001/* 002 * Licensed to the Apache Software Foundation (ASF) under one 003 * or more contributor license agreements. See the NOTICE file 004 * distributed with this work for additional information 005 * regarding copyright ownership. The ASF licenses this file 006 * to you under the Apache License, Version 2.0 (the 007 * "License"); you may not use this file except in compliance 008 * with the License. You may obtain a copy of the License at 009 * 010 * http://www.apache.org/licenses/LICENSE-2.0 011 * 012 * Unless required by applicable law or agreed to in writing, 013 * software distributed under the License is distributed on an 014 * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY 015 * KIND, either express or implied. See the License for the 016 * specific language governing permissions and limitations 017 * under the License. 018 */ 019package org.apache.commons.compress.compressors.snappy; 020 021import java.io.IOException; 022import java.io.OutputStream; 023 024import org.apache.commons.compress.compressors.CompressorOutputStream; 025import org.apache.commons.compress.compressors.lz77support.LZ77Compressor; 026import org.apache.commons.compress.compressors.lz77support.Parameters; 027import org.apache.commons.compress.utils.ByteUtils; 028 029/** 030 * CompressorOutputStream for the raw Snappy format. 031 * 032 * <p>This implementation uses an internal buffer in order to handle 033 * the back-references that are at the heart of the LZ77 algorithm. 034 * The size of the buffer must be at least as big as the biggest 035 * offset used in the compressed stream. The current version of the 036 * Snappy algorithm as defined by Google works on 32k blocks and 037 * doesn't contain offsets bigger than 32k which is the default block 038 * size used by this class.</p> 039 * 040 * <p>The raw Snappy format requires the uncompressed size to be 041 * written at the beginning of the stream using a varint 042 * representation, i.e. the number of bytes needed to write the 043 * information is not known before the uncompressed size is 044 * known. We've chosen to make the uncompressedSize a parameter of the 045 * constructor in favor of buffering the whole output until the size 046 * is known. When using the {@link FramedSnappyCompressorOutputStream} 047 * this limitation is taken care of by the warpping framing 048 * format.</p> 049 * 050 * @see <a href="https://github.com/google/snappy/blob/master/format_description.txt">Snappy compressed format description</a> 051 * @since 1.14 052 * @NotThreadSafe 053 */ 054public class SnappyCompressorOutputStream extends CompressorOutputStream { 055 private final LZ77Compressor compressor; 056 private final OutputStream os; 057 private final ByteUtils.ByteConsumer consumer; 058 059 // used in one-arg write method 060 private final byte[] oneByte = new byte[1]; 061 062 private boolean finished; 063 064 /** 065 * Constructor using the default block size of 32k. 066 * 067 * @param os the outputstream to write compressed data to 068 * @param uncompressedSize the uncompressed size of data 069 * @throws IOException if writing of the size fails 070 */ 071 public SnappyCompressorOutputStream(final OutputStream os, final long uncompressedSize) throws IOException { 072 this(os, uncompressedSize, SnappyCompressorInputStream.DEFAULT_BLOCK_SIZE); 073 } 074 075 /** 076 * Constructor using a configurable block size. 077 * 078 * @param os the outputstream to write compressed data to 079 * @param uncompressedSize the uncompressed size of data 080 * @param blockSize the block size used - must be a power of two 081 * @throws IOException if writing of the size fails 082 */ 083 public SnappyCompressorOutputStream(final OutputStream os, final long uncompressedSize, final int blockSize) 084 throws IOException { 085 this(os, uncompressedSize, createParameterBuilder(blockSize).build()); 086 } 087 088 /** 089 * Constructor providing full control over the underlying LZ77 compressor. 