1 /*
2  * Copyright (c) 1997, 2020, Oracle and/or its affiliates. All rights reserved.
3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4  *
5  * This code is free software; you can redistribute it and/or modify it
6  * under the terms of the GNU General Public License version 2 only, as
7  * published by the Free Software Foundation.  Oracle designates this
8  * particular file as subject to the "Classpath" exception as provided
9  * by Oracle in the LICENSE file that accompanied this code.
10  *
11  * This code is distributed in the hope that it will be useful, but WITHOUT
12  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
14  * version 2 for more details (a copy is included in the LICENSE file that
15  * accompanied this code).
16  *
17  * You should have received a copy of the GNU General Public License version
18  * 2 along with this work; if not, write to the Free Software Foundation,
19  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20  *
21  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
22  * or visit www.oracle.com if you need additional information or have any
23  * questions.
24  */
25 
26 /*
27  * (C) Copyright Taligent, Inc. 1996, 1997 - All Rights Reserved
28  * (C) Copyright IBM Corp. 1996-1998 - All Rights Reserved
29  *
30  *   The original version of this source code and documentation is copyrighted
31  * and owned by Taligent, Inc., a wholly-owned subsidiary of IBM. These
32  * materials are provided under terms of a License Agreement between Taligent
33  * and Sun. This technology is protected by multiple US and International
34  * patents. This notice and attribution to Taligent may not be removed.
35  *   Taligent is a registered trademark of Taligent, Inc.
36  *
37  */
38 
39 package java.text;
40 
41 import java.text.Normalizer;
42 import java.util.Vector;
43 import java.util.Locale;
44 
45 /**
46  * The {@code RuleBasedCollator} class is a concrete subclass of
47  * {@code Collator} that provides a simple, data-driven, table
48  * collator.  With this class you can create a customized table-based
49  * {@code Collator}.  {@code RuleBasedCollator} maps
50  * characters to sort keys.
51  *
52  * <p>
53  * {@code RuleBasedCollator} has the following restrictions
54  * for efficiency (other subclasses may be used for more complex languages) :
55  * <ol>
56  * <li>If a special collation rule controlled by a &lt;modifier&gt; is
57  *     specified it applies to the whole collator object.
58  * <li>All non-mentioned characters are at the end of the
59  *     collation order.
60  * </ol>
61  *
62  * <p>
63  * The collation table is composed of a list of collation rules, where each
64  * rule is of one of three forms:
65  * <pre>
66  *    &lt;modifier&gt;
67  *    &lt;relation&gt; &lt;text-argument&gt;
68  *    &lt;reset&gt; &lt;text-argument&gt;
69  * </pre>
70  * The definitions of the rule elements is as follows:
71  * <UL>
72  *    <LI><strong>Text-Argument</strong>: A text-argument is any sequence of
73  *        characters, excluding special characters (that is, common
74  *        whitespace characters [0009-000D, 0020] and rule syntax characters
75  *        [0021-002F, 003A-0040, 005B-0060, 007B-007E]). If those
76  *        characters are desired, you can put them in single quotes
77  *        (e.g. ampersand =&gt; '&amp;'). Note that unquoted white space characters
78  *        are ignored; e.g. {@code b c} is treated as {@code bc}.
79  *    <LI><strong>Modifier</strong>: There are currently two modifiers that
80  *        turn on special collation rules.
81  *        <UL>
82  *            <LI>'@' : Turns on backwards sorting of accents (secondary
83  *                      differences), as in French.
84  *            <LI>'!' : Turns on Thai/Lao vowel-consonant swapping.  If this
85  *                      rule is in force when a Thai vowel of the range
86  *                      &#92;U0E40-&#92;U0E44 precedes a Thai consonant of the range
87  *                      &#92;U0E01-&#92;U0E2E OR a Lao vowel of the range &#92;U0EC0-&#92;U0EC4
88  *                      precedes a Lao consonant of the range &#92;U0E81-&#92;U0EAE then
89  *                      the vowel is placed after the consonant for collation
90  *                      purposes.
