/core/src/main/java/com/twitter/elephantbird/mapreduce/input/Base64Codec.java

package com.twitter.elephantbird.mapreduce.input;

import java.util.Arrays;

import com.twitter.elephantbird.mapreduce.io.DecodeException;

/** A very fast and memory efficient class to encode and decode to and from BASE64 in full accordance
 * with RFC 2045.<br><br>
 * On Windows XP sp1 with 1.4.2_04 and later ;), this encoder and decoder is about 10 times faster
 * on small arrays (10 - 1000 bytes) and 2-3 times as fast on larger arrays (10000 - 1000000 bytes)
 * compared to <code>sun.misc.Encoder()/Decoder()</code>.<br><br>
 *
 * On byte arrays the encoder is about 20% faster than Jakarta Commons Base64 Codec for encode and
 * about 50% faster for decoding large arrays. This implementation is about twice as fast on very small
 * arrays (&lt 30 bytes). If source/destination is a <code>String</code> this
 * version is about three times as fast due to the fact that the Commons Codec result has to be recoded
 * to a <code>String</code> from <code>byte[]</code>, which is very expensive.<br><br>
 *
 * This encode/decode algorithm doesn't create any temporary arrays as many other codecs do, it only
 * allocates the resulting array. This produces less garbage and it is possible to handle arrays twice
 * as large as algorithms that create a temporary array. (E.g. Jakarta Commons Codec). It is unknown
 * whether Sun's <code>sun.misc.Encoder()/Decoder()</code> produce temporary arrays but since performance
 * is quite low it probably does.<br><br>
 *
 * The encoder produces the same output as the Sun one except that the Sun's encoder appends
 * a trailing line separator if the last character isn't a pad. Unclear why but it only adds to the
 * length and is probably a side effect. Both are in conformance with RFC 2045 though.<br>
 * Commons codec seem to always att a trailing line separator.<br><br>
 *
 * <b>Note!</b>
 * The encode/decode method pairs (types) come in three versions with the <b>exact</b> same algorithm and
 * thus a lot of code redundancy. This is to not create any temporary arrays for transcoding to/from different
 * format types. The methods not used can simply be commented out.<br><br>
 *
 * There is also a "fast" version of all decode methods that works the same way as the normal ones, but
 * har a few demands on the decoded input. Normally though, these fast verions should be used if the source if
 * the input is known and it hasn't bee tampered with.<br><br>
 *
 * If you find the code useful or you find a bug, please send me a note at base64 @ miginfocom . com.
 *
 * Licence (BSD):
 * ==============
 *
 * Copyright (c) 2004, Mikael Grev, MiG InfoCom AB. (base64 @ miginfocom . com)
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without modification,
 * are permitted provided that the following conditions are met:
 * Redistributions of source code must retain the above copyright notice, this list
 * of conditions and the following disclaimer.
 * Redistributions in binary form must reproduce the above copyright notice, this
 * list of conditions and the following disclaimer in the documentation and/or other
 * materials provided with the distribution.
 * Neither the name of the MiG InfoCom AB nor the names of its contributors may be
 * used to endorse or promote products derived from this software without specific
 * prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
 * OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
 * OF SUCH DAMAGE.
 *
 * @version 2.2
 * @author Mikael Grev
 *         Date: 2004-aug-02
 *         Time: 11:31:11
 */

public class Base64Codec
{
  private static final char[] CA = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/".toCharArray();
  private static final int[] IA = new int[256];
  static {
    Arrays.fill(IA, -1);
    for (int i = 0, iS = CA.length; i < iS; i++)
      IA[CA[i]] = i;
    IA['='] = 0;
  }

  // ****************************************************************************************
  // *  byte[] version
  // ****************************************************************************************

  /** Encodes a raw byte array into a BASE64 <code>byte[]</code> representation i accordance with RFC 2045.
   * @param sArr The bytes to convert. If <code>null</code> or length 0 an empty array will be returned.
   * @param lineSep Optional "\r\n" after 76 characters, unless end of file.<br>
   * No line separator will be in breach of RFC 2045 which specifies max 76 per line but will be a
   * little faster.
   * @return A BASE64 encoded array. Never <code>null</code>.
   */
  public final static byte[] encodeToByte(byte[] sArr, boolean lineSep)
  {
    return encodeToByte(sArr, 0, sArr != null ? sArr.length : 0, lineSep);
  }