090 * 091 * @param os the outputstream to write compressed data to 092 * @param uncompressedSize the uncompressed size of data 093 * @param params the parameters to use by the compressor - note 094 * that the format itself imposes some limits like a maximum match 095 * length of 64 bytes 096 * @throws IOException if writing of the size fails 097 */ 098 public SnappyCompressorOutputStream(final OutputStream os, final long uncompressedSize, final Parameters params) 099 throws IOException { 100 this.os = os; 101 consumer = new ByteUtils.OutputStreamByteConsumer(os); 102 compressor = new LZ77Compressor(params, block -> { 103 switch (block.getType()) { 104 case LITERAL: 105 writeLiteralBlock((LZ77Compressor.LiteralBlock) block); 106 break; 107 case BACK_REFERENCE: 108 writeBackReference((LZ77Compressor.BackReference) block); 109 break; 110 case EOD: 111 break; 112 } 113 }); 114 writeUncompressedSize(uncompressedSize); 115 } 116 117 @Override 118 public void write(final int b) throws IOException { 119 oneByte[0] = (byte) (b & 0xff); 120 write(oneByte); 121 } 122 123 @Override 124 public void write(final byte[] data, final int off, final int len) throws IOException { 125 compressor.compress(data, off, len); 126 } 127 128 @Override 129 public void close() throws IOException { 130 try { 131 finish(); 132 } finally { 133 os.close(); 134 } 135 } 136 137 /** 138 * Compresses all remaining data and writes it to the stream, 139 * doesn't close the underlying stream. 140 * @throws IOException if an error occurs 141 */ 142 public void finish() throws IOException { 143 if (!finished) { 144 compressor.finish(); 145 finished = true; 146 } 147 } 148 149 private void writeUncompressedSize(long uncompressedSize) throws IOException { 150 boolean more = false; 151 do { 152 int currentByte = (int) (uncompressedSize & 0x7F); 153 more = uncompressedSize > currentByte; 154 if (more) { 155 currentByte |= 0x80; 156 } 157 os.write(currentByte); 158 uncompressedSize >>= 7; 159 } while (more); 160 } 161 162 // literal length is stored as (len - 1) either inside the tag 163 // (six bits minus four flags) or in 1 to 4 bytes after the tag 164 private static final int MAX_LITERAL_SIZE_WITHOUT_SIZE_BYTES = 60; 165 private static final int MAX_LITERAL_SIZE_WITH_ONE_SIZE_BYTE = 1 << 8; 166 private static final int MAX_LITERAL_SIZE_WITH_TWO_SIZE_BYTES = 1 << 16; 167 private static final int MAX_LITERAL_SIZE_WITH_THREE_SIZE_BYTES = 1 << 24; 168 169 private static final int ONE_SIZE_BYTE_MARKER = 60 << 2; 170 private static final int TWO_SIZE_BYTE_MARKER = 61 << 2; 171 private static final int THREE_SIZE_BYTE_MARKER = 62 << 2; 172 private static final int FOUR_SIZE_BYTE_MARKER = 63 << 2; 173 174 private void writeLiteralBlock(final LZ77Compressor.LiteralBlock block) throws IOException { 175 final int len = block.getLength(); 176 if (len <= MAX_LITERAL_SIZE_WITHOUT_SIZE_BYTES) { 177 writeLiteralBlockNoSizeBytes(block, len); 178 } else if (len <= MAX_LITERAL_SIZE_WITH_ONE_SIZE_BYTE) { 179 writeLiteralBlockOneSizeByte(block, len); 180 } else if (len <= MAX_LITERAL_SIZE_WITH_TWO_SIZE_BYTES) { 181 writeLiteralBlockTwoSizeBytes(block, len); 182 } else if (len <= MAX_LITERAL_SIZE_WITH_THREE_SIZE_BYTES) { 183 writeLiteralBlockThreeSizeBytes(block, len); 184 } else { 185 writeLiteralBlockFourSizeBytes(block, len); 186 } 187 } 188 189 private void writeLiteralBlockNoSizeBytes(final LZ77Compressor.LiteralBlock block, final int len) throws IOException { 190 writeLiteralBlockWithSize(len - 1 << 2, 0, len, block); 191 } 192 193 private void writeLiteralBlockOneSizeByte(final LZ77Compressor.LiteralBlock block, final int len) throws IOException { 194 writeLiteralBlockWithSize(ONE_SIZE_BYTE_MARKER, 1, len, block); 195 } 196 197 private void writeLiteralBlockTwoSizeBytes(final LZ77Compressor.LiteralBlock block, final int len) throws IOException { 198 writeLiteralBlockWithSize(TWO_SIZE_BYTE_MARKER, 2, len, block); 199 } 200 201 private void