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, software
013 * distributed under the License is distributed on an "AS IS" BASIS,
014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
015 * See the License for the specific language governing permissions and
016 * limitations under the License.
017 */
018
019package org.apache.hadoop.io.compress;
020
021import java.io.BufferedInputStream;
022import java.io.IOException;
023import java.io.InputStream;
024import java.io.OutputStream;
025import java.nio.charset.StandardCharsets;
026
027import org.apache.hadoop.conf.Configurable;
028import org.apache.hadoop.conf.Configuration;
029
030import org.apache.hadoop.classification.InterfaceAudience;
031import org.apache.hadoop.classification.InterfaceStability;
032import org.apache.hadoop.fs.Seekable;
033import org.apache.hadoop.io.compress.bzip2.BZip2Constants;
034import org.apache.hadoop.io.compress.bzip2.CBZip2InputStream;
035import org.apache.hadoop.io.compress.bzip2.CBZip2OutputStream;
036import org.apache.hadoop.io.compress.bzip2.Bzip2Factory;
037
038import static org.apache.hadoop.fs.CommonConfigurationKeysPublic.IO_FILE_BUFFER_SIZE_DEFAULT;
039import static org.apache.hadoop.fs.CommonConfigurationKeysPublic.IO_FILE_BUFFER_SIZE_KEY;
040
041/**
042 * This class provides output and input streams for bzip2 compression
043 * and decompression.  It uses the native bzip2 library on the system
044 * if possible, else it uses a pure-Java implementation of the bzip2
045 * algorithm.  The configuration parameter
046 * io.compression.codec.bzip2.library can be used to control this
047 * behavior.
048 *
049 * In the pure-Java mode, the Compressor and Decompressor interfaces
050 * are not implemented.  Therefore, in that mode, those methods of
051 * CompressionCodec which have a Compressor or Decompressor type
052 * argument, throw UnsupportedOperationException.
053 *
054 * Currently, support for splittability is available only in the
055 * pure-Java mode; therefore, if a SplitCompressionInputStream is
056 * requested, the pure-Java implementation is used, regardless of the
057 * setting of the configuration parameter mentioned above.
058 */
059@InterfaceAudience.Public
060@InterfaceStability.Evolving
061public class BZip2Codec implements Configurable, SplittableCompressionCodec {
062
063  private static final String HEADER = "BZ";
064  private static final int HEADER_LEN = HEADER.length();
065  private static final String SUB_HEADER = "h9";
066  private static final int SUB_HEADER_LEN = SUB_HEADER.length();
067
068  private Configuration conf;
069  
070  /**
071   * Set the configuration to be used by this object.
072   *
073   * @param conf the configuration object.
074   */
075  @Override
076  public void setConf(Configuration conf) {
077    this.conf = conf;
078  }
079  
080  /**
081   * Return the configuration used by this object.
082   *
083   * @return the configuration object used by this objec.
084   */
085  @Override
086  public Configuration getConf() {
087    return conf;
088  }
089  
090  /**
091  * Creates a new instance of BZip2Codec.
092  */
093  public BZip2Codec() { }
094
095  /**
096   * Create a {@link CompressionOutputStream} that will write to the given
097   * {@link OutputStream}.
098   *
099   * @param out        the location for the final output stream
100   * @return a stream the user can write uncompressed data to, to have it 
101   *         compressed
102   * @throws IOException
103   */
104  @Override
105  public CompressionOutputStream createOutputStream(OutputStream out)
106      throws IOException {
107    return CompressionCodec.Util.
108        createOutputStreamWithCodecPool(this, conf, out);
109  }
110
111  /**
112   * Create a {@link CompressionOutputStream} that will write to the given
113   * {@link OutputStream} with the given {@link Compressor}.
