com.dandelion.memberapp.util.Base64.java Source code

Java tutorial

Introduction

Here is the source code for com.dandelion.memberapp.util.Base64.java

Source

package com.dandelion.memberapp.util;

/*
 * Copyright (C) 2010 The Android Open Source Project
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

import java.io.UnsupportedEncodingException;

import org.apache.commons.codec.binary.Base64OutputStream;

/**
 * Utilities for encoding and decoding the Base64 representation of binary data.
 * See RFCs <a href="http://www.ietf.org/rfc/rfc2045.txt">2045</a> and <a
 * href="http://www.ietf.org/rfc/rfc3548.txt">3548</a>.
 */
public class Base64 {
    /**
     * Default values for encoder/decoder flags.
     */
    public static final int DEFAULT = 0;

    /**
     * Encoder flag bit to omit the padding '=' characters at the end of the
     * output (if any).
     */
    public static final int NO_PADDING = 1;

    /**
     * Encoder flag bit to omit all line terminators (i.e., the output will be
     * on one long line).
     */
    public static final int NO_WRAP = 2;

    /**
     * Encoder flag bit to indicate lines should be terminated with a CRLF pair
     * instead of just an LF. Has no effect if {@code NO_WRAP} is specified as
     * well.
     */
    public static final int CRLF = 4;

    /**
     * Encoder/decoder flag bit to indicate using the "URL and filename safe"
     * variant of Base64 (see RFC 3548 section 4) where {@code -} and {@code _}
     * are used in place of {@code +} and {@code /}.
     */
    public static final int URL_SAFE = 8;

    /**
     * Flag to pass to {@link Base64OutputStream} to indicate that it should not
     * close the output stream it is wrapping when it itself is closed.
     */
    public static final int NO_CLOSE = 16;

    // --------------------------------------------------------
    // shared code
    // --------------------------------------------------------

    /* package */static abstract class Coder {
        public byte[] output;
        public int op;

        /**
         * Encode/decode another block of input data. this.output is provided by
         * the caller, and must be big enough to hold all the coded data. On
         * exit, this.opwill be set to the length of the coded data.
         * 
         * @param finish
         *            true if this is the final call to process for this object.
         *            Will finalize the coder state and include any final bytes
         *            in the output.
         * 
         * @return true if the input so far is good; false if some error has
         *         been detected in the input stream..
         */
        public abstract boolean process(byte[] input, int offset, int len, boolean finish);

        /**
         * @return the maximum number of bytes a call to process() could produce
         *         for the given number of input bytes. This may be an
         *         overestimate.
         */
        public abstract int maxOutputSize(int len);
    }

    // --------------------------------------------------------
    // decoding
    // --------------------------------------------------------

    /**
     * Decode the Base64-encoded data in input and return the data in a new byte
     * array.
     * 
     * <p>
     * The padding '=' characters at the end are considered optional, but if any
     * are present, there must be the correct number of them.
     * 
     * @param str
     *            the input String to decode, which is converted to bytes using
     *            the default charset
     * @param flags
     *            controls certain features of the decoded output. Pass
     *            {@code DEFAULT} to decode standard Base64.
     * 
     * @throws IllegalArgumentException
     *             if the input contains incorrect padding
     */
    public static byte[] decode(String str, int flags) {
        return decode(str.getBytes(), flags);
    }

    public static String decodeToString(String str, int flags, String charset) {
        try {
            return new String(decode(str, flags), charset);
        } catch (UnsupportedEncodingException e) {
            e.printStackTrace();
            throw new RuntimeException(e);
        }
    }

    /**
     * Decode the Base64-encoded data in input and return the data in a new byte
     * array.
     * 
     * <p>
     * The padding '=' characters at the end are considered optional, but if any
     * are present, there must be the correct number of them.
     * 
     * @param input
     *            the input array to decode
     * @param flags
     *            controls certain features of the decoded output. Pass
     *            {@code DEFAULT} to decode standard Base64.
     * 
     * @throws IllegalArgumentException
     *             if the input contains incorrect padding
     */
    public static byte[] decode(byte[] input, int flags) {
        return decode(input, 0, input.length, flags);
    }