  /** Encodes a raw byte array into a BASE64 <code>byte[]</code> representation i accordance with RFC 2045.
   * @param sArr The bytes to convert. If <code>null</code> an empty array will be returned.
   * @param sOff The starting position in the bytes to convert.
   * @param sLen The number of bytes to convert. If 0 an empty array will be returned.
   * @param lineSep Optional "\r\n" after 76 characters, unless end of file.<br>
   * No line separator will be in breach of RFC 2045 which specifies max 76 per line but will be a
   * little faster.
   * @return A BASE64 encoded array. Never <code>null</code>.
   */
  public final static byte[] encodeToByte(byte[] sArr, int sOff, int sLen, boolean lineSep)
  {
    // Check special case
    if (sArr == null || sLen == 0)
      return new byte[0];

    int eLen = (sLen / 3) * 3;                              // Length of even 24-bits.
    int cCnt = ((sLen - 1) / 3 + 1) << 2;                   // Returned character count
    int dLen = cCnt + (lineSep ? (cCnt - 1) / 76 << 1 : 0); // Length of returned array
    byte[] dArr = new byte[dLen];

    // Encode even 24-bits
    for (int s = sOff, d = 0, cc = 0; s < sOff + eLen;) {
      // Copy next three bytes into lower 24 bits of int, paying attension to sign.
      int i = (sArr[s++] & 0xff) << 16 | (sArr[s++] & 0xff) << 8 | (sArr[s++] & 0xff);

      // Encode the int into four chars
      dArr[d++] = (byte) CA[(i >>> 18) & 0x3f];
      dArr[d++] = (byte) CA[(i >>> 12) & 0x3f];
      dArr[d++] = (byte) CA[(i >>> 6) & 0x3f];
      dArr[d++] = (byte) CA[i & 0x3f];

      // Add optional line separator
      if (lineSep && ++cc == 19 && d < dLen - 2) {
        dArr[d++] = '\r';
        dArr[d++] = '\n';
        cc = 0;
      }
    }

    // Pad and encode last bits if source isn't an even 24 bits.
    int left = sLen - eLen; // 0 - 2.
    if (left > 0) {
      // Prepare the int
      int i = ((sArr[sOff + eLen] & 0xff) << 10) | (left == 2 ? ((sArr[sOff + sLen - 1] & 0xff) << 2) : 0);

      // Set last four chars
      dArr[dLen - 4] = (byte) CA[i >> 12];
      dArr[dLen - 3] = (byte) CA[(i >>> 6) & 0x3f];
      dArr[dLen - 2] = left == 2 ? (byte) CA[i & 0x3f] : (byte) '=';
      dArr[dLen - 1] = '=';
    }
    return dArr;
  }

  /** Decodes a BASE64 encoded byte array. All illegal characters will be ignored and can handle both arrays with
   * and without line separators.
   * @param sArr The source array. Length 0 will return an empty array. <code>null</code> will throw an exception.
   * @return The decoded array of bytes. May be of length 0. Will be <code>null</code> if the legal characters
   * (including '=') isn't divideable by 4. (I.e. definitely corrupted).
   */
  public final static byte[] decode(byte[] sArr) throws DecodeException
  {
    return decode(sArr, 0, sArr.length);
  }

  /** Decodes a BASE64 encoded byte array. All illegal characters will be ignored and can handle both arrays with
   * and without line separators.
   * @param sArr The source array. <code>null</code> will throw an exception.
   * @param sOff The starting position in the source array.
   * @param sLen The number of bytes to decode from the source array. Length 0 will return an empty array.
   * @return The decoded array of bytes. May be of length 0. Will be <code>null</code> if the legal characters
   * (including '=') isn't divideable by 4. (I.e. definitely corrupted).
   */
  public final static byte[] decode(byte[] sArr, int sOff, int sLen) throws DecodeException
  {
    try {
      return doDecode(sArr, sOff, sLen);
    } catch (RuntimeException e) {
      throw new DecodeException(e);
    }
  }

  private final static byte[] doDecode(byte[] sArr, int sOff, int sLen) {
    // Count illegal characters (including '\r', '\n') to know what size the returned array will be,
    // so we don't have to reallocate & copy it later.
    int sepCnt = 0; // Number of separator characters. (Actually illegal characters, but that's a bonus...)
    for (int i = 0; i < sLen; i++)      // If input is "pure" (I.e. no line separators or illegal chars) base64 this loop can be commented out.
      if (IA[sArr[sOff + i] & 0xff] < 0)
        sepCnt++;