114   *
115   * @param out        the location for the final output stream
116   * @param compressor compressor to use
117   * @return a stream the user can write uncompressed data to, to have it 
118   *         compressed
119   * @throws IOException
120   */
121  @Override
122  public CompressionOutputStream createOutputStream(OutputStream out,
123      Compressor compressor) throws IOException {
124    return Bzip2Factory.isNativeBzip2Loaded(conf) ?
125      new CompressorStream(out, compressor, 
126                           conf.getInt(IO_FILE_BUFFER_SIZE_KEY,
127                                   IO_FILE_BUFFER_SIZE_DEFAULT)) :
128      new BZip2CompressionOutputStream(out);
129  }
130
131  /**
132   * Get the type of {@link Compressor} needed by this {@link CompressionCodec}.
133   *
134   * @return the type of compressor needed by this codec.
135   */
136  @Override
137  public Class<? extends Compressor> getCompressorType() {
138    return Bzip2Factory.getBzip2CompressorType(conf);
139  }
140
141  /**
142   * Create a new {@link Compressor} for use by this {@link CompressionCodec}.
143   *
144   * @return a new compressor for use by this codec
145   */
146  @Override
147  public Compressor createCompressor() {
148    return Bzip2Factory.getBzip2Compressor(conf);
149  }
150
151  /**
152   * Create a {@link CompressionInputStream} that will read from the given
153   * input stream and return a stream for uncompressed data.
154   *
155   * @param in the stream to read compressed bytes from
156   * @return a stream to read uncompressed bytes from
157   * @throws IOException
158   */
159  @Override
160  public CompressionInputStream createInputStream(InputStream in)
161      throws IOException {
162    return CompressionCodec.Util.
163        createInputStreamWithCodecPool(this, conf, in);
164  }
165
166  /**
167   * Create a {@link CompressionInputStream} that will read from the given
168   * {@link InputStream} with the given {@link Decompressor}, and return a 
169   * stream for uncompressed data.
170   *
171   * @param in           the stream to read compressed bytes from
172   * @param decompressor decompressor to use
173   * @return a stream to read uncompressed bytes from
174   * @throws IOException
175   */
176  @Override
177  public CompressionInputStream createInputStream(InputStream in,
178      Decompressor decompressor) throws IOException {
179    return Bzip2Factory.isNativeBzip2Loaded(conf) ? 
180      new DecompressorStream(in, decompressor,
181                             conf.getInt(IO_FILE_BUFFER_SIZE_KEY,
182                                 IO_FILE_BUFFER_SIZE_DEFAULT)) :
183      new BZip2CompressionInputStream(in);
184  }
185
186  /**
187   * Creates CompressionInputStream to be used to read off uncompressed data
188   * in one of the two reading modes. i.e. Continuous or Blocked reading modes
189   *
190   * @param seekableIn The InputStream
191   * @param start The start offset into the compressed stream
192   * @param end The end offset into the compressed stream
193   * @param readMode Controls whether progress is reported continuously or
194   *                 only at block boundaries.
195   *
196   * @return CompressionInputStream for BZip2 aligned at block boundaries
197   */
198  public SplitCompressionInputStream createInputStream(InputStream seekableIn,
199      Decompressor decompressor, long start, long end, READ_MODE readMode)
200      throws IOException {
201
202    if (!(seekableIn instanceof Seekable)) {
203      throw new IOException("seekableIn must be an instance of " +
204          Seekable.class.getName());
205    }
206
207    //find the position of first BZip2 start up marker
208    ((Seekable)seekableIn).seek(0);
209
210    // BZip2 start of block markers are of 6 bytes.  But the very first block
211    // also has "BZh9", making it 10 bytes.  This is the common case.  But at
212    // time stream might start without a leading BZ.
213    final long FIRST_BZIP2_BLOCK_MARKER_POSITION =
214      CBZip2InputStream.numberOfBytesTillNextMarker(seekableIn);
215    long adjStart = 0L;
216    if (start != 0) {
217      // Other than the first of file, the marker size is 6 bytes.