91  *        </UL>
92  *        <p>'@' : Indicates that accents are sorted backwards, as in French.
93  *    <LI><strong>Relation</strong>: The relations are the following:
94  *        <UL>
95  *            <LI>'&lt;' : Greater, as a letter difference (primary)
96  *            <LI>';' : Greater, as an accent difference (secondary)
97  *            <LI>',' : Greater, as a case difference (tertiary)
98  *            <LI>'=' : Equal
99  *        </UL>
100  *    <LI><strong>Reset</strong>: There is a single reset
101  *        which is used primarily for contractions and expansions, but which
102  *        can also be used to add a modification at the end of a set of rules.
103  *        <p>'&amp;' : Indicates that the next rule follows the position to where
104  *            the reset text-argument would be sorted.
105  * </UL>
106  *
107  * <p>
108  * This sounds more complicated than it is in practice. For example, the
109  * following are equivalent ways of expressing the same thing:
110  * <blockquote>
111  * <pre>
112  * a &lt; b &lt; c
113  * a &lt; b &amp; b &lt; c
114  * a &lt; c &amp; a &lt; b
115  * </pre>
116  * </blockquote>
117  * Notice that the order is important, as the subsequent item goes immediately
118  * after the text-argument. The following are not equivalent:
119  * <blockquote>
120  * <pre>
121  * a &lt; b &amp; a &lt; c
122  * a &lt; c &amp; a &lt; b
123  * </pre>
124  * </blockquote>
125  * Either the text-argument must already be present in the sequence, or some
126  * initial substring of the text-argument must be present. (e.g. "a &lt; b &amp; ae &lt;
127  * e" is valid since "a" is present in the sequence before "ae" is reset). In
128  * this latter case, "ae" is not entered and treated as a single character;
129  * instead, "e" is sorted as if it were expanded to two characters: "a"
130  * followed by an "e". This difference appears in natural languages: in
131  * traditional Spanish "ch" is treated as though it contracts to a single
132  * character (expressed as "c &lt; ch &lt; d"), while in traditional German
133  * a-umlaut is treated as though it expanded to two characters
134  * (expressed as "a,A &lt; b,B ... &amp;ae;&#92;u00e3&amp;AE;&#92;u00c3").
135  * [&#92;u00e3 and &#92;u00c3 are, of course, the escape sequences for a-umlaut.]
136  * <p>
137  * <strong>Ignorable Characters</strong>
138  * <p>
139  * For ignorable characters, the first rule must start with a relation (the
140  * examples we have used above are really fragments; "a &lt; b" really should be
141  * "&lt; a &lt; b"). If, however, the first relation is not "&lt;", then all the all
142  * text-arguments up to the first "&lt;" are ignorable. For example, ", - &lt; a &lt; b"
143  * makes "-" an ignorable character, as we saw earlier in the word
144  * "black-birds". In the samples for different languages, you see that most
145  * accents are ignorable.
146  *
147  * <p><strong>Normalization and Accents</strong>
148  * <p>
149  * {@code RuleBasedCollator} automatically processes its rule table to
150  * include both pre-composed and combining-character versions of
151  * accented characters.  Even if the provided rule string contains only
152  * base characters and separate combining accent characters, the pre-composed
153  * accented characters matching all canonical combinations of characters from
154  * the rule string will be entered in the table.
155  * <p>
156  * This allows you to use a RuleBasedCollator to compare accented strings
157  * even when the collator is set to NO_DECOMPOSITION.  There are two caveats,
158  * however.  First, if the strings to be collated contain combining
159  * sequences that may not be in canonical order, you should set the collator to
160  * CANONICAL_DECOMPOSITION or FULL_DECOMPOSITION to enable sorting of
161  * combining sequences.  Second, if the strings contain characters with
162  * compatibility decompositions (such as full-width and half-width forms),
163  * you must use FULL_DECOMPOSITION, since the rule tables only include
164  * canonical mappings.
