASCII characters is already mail-safe. If the
 * String contains non US-ASCII characters, it must be encoded. An
 * additional complexity in this step is that since Unicode is not
 * yet a widely used charset, one might want to first charset-encode
 * the String into another charset and then do the transfer-encoding.
 * <p>
 * Note that to get the actual bytes of a mail-safe String (say,
 * for sending over SMTP), one must do
 * <p><blockquote><pre>
 *
 *      byte[] bytes = string.getBytes("iso-8859-1");
 *
 * </pre></blockquote><p>
 *
 * The <code>setHeader</code> and <code>addHeader</code> methods
 * on MimeMessage and MimeBodyPart assume that the given header values
 * are Unicode strings that contain only US-ASCII characters. Hence
 * the callers of those methods must insure that the values they pass
 * do not contain non US-ASCII characters. The methods in this class
 * help do this. <p>
 *
 * The <code>getHeader</code> family of methods on MimeMessage and
 * MimeBodyPart return the raw header value. These might be encoded
 * as per RFC 2047, and if so, must be decoded into Unicode Strings.
 * The methods in this class help to do this. <p>
 *
 * Several System properties control strict conformance to the MIME
 * spec.  Note that these are not session properties but must be set
 * globally as System properties. <p>
 *
 * The <code>mail.mime.decodetext.strict</code> property controls
 * decoding of MIME encoded words.  The MIME spec requires that encoded
 * words start at the beginning of a whitespace separated word.  Some
 * mailers incorrectly include encoded words in the middle of a word.
 * If the <code>mail.mime.decodetext.strict</code> System property is
 * set to <code>"false"</code>, an attempt will be made to decode these
 * illegal encoded words. The default is true. <p>
 *
 * The <code>mail.mime.encodeeol.strict</code> property controls the
 * choice of Content-Transfer-Encoding for MIME parts that are not of
 * type "text".  Often such parts will contain textual data for which
 * an encoding that allows normal end of line conventions is appropriate.
 * In rare cases, such a part will appear to contain entirely textual
 * data, but will require an encoding that preserves CR and LF characters
 * without change.  If the <code>mail.mime.decodetext.strict</code>
 * System property is set to <code>"true"</code>, such an encoding will
 * be used when necessary.  The default is false. <p>
 *
 * In addition, the <code>mail.mime.charset</code> System property can
 * be used to specify the default MIME charset to use for encoded words
 * and text parts that don't otherwise specify a charset.  Normally, the
 * default MIME charset is derived from the default Java charset, as
 * specified in the <code>file.encoding</code> System property.  Most
 * applications will have no need to explicitly set the default MIME
 * charset.  In cases where the default MIME charset to be used for
 * mail messages is different than the charset used for files stored on
 * the system, this property should be set.
 *
 * @version 1.45, 03/03/10
 * @author  John Mani
 * @author  Bill Shannon
 */

public class MimeUtility {

    // This class cannot be instantiated
    private MimeUtility() { }

    public static final int ALL = -1;

    private static final int BUFFER_SIZE = 1024;
    private static boolean decodeStrict = true;
    private static boolean encodeEolStrict = false;
    private static boolean foldEncodedWords = false;
    private static boolean foldText = true;

    static {
        try {
            String s = SAAJUtil.getSystemProperty("mail.mime.decodetext.strict");
            // default to true
            decodeStrict = s == null || !s.equalsIgnoreCase("false");
            s = SAAJUtil.getSystemProperty("mail.mime.encodeeol.strict");
            // default to false
            encodeEolStrict = s != null && s.equalsIgnoreCase("true");
            s = SAAJUtil.getSystemProperty("mail.mime.foldencodedwords");
            // default to false
            foldEncodedWords = s != null && s.equalsIgnoreCase("true");
            s = SAAJUtil.getSystemProperty("mail.mime.foldtext");
            // default to true
            foldText = s == null || !s.equalsIgnoreCase("false");
        } catch (SecurityException sex) {
            // ignore it
        }
    }


    /**
     * Get the content-transfer-encoding that should be applied
     * to the input stream of this datasource, to make it mailsafe. <p>
     *
     * The algorithm used here is: <br>
     * <ul>
     * <li>
     * If the primary type of this datasource is "text" and if all
     * the bytes in its input stream are US-ASCII, then the encoding
     * is "7bit". If more than half of the bytes are non-US-ASCII, then
     * the encoding is "base64". If less than half of the bytes are
     * non-US-ASCII, then the encoding is "quoted-printable".
     * <li>
     * If the primary type of this datasource is not "text", then if
     * all the bytes of its input stream are US-ASCII, the encoding
     * is "7bit". If there is even one non-US-ASCII character, the
     * encoding is "base64".
     * </ul>
     *
     * @param   ds      DataSource
     * @return          the encoding. This is either "7bit",
     *                  "quoted-printable" or "base64"
     */
    public static String getEncoding(DataSource ds) {
        ContentType cType = null;
        InputStream is = null;
        String encoding = null;

        try {
            cType = new ContentType(ds.getContentType());
            is = ds.getInputStream();
        } catch (Exception ex) {
            return "base64"; // what else ?!
        }

        boolean isText = cType.match("text/*");
        // if not text, stop processing when we see non-ASCII
        int i = checkAscii(is, ALL, !isText);
        switch (i) {
        case ALL_ASCII:
            encoding = "7bit"; // all ascii
            break;
        case MOSTLY_ASCII:
            encoding = "quoted-printable"; // mostly ascii
            break;
        default:
            encoding = "base64"; // mostly binary
            break;
        }

        // Close the input stream
        try {
            is.close();
        } catch (IOException ioex) { }

        return encoding;
    }

    /**
     * Same as <code>getEncoding(DataSource)</code> except that instead
     * of reading the data from an <code>InputStream</code> it uses the
     * <code>writeTo</code> method to examine the data.  This is more
     * efficient in the common case of a <code>DataHandler</code>
     * created with an object and a MIME type (for example, a
     * "text/plain" String) because all the I/O is done in this
     * thread.  In the case requiring an <code>InputStream</code> the
     * <code>DataHandler</code> uses a thread, a pair of pipe streams,
     * and the <code>writeTo</code> method to produce the data. <p>
     *
     * @since   JavaMail 1.2
     */
    public static String getEncoding(DataHandler dh) {
        ContentType cType = null;
        String encoding = null;