    /**
     * Decode the Base64-encoded data in input and return the data in a new byte
     * array.
     * 
     * <p>
     * The padding '=' characters at the end are considered optional, but if any
     * are present, there must be the correct number of them.
     * 
     * @param input
     *            the data to decode
     * @param offset
     *            the position within the input array at which to start
     * @param len
     *            the number of bytes of input to decode
     * @param flags
     *            controls certain features of the decoded output. Pass
     *            {@code DEFAULT} to decode standard Base64.
     * 
     * @throws IllegalArgumentException
     *             if the input contains incorrect padding
     */
    public static byte[] decode(byte[] input, int offset, int len, int flags) {
        // Allocate space for the most data the input could represent.
        // (It could contain less if it contains whitespace, etc.)
        Decoder decoder = new Decoder(flags, new byte[len * 3 / 4]);

        if (!decoder.process(input, offset, len, true)) {
            throw new IllegalArgumentException("bad base-64");
        }

        // Maybe we got lucky and allocated exactly enough output space.
        if (decoder.op == decoder.output.length) {
            return decoder.output;
        }

        // Need to shorten the array, so allocate a new one of the
        // right size and copy.
        byte[] temp = new byte[decoder.op];
        System.arraycopy(decoder.output, 0, temp, 0, decoder.op);
        return temp;
    }

    /* package */static class Decoder extends Coder {
        /**
         * Lookup table for turning bytes into their position in the Base64
         * alphabet.
         */
        private static final int DECODE[] = { -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, 62, -1, -1, -1, 63, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, -1, -1, -1, -2, -1, -1, -1, 0, 1, 2,
                3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, -1, -1, -1, -1,
                -1, -1, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48,
                49, 50, 51, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, };

        /**
         * Decode lookup table for the "web safe" variant (RFC 3548 sec. 4)
         * where - and _ replace + and /.
         */
        private static final int DECODE_WEBSAFE[] = { -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, 62, -1, -1, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, -1, -1, -1, -2, -1, -1, -1,
                0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, -1,
                -1, -1, -1, 63, -1, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45,
                46, 47, 48, 49, 50, 51, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
                -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, };

        /** Non-data values in the DECODE arrays. */
        private static final int SKIP = -1;
        private static final int EQUALS = -2;

        /**
         * States 0-3 are reading through the next input tuple. State 4 is
         * having read one '=' and expecting exactly one more. State 5 is
         * expecting no more data or padding characters in the input. State 6 is
         * the error state; an error has been detected in the input and no
         * future input can "fix" it.
         */
        private int state; // state number (0 to 6)
        private int value;

        final private int[] alphabet;

        public Decoder(int flags, byte[] output) {
            this.output = output;

            alphabet = ((flags & URL_SAFE) == 0) ? DECODE : DECODE_WEBSAFE;
            state = 0;
            value = 0;
        }

        /**
         * @return an overestimate for the number of bytes {@code len} bytes
         *         could decode to.
         */
        public int maxOutputSize(int len) {
            return len * 3 / 4 + 10;
        }

        /**
         * Decode another block of input data.
         * 
         * @return true if the state machine is still healthy. false if bad
         *         base-64 data has been detected in the input stream.
         */
        public boolean process(byte[] input, int offset, int len, boolean finish) {
            if (this.state == 6)
                return false;

            int p = offset;
            len += offset;

            // Using local variables makes the decoder about 12%
            // faster than if we manipulate the member variables in
            // the loop. (Even alphabet makes a measurable
            // difference, which is somewhat surprising to me since
            // the member variable is final.)
            int state = this.state;
            int value = this.value;
            int op = 0;
            final byte[] output = this.output;
            final int[] alphabet = this.alphabet;

            while (p < len) {
                // Try the fast path: we're starting a new tuple and the
                // next four bytes of the input stream are all data
                // bytes. This corresponds to going through states
                // 0-1-2-3-0. We expect to use this method for most of
                // the data.
                //
                // If any of the next four bytes of input are non-data
                // (whitespace, etc.), value will end up negative. (All
                // the non-data values in decode are small negative
                // numbers, so shifting any of them up and or'ing them
                // together will result in a value with its top bit set.)
                //
                // You can remove this whole block and the output should
                // be the same, just slower.
                if (state == 0) {
                    while (p + 4 <= len && (value = ((alphabet[input[p] & 0xff] << 18)
                            | (alphabet[input[p + 1] & 0xff] << 12) | (alphabet[input[p + 2] & 0xff] << 6)
                            | (alphabet[input[p + 3] & 0xff]))) >= 0) {
                        output[op + 2] = (byte) value;
                        output[op + 1] = (byte) (value >> 8);
                        output[op] = (byte) (value >> 16);
                        op += 3;
                        p += 4;
                    }
                    if (p >= len)
                        break;
                }