165  *
166  * <p><strong>Errors</strong>
167  * <p>
168  * The following are errors:
169  * <UL>
170  *     <LI>A text-argument contains unquoted punctuation symbols
171  *        (e.g. "a &lt; b-c &lt; d").
172  *     <LI>A relation or reset character not followed by a text-argument
173  *        (e.g. "a &lt; ,b").
174  *     <LI>A reset where the text-argument (or an initial substring of the
175  *         text-argument) is not already in the sequence.
176  *         (e.g. "a &lt; b &amp; e &lt; f")
177  * </UL>
178  * If you produce one of these errors, a {@code RuleBasedCollator} throws
179  * a {@code ParseException}.
180  *
181  * <p><strong>Examples</strong>
182  * <p>Simple:     "&lt; a &lt; b &lt; c &lt; d"
183  * <p>Norwegian:  "&lt; a, A &lt; b, B &lt; c, C &lt; d, D &lt; e, E &lt; f, F
184  *                 &lt; g, G &lt; h, H &lt; i, I &lt; j, J &lt; k, K &lt; l, L
185  *                 &lt; m, M &lt; n, N &lt; o, O &lt; p, P &lt; q, Q &lt; r, R
186  *                 &lt; s, S &lt; t, T &lt; u, U &lt; v, V &lt; w, W &lt; x, X
187  *                 &lt; y, Y &lt; z, Z
188  *                 &lt; &#92;u00E6, &#92;u00C6
189  *                 &lt; &#92;u00F8, &#92;u00D8
190  *                 &lt; &#92;u00E5 = a&#92;u030A, &#92;u00C5 = A&#92;u030A;
191  *                      aa, AA"
192  *
193  * <p>
194  * To create a {@code RuleBasedCollator} object with specialized
195  * rules tailored to your needs, you construct the {@code RuleBasedCollator}
196  * with the rules contained in a {@code String} object. For example:
197  * <blockquote>
198  * <pre>
199  * String simple = "&lt; a&lt; b&lt; c&lt; d";
200  * RuleBasedCollator mySimple = new RuleBasedCollator(simple);
201  * </pre>
202  * </blockquote>
203  * Or:
204  * <blockquote>
205  * <pre>
206  * String Norwegian = "&lt; a, A &lt; b, B &lt; c, C &lt; d, D &lt; e, E &lt; f, F &lt; g, G &lt; h, H &lt; i, I" +
207  *                    "&lt; j, J &lt; k, K &lt; l, L &lt; m, M &lt; n, N &lt; o, O &lt; p, P &lt; q, Q &lt; r, R" +
208  *                    "&lt; s, S &lt; t, T &lt; u, U &lt; v, V &lt; w, W &lt; x, X &lt; y, Y &lt; z, Z" +
209  *                    "&lt; &#92;u00E6, &#92;u00C6" +     // Latin letter ae &amp; AE
210  *                    "&lt; &#92;u00F8, &#92;u00D8" +     // Latin letter o &amp; O with stroke
211  *                    "&lt; &#92;u00E5 = a&#92;u030A," +  // Latin letter a with ring above
212  *                    "  &#92;u00C5 = A&#92;u030A;" +  // Latin letter A with ring above
213  *                    "  aa, AA";
214  * RuleBasedCollator myNorwegian = new RuleBasedCollator(Norwegian);
215  * </pre>
216  * </blockquote>
217  *
218  * <p>
219  * A new collation rules string can be created by concatenating rules
220  * strings. For example, the rules returned by {@link #getRules()} could
221  * be concatenated to combine multiple {@code RuleBasedCollator}s.
