001/* Deflater.java - Compress a data stream
002   Copyright (C) 1999, 2000, 2001, 2004, 2005 Free Software Foundation, Inc.
003
004This file is part of GNU Classpath.
005
006GNU Classpath is free software; you can redistribute it and/or modify
007it under the terms of the GNU General Public License as published by
008the Free Software Foundation; either version 2, or (at your option)
009any later version.
010
011GNU Classpath is distributed in the hope that it will be useful, but
012WITHOUT ANY WARRANTY; without even the implied warranty of
013MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
014General Public License for more details.
015
016You should have received a copy of the GNU General Public License
017along with GNU Classpath; see the file COPYING.  If not, write to the
018Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
01902110-1301 USA.
020
021Linking this library statically or dynamically with other modules is
022making a combined work based on this library.  Thus, the terms and
023conditions of the GNU General Public License cover the whole
024combination.
025
026As a special exception, the copyright holders of this library give you
027permission to link this library with independent modules to produce an
028executable, regardless of the license terms of these independent
029modules, and to copy and distribute the resulting executable under
030terms of your choice, provided that you also meet, for each linked
031independent module, the terms and conditions of the license of that
032module.  An independent module is a module which is not derived from
033or based on this library.  If you modify this library, you may extend
034this exception to your version of the library, but you are not
035obligated to do so.  If you do not wish to do so, delete this
036exception statement from your version. */
037
038package java.util.zip;
039
040/**
041 * This is the Deflater class.  The deflater class compresses input
042 * with the deflate algorithm described in RFC 1951.  It has several
043 * compression levels and three different strategies described below.
044 *
045 * This class is <i>not</i> thread safe.  This is inherent in the API, due
046 * to the split of deflate and setInput.
047 *
048 * @author Jochen Hoenicke
049 * @author Tom Tromey
050 */
051public class Deflater
052{
053  /**
054   * The best and slowest compression level.  This tries to find very
055   * long and distant string repetitions.
056   */
057  public static final int BEST_COMPRESSION = 9;
058  /**
059   * The worst but fastest compression level.
060   */
061  public static final int BEST_SPEED = 1;
062  /**
063   * The default compression level.
064   */
065  public static final int DEFAULT_COMPRESSION = -1;
066  /**
067   * This level won't compress at all but output uncompressed blocks.
068   */
069  public static final int NO_COMPRESSION = 0;
070
071  /**
072   * The default strategy.
073   */
074  public static final int DEFAULT_STRATEGY = 0;
075  /**
076   * This strategy will only allow longer string repetitions.  It is
077   * useful for random data with a small character set.
078   */
079  public static final int FILTERED = 1;
080
081  /**
082   * This strategy will not look for string repetitions at all.  It
083   * only encodes with Huffman trees (which means, that more common
084   * characters get a smaller encoding.
085   */
086  public static final int HUFFMAN_ONLY = 2;
087
088  /**
089   * The compression method.  This is the only method supported so far.
090   * There is no need to use this constant at all.
091   */
092  public static final int DEFLATED = 8;
093
094  /*
095   * The Deflater can do the following state transitions:
096   *
097   * (1) -> INIT_STATE   ----> INIT_FINISHING_STATE ---.
098   *        /  | (2)      (5)                         |
099   *       /   v          (5)                         |
100   *   (3)| SETDICT_STATE ---> SETDICT_FINISHING_STATE |(3)
101   *       \   | (3)                 |        ,-------'
102   *        |  |                     | (3)   /
103   *        v  v          (5)        v      v
104   * (1) -> BUSY_STATE   ----> FINISHING_STATE
105   *                                | (6)
106   *                                v
107   *                           FINISHED_STATE
108   *    \_____________________________________/
109   *          | (7)
110   *          v
111   *        CLOSED_STATE
112   *
113   * (1) If we should produce a header we start in INIT_STATE, otherwise
114   *     we start in BUSY_STATE.
115   * (2) A dictionary may be set only when we are in INIT_STATE, then
116   *     we change the state as indicated.
117   * (3) Whether a dictionary is set or not, on the first call of deflate
118   *     we change to BUSY_STATE.
119   * (4) -- intentionally left blank -- :)
120   * (5) FINISHING_STATE is entered, when flush() is called to indicate that
121   *     there is no more INPUT.  There are also states indicating, that
122   *     the header wasn't written yet.