                // The fast path isn't available -- either we've read a
                // partial tuple, or the next four input bytes aren't all
                // data, or whatever. Fall back to the slower state
                // machine implementation.

                int d = alphabet[input[p++] & 0xff];

                switch (state) {
                case 0:
                    if (d >= 0) {
                        value = d;
                        ++state;
                    } else if (d != SKIP) {
                        this.state = 6;
                        return false;
                    }
                    break;

                case 1:
                    if (d >= 0) {
                        value = (value << 6) | d;
                        ++state;
                    } else if (d != SKIP) {
                        this.state = 6;
                        return false;
                    }
                    break;

                case 2:
                    if (d >= 0) {
                        value = (value << 6) | d;
                        ++state;
                    } else if (d == EQUALS) {
                        // Emit the last (partial) output tuple;
                        // expect exactly one more padding character.
                        output[op++] = (byte) (value >> 4);
                        state = 4;
                    } else if (d != SKIP) {
                        this.state = 6;
                        return false;
                    }
                    break;

                case 3:
                    if (d >= 0) {
                        // Emit the output triple and return to state 0.
                        value = (value << 6) | d;
                        output[op + 2] = (byte) value;
                        output[op + 1] = (byte) (value >> 8);
                        output[op] = (byte) (value >> 16);
                        op += 3;
                        state = 0;
                    } else if (d == EQUALS) {
                        // Emit the last (partial) output tuple;
                        // expect no further data or padding characters.
                        output[op + 1] = (byte) (value >> 2);
                        output[op] = (byte) (value >> 10);
                        op += 2;
                        state = 5;
                    } else if (d != SKIP) {
                        this.state = 6;
                        return false;
                    }
                    break;

                case 4:
                    if (d == EQUALS) {
                        ++state;
                    } else if (d != SKIP) {
                        this.state = 6;
                        return false;
                    }
                    break;

                case 5:
                    if (d != SKIP) {
                        this.state = 6;
                        return false;
                    }
                    break;
                }
            }

            if (!finish) {
                // We're out of input, but a future call could provide
                // more.
                this.state = state;
                this.value = value;
                this.op = op;
                return true;
            }

            // Done reading input. Now figure out where we are left in
            // the state machine and finish up.

            switch (state) {
            case 0:
                // Output length is a multiple of three. Fine.
                break;
            case 1:
                // Read one extra input byte, which isn't enough to
                // make another output byte. Illegal.
                this.state = 6;
                return false;
            case 2:
                // Read two extra input bytes, enough to emit 1 more
                // output byte. Fine.
                output[op++] = (byte) (value >> 4);
                break;
            case 3:
                // Read three extra input bytes, enough to emit 2 more
                // output bytes. Fine.
                output[op++] = (byte) (value >> 10);
                output[op++] = (byte) (value >> 2);
                break;
            case 4:
                // Read one padding '=' when we expected 2. Illegal.
                this.state = 6;
                return false;
            case 5:
                // Read all the padding '='s we expected and no more.
                // Fine.
                break;
            }

            this.state = state;
            this.op = op;
            return true;
        }
    }

    // --------------------------------------------------------
    // encoding
    // --------------------------------------------------------

    /**
     * Base64-encode the given data and return a newly allocated String with the
     * result.
     * 
     * @param input
     *            the data to encode
     * @param flags
     *            controls certain features of the encoded output. Passing
     *            {@code DEFAULT} results in output that adheres to RFC 2045.
     */
    public static String encodeToString(byte[] input, int flags) {
        try {
            return new String(encode(input, flags), "US-ASCII");
        } catch (UnsupportedEncodingException e) {
            // US-ASCII is guaranteed to be available.
            throw new AssertionError(e);
        }
    }

    public static String encodeToString(String input, int flags, String charset) {
        try {
            return encodeToString(input.getBytes(charset), flags);
        } catch (UnsupportedEncodingException e) {
            e.printStackTrace();
            throw new RuntimeException(e);
        }
    }