222  *
223  * <p>
224  * The following example demonstrates how to change the order of
225  * non-spacing accents,
226  * <blockquote>
227  * <pre>
228  * // old rule
229  * String oldRules = "=&#92;u0301;&#92;u0300;&#92;u0302;&#92;u0308"    // main accents
230  *                 + ";&#92;u0327;&#92;u0303;&#92;u0304;&#92;u0305"    // main accents
231  *                 + ";&#92;u0306;&#92;u0307;&#92;u0309;&#92;u030A"    // main accents
232  *                 + ";&#92;u030B;&#92;u030C;&#92;u030D;&#92;u030E"    // main accents
233  *                 + ";&#92;u030F;&#92;u0310;&#92;u0311;&#92;u0312"    // main accents
234  *                 + "&lt; a , A ; ae, AE ; &#92;u00e6 , &#92;u00c6"
235  *                 + "&lt; b , B &lt; c, C &lt; e, E &amp; C &lt; d, D";
236  * // change the order of accent characters
237  * String addOn = "&amp; &#92;u0300 ; &#92;u0308 ; &#92;u0302";
238  * RuleBasedCollator myCollator = new RuleBasedCollator(oldRules + addOn);
239  * </pre>
240  * </blockquote>
241  *
242  * @see        Collator
243  * @see        CollationElementIterator
244  * @author     Helena Shih, Laura Werner, Richard Gillam
245  * @since 1.1
246  */
247 public class RuleBasedCollator extends Collator{
248     // IMPLEMENTATION NOTES:  The implementation of the collation algorithm is
249     // divided across three classes: RuleBasedCollator, RBCollationTables, and
250     // CollationElementIterator.  RuleBasedCollator contains the collator's
251     // transient state and includes the code that uses the other classes to
252     // implement comparison and sort-key building.  RuleBasedCollator also
253     // contains the logic to handle French secondary accent sorting.
254     // A RuleBasedCollator has two CollationElementIterators.  State doesn't
255     // need to be preserved in these objects between calls to compare() or
256     // getCollationKey(), but the objects persist anyway to avoid wasting extra
257     // creation time.  compare() and getCollationKey() are synchronized to ensure
258     // thread safety with this scheme.  The CollationElementIterator is responsible
259     // for generating collation elements from strings and returning one element at
260     // a time (sometimes there's a one-to-many or many-to-one mapping between
261     // characters and collation elements-- this class handles that).
262     // CollationElementIterator depends on RBCollationTables, which contains the
263     // collator's static state.  RBCollationTables contains the actual data
264     // tables specifying the collation order of characters for a particular locale
265     // or use.  It also contains the base logic that CollationElementIterator
266     // uses to map from characters to collation elements.  A single RBCollationTables
267     // object is shared among all RuleBasedCollators for the same locale, and
268     // thus by all the CollationElementIterators they create.
269 
270     /**
271      * RuleBasedCollator constructor.  This takes the table rules and builds
272      * a collation table out of them.  Please see RuleBasedCollator class
273      * description for more details on the collation rule syntax.
274      * @see java.util.Locale
275      * @param rules the collation rules to build the collation table from.
276      * @throws    ParseException A format exception
277      * will be thrown if the build process of the rules fails. For
278      * example, build rule "a &lt; ? &lt; d" will cause the constructor to
279      * throw the ParseException because the '?' is not quoted.
280      */
RuleBasedCollator(String rules)281     public RuleBasedCollator(String rules) throws ParseException {
282         this(rules, Collator.CANONICAL_DECOMPOSITION);
283     }
284 
285     /**
286      * RuleBasedCollator constructor.  This takes the table rules and builds
287      * a collation table out of them.  Please see RuleBasedCollator class
288      * description for more details on the collation rule syntax.
289      * @see java.util.Locale
290      * @param rules the collation rules to build the collation table from.
291      * @param decomp the decomposition strength used to build the
292      * collation table and to perform comparisons.
293      * @throws    ParseException A format exception
294      * will be thrown if the build process of the rules fails. For
295      * example, build rule "a < ? < d" will cause the constructor to
296      * throw the ParseException because the '?' is not quoted.
