2 *******************************************************************************
3 * Copyright (C) 1996-2009, International Business Machines Corporation and *
4 * others. All Rights Reserved. *
5 *******************************************************************************
7 package com.ibm.icu.impl;
9 import com.ibm.icu.text.UCharacterIterator;
12 * <p>Binary Ordered Compression for Unicode</p>
14 * <p>Users are strongly encouraged to read the ICU paper on
15 * <a href="http://www.icu-project.org/docs/papers/binary_ordered_compression_for_unicode.html">
16 * BOCU</a> before attempting to use this class.</p>
18 * <p>BOCU is used to compress unicode text into a stream of unsigned
19 * bytes. For many kinds of text the compression compares favorably
20 * to UTF-8, and for some kinds of text (such as CJK) it does better.
21 * The resulting bytes will compare in the same order as the original
22 * code points. The byte stream does not contain the values 0, 1, or
25 * <p>One example of a use of BOCU is in
26 * com.ibm.icu.text.Collator#getCollationKey(String) for a RuleBasedCollator object with
27 * collation strength IDENTICAL. The result CollationKey will consist of the
28 * collation order of the source string followed by the BOCU result of the
32 * <p>Unlike a UTF encoding, BOCU-compressed text is not suitable for
35 * <p>Method: Slope Detection<br> Remember the previous code point
36 * (initial 0). For each code point in the string, encode the
37 * difference with the previous one. Similar to a UTF, the length of
38 * the byte sequence is encoded in the lead bytes. Unlike a UTF, the
39 * trail byte values may overlap with lead/single byte values. The
40 * signedness of the difference must be encoded as the most
41 * significant part.</p>
43 * <p>We encode differences with few bytes if their absolute values
44 * are small. For correct ordering, we must treat the entire value
45 * range -10ffff..+10ffff in ascending order, which forbids encoding
46 * the sign and the absolute value separately. Instead, we split the
47 * lead byte range in the middle and encode non-negative values going
48 * up and negative values going down.</p>
50 * <p>For very small absolute values, the difference is added to a
51 * middle byte value for single-byte encoded differences. For
52 * somewhat larger absolute values, the difference is divided by the
53 * number of byte values available, the modulo is used for one trail
54 * byte, and the remainder is added to a lead byte avoiding the
55 * single-byte range. For large absolute values, the difference is
56 * similarly encoded in three bytes. (Syn Wee, I need examples
59 * <p>BOCU does not use byte values 0, 1, or 2, but uses all other
60 * byte values for lead and single bytes, so that the middle range of
61 * single bytes is as large as possible.</p>
63 * <p>Note that the lead byte ranges overlap some, but that the
64 * sequences as a whole are well ordered. I.e., even if the lead byte
65 * is the same for sequences of different lengths, the trail bytes
66 * establish correct order. It would be possible to encode slightly
67 * larger ranges for each length (>1) by subtracting the lower bound
68 * of the range. However, that would also slow down the calculation.
69 * (Syn Wee, need an example).</p>
71 * <p>For the actual string encoding, an optimization moves the
72 * previous code point value to the middle of its Unicode script block
73 * to minimize the differences in same-script text runs. (Syn Wee,
74 * need an example.)</p>
76 * @author Syn Wee Quek
77 * @since release 2.2, May 3rd 2002
81 // public constructors --------------------------------------------------
83 // public methods -------------------------------------------------------
86 * <p>Encode the code points of a string as a sequence of bytes,
87 * preserving lexical order.</p>
88 * <p>The minimum size of buffer required for the compression can be
89 * preflighted by getCompressionLength(String).</p>
90 * @param source text source
91 * @param buffer output buffer
92 * @param offset to start writing to
93 * @return end offset where the writing stopped
94 * @see #getCompressionLength(String)
95 * @exception ArrayIndexOutOfBoundsException thrown if size of buffer is
96 * too small for the output.
98 public static int compress(String source, byte buffer[], int offset)
101 UCharacterIterator iterator = UCharacterIterator.getInstance(source);
102 int codepoint = iterator.nextCodePoint();
103 while (codepoint != UCharacterIterator.DONE) {
104 if (prev < 0x4e00 || prev >= 0xa000) {
105 prev = (prev & ~0x7f) - SLOPE_REACH_NEG_1_;
108 // Unihan U+4e00..U+9fa5:
109 // double-bytes down from the upper end
110 prev = 0x9fff - SLOPE_REACH_POS_2_;
113 offset = writeDiff(codepoint - prev, buffer, offset);
115 codepoint = iterator.nextCodePoint();
121 * Return the number of bytes that compress() would write.