    // Check so that legal chars (including '=') are evenly divideable by 4 as specified in RFC 2045.
    if ((sLen - sepCnt) % 4 != 0)
      return null;

    int pad = 0;
    for (int i = sLen; i > 1 && IA[sArr[sOff + --i] & 0xff] <= 0;)
      if (sArr[sOff + i] == '=')
        pad++;

    int len = ((sLen - sepCnt) * 6 >> 3) - pad;

    byte[] dArr = new byte[len];       // Preallocate byte[] of exact length

    for (int s = 0, d = 0; d < len;) {
      // Assemble three bytes into an int from four "valid" characters.
      int i = 0;
      for (int j = 0; j < 4; j++) {   // j only increased if a valid char was found.
        int c = IA[sArr[sOff + s++] & 0xff];
        if (c >= 0)
            i |= c << (18 - j * 6);
        else
          j--;
      }

      // Add the bytes
      dArr[d++] = (byte) (i >> 16);
      if (d < len) {
        dArr[d++]= (byte) (i >> 8);
        if (d < len)
          dArr[d++] = (byte) i;
      }
    }

    return dArr;
  }


  /** Decodes a BASE64 encoded byte array that is known to be resonably well formatted. The method is about twice as
   * fast as {@link #decode(byte[])}. The preconditions are:<br>
   * + The array must have a line length of 76 chars OR no line separators at all (one line).<br>
   * + Line separator must be "\r\n", as specified in RFC 2045
   * + The array must not contain illegal characters within the encoded string<br>
   * + The array CAN have illegal characters at the beginning and end, those will be dealt with appropriately.<br>
   * @param sArr The source array. Length 0 will return an empty array. <code>null</code> will throw an exception.
   * @return The decoded array of bytes. May be of length 0.
   */
  public final static byte[] decodeFast(byte[] sArr, int sLen) throws DecodeException
  {
    try {
      return doDecodeFast(sArr, sLen);
    } catch (RuntimeException e) {
      throw new DecodeException(e);
    }
  }

  private final static byte[] doDecodeFast(byte[] sArr, int sLen) {
    // Check special case
    if (sLen == 0)
      return new byte[0];

    int sIx = 0, eIx = sLen - 1;    // Start and end index after trimming.

    // Trim illegal chars from start
    while (sIx < eIx && IA[sArr[sIx] & 0xff] < 0)
      sIx++;

    // Trim illegal chars from end
    while (eIx > 0 && IA[sArr[eIx] & 0xff] < 0)
      eIx--;

    // get the padding count (=) (0, 1 or 2)
    int pad = sArr[eIx] == '=' ? (sArr[eIx - 1] == '=' ? 2 : 1) : 0;  // Count '=' at end.
    int cCnt = eIx - sIx + 1;   // Content count including possible separators
    int sepCnt = sLen > 76 ? (sArr[76] == '\r' ? cCnt / 78 : 0) << 1 : 0;

    int len = ((cCnt - sepCnt) * 6 >> 3) - pad; // The number of decoded bytes
    byte[] dArr = new byte[len];       // Preallocate byte[] of exact length

    // Decode all but the last 0 - 2 bytes.
    int d = 0;
    for (int cc = 0, eLen = (len / 3) * 3; d < eLen;) {
      // Assemble three bytes into an int from four "valid" characters.
      int i = IA[sArr[sIx++]] << 18 | IA[sArr[sIx++]] << 12 | IA[sArr[sIx++]] << 6 | IA[sArr[sIx++]];

      // Add the bytes
      dArr[d++] = (byte) (i >> 16);
      dArr[d++] = (byte) (i >> 8);
      dArr[d++] = (byte) i;

      // If line separator, jump over it.
      if (sepCnt > 0 && ++cc == 19) {
        sIx += 2;
        cc = 0;
      }
    }

    if (d < len) {
      // Decode last 1-3 bytes (incl '=') into 1-3 bytes
      int i = 0;
      for (int j = 0; sIx <= eIx - pad; j++)
        i |= IA[sArr[sIx++]] << (18 - j * 6);

      for (int r = 16; d < len; r -= 8)
        dArr[d++] = (byte) (i >> r);
    }

    return dArr;
  }

}