297      */
RuleBasedCollator(String rules, int decomp)298     RuleBasedCollator(String rules, int decomp) throws ParseException {
299         setStrength(Collator.TERTIARY);
300         setDecomposition(decomp);
301         tables = new RBCollationTables(rules, decomp);
302     }
303 
304     /**
305      * "Copy constructor."  Used in clone() for performance.
306      */
RuleBasedCollator(RuleBasedCollator that)307     private RuleBasedCollator(RuleBasedCollator that) {
308         setStrength(that.getStrength());
309         setDecomposition(that.getDecomposition());
310         tables = that.tables;
311     }
312 
313     /**
314      * Gets the table-based rules for the collation object.
315      * @return returns the collation rules that the table collation object
316      * was created from.
317      */
getRules()318     public String getRules()
319     {
320         return tables.getRules();
321     }
322 
323     /**
324      * Returns a CollationElementIterator for the given String.
325      *
326      * @param source the string to be collated
327      * @return a {@code CollationElementIterator} object
328      * @see java.text.CollationElementIterator
329      */
getCollationElementIterator(String source)330     public CollationElementIterator getCollationElementIterator(String source) {
331         return new CollationElementIterator( source, this );
332     }
333 
334     /**
335      * Returns a CollationElementIterator for the given CharacterIterator.
336      *
337      * @param source the character iterator to be collated
338      * @return a {@code CollationElementIterator} object
339      * @see java.text.CollationElementIterator
340      * @since 1.2
341      */
getCollationElementIterator( CharacterIterator source)342     public CollationElementIterator getCollationElementIterator(
343                                                 CharacterIterator source) {
344         return new CollationElementIterator( source, this );
345     }
346 
347     /**
348      * Compares the character data stored in two different strings based on the
349      * collation rules.  Returns information about whether a string is less
350      * than, greater than or equal to another string in a language.
351      * This can be overridden in a subclass.
352      *
353      * @throws    NullPointerException if {@code source} or {@code target} is null.
354      */
compare(String source, String target)355     public synchronized int compare(String source, String target)
356     {
357         if (source == null || target == null) {
358             throw new NullPointerException();
359         }
360 
361         // The basic algorithm here is that we use CollationElementIterators
362         // to step through both the source and target strings.  We compare each
363         // collation element in the source string against the corresponding one
364         // in the target, checking for differences.
365         //
366         // If a difference is found, we set <result> to LESS or GREATER to
367         // indicate whether the source string is less or greater than the target.
368         //
369         // However, it's not that simple.  If we find a tertiary difference
370         // (e.g. 'A' vs. 'a') near the beginning of a string, it can be
371         // overridden by a primary difference (e.g. "A" vs. "B") later in
372         // the string.  For example, "AA" < "aB", even though 'A' > 'a'.
373         //
374         // To keep track of this, we use strengthResult to keep track of the
375         // strength of the most significant difference that has been found
376         // so far.  When we find a difference whose strength is greater than
377         // strengthResult, it overrides the last difference (if any) that
378         // was found.
379 
380         int result = Collator.EQUAL;
381 
382         if (sourceCursor == null) {
383             sourceCursor = getCollationElementIterator(source);
384         } else {
385             sourceCursor.setText(source);
386         }
387         if (targetCursor == null) {
388             targetCursor = getCollationElementIterator(target);
389         } else {
390             targetCursor.setText(target);
391         }
392 
393         int sOrder = 0, tOrder = 0;
394 
395         boolean initialCheckSecTer = getStrength() >= Collator.SECONDARY;
396         boolean checkSecTer = initialCheckSecTer;
397         boolean checkTertiary = getStrength() >= Collator.TERTIARY;
398 
399         boolean gets = true, gett = true;
400 
401         while(true) {
402             // Get the next collation element in each of the strings, unless
403             // we've been requested to skip it.