    /**
     * Base64-encode the given data and return a newly allocated String with the
     * result.
     * 
     * @param input
     *            the data to encode
     * @param offset
     *            the position within the input array at which to start
     * @param len
     *            the number of bytes of input to encode
     * @param flags
     *            controls certain features of the encoded output. Passing
     *            {@code DEFAULT} results in output that adheres to RFC 2045.
     */
    public static String encodeToString(byte[] input, int offset, int len, int flags) {
        try {
            return new String(encode(input, offset, len, flags), "US-ASCII");
        } catch (UnsupportedEncodingException e) {
            // US-ASCII is guaranteed to be available.
            throw new AssertionError(e);
        }
    }

    /**
     * Base64-encode the given data and return a newly allocated byte[] with the
     * result.
     * 
     * @param input
     *            the data to encode
     * @param flags
     *            controls certain features of the encoded output. Passing
     *            {@code DEFAULT} results in output that adheres to RFC 2045.
     */
    public static byte[] encode(byte[] input, int flags) {
        return encode(input, 0, input.length, flags);
    }

    /**
     * Base64-encode the given data and return a newly allocated byte[] with the
     * result.
     * 
     * @param input
     *            the data to encode
     * @param offset
     *            the position within the input array at which to start
     * @param len
     *            the number of bytes of input to encode
     * @param flags
     *            controls certain features of the encoded output. Passing
     *            {@code DEFAULT} results in output that adheres to RFC 2045.
     */
    public static byte[] encode(byte[] input, int offset, int len, int flags) {
        Encoder encoder = new Encoder(flags, null);

        // Compute the exact length of the array we will produce.
        int output_len = len / 3 * 4;

        // Account for the tail of the data and the padding bytes, if any.
        if (encoder.do_padding) {
            if (len % 3 > 0) {
                output_len += 4;
            }
        } else {
            switch (len % 3) {
            case 0:
                break;
            case 1:
                output_len += 2;
                break;
            case 2:
                output_len += 3;
                break;
            }
        }

        // Account for the newlines, if any.
        if (encoder.do_newline && len > 0) {
            output_len += (((len - 1) / (3 * Encoder.LINE_GROUPS)) + 1) * (encoder.do_cr ? 2 : 1);
        }

        encoder.output = new byte[output_len];
        encoder.process(input, offset, len, true);

        assert encoder.op == output_len;

        return encoder.output;
    }

    /* package */static class Encoder extends Coder {
        /**
         * Emit a new line every this many output tuples. Corresponds to a
         * 76-character line length (the maximum allowable according to <a
         * href="http://www.ietf.org/rfc/rfc2045.txt">RFC 2045</a>).
         */
        public static final int LINE_GROUPS = 19;

        /**
         * Lookup table for turning Base64 alphabet positions (6 bits) into
         * output bytes.
         */
        private static final byte ENCODE[] = { 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N',
                'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h',
                'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z', '0', '1',
                '2', '3', '4', '5', '6', '7', '8', '9', '+', '/', };

        /**
         * Lookup table for turning Base64 alphabet positions (6 bits) into
         * output bytes.
         */
        private static final byte ENCODE_WEBSAFE[] = { 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L',
                'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f',
                'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z',
                '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '-', '_', };

        final private byte[] tail;
        /* package */int tailLen;
        private int count;

        final public boolean do_padding;
        final public boolean do_newline;
        final public boolean do_cr;
        final private byte[] alphabet;

        public Encoder(int flags, byte[] output) {
            this.output = output;

            do_padding = (flags & NO_PADDING) == 0;
            do_newline = (flags & NO_WRAP) == 0;
            do_cr = (flags & CRLF) != 0;
            alphabet = ((flags & URL_SAFE) == 0) ? ENCODE : ENCODE_WEBSAFE;

            tail = new byte[2];
            tailLen = 0;

            count = do_newline ? LINE_GROUPS : -1;
        }

        /**
         * @return an overestimate for the number of bytes {@code len} bytes
         *         could encode to.
         */
        public int maxOutputSize(int len) {
            return len * 8 / 5 + 10;
        }

        public boolean process(byte[] input, int offset, int len, boolean finish) {
            // Using local variables makes the encoder about 9% faster.
            final byte[] alphabet = this.alphabet;
            final byte[] output = this.output;
            int op = 0;
            int count = this.count;

            int p = offset;
            len += offset;
            int v = -1;