123   * (6) FINISHED_STATE is entered, when everything has been flushed to the
124   *     internal pending output buffer.
125   * (7) At any time (7)
126   *
127   */
128
129  private static final int IS_SETDICT              = 0x01;
130  private static final int IS_FLUSHING             = 0x04;
131  private static final int IS_FINISHING            = 0x08;
132
133  private static final int INIT_STATE              = 0x00;
134  private static final int SETDICT_STATE           = 0x01;
135  private static final int INIT_FINISHING_STATE    = 0x08;
136  private static final int SETDICT_FINISHING_STATE = 0x09;
137  private static final int BUSY_STATE              = 0x10;
138  private static final int FLUSHING_STATE          = 0x14;
139  private static final int FINISHING_STATE         = 0x1c;
140  private static final int FINISHED_STATE          = 0x1e;
141  private static final int CLOSED_STATE            = 0x7f;
142
143  /** Compression level. */
144  private int level;
145
146  /** should we include a header. */
147  private boolean noHeader;
148
149  /** The current state. */
150  private int state;
151
152  /** The total bytes of output written. */
153  private long totalOut;
154
155  /** The pending output. */
156  private DeflaterPending pending;
157
158  /** The deflater engine. */
159  private DeflaterEngine engine;
160
161  /**
162   * Creates a new deflater with default compression level.
163   */
164  public Deflater()
165  {
166    this(DEFAULT_COMPRESSION, false);
167  }
168
169  /**
170   * Creates a new deflater with given compression level.
171   * @param lvl the compression level, a value between NO_COMPRESSION
172   * and BEST_COMPRESSION, or DEFAULT_COMPRESSION.
173   * @exception IllegalArgumentException if lvl is out of range.
174   */
175  public Deflater(int lvl)
176  {
177    this(lvl, false);
178  }
179
180  /**
181   * Creates a new deflater with given compression level.
182   * @param lvl the compression level, a value between NO_COMPRESSION
183   * and BEST_COMPRESSION.
184   * @param nowrap true, iff we should suppress the deflate header at the
185   * beginning and the adler checksum at the end of the output.  This is
186   * useful for the GZIP format.
187   * @exception IllegalArgumentException if lvl is out of range.
188   */
189  public Deflater(int lvl, boolean nowrap)
190  {
191    if (lvl == DEFAULT_COMPRESSION)
192      lvl = 6;
193    else if (lvl < NO_COMPRESSION || lvl > BEST_COMPRESSION)
194      throw new IllegalArgumentException();
195
196    pending = new DeflaterPending();
197    engine = new DeflaterEngine(pending);
198    this.noHeader = nowrap;
199    setStrategy(DEFAULT_STRATEGY);
200    setLevel(lvl);
201    reset();
202  }
203
204  /**
205   * Resets the deflater.  The deflater acts afterwards as if it was
206   * just created with the same compression level and strategy as it
207   * had before.
208   */
209  public void reset()
210  {
211    state = (noHeader ? BUSY_STATE : INIT_STATE);
212    totalOut = 0;
213    pending.reset();
214    engine.reset();
215  }
216
217  /**
218   * Frees all objects allocated by the compressor.  There's no
219   * reason to call this, since you can just rely on garbage
220   * collection.  Exists only for compatibility against Sun's JDK,
221   * where the compressor allocates native memory.
222   * If you call any method (even reset) afterwards the behaviour is
223   * <i>undefined</i>.
224   */
225  public void end()
226  {
227    engine = null;
228    pending = null;
229    state = CLOSED_STATE;
230  }
231
232  /**
233   * Gets the current adler checksum of the data that was processed so
234   * far.
235   */
236  public int getAdler()
237  {
238    return engine.getAdler();
239  }
240
241  /**
242   * Gets the number of input bytes processed so far.
243   */
244  public int getTotalIn()
245  {
246    return (int) engine.getTotalIn();
247  }
248
249  /**
250   * Gets the number of input bytes processed so far.
251   * @since 1.5
252   */
253  public long getBytesRead()
254  {
255    return engine.getTotalIn();
256  }
257
258  /**
259   * Gets the number of output bytes so far.
260   */
261  public int getTotalOut()
262  {
263    return (int) totalOut;
264  }
265
266  /**
267   * Gets the number of output bytes so far.