        /*
         * Try to pick the most efficient means of determining the
         * encoding.  If this DataHandler was created using a DataSource,
         * the getEncoding(DataSource) method is typically faster.  If
         * the DataHandler was created with an object, this method is
         * much faster.  To distinguish the two cases, we use a heuristic.
         * A DataHandler created with an object will always have a null name.
         * A DataHandler created with a DataSource will usually have a
         * non-null name.
         *
         * XXX - This is actually quite a disgusting hack, but it makes
         *       a common case run over twice as fast.
         */
        if (dh.getName() != null)
            return getEncoding(dh.getDataSource());

        try {
            cType = new ContentType(dh.getContentType());
        } catch (Exception ex) {
            return "base64"; // what else ?!
        }

        if (cType.match("text/*")) {
            // Check all of the available bytes
            AsciiOutputStream aos = new AsciiOutputStream(false, false);
            try {
                dh.writeTo(aos);
            } catch (IOException ex) { }        // ignore it
            switch (aos.getAscii()) {
            case ALL_ASCII:
                encoding = "7bit"; // all ascii
                break;
            case MOSTLY_ASCII:
                encoding = "quoted-printable"; // mostly ascii
                break;
            default:
                encoding = "base64"; // mostly binary
                break;
            }
        } else { // not "text"
            // Check all of available bytes, break out if we find
            // at least one non-US-ASCII character
            AsciiOutputStream aos =
                        new AsciiOutputStream(true, encodeEolStrict);
            try {
                dh.writeTo(aos);
            } catch (IOException ex) { }        // ignore it
            if (aos.getAscii() == ALL_ASCII) // all ascii
                encoding = "7bit";
            else // found atleast one non-ascii character, use b64
                encoding = "base64";
        }

        return encoding;
    }

    /**
     * Decode the given input stream. The Input stream returned is
     * the decoded input stream. All the encodings defined in RFC 2045
     * are supported here. They include "base64", "quoted-printable",
     * "7bit", "8bit", and "binary". In addition, "uuencode" is also
     * supported.
     *
     * @param   is              input stream
     * @param   encoding        the encoding of the stream.
     * @return                  decoded input stream.
     */
    public static InputStream decode(InputStream is, String encoding)
                throws MessagingException {
        if (encoding.equalsIgnoreCase("base64"))
            return new BASE64DecoderStream(is);
        else if (encoding.equalsIgnoreCase("quoted-printable"))
            return new QPDecoderStream(is);
        else if (encoding.equalsIgnoreCase("uuencode") ||
                 encoding.equalsIgnoreCase("x-uuencode") ||
                 encoding.equalsIgnoreCase("x-uue"))
            return new UUDecoderStream(is);
        else if (encoding.equalsIgnoreCase("binary") ||
                 encoding.equalsIgnoreCase("7bit") ||
                 encoding.equalsIgnoreCase("8bit"))
            return is;
        else
            throw new MessagingException("Unknown encoding: " + encoding);
    }

    /**
     * Wrap an encoder around the given output stream.
     * All the encodings defined in RFC 2045 are supported here.
     * They include "base64", "quoted-printable", "7bit", "8bit" and
     * "binary". In addition, "uuencode" is also supported.
     *
     * @param   os              output stream
     * @param   encoding        the encoding of the stream.
     * @return                  output stream that applies the
     *                          specified encoding.
     */
    public static OutputStream encode(OutputStream os, String encoding)
                throws MessagingException {
        if (encoding == null)
            return os;
        else if (encoding.equalsIgnoreCase("base64"))
            return new BASE64EncoderStream(os);
        else if (encoding.equalsIgnoreCase("quoted-printable"))
            return new QPEncoderStream(os);
        else if (encoding.equalsIgnoreCase("uuencode") ||
                 encoding.equalsIgnoreCase("x-uuencode") ||
                 encoding.equalsIgnoreCase("x-uue"))
            return new UUEncoderStream(os);
        else if (encoding.equalsIgnoreCase("binary") ||
                 encoding.equalsIgnoreCase("7bit") ||
                 encoding.equalsIgnoreCase("8bit"))
            return os;
        else
            throw new MessagingException("Unknown encoding: " +encoding);
    }

    /**
     * Wrap an encoder around the given output stream.
     * All the encodings defined in RFC 2045 are supported here.
     * They include "base64", "quoted-printable", "7bit", "8bit" and
     * "binary". In addition, "uuencode" is also supported.
     * The <code>filename</code> parameter is used with the "uuencode"
     * encoding and is included in the encoded output.
     *
     * @param   os              output stream
     * @param   encoding        the encoding of the stream.
     * @param   filename        name for the file being encoded (only used
     *                          with uuencode)
     * @return                  output stream that applies the
     *                          specified encoding.
     * @since                   JavaMail 1.2
     */
    public static OutputStream encode(OutputStream os, String encoding,
                                      String filename)
                throws MessagingException {
        if (encoding == null)
            return os;
        else if (encoding.equalsIgnoreCase("base64"))
            return new BASE64EncoderStream(os);
        else if (encoding.equalsIgnoreCase("quoted-printable"))
            return new QPEncoderStream(os);
        else if (encoding.equalsIgnoreCase("uuencode") ||
                 encoding.equalsIgnoreCase("x-uuencode") ||
                 encoding.equalsIgnoreCase("x-uue"))
            return new UUEncoderStream(os, filename);
        else if (encoding.equalsIgnoreCase("binary") ||
                 encoding.equalsIgnoreCase("7bit") ||
                 encoding.equalsIgnoreCase("8bit"))
            return os;
        else
            throw new MessagingException("Unknown encoding: " +encoding);
    }