218      adjStart = Math.max(0L, start - (FIRST_BZIP2_BLOCK_MARKER_POSITION
219          - (HEADER_LEN + SUB_HEADER_LEN)));
220    }
221
222    ((Seekable)seekableIn).seek(adjStart);
223    SplitCompressionInputStream in =
224      new BZip2CompressionInputStream(seekableIn, adjStart, end, readMode);
225
226
227    // The following if clause handles the following case:
228    // Assume the following scenario in BZip2 compressed stream where
229    // . represent compressed data.
230    // .....[48 bit Block].....[48 bit   Block].....[48 bit Block]...
231    // ........................[47 bits][1 bit].....[48 bit Block]...
232    // ................................^[Assume a Byte alignment here]
233    // ........................................^^[current position of stream]
234    // .....................^^[We go back 10 Bytes in stream and find a Block marker]
235    // ........................................^^[We align at wrong position!]
236    // ...........................................................^^[While this pos is correct]
237
238    if (in.getPos() < start) {
239      ((Seekable)seekableIn).seek(start);
240      in = new BZip2CompressionInputStream(seekableIn, start, end, readMode);
241    }
242
243    return in;
244  }
245
246  /**
247   * Get the type of {@link Decompressor} needed by this {@link CompressionCodec}.
248   *
249   * @return the type of decompressor needed by this codec.
250   */
251  @Override
252  public Class<? extends Decompressor> getDecompressorType() {
253    return Bzip2Factory.getBzip2DecompressorType(conf);
254  }
255
256  /**
257   * Create a new {@link Decompressor} for use by this {@link CompressionCodec}.
258   *
259   * @return a new decompressor for use by this codec
260   */
261  @Override
262  public Decompressor createDecompressor() {
263    return Bzip2Factory.getBzip2Decompressor(conf);
264  }
265
266  /**
267  * .bz2 is recognized as the default extension for compressed BZip2 files
268  *
269  * @return A String telling the default bzip2 file extension
270  */
271  @Override
272  public String getDefaultExtension() {
273    return ".bz2";
274  }
275
276  private static class BZip2CompressionOutputStream extends
277      CompressionOutputStream {
278
279    // class data starts here//
280    private CBZip2OutputStream output;
281    private boolean needsReset; 
282    // class data ends here//
283
284    public BZip2CompressionOutputStream(OutputStream out)
285        throws IOException {
286      super(out);
287      needsReset = true;
288    }
289
290    private void writeStreamHeader() throws IOException {
291      if (super.out != null) {
292        // The compressed bzip2 stream should start with the
293        // identifying characters BZ. Caller of CBZip2OutputStream
294        // i.e. this class must write these characters.
295        out.write(HEADER.getBytes(StandardCharsets.UTF_8));
296      }
297    }
298
299    public void finish() throws IOException {
300      if (needsReset) {
301        // In the case that nothing is written to this stream, we still need to
302        // write out the header before closing, otherwise the stream won't be
303        // recognized by BZip2CompressionInputStream.
304        internalReset();
305      }
306      this.output.finish();
307      needsReset = true;
308    }
309
310    private void internalReset() throws IOException {
311      if (needsReset) {
312        needsReset = false;
313        writeStreamHeader();
314        this.output = new CBZip2OutputStream(out);
315      }
316    }    
317    
318    public void resetState() throws IOException {
319      // Cannot write to out at this point because out might not be ready
320      // yet, as in SequenceFile.Writer implementation.
321      needsReset = true;
322    }
323
324    public void write(int b) throws IOException {
325      if (needsReset) {
326        internalReset();
327      }
328      this.output.write(b);
329    }
330
331    public void write(byte[] b, int off, int len) throws IOException {
332      if (needsReset) {
333        internalReset();
334      }
335      this.output.write(b, off, len);
336    }
337
338    public void close() throws IOException {
339      if (needsReset) {
340        // In the case that nothing is written to this stream, we still need to
341        // write out the header before closing, otherwise the stream won't be
342        // recognized by BZip2CompressionInputStream.