404             if (gets) sOrder = sourceCursor.next(); else gets = true;
405             if (gett) tOrder = targetCursor.next(); else gett = true;
406 
407             // If we've hit the end of one of the strings, jump out of the loop
408             if ((sOrder == CollationElementIterator.NULLORDER)||
409                 (tOrder == CollationElementIterator.NULLORDER))
410                 break;
411 
412             int pSOrder = CollationElementIterator.primaryOrder(sOrder);
413             int pTOrder = CollationElementIterator.primaryOrder(tOrder);
414 
415             // If there's no difference at this position, we can skip it
416             if (sOrder == tOrder) {
417                 if (tables.isFrenchSec() && pSOrder != 0) {
418                     if (!checkSecTer) {
419                         // in french, a secondary difference more to the right is stronger,
420                         // so accents have to be checked with each base element
421                         checkSecTer = initialCheckSecTer;
422                         // but tertiary differences are less important than the first
423                         // secondary difference, so checking tertiary remains disabled
424                         checkTertiary = false;
425                     }
426                 }
427                 continue;
428             }
429 
430             // Compare primary differences first.
431             if ( pSOrder != pTOrder )
432             {
433                 if (sOrder == 0) {
434                     // The entire source element is ignorable.
435                     // Skip to the next source element, but don't fetch another target element.
436                     gett = false;
437                     continue;
438                 }
439                 if (tOrder == 0) {
440                     gets = false;
441                     continue;
442                 }
443 
444                 // The source and target elements aren't ignorable, but it's still possible
445                 // for the primary component of one of the elements to be ignorable....
446 
447                 if (pSOrder == 0)  // primary order in source is ignorable
448                 {
449                     // The source's primary is ignorable, but the target's isn't.  We treat ignorables
450                     // as a secondary difference, so remember that we found one.
451                     if (checkSecTer) {
452                         result = Collator.GREATER;  // (strength is SECONDARY)
453                         checkSecTer = false;
454                     }
455                     // Skip to the next source element, but don't fetch another target element.
456                     gett = false;
457                 }
458                 else if (pTOrder == 0)
459                 {
460                     // record differences - see the comment above.
461                     if (checkSecTer) {
462                         result = Collator.LESS;  // (strength is SECONDARY)
463                         checkSecTer = false;
464                     }
465                     // Skip to the next source element, but don't fetch another target element.
466                     gets = false;
467                 } else {
468                     // Neither of the orders is ignorable, and we already know that the primary
469                     // orders are different because of the (pSOrder != pTOrder) test above.
470                     // Record the difference and stop the comparison.
471                     if (pSOrder < pTOrder) {
472                         return Collator.LESS;  // (strength is PRIMARY)
473                     } else {
474                         return Collator.GREATER;  // (strength is PRIMARY)
475                     }
476                 }
477             } else { // else of if ( pSOrder != pTOrder )
478                 // primary order is the same, but complete order is different. So there
479                 // are no base elements at this point, only ignorables (Since the strings are
480                 // normalized)
481 
482                 if (checkSecTer) {
483                     // a secondary or tertiary difference may still matter
484                     short secSOrder = CollationElementIterator.secondaryOrder(sOrder);
485                     short secTOrder = CollationElementIterator.secondaryOrder(tOrder);
486                     if (secSOrder != secTOrder) {
487                         // there is a secondary difference
488                         result = (secSOrder < secTOrder) ? Collator.LESS : Collator.GREATER;
489                                                 // (strength is SECONDARY)
490                         checkSecTer = false;
491                         // (even in french, only the first secondary difference within
492                         //  a base character matters)
493                     } else {
494                         if (checkTertiary) {
495                             // a tertiary difference may still matter
496                             short terSOrder = CollationElementIterator.tertiaryOrder(sOrder);
497                             short terTOrder = CollationElementIterator.tertiaryOrder(tOrder);
498                             if (terSOrder != terTOrder) {
499                                 // there is a tertiary difference
500                                 result = (terSOrder < terTOrder) ? Collator.LESS : Collator.GREATER;
501                                                 // (strength is TERTIARY)
502                                 checkTertiary = false;
503                             }
504                         }
505                     }
506                 } // if (checkSecTer)
507 
508             }  // if ( pSOrder != pTOrder )
509         } // while()
510 
511         if (sOrder != CollationElementIterator.NULLORDER) {
512             // (tOrder must be CollationElementIterator::NULLORDER,
513             //  since this point is only reached when sOrder or tOrder is NULLORDER.)