    /**
     * Encode a RFC 822 "text" token into mail-safe form as per
     * RFC 2047. <p>
     *
     * The given Unicode string is examined for non US-ASCII
     * characters. If the string contains only US-ASCII characters,
     * it is returned as-is.  If the string contains non US-ASCII
     * characters, it is first character-encoded using the platform's
     * default charset, then transfer-encoded using either the B or
     * Q encoding. The resulting bytes are then returned as a Unicode
     * string containing only ASCII  characters. <p>
     *
     * Note that this method should be used to encode only
     * "unstructured" RFC 822 headers. <p>
     *
     * Example of usage:
     * <p><blockquote><pre>
     *
     *  MimeBodyPart part = ...
     *  String rawvalue = "FooBar Mailer, Japanese version 1.1"
     *  try {
     *    // If we know for sure that rawvalue contains only US-ASCII
     *    // characters, we can skip the encoding part
     *    part.setHeader("X-mailer", MimeUtility.encodeText(rawvalue));
     *  } catch (UnsupportedEncodingException e) {
     *    // encoding failure
     *  } catch (MessagingException me) {
     *   // setHeader() failure
     *  }
     *
     * </pre></blockquote><p>
     *
     * @param   text    unicode string
     * @return  Unicode string containing only US-ASCII characters
     * @exception UnsupportedEncodingException if the encoding fails
     */
    public static String encodeText(String text)
                        throws UnsupportedEncodingException {
        return encodeText(text, null, null);
    }

    /**
     * Encode a RFC 822 "text" token into mail-safe form as per
     * RFC 2047. <p>
     *
     * The given Unicode string is examined for non US-ASCII
     * characters. If the string contains only US-ASCII characters,
     * it is returned as-is.  If the string contains non US-ASCII
     * characters, it is first character-encoded using the specified
     * charset, then transfer-encoded using either the B or Q encoding.
     * The resulting bytes are then returned as a Unicode string
     * containing only ASCII characters. <p>
     *
     * Note that this method should be used to encode only
     * "unstructured" RFC 822 headers.
     *
     * @param   text    the header value
     * @param   charset the charset. If this parameter is null, the
     *          platform's default chatset is used.
     * @param   encoding the encoding to be used. Currently supported
     *          values are "B" and "Q". If this parameter is null, then
     *          the "Q" encoding is used if most of characters to be
     *          encoded are in the ASCII charset, otherwise "B" encoding
     *          is used.
     * @return  Unicode string containing only US-ASCII characters
     */
    public static String encodeText(String text, String charset,
                                    String encoding)
                        throws UnsupportedEncodingException {
        return encodeWord(text, charset, encoding, false);
    }

    /**
     * Decode "unstructured" headers, that is, headers that are defined
     * as '*text' as per RFC 822. <p>
     *
     * The string is decoded using the algorithm specified in
     * RFC 2047, Section 6.1.1. If the charset-conversion fails
     * for any sequence, an UnsupportedEncodingException is thrown.
     * If the String is not an RFC 2047 style encoded header, it is
     * returned as-is <p>
     *
     * Example of usage:
     * <p><blockquote><pre>
     *
     *  MimeBodyPart part = ...
     *  String rawvalue = null;
     *  String  value = null;
     *  try {
     *    if ((rawvalue = part.getHeader("X-mailer")[0]) != null)
     *      value = MimeUtility.decodeText(rawvalue);
     *  } catch (UnsupportedEncodingException e) {
     *      // Don't care
     *      value = rawvalue;
     *  } catch (MessagingException me) { }
     *
     *  return value;
     *
     * </pre></blockquote><p>
     *
     * @param   etext   the possibly encoded value
     * @exception       UnsupportedEncodingException if the charset
     *                  conversion failed.
     */
    public static String decodeText(String etext)
                throws UnsupportedEncodingException {
        /*
         * We look for sequences separated by "linear-white-space".
         * (as per RFC 2047, Section 6.1.1)
         * RFC 822 defines "linear-white-space" as SPACE | HT | CR | NL.
         */
        String lwsp = " \t\n\r";
        StringTokenizer st;

        /*
         * First, lets do a quick run thru the string and check
         * whether the sequence "=?"  exists at all. If none exists,
         * we know there are no encoded-words in here and we can just
         * return the string as-is, without suffering thru the later
         * decoding logic.
         * This handles the most common case of unencoded headers
         * efficiently.
         */
        if (etext.indexOf("=?") == -1)
            return etext;

        // Encoded words found. Start decoding ...

        st = new StringTokenizer(etext, lwsp, true);
        StringBuffer sb = new StringBuffer();  // decode buffer
        StringBuffer wsb = new StringBuffer(); // white space buffer
        boolean prevWasEncoded = false;

        while (st.hasMoreTokens()) {
            char c;
            String s = st.nextToken();
            // If whitespace, append it to the whitespace buffer
            if (((c = s.charAt(0)) == ' ') || (c == '\t') ||
                (c == '\r') || (c == '\n'))
                wsb.append(c);
            else {
                // Check if token is an 'encoded-word' ..
                String word;
                try {
                    word = decodeWord(s);
                    // Yes, this IS an 'encoded-word'.
                    if (!prevWasEncoded && wsb.length() > 0) {
                        // if the previous word was also encoded, we
                        // should ignore the collected whitespace. Else
                        // we include the whitespace as well.
                        sb.append(wsb);
                    }
                    prevWasEncoded = true;
                } catch (ParseException pex) {
                    // This is NOT an 'encoded-word'.
                    word = s;
                    // possibly decode inner encoded words
                    if (!decodeStrict)
                        word = decodeInnerWords(word);
                    // include colleced whitespace ..
                    if (wsb.length() > 0)
                        sb.append(wsb);
                    prevWasEncoded = false;
                }
                sb.append(word); // append the actual word
                wsb.setLength(0); // reset wsb for reuse
            }
        }
        return sb.toString();
    }

    /**
     * Encode a RFC 822 "word" token into mail-safe form as per
     * RFC 2047. <p>
     *
     * The given Unicode string is examined for non US-ASCII
     * characters. If the string contains only US-ASCII characters,
     * it is returned as-is.  If the string contains non US-ASCII
     * characters, it is first character-encoded using the platform's
     * default charset, then transfer-encoded using either the B or
     * Q encoding. The resulting bytes are then returned as a Unicode
     * string containing only ASCII  characters. <p>
     *
     * This method is meant to be used when creating RFC 822 "phrases".
     * The InternetAddress class, for example, uses this to encode
     * it's 'phrase' component.
     *
     * @param   text    unicode string
     * @return  Array of Unicode strings containing only US-ASCII
     *          characters.
     * @exception UnsupportedEncodingException if the encoding fails
     */
    public static String encodeWord(String word)
                        throws UnsupportedEncodingException {
        return encodeWord(word, null, null);
    }