343        internalReset();
344      }
345      this.output.flush();
346      this.output.close();
347      needsReset = true;
348    }
349
350  }// end of class BZip2CompressionOutputStream
351
352  /**
353   * This class is capable to de-compress BZip2 data in two modes;
354   * CONTINOUS and BYBLOCK.  BYBLOCK mode makes it possible to
355   * do decompression starting any arbitrary position in the stream.
356   *
357   * So this facility can easily be used to parallelize decompression
358   * of a large BZip2 file for performance reasons.  (It is exactly
359   * done so for Hadoop framework.  See LineRecordReader for an
360   * example).  So one can break the file (of course logically) into
361   * chunks for parallel processing.  These "splits" should be like
362   * default Hadoop splits (e.g as in FileInputFormat getSplit metod).
363   * So this code is designed and tested for FileInputFormat's way
364   * of splitting only.
365   */
366
367  private static class BZip2CompressionInputStream extends
368      SplitCompressionInputStream {
369
370    // class data starts here//
371    private CBZip2InputStream input;
372    boolean needsReset;
373    private BufferedInputStream bufferedIn;
374    private boolean isHeaderStripped = false;
375    private boolean isSubHeaderStripped = false;
376    private READ_MODE readMode = READ_MODE.CONTINUOUS;
377    private long startingPos = 0L;
378
379    // Following state machine handles different states of compressed stream
380    // position
381    // HOLD : Don't advertise compressed stream position
382    // ADVERTISE : Read 1 more character and advertise stream position
383    // See more comments about it before updatePos method.
384    private enum POS_ADVERTISEMENT_STATE_MACHINE {
385      HOLD, ADVERTISE
386    };
387
388    POS_ADVERTISEMENT_STATE_MACHINE posSM = POS_ADVERTISEMENT_STATE_MACHINE.HOLD;
389    long compressedStreamPosition = 0;
390
391    // class data ends here//
392
393    public BZip2CompressionInputStream(InputStream in) throws IOException {
394      this(in, 0L, Long.MAX_VALUE, READ_MODE.CONTINUOUS);
395    }
396
397    public BZip2CompressionInputStream(InputStream in, long start, long end,
398        READ_MODE readMode) throws IOException {
399      super(in, start, end);
400      needsReset = false;
401      bufferedIn = new BufferedInputStream(super.in);
402      this.startingPos = super.getPos();
403      this.readMode = readMode;
404      if (this.startingPos == 0) {
405        // We only strip header if it is start of file
406        bufferedIn = readStreamHeader();
407      }
408      input = new CBZip2InputStream(bufferedIn, readMode);
409      if (this.isHeaderStripped) {
410        input.updateReportedByteCount(HEADER_LEN);
411      }
412
413      if (this.isSubHeaderStripped) {
414        input.updateReportedByteCount(SUB_HEADER_LEN);
415      }
416
417      this.updatePos(false);
418    }
419
420    private BufferedInputStream readStreamHeader() throws IOException {
421      // We are flexible enough to allow the compressed stream not to
422      // start with the header of BZ. So it works fine either we have
423      // the header or not.
424      if (super.in != null) {
425        bufferedIn.mark(HEADER_LEN);
426        byte[] headerBytes = new byte[HEADER_LEN];
427        int actualRead = bufferedIn.read(headerBytes, 0, HEADER_LEN);
428        if (actualRead != -1) {
429          String header = new String(headerBytes, StandardCharsets.UTF_8);
430          if (header.compareTo(HEADER) != 0) {
431            bufferedIn.reset();
432          } else {
433            this.isHeaderStripped = true;
434            // In case of BYBLOCK mode, we also want to strip off
435            // remaining two character of the header.