writeLiteralBlockThreeSizeBytes(final LZ77Compressor.LiteralBlock block, final int len) throws IOException { 202 writeLiteralBlockWithSize(THREE_SIZE_BYTE_MARKER, 3, len, block); 203 } 204 205 private void writeLiteralBlockFourSizeBytes(final LZ77Compressor.LiteralBlock block, final int len) throws IOException { 206 writeLiteralBlockWithSize(FOUR_SIZE_BYTE_MARKER, 4, len, block); 207 } 208 209 private void writeLiteralBlockWithSize(final int tagByte, final int sizeBytes, final int len, final LZ77Compressor.LiteralBlock block) 210 throws IOException { 211 os.write(tagByte); 212 writeLittleEndian(sizeBytes, len - 1); 213 os.write(block.getData(), block.getOffset(), len); 214 } 215 216 private void writeLittleEndian(final int numBytes, final int num) throws IOException { 217 ByteUtils.toLittleEndian(consumer, num, numBytes); 218 } 219 220 // Back-references ("copies") have their offset/size information 221 // in two, three or five bytes. 222 private static final int MIN_MATCH_LENGTH_WITH_ONE_OFFSET_BYTE = 4; 223 private static final int MAX_MATCH_LENGTH_WITH_ONE_OFFSET_BYTE = 11; 224 private static final int MAX_OFFSET_WITH_ONE_OFFSET_BYTE = 1 << 11 - 1; 225 private static final int MAX_OFFSET_WITH_TWO_OFFSET_BYTES = 1 << 16 - 1; 226 227 private static final int ONE_BYTE_COPY_TAG = 1; 228 private static final int TWO_BYTE_COPY_TAG = 2; 229 private static final int FOUR_BYTE_COPY_TAG = 3; 230 231 private void writeBackReference(final LZ77Compressor.BackReference block) throws IOException { 232 final int len = block.getLength(); 233 final int offset = block.getOffset(); 234 if (len >= MIN_MATCH_LENGTH_WITH_ONE_OFFSET_BYTE && len <= MAX_MATCH_LENGTH_WITH_ONE_OFFSET_BYTE 235 && offset <= MAX_OFFSET_WITH_ONE_OFFSET_BYTE) { 236 writeBackReferenceWithOneOffsetByte(len, offset); 237 } else if (offset < MAX_OFFSET_WITH_TWO_OFFSET_BYTES) { 238 writeBackReferenceWithTwoOffsetBytes(len, offset); 239 } else { 240 writeBackReferenceWithFourOffsetBytes(len, offset); 241 } 242 } 243 244 private void writeBackReferenceWithOneOffsetByte(final int len, final int offset) throws IOException { 245 os.write(ONE_BYTE_COPY_TAG | ((len - 4) << 2) | ((offset & 0x700) >> 3)); 246 os.write(offset & 0xff); 247 } 248 249 private void writeBackReferenceWithTwoOffsetBytes(final int len, final int offset) throws IOException { 250 writeBackReferenceWithLittleEndianOffset(TWO_BYTE_COPY_TAG, 2, len, offset); 251 } 252 253 private void writeBackReferenceWithFourOffsetBytes(final int len, final int offset) throws IOException { 254 writeBackReferenceWithLittleEndianOffset(FOUR_BYTE_COPY_TAG, 4, len, offset); 255 } 256 257 private void writeBackReferenceWithLittleEndianOffset(final int tag, final int offsetBytes, final int len, final int offset) 258 throws IOException { 259 os.write(tag | ((len - 1) << 2)); 260 writeLittleEndian(offsetBytes, offset); 261 } 262 263 // technically the format could use shorter matches but with a 264 // length of three the offset would be encoded as at least two 265 // bytes in addition to the tag, so yield no compression at all 266 private static final int MIN_MATCH_LENGTH = 4; 267 // Snappy stores the match length in six bits of the tag 268 private static final int MAX_MATCH_LENGTH = 64; 269 270 /** 271 * Returns a builder correctly configured for the Snappy algorithm using the gven block size. 272 * @param blockSize the block size. 273 * @return a builder correctly configured for the Snappy algorithm using the gven block size 274 */ 275 public static Parameters.Builder createParameterBuilder(final int blockSize) { 276 // the max offset and max literal length defined by the format 277 // are 2^32 - 1 and 2^32 respectively - with blockSize being 278 // an integer we will never exceed that 279 return Parameters.builder(blockSize) 280 .withMinBackReferenceLength(MIN_MATCH_LENGTH) 281 .withMaxBackReferenceLength(MAX_MATCH_LENGTH) 282 .withMaxOffset(blockSize) 283 .withMaxLiteralLength(blockSize); 284 } 285}