    /**
     * Encode a RFC 822 "word" token into mail-safe form as per
     * RFC 2047. <p>
     *
     * The given Unicode string is examined for non US-ASCII
     * characters. If the string contains only US-ASCII characters,
     * it is returned as-is.  If the string contains non US-ASCII
     * characters, it is first character-encoded using the specified
     * charset, then transfer-encoded using either the B or Q encoding.
     * The resulting bytes are then returned as a Unicode string
     * containing only ASCII characters. <p>
     *
     * @param   text    unicode string
     * @param   charset the MIME charset
     * @param   encoding the encoding to be used. Currently supported
     *          values are "B" and "Q". If this parameter is null, then
     *          the "Q" encoding is used if most of characters to be
     *          encoded are in the ASCII charset, otherwise "B" encoding
     *          is used.
     * @return  Unicode string containing only US-ASCII characters
     * @exception UnsupportedEncodingException if the encoding fails
     */
    public static String encodeWord(String word, String charset,
                                    String encoding)
                        throws UnsupportedEncodingException {
        return encodeWord(word, charset, encoding, true);
    }

    /*
     * Encode the given string. The parameter 'encodingWord' should
     * be true if a RFC 822 "word" token is being encoded and false if a
     * RFC 822 "text" token is being encoded. This is because the
     * "Q" encoding defined in RFC 2047 has more restrictions when
     * encoding "word" tokens. (Sigh)
     */
    private static String encodeWord(String string, String charset,
                                     String encoding, boolean encodingWord)
                        throws UnsupportedEncodingException {

        // If 'string' contains only US-ASCII characters, just
        // return it.
        int ascii = checkAscii(string);
        if (ascii == ALL_ASCII)
            return string;

        // Else, apply the specified charset conversion.
        String jcharset;
        if (charset == null) { // use default charset
            jcharset = getDefaultJavaCharset(); // the java charset
            charset = getDefaultMIMECharset(); // the MIME equivalent
        } else // MIME charset -> java charset
            jcharset = javaCharset(charset);

        // If no transfer-encoding is specified, figure one out.
        if (encoding == null) {
            if (ascii != MOSTLY_NONASCII)
                encoding = "Q";
            else
                encoding = "B";
        }

        boolean b64;
        if (encoding.equalsIgnoreCase("B"))
            b64 = true;
        else if (encoding.equalsIgnoreCase("Q"))
            b64 = false;
        else
            throw new UnsupportedEncodingException(
                        "Unknown transfer encoding: " + encoding);

        StringBuffer outb = new StringBuffer(); // the output buffer
        doEncode(string, b64, jcharset,
                 // As per RFC 2047, size of an encoded string should not
                 // exceed 75 bytes.
                 // 7 = size of "=?", '?', 'B'/'Q', '?', "?="
                 75 - 7 - charset.length(), // the available space
                 "=?" + charset + "?" + encoding + "?", // prefix
                 true, encodingWord, outb);

        return outb.toString();
    }

    private static void doEncode(String string, boolean b64,
                String jcharset, int avail, String prefix,
                boolean first, boolean encodingWord, StringBuffer buf)
                        throws UnsupportedEncodingException {

        // First find out what the length of the encoded version of
        // 'string' would be.
        byte[] bytes = string.getBytes(jcharset);
        int len;
        if (b64) // "B" encoding
            len = BEncoderStream.encodedLength(bytes);
        else // "Q"
            len = QEncoderStream.encodedLength(bytes, encodingWord);

        int size;
        if ((len > avail) && ((size = string.length()) > 1)) {
            // If the length is greater than 'avail', split 'string'
            // into two and recurse.
            doEncode(string.substring(0, size/2), b64, jcharset,
                     avail, prefix, first, encodingWord, buf);
            doEncode(string.substring(size/2, size), b64, jcharset,
                     avail, prefix, false, encodingWord, buf);
        } else {
            // length <= than 'avail'. Encode the given string
            ByteArrayOutputStream os = new ByteArrayOutputStream(BUFFER_SIZE);
            OutputStream eos; // the encoder
            if (b64) // "B" encoding
                eos = new BEncoderStream(os);
            else // "Q" encoding
                eos = new QEncoderStream(os, encodingWord);

            try { // do the encoding
                eos.write(bytes);
                eos.close();
            } catch (IOException ioex) { }

            byte[] encodedBytes = os.toByteArray(); // the encoded stuff
            // Now write out the encoded (all ASCII) bytes into our
            // StringBuffer
            if (!first) // not the first line of this sequence
                if (foldEncodedWords)
                    buf.append("\r\n "); // start a continuation line
                else
                    buf.append(" "); // line will be folded later

            buf.append(prefix);
            for (int i = 0; i < encodedBytes.length; i++)
                buf.append((char)encodedBytes[i]);
            buf.append("?="); // terminate the current sequence
        }
    }

    /**
     * The string is parsed using the rules in RFC 2047 for parsing
     * an "encoded-word". If the parse fails, a ParseException is
     * thrown. Otherwise, it is transfer-decoded, and then
     * charset-converted into Unicode. If the charset-conversion
     * fails, an UnsupportedEncodingException is thrown.<p>
     *
     * @param   eword   the possibly encoded value
     * @exception       ParseException if the string is not an
     *                  encoded-word as per RFC 2047.
     * @exception       UnsupportedEncodingException if the charset
     *                  conversion failed.
     */
    public static String decodeWord(String eword)
                throws ParseException, UnsupportedEncodingException {

        if (!eword.startsWith("=?")) // not an encoded word
            throw new ParseException();

        // get charset
        int start = 2; int pos;
        if ((pos = eword.indexOf('?', start)) == -1)
            throw new ParseException();
        String charset = javaCharset(eword.substring(start, pos));

        // get encoding
        start = pos+1;
        if ((pos = eword.indexOf('?', start)) == -1)
            throw new ParseException();
        String encoding = eword.substring(start, pos);