514             // The source string has more elements, but the target string hasn't.
515             do {
516                 if (CollationElementIterator.primaryOrder(sOrder) != 0) {
517                     // We found an additional non-ignorable base character in the source string.
518                     // This is a primary difference, so the source is greater
519                     return Collator.GREATER; // (strength is PRIMARY)
520                 }
521                 else if (CollationElementIterator.secondaryOrder(sOrder) != 0) {
522                     // Additional secondary elements mean the source string is greater
523                     if (checkSecTer) {
524                         result = Collator.GREATER;  // (strength is SECONDARY)
525                         checkSecTer = false;
526                     }
527                 }
528             } while ((sOrder = sourceCursor.next()) != CollationElementIterator.NULLORDER);
529         }
530         else if (tOrder != CollationElementIterator.NULLORDER) {
531             // The target string has more elements, but the source string hasn't.
532             do {
533                 if (CollationElementIterator.primaryOrder(tOrder) != 0)
534                     // We found an additional non-ignorable base character in the target string.
535                     // This is a primary difference, so the source is less
536                     return Collator.LESS; // (strength is PRIMARY)
537                 else if (CollationElementIterator.secondaryOrder(tOrder) != 0) {
538                     // Additional secondary elements in the target mean the source string is less
539                     if (checkSecTer) {
540                         result = Collator.LESS;  // (strength is SECONDARY)
541                         checkSecTer = false;
542                     }
543                 }
544             } while ((tOrder = targetCursor.next()) != CollationElementIterator.NULLORDER);
545         }
546 
547         // For IDENTICAL comparisons, we use a bitwise character comparison
548         // as a tiebreaker if all else is equal
549         if (result == 0 && getStrength() == IDENTICAL) {
550             int mode = getDecomposition();
551             Normalizer.Form form;
552             if (mode == CANONICAL_DECOMPOSITION) {
553                 form = Normalizer.Form.NFD;
554             } else if (mode == FULL_DECOMPOSITION) {
555                 form = Normalizer.Form.NFKD;
556             } else {
557                 return source.compareTo(target);
558             }
559 
560             String sourceDecomposition = Normalizer.normalize(source, form);
561             String targetDecomposition = Normalizer.normalize(target, form);
562             return sourceDecomposition.compareTo(targetDecomposition);
563         }
564         return result;
565     }
566 
567     /**
568      * Transforms the string into a series of characters that can be compared
569      * with CollationKey.compareTo. This overrides java.text.Collator.getCollationKey.
570      * It can be overridden in a subclass.
571      */
getCollationKey(String source)572     public synchronized CollationKey getCollationKey(String source)
573     {
574         //
575         // The basic algorithm here is to find all of the collation elements for each
576         // character in the source string, convert them to a char representation,
577         // and put them into the collation key.  But it's trickier than that.
578         // Each collation element in a string has three components: primary (A vs B),
579         // secondary (A vs A-acute), and tertiary (A' vs a); and a primary difference
580         // at the end of a string takes precedence over a secondary or tertiary
581         // difference earlier in the string.
582         //
583         // To account for this, we put all of the primary orders at the beginning of the
584         // string, followed by the secondary and tertiary orders, separated by nulls.
585         //
586         // Here's a hypothetical example, with the collation element represented as
587         // a three-digit number, one digit for primary, one for secondary, etc.
588         //
589         // String:              A     a     B   \u00e9 <--(e-acute)
590         // Collation Elements: 101   100   201  510
591         //
592         // Collation Key:      1125<null>0001<null>1010
593         //
594         // To make things even trickier, secondary differences (accent marks) are compared
595         // starting at the *end* of the string in languages with French secondary ordering.