            // First we need to concatenate the tail of the previous call
            // with any input bytes available now and see if we can empty
            // the tail.

            switch (tailLen) {
            case 0:
                // There was no tail.
                break;

            case 1:
                if (p + 2 <= len) {
                    // A 1-byte tail with at least 2 bytes of
                    // input available now.
                    v = ((tail[0] & 0xff) << 16) | ((input[p++] & 0xff) << 8) | (input[p++] & 0xff);
                    tailLen = 0;
                }
                ;
                break;

            case 2:
                if (p + 1 <= len) {
                    // A 2-byte tail with at least 1 byte of input.
                    v = ((tail[0] & 0xff) << 16) | ((tail[1] & 0xff) << 8) | (input[p++] & 0xff);
                    tailLen = 0;
                }
                break;
            }

            if (v != -1) {
                output[op++] = alphabet[(v >> 18) & 0x3f];
                output[op++] = alphabet[(v >> 12) & 0x3f];
                output[op++] = alphabet[(v >> 6) & 0x3f];
                output[op++] = alphabet[v & 0x3f];
                if (--count == 0) {
                    if (do_cr)
                        output[op++] = '\r';
                    output[op++] = '\n';
                    count = LINE_GROUPS;
                }
            }

            // At this point either there is no tail, or there are fewer
            // than 3 bytes of input available.

            // The main loop, turning 3 input bytes into 4 output bytes on
            // each iteration.
            while (p + 3 <= len) {
                v = ((input[p] & 0xff) << 16) | ((input[p + 1] & 0xff) << 8) | (input[p + 2] & 0xff);
                output[op] = alphabet[(v >> 18) & 0x3f];
                output[op + 1] = alphabet[(v >> 12) & 0x3f];
                output[op + 2] = alphabet[(v >> 6) & 0x3f];
                output[op + 3] = alphabet[v & 0x3f];
                p += 3;
                op += 4;
                if (--count == 0) {
                    if (do_cr)
                        output[op++] = '\r';
                    output[op++] = '\n';
                    count = LINE_GROUPS;
                }
            }

            if (finish) {
                // Finish up the tail of the input. Note that we need to
                // consume any bytes in tail before any bytes
                // remaining in input; there should be at most two bytes
                // total.

                if (p - tailLen == len - 1) {
                    int t = 0;
                    v = ((tailLen > 0 ? tail[t++] : input[p++]) & 0xff) << 4;
                    tailLen -= t;
                    output[op++] = alphabet[(v >> 6) & 0x3f];
                    output[op++] = alphabet[v & 0x3f];
                    if (do_padding) {
                        output[op++] = '=';
                        output[op++] = '=';
                    }
                    if (do_newline) {
                        if (do_cr)
                            output[op++] = '\r';
                        output[op++] = '\n';
                    }
                } else if (p - tailLen == len - 2) {
                    int t = 0;
                    v = (((tailLen > 1 ? tail[t++] : input[p++]) & 0xff) << 10)
                            | (((tailLen > 0 ? tail[t++] : input[p++]) & 0xff) << 2);
                    tailLen -= t;
                    output[op++] = alphabet[(v >> 12) & 0x3f];
                    output[op++] = alphabet[(v >> 6) & 0x3f];
                    output[op++] = alphabet[v & 0x3f];
                    if (do_padding) {
                        output[op++] = '=';
                    }
                    if (do_newline) {
                        if (do_cr)
                            output[op++] = '\r';
                        output[op++] = '\n';
                    }
                } else if (do_newline && op > 0 && count != LINE_GROUPS) {
                    if (do_cr)
                        output[op++] = '\r';
                    output[op++] = '\n';
                }

                assert tailLen == 0;
                assert p == len;
            } else {
                // Save the leftovers in tail to be consumed on the next
                // call to encodeInternal.

                if (p == len - 1) {
                    tail[tailLen++] = input[p];
                } else if (p == len - 2) {
                    tail[tailLen++] = input[p];
                    tail[tailLen++] = input[p + 1];
                }
            }

            this.op = op;
            this.count = count;

            return true;
        }
    }

    private Base64() {
    } // don't instantiate
}