        // get encoded-sequence
        start = pos+1;
        if ((pos = eword.indexOf("?=", start)) == -1)
            throw new ParseException();
        String word = eword.substring(start, pos);

        try {
            // Extract the bytes from word
            ByteArrayInputStream bis =
                new ByteArrayInputStream(ASCIIUtility.getBytes(word));

            // Get the appropriate decoder
            InputStream is;
            if (encoding.equalsIgnoreCase("B"))
                is = new BASE64DecoderStream(bis);
            else if (encoding.equalsIgnoreCase("Q"))
                is = new QDecoderStream(bis);
            else
                throw new UnsupportedEncodingException(
                                "unknown encoding: " + encoding);

            // For b64 & q, size of decoded word <= size of word. So
            // the decoded bytes must fit into the 'bytes' array. This
            // is certainly more efficient than writing bytes into a
            // ByteArrayOutputStream and then pulling out the byte[]
            // from it.
            int count = bis.available();
            byte[] bytes = new byte[count];
            // count is set to the actual number of decoded bytes
            count = is.read(bytes, 0, count);

            // Finally, convert the decoded bytes into a String using
            // the specified charset
            String s = new String(bytes, 0, count, charset);
            if (pos + 2 < eword.length()) {
                // there's still more text in the string
                String rest = eword.substring(pos + 2);
                if (!decodeStrict)
                    rest = decodeInnerWords(rest);
                s += rest;
            }
            return s;
        } catch (UnsupportedEncodingException uex) {
            // explicitly catch and rethrow this exception, otherwise
            // the below IOException catch will swallow this up!
            throw uex;
        } catch (IOException ioex) {
            // Shouldn't happen.
            throw new ParseException();
        } catch (IllegalArgumentException iex) {
            /* An unknown charset of the form ISO-XXX-XXX, will cause
             * the JDK to throw an IllegalArgumentException ... Since the
             * JDK will attempt to create a classname using this string,
             * but valid classnames must not contain the character '-',
             * and this results in an IllegalArgumentException, rather than
             * the expected UnsupportedEncodingException. Yikes
             */
            throw new UnsupportedEncodingException();
        }
    }

    /**
     * Look for encoded words within a word.  The MIME spec doesn't
     * allow this, but many broken mailers, especially Japanese mailers,
     * produce such incorrect encodings.
     */
    private static String decodeInnerWords(String word)
                                throws UnsupportedEncodingException {
        int start = 0, i;
        StringBuffer buf = new StringBuffer();
        while ((i = word.indexOf("=?", start)) >= 0) {
            buf.append(word.substring(start, i));
            int end = word.indexOf("?=", i);
            if (end < 0)
                break;
            String s = word.substring(i, end + 2);
            try {
                s = decodeWord(s);
            } catch (ParseException pex) {
                // ignore it, just use the original string
            }
            buf.append(s);
            start = end + 2;
        }
        if (start == 0)
            return word;
        if (start < word.length())
            buf.append(word.substring(start));
        return buf.toString();
    }

    /**
     * A utility method to quote a word, if the word contains any
     * characters from the specified 'specials' list.<p>
     *
     * The <code>HeaderTokenizer</code> class defines two special
     * sets of delimiters - MIME and RFC 822. <p>
     *
     * This method is typically used during the generation of
     * RFC 822 and MIME header fields.
     *
     * @param   word    word to be quoted
     * @param   specials the set of special characters
     * @return          the possibly quoted word
     * @see     javax.mail.internet.HeaderTokenizer#MIME
     * @see     javax.mail.internet.HeaderTokenizer#RFC822
     */
    public static String quote(String word, String specials) {
        int len = word.length();

        /*
         * Look for any "bad" characters, Escape and
         *  quote the entire string if necessary.
         */
        boolean needQuoting = false;
        for (int i = 0; i < len; i++) {
            char c = word.charAt(i);
            if (c == '"' || c == '\\' || c == '\r' || c == '\n') {
                // need to escape them and then quote the whole string
                StringBuffer sb = new StringBuffer(len + 3);
                sb.append('"');
                sb.append(word.substring(0, i));
                int lastc = 0;
                for (int j = i; j < len; j++) {
                    char cc = word.charAt(j);
                    if ((cc == '"') || (cc == '\\') ||
                        (cc == '\r') || (cc == '\n'))
                        if (cc == '\n' && lastc == '\r')
                            ;   // do nothing, CR was already escaped
                        else
                            sb.append('\\');    // Escape the character
                    sb.append(cc);
                    lastc = cc;
                }
                sb.append('"');
                return sb.toString();
            } else if (c < 040 || c >= 0177 || specials.indexOf(c) >= 0)
                // These characters cause the string to be quoted
                needQuoting = true;
        }

        if (needQuoting) {
            StringBuffer sb = new StringBuffer(len + 2);
            sb.append('"').append(word).append('"');
            return sb.toString();
        } else
            return word;
    }

    /**
     * Fold a string at linear whitespace so that each line is no longer
     * than 76 characters, if possible.  If there are more than 76
     * non-whitespace characters consecutively, the string is folded at
     * the first whitespace after that sequence.  The parameter
     * <code>used</code> indicates how many characters have been used in
     * the current line; it is usually the length of the header name. <p>
     *
     * Note that line breaks in the string aren't escaped; they probably
     * should be.
     *
     * @param   used    characters used in line so far
     * @param   s       the string to fold
     * @return          the folded string
     */
    /*public*/ static String fold(int used, String s) {
        if (!foldText)
            return s;

        int end;
        char c;
        // Strip trailing spaces
        for (end = s.length() - 1; end >= 0; end--) {
            c = s.charAt(end);
            if (c != ' ' && c != '\t')
                break;
        }
        if (end != s.length() - 1)
            s = s.substring(0, end + 1);

        // if the string fits now, just return it
        if (used + s.length() <= 76)
            return s;

        // have to actually fold the string
        StringBuffer sb = new StringBuffer(s.length() + 4);
        char lastc = 0;
        while (used + s.length() > 76) {
            int lastspace = -1;
            for (int i = 0; i < s.length(); i++) {
                if (lastspace != -1 && used + i > 76)
                    break;
                c = s.charAt(i);
                if (c == ' ' || c == '\t')
                    if (!(lastc == ' ' || lastc == '\t'))
                        lastspace = i;
                lastc = c;
            }
            if (lastspace == -1) {
                // no space, use the whole thing
                sb.append(s);
                s = "";
                used = 0;
                break;
            }
            sb.append(s.substring(0, lastspace));
            sb.append("\r\n");
            lastc = s.charAt(lastspace);
            sb.append(lastc);
            s = s.substring(lastspace + 1);
            used = 1;
        }
        sb.append(s);
        return sb.toString();
    }