122 * @param source text source string
123 * @return the length of the BOCU result
124 * @see #compress(String, byte[], int)
126 public static int getCompressionLength(String source)
130 UCharacterIterator iterator = UCharacterIterator.getInstance(source);
131 int codepoint = iterator.nextCodePoint();
132 while (codepoint != UCharacterIterator.DONE) {
133 if (prev < 0x4e00 || prev >= 0xa000) {
134 prev = (prev & ~0x7f) - SLOPE_REACH_NEG_1_;
137 // Unihan U+4e00..U+9fa5:
138 // double-bytes down from the upper end
139 prev = 0x9fff - SLOPE_REACH_POS_2_;
142 codepoint = iterator.nextCodePoint();
143 result += lengthOfDiff(codepoint - prev);
149 // public setter methods -------------------------------------------------
151 // public getter methods ------------------------------------------------
153 // public other methods -------------------------------------------------
155 // protected constructor ------------------------------------------------
157 // protected data members ------------------------------------------------
159 // protected methods -----------------------------------------------------
161 // private data members --------------------------------------------------
164 * Do not use byte values 0, 1, 2 because they are separators in sort keys.
166 private static final int SLOPE_MIN_ = 3;
167 private static final int SLOPE_MAX_ = 0xff;
168 private static final int SLOPE_MIDDLE_ = 0x81;
169 private static final int SLOPE_TAIL_COUNT_ = SLOPE_MAX_ - SLOPE_MIN_ + 1;
170 //private static final int SLOPE_MAX_BYTES_ = 4;
173 * Number of lead bytes:
174 * 1 middle byte for 0
175 * 2*80=160 single bytes for !=0
176 * 2*42=84 for double-byte values
177 * 2*3=6 for 3-byte values
178 * 2*1=2 for 4-byte values
180 * The sum must be <=SLOPE_TAIL_COUNT.
183 * - There should be >=128 single-byte values to cover 128-blocks
184 * with small scripts.
185 * - There should be >=20902 single/double-byte values to cover Unihan.
186 * - It helps CJK Extension B some if there are 3-byte values that cover
187 * the distance between them and Unihan.
188 * This also helps to jump among distant places in the BMP.
189 * - Four-byte values are necessary to cover the rest of Unicode.
191 * Symmetrical lead byte counts are for convenience.
192 * With an equal distribution of even and odd differences there is also
193 * no advantage to asymmetrical lead byte counts.
195 private static final int SLOPE_SINGLE_ = 80;
196 private static final int SLOPE_LEAD_2_ = 42;
197 private static final int SLOPE_LEAD_3_ = 3;
198 //private static final int SLOPE_LEAD_4_ = 1;
201 * The difference value range for single-byters.
203 private static final int SLOPE_REACH_POS_1_ = SLOPE_SINGLE_;
204 private static final int SLOPE_REACH_NEG_1_ = (-SLOPE_SINGLE_);
207 * The difference value range for double-byters.
209 private static final int SLOPE_REACH_POS_2_ =
210 SLOPE_LEAD_2_ * SLOPE_TAIL_COUNT_ + SLOPE_LEAD_2_ - 1;
211 private static final int SLOPE_REACH_NEG_2_ = (-SLOPE_REACH_POS_2_ - 1);
214 * The difference value range for 3-byters.
216 private static final int SLOPE_REACH_POS_3_ = SLOPE_LEAD_3_
219 + (SLOPE_LEAD_3_ - 1)
220 * SLOPE_TAIL_COUNT_ +
221 (SLOPE_TAIL_COUNT_ - 1);
222 private static final int SLOPE_REACH_NEG_3_ = (-SLOPE_REACH_POS_3_ - 1);
225 * The lead byte start values.