268   * @since 1.5
269   */
270  public long getBytesWritten()
271  {
272    return totalOut;
273  }
274
275  /**
276   * Finalizes this object.
277   */
278  protected void finalize()
279  {
280    /* Exists solely for compatibility.  We don't have any native state. */
281  }
282
283  /**
284   * Flushes the current input block.  Further calls to deflate() will
285   * produce enough output to inflate everything in the current input
286   * block.  This is not part of Sun's JDK so I have made it package
287   * private.  It is used by DeflaterOutputStream to implement
288   * flush().
289   */
290  void flush() {
291    state |= IS_FLUSHING;
292  }
293
294  /**
295   * Finishes the deflater with the current input block.  It is an error
296   * to give more input after this method was called.  This method must
297   * be called to force all bytes to be flushed.
298   */
299  public void finish() {
300    state |= IS_FLUSHING | IS_FINISHING;
301  }
302
303  /**
304   * Returns true iff the stream was finished and no more output bytes
305   * are available.
306   */
307  public boolean finished()
308  {
309    return state == FINISHED_STATE && pending.isFlushed();
310  }
311
312  /**
313   * Returns true, if the input buffer is empty.
314   * You should then call setInput(). <br>
315   *
316   * <em>NOTE</em>: This method can also return true when the stream
317   * was finished.
318   */
319  public boolean needsInput()
320  {
321    return engine.needsInput();
322  }
323
324  /**
325   * Sets the data which should be compressed next.  This should be only
326   * called when needsInput indicates that more input is needed.
327   * If you call setInput when needsInput() returns false, the
328   * previous input that is still pending will be thrown away.
329   * The given byte array should not be changed, before needsInput() returns
330   * true again.
331   * This call is equivalent to <code>setInput(input, 0, input.length)</code>.
332   * @param input the buffer containing the input data.
333   * @exception IllegalStateException if the buffer was finished() or ended().
334   */
335  public void setInput(byte[] input)
336  {
337    setInput(input, 0, input.length);
338  }
339
340  /**
341   * Sets the data which should be compressed next.  This should be
342   * only called when needsInput indicates that more input is needed.
343   * The given byte array should not be changed, before needsInput() returns
344   * true again.
345   * @param input the buffer containing the input data.
346   * @param off the start of the data.
347   * @param len the length of the data.
348   * @exception IllegalStateException if the buffer was finished() or ended()
349   * or if previous input is still pending.
350   */
351  public void setInput(byte[] input, int off, int len)
352  {
353    if ((state & IS_FINISHING) != 0)
354      throw new IllegalStateException("finish()/end() already called");
355    engine.setInput(input, off, len);
356  }
357
358  /**
359   * Sets the compression level.  There is no guarantee of the exact
360   * position of the change, but if you call this when needsInput is
361   * true the change of compression level will occur somewhere near
362   * before the end of the so far given input.
363   * @param lvl the new compression level.
364   */
365  public void setLevel(int lvl)
366  {
367    if (lvl == DEFAULT_COMPRESSION)
368      lvl = 6;
369    else if (lvl < NO_COMPRESSION || lvl > BEST_COMPRESSION)
370      throw new IllegalArgumentException();
371
372
373    if (level != lvl)
374      {
375        level = lvl;
376        engine.setLevel(lvl);
377      }
378  }
379
380  /**
381   * Sets the compression strategy. Strategy is one of
382   * DEFAULT_STRATEGY, HUFFMAN_ONLY and FILTERED.  For the exact
383   * position where the strategy is changed, the same as for
384   * setLevel() applies.
385   * @param stgy the new compression strategy.
386   */
387  public void setStrategy(int stgy)
388  {
389    if (stgy != DEFAULT_STRATEGY && stgy != FILTERED
390        && stgy != HUFFMAN_ONLY)
391      throw new IllegalArgumentException();
392    engine.setStrategy(stgy);
393  }
394
395  /**
396   * Deflates the current input block to the given array.  It returns
397   * the number of bytes compressed, or 0 if either
398   * needsInput() or finished() returns true or length is zero.
399   * @param output the buffer where to write the compressed data.
400   */
401  public int deflate(byte[] output)
402  {
403    return deflate(output, 0, output.length);
404  }
405
406  /**
407   * Deflates the current input block to the given array.  It returns
408   * the number of bytes compressed, or 0 if either
409   * needsInput() or finished() returns true or length is zero.