    /**
     * Unfold a folded header.  Any line breaks that aren't escaped and
     * are followed by whitespace are removed.
     *
     * @param   s       the string to unfold
     * @return          the unfolded string
     */
    /*public*/ static String unfold(String s) {
        if (!foldText)
            return s;

        StringBuffer sb = null;
        int i;
        while ((i = indexOfAny(s, "\r\n")) >= 0) {
            int start = i;
            int l = s.length();
            i++;                // skip CR or NL
            if (i < l && s.charAt(i - 1) == '\r' && s.charAt(i) == '\n')
                i++;    // skip LF
            if (start == 0 || s.charAt(start - 1) != '\\') {
                char c;
                // if next line starts with whitespace, skip all of it
                // XXX - always has to be true?
                if (i < l && ((c = s.charAt(i)) == ' ' || c == '\t')) {
                    i++;        // skip whitespace
                    while (i < l && ((c = s.charAt(i)) == ' ' || c == '\t'))
                        i++;
                    if (sb == null)
                        sb = new StringBuffer(s.length());
                    if (start != 0) {
                        sb.append(s.substring(0, start));
                        sb.append(' ');
                    }
                    s = s.substring(i);
                    continue;
                }
                // it's not a continuation line, just leave it in
                if (sb == null)
                    sb = new StringBuffer(s.length());
                sb.append(s.substring(0, i));
                s = s.substring(i);
            } else {
                // there's a backslash at "start - 1"
                // strip it out, but leave in the line break
                if (sb == null)
                    sb = new StringBuffer(s.length());
                sb.append(s.substring(0, start - 1));
                sb.append(s.substring(start, i));
                s = s.substring(i);
            }
        }
        if (sb != null) {
            sb.append(s);
            return sb.toString();
        } else
            return s;
    }

    /**
     * Return the first index of any of the characters in "any" in "s",
     * or -1 if none are found.
     *
     * This should be a method on String.
     */
    private static int indexOfAny(String s, String any) {
        return indexOfAny(s, any, 0);
    }

    private static int indexOfAny(String s, String any, int start) {
        try {
            int len = s.length();
            for (int i = start; i < len; i++) {
                if (any.indexOf(s.charAt(i)) >= 0)
                    return i;
            }
            return -1;
        } catch (StringIndexOutOfBoundsException e) {
            return -1;
        }
    }

    /**
     * Convert a MIME charset name into a valid Java charset name. <p>
     *
     * @param charset   the MIME charset name
     * @return  the Java charset equivalent. If a suitable mapping is
     *          not available, the passed in charset is itself returned.
     */
    public static String javaCharset(String charset) {
        if (mime2java == null || charset == null)
            // no mapping table, or charset parameter is null
            return charset;

        String alias = (String)mime2java.get(charset.toLowerCase());
        return alias == null ? charset : alias;
    }

    /**
     * Convert a java charset into its MIME charset name. <p>
     *
     * Note that a future version of JDK (post 1.2) might provide
     * this functionality, in which case, we may deprecate this
     * method then.
     *
     * @param   charset    the JDK charset
     * @return          the MIME/IANA equivalent. If a mapping
     *                  is not possible, the passed in charset itself
     *                  is returned.
     * @since           JavaMail 1.1
     */
    public static String mimeCharset(String charset) {
        if (java2mime == null || charset == null)
            // no mapping table or charset param is null
            return charset;

        String alias = (String)java2mime.get(charset.toLowerCase());
        return alias == null ? charset : alias;
    }

    private static String defaultJavaCharset;
    private static String defaultMIMECharset;

    /**
     * Get the default charset corresponding to the system's current
     * default locale.  If the System property <code>mail.mime.charset</code>
     * is set, a system charset corresponding to this MIME charset will be
     * returned. <p>
     *
     * @return  the default charset of the system's default locale,
     *          as a Java charset. (NOT a MIME charset)
     * @since   JavaMail 1.1
     */
    public static String getDefaultJavaCharset() {
        if (defaultJavaCharset == null) {
            /*
             * If mail.mime.charset is set, it controls the default
             * Java charset as well.
             */
            String mimecs = null;

            mimecs = SAAJUtil.getSystemProperty("mail.mime.charset");

            if (mimecs != null && mimecs.length() > 0) {
                defaultJavaCharset = javaCharset(mimecs);
                return defaultJavaCharset;
            }

            try {
                defaultJavaCharset = System.getProperty("file.encoding",
                                                        "8859_1");
            } catch (SecurityException sex) {

                class NullInputStream extends InputStream {
                    public int read() {
                        return 0;
                    }
                }
                InputStreamReader reader =
                        new InputStreamReader(new NullInputStream());
                defaultJavaCharset = reader.getEncoding();
                if (defaultJavaCharset == null)
                    defaultJavaCharset = "8859_1";
            }
        }

        return defaultJavaCharset;
    }

    /*
     * Get the default MIME charset for this locale.
     */
    static String getDefaultMIMECharset() {
        if (defaultMIMECharset == null) {
                defaultMIMECharset = SAAJUtil.getSystemProperty("mail.mime.charset");
        }
        if (defaultMIMECharset == null)
            defaultMIMECharset = mimeCharset(getDefaultJavaCharset());
        return defaultMIMECharset;
    }

    // Tables to map MIME charset names to Java names and vice versa.
    // XXX - Should eventually use J2SE 1.4 java.nio.charset.Charset
    private static Hashtable mime2java;
    private static Hashtable java2mime;

    static {
        java2mime = new Hashtable(40);
        mime2java = new Hashtable(10);

        try {
            // Use this class's classloader to load the mapping file
            // XXX - we should use SecuritySupport, but it's in another package
            InputStream is =
                    com.sun.xml.internal.messaging.saaj.packaging.mime.internet.MimeUtility.class.getResourceAsStream(
                    "/META-INF/javamail.charset.map");

            if (is != null) {
                is = new LineInputStream(is);