436            if (this.readMode == READ_MODE.BYBLOCK) {
437              actualRead = bufferedIn.read(headerBytes, 0,
438                  SUB_HEADER_LEN);
439              if (actualRead != -1) {
440                this.isSubHeaderStripped = true;
441              }
442            }
443          }
444        }
445      }
446
447      if (bufferedIn == null) {
448        throw new IOException("Failed to read bzip2 stream.");
449      }
450
451      return bufferedIn;
452
453    }// end of method
454
455    public void close() throws IOException {
456      if (!needsReset) {
457        input.close();
458        needsReset = true;
459      }
460    }
461
462    /**
463    * This method updates compressed stream position exactly when the
464    * client of this code has read off at least one byte passed any BZip2
465    * end of block marker.
466    *
467    * This mechanism is very helpful to deal with data level record
468    * boundaries. Please see constructor and next methods of
469    * org.apache.hadoop.mapred.LineRecordReader as an example usage of this
470    * feature.  We elaborate it with an example in the following:
471    *
472    * Assume two different scenarios of the BZip2 compressed stream, where
473    * [m] represent end of block, \n is line delimiter and . represent compressed
474    * data.
475    *
476    * ............[m]......\n.......
477    *
478    * ..........\n[m]......\n.......
479    *
480    * Assume that end is right after [m].  In the first case the reading
481    * will stop at \n and there is no need to read one more line.  (To see the
482    * reason of reading one more line in the next() method is explained in LineRecordReader.)
483    * While in the second example LineRecordReader needs to read one more line
484    * (till the second \n).  Now since BZip2Codecs only update position
485    * at least one byte passed a maker, so it is straight forward to differentiate
486    * between the two cases mentioned.
487    *
488    */
489
490    public int read(byte[] b, int off, int len) throws IOException {
491      if (needsReset) {
492        internalReset();
493      }
494
495      int result = 0;
496      result = this.input.read(b, off, len);
497      if (result == BZip2Constants.END_OF_BLOCK) {
498        this.posSM = POS_ADVERTISEMENT_STATE_MACHINE.ADVERTISE;
499      }
500
501      if (this.posSM == POS_ADVERTISEMENT_STATE_MACHINE.ADVERTISE) {
502        result = this.input.read(b, off, off + 1);
503        // This is the precise time to update compressed stream position
504        // to the client of this code.
505        this.updatePos(true);
506        this.posSM = POS_ADVERTISEMENT_STATE_MACHINE.HOLD;
507      }
508
509      return result;
510
511    }
512
513    public int read() throws IOException {
514      byte b[] = new byte[1];
515      int result = this.read(b, 0, 1);
516      return (result < 0) ? result : (b[0] & 0xff);
517    }
518
519    private void internalReset() throws IOException {
520      if (needsReset) {
521        needsReset = false;
522        BufferedInputStream bufferedIn = readStreamHeader();
523        input = new CBZip2InputStream(bufferedIn, this.readMode);
524      }
525    }    
526    
527    public void resetState() throws IOException {
528      // Cannot read from bufferedIn at this point because bufferedIn
529      // might not be ready
530      // yet, as in SequenceFile.Reader implementation.
531      needsReset = true;
532    }
533
534    public long getPos() {
535      return this.compressedStreamPosition;
536      }
537
538    /*
539     * As the comments before read method tell that
540     * compressed stream is advertised when at least
541     * one byte passed EOB have been read off.  But
542     * there is an exception to this rule.  When we
543     * construct the stream we advertise the position
544     * exactly at EOB.  In the following method
545     * shouldAddOn boolean captures this exception.
546     *
547     */
548    private void updatePos(boolean shouldAddOn) {
549      int addOn = shouldAddOn ? 1 : 0;
550      this.compressedStreamPosition = this.startingPos
551          + this.input.getProcessedByteCount() + addOn;
552    }
553
554  }// end of BZip2CompressionInputStream
555
556}