227 private static final int SLOPE_START_POS_2_ = SLOPE_MIDDLE_
229 private static final int SLOPE_START_POS_3_ = SLOPE_START_POS_2_
231 private static final int SLOPE_START_NEG_2_ = SLOPE_MIDDLE_ +
233 private static final int SLOPE_START_NEG_3_ = SLOPE_START_NEG_2_
236 // private constructor ---------------------------------------------------
239 * Constructor private to prevent initialization
247 // private methods -------------------------------------------------------
250 * Integer division and modulo with negative numerators
251 * yields negative modulo results and quotients that are one more than
253 * @param number which operations are to be performed on
254 * @param factor the factor to use for division
255 * @return (result of division) << 32 | modulo
257 private static final long getNegDivMod(int number, int factor)
259 int modulo = number % factor;
260 long result = number / factor;
265 return (result << 32) | modulo;
269 * Encode one difference value -0x10ffff..+0x10ffff in 1..3 bytes,
270 * preserving lexical order
272 * @param buffer byte buffer to append to
273 * @param offset to the byte buffer to start appending
274 * @return end offset where the appending stops
276 private static final int writeDiff(int diff, byte buffer[], int offset)
278 if (diff >= SLOPE_REACH_NEG_1_) {
279 if (diff <= SLOPE_REACH_POS_1_) {
280 buffer[offset ++] = (byte)(SLOPE_MIDDLE_ + diff);
282 else if (diff <= SLOPE_REACH_POS_2_) {
283 buffer[offset ++] = (byte)(SLOPE_START_POS_2_
284 + (diff / SLOPE_TAIL_COUNT_));
285 buffer[offset ++] = (byte)(SLOPE_MIN_ +
286 (diff % SLOPE_TAIL_COUNT_));
288 else if (diff <= SLOPE_REACH_POS_3_) {
289 buffer[offset + 2] = (byte)(SLOPE_MIN_
290 + (diff % SLOPE_TAIL_COUNT_));
291 diff /= SLOPE_TAIL_COUNT_;
292 buffer[offset + 1] = (byte)(SLOPE_MIN_
293 + (diff % SLOPE_TAIL_COUNT_));
294 buffer[offset] = (byte)(SLOPE_START_POS_3_
295 + (diff / SLOPE_TAIL_COUNT_));
299 buffer[offset + 3] = (byte)(SLOPE_MIN_
300 + diff % SLOPE_TAIL_COUNT_);
301 diff /= SLOPE_TAIL_COUNT_;
302 buffer[offset] = (byte)(SLOPE_MIN_
303 + diff % SLOPE_TAIL_COUNT_);
304 diff /= SLOPE_TAIL_COUNT_;
305 buffer[offset + 1] = (byte)(SLOPE_MIN_
306 + diff % SLOPE_TAIL_COUNT_);
307 buffer[offset] = (byte)SLOPE_MAX_;
312 long division = getNegDivMod(diff, SLOPE_TAIL_COUNT_);
313 int modulo = (int)division;
314 if (diff >= SLOPE_REACH_NEG_2_) {
315 diff = (int)(division >> 32);
316 buffer[offset ++] = (byte)(SLOPE_START_NEG_2_ + diff);
317 buffer[offset ++] = (byte)(SLOPE_MIN_ + modulo);
319 else if (diff >= SLOPE_REACH_NEG_3_) {
320 buffer[offset + 2] = (byte)(SLOPE_MIN_ + modulo);
321 diff = (int)(division >> 32);
322 division = getNegDivMod(diff, SLOPE_TAIL_COUNT_);
323 modulo = (int)division;
324 diff = (int)(division >> 32);
325 buffer[offset + 1] = (byte)(SLOPE_MIN_ + modulo);
326 buffer[offset] = (byte)(SLOPE_START_NEG_3_ + diff);
330 buffer[offset + 3] = (byte)(SLOPE_MIN_ + modulo);
331 diff = (int)(division >> 32);
332 division = getNegDivMod(diff, SLOPE_TAIL_COUNT_);
333 modulo = (int)division;
334 diff = (int)(division >> 32);
335 buffer[offset + 2] = (byte)(SLOPE_MIN_ + modulo);
336 division = getNegDivMod(diff, SLOPE_TAIL_COUNT_);
337 modulo = (int)division;
338 buffer[offset + 1] = (byte)(SLOPE_MIN_ + modulo);
339 buffer[offset] = SLOPE_MIN_;
347 * How many bytes would writeDiff() write?
350 private static final int lengthOfDiff(int diff)
352 if (diff >= SLOPE_REACH_NEG_1_) {
353 if (diff <= SLOPE_REACH_POS_1_) {
356 else if (diff <= SLOPE_REACH_POS_2_) {
359 else if(diff <= SLOPE_REACH_POS_3_) {
367 if (diff >= SLOPE_REACH_NEG_2_) {
370 else if (diff >= SLOPE_REACH_NEG_3_) {