                // Load the JDK-to-MIME charset mapping table
                loadMappings((LineInputStream)is, java2mime);

                // Load the MIME-to-JDK charset mapping table
                loadMappings((LineInputStream)is, mime2java);
            }
        } catch (Exception ex) { }

        // If we didn't load the tables, e.g., because we didn't have
        // permission, load them manually.  The entries here should be
        // the same as the default javamail.charset.map.
        if (java2mime.isEmpty()) {
            java2mime.put("8859_1", "ISO-8859-1");
            java2mime.put("iso8859_1", "ISO-8859-1");
            java2mime.put("ISO8859-1", "ISO-8859-1");

            java2mime.put("8859_2", "ISO-8859-2");
            java2mime.put("iso8859_2", "ISO-8859-2");
            java2mime.put("ISO8859-2", "ISO-8859-2");

            java2mime.put("8859_3", "ISO-8859-3");
            java2mime.put("iso8859_3", "ISO-8859-3");
            java2mime.put("ISO8859-3", "ISO-8859-3");

            java2mime.put("8859_4", "ISO-8859-4");
            java2mime.put("iso8859_4", "ISO-8859-4");
            java2mime.put("ISO8859-4", "ISO-8859-4");

            java2mime.put("8859_5", "ISO-8859-5");
            java2mime.put("iso8859_5", "ISO-8859-5");
            java2mime.put("ISO8859-5", "ISO-8859-5");

            java2mime.put("8859_6", "ISO-8859-6");
            java2mime.put("iso8859_6", "ISO-8859-6");
            java2mime.put("ISO8859-6", "ISO-8859-6");

            java2mime.put("8859_7", "ISO-8859-7");
            java2mime.put("iso8859_7", "ISO-8859-7");
            java2mime.put("ISO8859-7", "ISO-8859-7");

            java2mime.put("8859_8", "ISO-8859-8");
            java2mime.put("iso8859_8", "ISO-8859-8");
            java2mime.put("ISO8859-8", "ISO-8859-8");

            java2mime.put("8859_9", "ISO-8859-9");
            java2mime.put("iso8859_9", "ISO-8859-9");
            java2mime.put("ISO8859-9", "ISO-8859-9");

            java2mime.put("SJIS", "Shift_JIS");
            java2mime.put("MS932", "Shift_JIS");
            java2mime.put("JIS", "ISO-2022-JP");
            java2mime.put("ISO2022JP", "ISO-2022-JP");
            java2mime.put("EUC_JP", "euc-jp");
            java2mime.put("KOI8_R", "koi8-r");
            java2mime.put("EUC_CN", "euc-cn");
            java2mime.put("EUC_TW", "euc-tw");
            java2mime.put("EUC_KR", "euc-kr");
        }
        if (mime2java.isEmpty()) {
            mime2java.put("iso-2022-cn", "ISO2022CN");
            mime2java.put("iso-2022-kr", "ISO2022KR");
            mime2java.put("utf-8", "UTF8");
            mime2java.put("utf8", "UTF8");
            mime2java.put("ja_jp.iso2022-7", "ISO2022JP");
            mime2java.put("ja_jp.eucjp", "EUCJIS");
            mime2java.put("euc-kr", "KSC5601");
            mime2java.put("euckr", "KSC5601");
            mime2java.put("us-ascii", "ISO-8859-1");
            mime2java.put("x-us-ascii", "ISO-8859-1");
        }
    }

    private static void loadMappings(LineInputStream is, Hashtable table) {
        String currLine;

        while (true) {
            try {
                currLine = is.readLine();
            } catch (IOException ioex) {
                break; // error in reading, stop
            }

            if (currLine == null) // end of file, stop
                break;
            if (currLine.startsWith("--") && currLine.endsWith("--"))
                // end of this table
                break;

            // ignore empty lines and comments
            if (currLine.trim().length() == 0 || currLine.startsWith("#"))
                continue;

            // A valid entry is of the form <key><separator><value>
            // where, <separator> := SPACE | HT. Parse this
            StringTokenizer tk = new StringTokenizer(currLine, " \t");
            try {
                String key = tk.nextToken();
                String value = tk.nextToken();
                table.put(key.toLowerCase(), value);
            } catch (NoSuchElementException nex) { }
        }
    }

    static final int ALL_ASCII          = 1;
    static final int MOSTLY_ASCII       = 2;
    static final int MOSTLY_NONASCII    = 3;

    /**
     * Check if the given string contains non US-ASCII characters.
     * @param   s       string
     * @return          ALL_ASCII if all characters in the string
     *                  belong to the US-ASCII charset. MOSTLY_ASCII
     *                  if more than half of the available characters
     *                  are US-ASCII characters. Else MOSTLY_NONASCII.
     */
    static int checkAscii(String s) {
        int ascii = 0, non_ascii = 0;
        int l = s.length();

        for (int i = 0; i < l; i++) {
            if (nonascii((int)s.charAt(i))) // non-ascii
                non_ascii++;
            else
                ascii++;
        }

        if (non_ascii == 0)
            return ALL_ASCII;
        if (ascii > non_ascii)
            return MOSTLY_ASCII;

        return MOSTLY_NONASCII;
    }

    /**
     * Check if the given byte array contains non US-ASCII characters.
     * @param   b       byte array
     * @return          ALL_ASCII if all characters in the string
     *                  belong to the US-ASCII charset. MOSTLY_ASCII
     *                  if more than half of the available characters
     *                  are US-ASCII characters. Else MOSTLY_NONASCII.
     *
     * XXX - this method is no longer used
     */
    static int checkAscii(byte[] b) {
        int ascii = 0, non_ascii = 0;

        for (int i=0; i < b.length; i++) {
            // The '&' operator automatically causes b[i] to be promoted
            // to an int, and we mask out the higher bytes in the int
            // so that the resulting value is not a negative integer.
            if (nonascii(b[i] & 0xff)) // non-ascii
                non_ascii++;
            else
                ascii++;
        }

        if (non_ascii == 0)
            return ALL_ASCII;
        if (ascii > non_ascii)
            return MOSTLY_ASCII;