410   * @param output the buffer where to write the compressed data.
411   * @param offset the offset into the output array.
412   * @param length the maximum number of bytes that may be written.
413   * @exception IllegalStateException if end() was called.
414   * @exception IndexOutOfBoundsException if offset and/or length
415   * don't match the array length.
416   */
417  public int deflate(byte[] output, int offset, int length)
418  {
419    int origLength = length;
420
421    if (state == CLOSED_STATE)
422      throw new IllegalStateException("Deflater closed");
423
424    if (state < BUSY_STATE)
425      {
426        /* output header */
427        int header = (DEFLATED +
428                      ((DeflaterConstants.MAX_WBITS - 8) << 4)) << 8;
429        int level_flags = (level - 1) >> 1;
430        if (level_flags < 0 || level_flags > 3)
431          level_flags = 3;
432        header |= level_flags << 6;
433        if ((state & IS_SETDICT) != 0)
434          /* Dictionary was set */
435          header |= DeflaterConstants.PRESET_DICT;
436        header += 31 - (header % 31);
437
438        pending.writeShortMSB(header);
439        if ((state & IS_SETDICT) != 0)
440          {
441            int chksum = engine.getAdler();
442            engine.resetAdler();
443            pending.writeShortMSB(chksum >> 16);
444            pending.writeShortMSB(chksum & 0xffff);
445          }
446
447        state = BUSY_STATE | (state & (IS_FLUSHING | IS_FINISHING));
448      }
449
450    for (;;)
451      {
452        int count = pending.flush(output, offset, length);
453        offset += count;
454        totalOut += count;
455        length -= count;
456        if (length == 0 || state == FINISHED_STATE)
457          break;
458
459        if (!engine.deflate((state & IS_FLUSHING) != 0,
460                            (state & IS_FINISHING) != 0))
461          {
462            if (state == BUSY_STATE)
463              /* We need more input now */
464              return origLength - length;
465            else if (state == FLUSHING_STATE)
466              {
467                if (level != NO_COMPRESSION)
468                  {
469                    /* We have to supply some lookahead.  8 bit lookahead
470                     * are needed by the zlib inflater, and we must fill
471                     * the next byte, so that all bits are flushed.
472                     */
473                    int neededbits = 8 + ((-pending.getBitCount()) & 7);
474                    while (neededbits > 0)
475                      {
476                        /* write a static tree block consisting solely of
477                         * an EOF:
478                         */
479                        pending.writeBits(2, 10);
480                        neededbits -= 10;
481                      }
482                  }
483                state = BUSY_STATE;
484              }
485            else if (state == FINISHING_STATE)
486              {
487                pending.alignToByte();
488                /* We have completed the stream */
489                if (!noHeader)
490                  {
491                    int adler = engine.getAdler();
492                    pending.writeShortMSB(adler >> 16);
493                    pending.writeShortMSB(adler & 0xffff);
494                  }
495                state = FINISHED_STATE;
496              }
497          }
498      }
499
500    return origLength - length;
501  }
502
503  /**
504   * Sets the dictionary which should be used in the deflate process.
505   * This call is equivalent to <code>setDictionary(dict, 0,
506   * dict.length)</code>.
507   * @param dict the dictionary.
508   * @exception IllegalStateException if setInput () or deflate ()
509   * were already called or another dictionary was already set.
510   */
511  public void setDictionary(byte[] dict)
512  {
513    setDictionary(dict, 0, dict.length);
514  }
515
516  /**
517   * Sets the dictionary which should be used in the deflate process.
518   * The dictionary should be a byte array containing strings that are
519   * likely to occur in the data which should be compressed.  The
520   * dictionary is not stored in the compressed output, only a
521   * checksum.  To decompress the output you need to supply the same
522   * dictionary again.
523   * @param dict the dictionary.
524   * @param offset an offset into the dictionary.
525   * @param length the length of the dictionary.
526   * @exception IllegalStateException if setInput () or deflate () were
527   * already called or another dictionary was already set.
528   */
529  public void setDictionary(byte[] dict, int offset, int length)
530  {
531    if (state != INIT_STATE)
532      throw new IllegalStateException();
533
534    state = SETDICT_STATE;
535    engine.setDictionary(dict, offset, length);
536  }
537}