        return MOSTLY_NONASCII;
    }

    /**
     * Check if the given input stream contains non US-ASCII characters.
     * Upto <code>max</code> bytes are checked. If <code>max</code> is
     * set to <code>ALL</code>, then all the bytes available in this
     * input stream are checked. If <code>breakOnNonAscii</code> is true
     * the check terminates when the first non-US-ASCII character is
     * found and MOSTLY_NONASCII is returned. Else, the check continues
     * till <code>max</code> bytes or till the end of stream.
     *
     * @param   is      the input stream
     * @param   max     maximum bytes to check for. The special value
     *                  ALL indicates that all the bytes in this input
     *                  stream must be checked.
     * @param   breakOnNonAscii if <code>true</code>, then terminate the
     *                  the check when the first non-US-ASCII character
     *                  is found.
     * @return          ALL_ASCII if all characters in the string
     *                  belong to the US-ASCII charset. MOSTLY_ASCII
     *                  if more than half of the available characters
     *                  are US-ASCII characters. Else MOSTLY_NONASCII.
     */
    static int checkAscii(InputStream is, int max, boolean breakOnNonAscii) {
        int ascii = 0, non_ascii = 0;
        int len;
        int block = 4096;
        int linelen = 0;
        boolean longLine = false, badEOL = false;
        boolean checkEOL = encodeEolStrict && breakOnNonAscii;
        byte buf[] = null;
        if (max != 0) {
            block = (max == ALL) ? 4096 : Math.min(max, 4096);
            buf = new byte[block];
        }
        while (max != 0) {
            try {
                if ((len = is.read(buf, 0, block)) == -1)
                    break;
                int lastb = 0;
                for (int i = 0; i < len; i++) {
                    // The '&' operator automatically causes b[i] to
                    // be promoted to an int, and we mask out the higher
                    // bytes in the int so that the resulting value is
                    // not a negative integer.
                    int b = buf[i] & 0xff;
                    if (checkEOL &&
                            ((lastb == '\r' && b != '\n') ||
                            (lastb != '\r' && b == '\n')))
                        badEOL = true;
                    if (b == '\r' || b == '\n')
                        linelen = 0;
                    else {
                        linelen++;
                        if (linelen > 998)      // 1000 - CRLF
                            longLine = true;
                    }
                    if (nonascii(b)) {  // non-ascii
                        if (breakOnNonAscii) // we are done
                            return MOSTLY_NONASCII;
                        else
                            non_ascii++;
                    } else
                        ascii++;
                    lastb = b;
                }
            } catch (IOException ioex) {
                break;
            }
            if (max != ALL)
                max -= len;
        }

        if (max == 0 && breakOnNonAscii)
            // We have been told to break on the first non-ascii character.
            // We haven't got any non-ascii character yet, but then we
            // have not checked all of the available bytes either. So we
            // cannot say for sure that this input stream is ALL_ASCII,
            // and hence we must play safe and return MOSTLY_NONASCII

            return MOSTLY_NONASCII;

        if (non_ascii == 0) { // no non-us-ascii characters so far
            // If we're looking at non-text data, and we saw CR without LF
            // or vice versa, consider this mostly non-ASCII so that it
            // will be base64 encoded (since the quoted-printable encoder
            // doesn't encode this case properly).
            if (badEOL)
                return MOSTLY_NONASCII;
            // if we've seen a long line, we degrade to mostly ascii
            else if (longLine)
                return MOSTLY_ASCII;
            else
                return ALL_ASCII;
        }
        if (ascii > non_ascii) // mostly ascii
            return MOSTLY_ASCII;
        return MOSTLY_NONASCII;
    }

    static final boolean nonascii(int b) {
        return b >= 0177 || (b < 040 && b != '\r' && b != '\n' && b != '\t');
    }
}

/**
 * An OutputStream that determines whether the data written to
 * it is all ASCII, mostly ASCII, or mostly non-ASCII.
 */
class AsciiOutputStream extends OutputStream {
    private boolean breakOnNonAscii;
    private int ascii = 0, non_ascii = 0;
    private int linelen = 0;
    private boolean longLine = false;
    private boolean badEOL = false;
    private boolean checkEOL = false;
    private int lastb = 0;
    private int ret = 0;

    public AsciiOutputStream(boolean breakOnNonAscii, boolean encodeEolStrict) {
        this.breakOnNonAscii = breakOnNonAscii;
        checkEOL = encodeEolStrict && breakOnNonAscii;
    }

    public void write(int b) throws IOException {
        check(b);
    }

    public void write(byte b[]) throws IOException {
        write(b, 0, b.length);
    }

    public void write(byte b[], int off, int len) throws IOException {
        len += off;
        for (int i = off; i < len ; i++)
            check(b[i]);
    }

    private final void check(int b) throws IOException {
        b &= 0xff;
        if (checkEOL &&
                ((lastb == '\r' && b != '\n') || (lastb != '\r' && b == '\n')))
            badEOL = true;
        if (b == '\r' || b == '\n')
            linelen = 0;
        else {
            linelen++;
            if (linelen > 998)  // 1000 - CRLF
                longLine = true;
        }
        if (MimeUtility.nonascii(b)) { // non-ascii
            non_ascii++;
            if (breakOnNonAscii) {      // we are done
                ret = MimeUtility.MOSTLY_NONASCII;
                throw new EOFException();
            }
        } else
            ascii++;
        lastb = b;
    }

    /**
     * Return ASCII-ness of data stream.
     */
    public int getAscii() {
        if (ret != 0)
            return ret;
        // If we're looking at non-text data, and we saw CR without LF
        // or vice versa, consider this mostly non-ASCII so that it
        // will be base64 encoded (since the quoted-printable encoder
        // doesn't encode this case properly).
        if (badEOL)
            return MimeUtility.MOSTLY_NONASCII;
        else if (non_ascii == 0) { // no non-us-ascii characters so far
            // if we've seen a long line, we degrade to mostly ascii
            if (longLine)
                return MimeUtility.MOSTLY_ASCII;
            else
                return MimeUtility.ALL_ASCII;
        }
        if (ascii > non_ascii) // mostly ascii
            return MimeUtility.MOSTLY_ASCII;
        return MimeUtility.MOSTLY_NONASCII;
    }
}