596         // But when comparing the accent marks on a single base character, they are compared
597         // from the beginning.  To handle this, we reverse all of the accents that belong
598         // to each base character, then we reverse the entire string of secondary orderings
599         // at the end.  Taking the same example above, a French collator might return
600         // this instead:
601         //
602         // Collation Key:      1125<null>1000<null>1010
603         //
604         if (source == null)
605             return null;
606 
607         if (primResult == null) {
608             primResult = new StringBuffer();
609             secResult = new StringBuffer();
610             terResult = new StringBuffer();
611         } else {
612             primResult.setLength(0);
613             secResult.setLength(0);
614             terResult.setLength(0);
615         }
616         int order = 0;
617         boolean compareSec = (getStrength() >= Collator.SECONDARY);
618         boolean compareTer = (getStrength() >= Collator.TERTIARY);
619         int secOrder = CollationElementIterator.NULLORDER;
620         int terOrder = CollationElementIterator.NULLORDER;
621         int preSecIgnore = 0;
622 
623         if (sourceCursor == null) {
624             sourceCursor = getCollationElementIterator(source);
625         } else {
626             sourceCursor.setText(source);
627         }
628 
629         // walk through each character
630         while ((order = sourceCursor.next()) !=
631                CollationElementIterator.NULLORDER)
632         {
633             secOrder = CollationElementIterator.secondaryOrder(order);
634             terOrder = CollationElementIterator.tertiaryOrder(order);
635             if (!CollationElementIterator.isIgnorable(order))
636             {
637                 primResult.append((char) (CollationElementIterator.primaryOrder(order)
638                                     + COLLATIONKEYOFFSET));
639 
640                 if (compareSec) {
641                     //
642                     // accumulate all of the ignorable/secondary characters attached
643                     // to a given base character
644                     //
645                     if (tables.isFrenchSec() && preSecIgnore < secResult.length()) {
646                         //
647                         // We're doing reversed secondary ordering and we've hit a base
648                         // (non-ignorable) character.  Reverse any secondary orderings
649                         // that applied to the last base character.  (see block comment above.)
650                         //
651                         RBCollationTables.reverse(secResult, preSecIgnore, secResult.length());
652                     }
653                     // Remember where we are in the secondary orderings - this is how far
654                     // back to go if we need to reverse them later.
655                     secResult.append((char)(secOrder+ COLLATIONKEYOFFSET));
656                     preSecIgnore = secResult.length();
657                 }
658                 if (compareTer) {
659                     terResult.append((char)(terOrder+ COLLATIONKEYOFFSET));
660                 }
661             }
662             else
663             {
664                 if (compareSec && secOrder != 0)
665                     secResult.append((char)
666                         (secOrder + tables.getMaxSecOrder() + COLLATIONKEYOFFSET));
667                 if (compareTer && terOrder != 0)
668                     terResult.append((char)
669                         (terOrder + tables.getMaxTerOrder() + COLLATIONKEYOFFSET));
670             }
671         }
672         if (tables.isFrenchSec())
673         {
674             if (preSecIgnore < secResult.length()) {
675                 // If we've accumulated any secondary characters after the last base character,
676                 // reverse them.
677                 RBCollationTables.reverse(secResult, preSecIgnore, secResult.length());
678             }
679             // And now reverse the entire secResult to get French secondary ordering.
680             RBCollationTables.reverse(secResult, 0, secResult.length());
681         }
682         primResult.append((char)0);
683         secResult.append((char)0);
684         secResult.append(terResult.toString());
685         primResult.append(secResult.toString());
686 
687         if (getStrength() == IDENTICAL) {
688             primResult.append((char)0);
689             int mode = getDecomposition();
690             if (mode == CANONICAL_DECOMPOSITION) {
691                 primResult.append(Normalizer.normalize(source, Normalizer.Form.NFD));
692             } else if (mode == FULL_DECOMPOSITION) {
693                 primResult.append(Normalizer.normalize(source, Normalizer.Form.NFKD));
694             } else {
695                 primResult.append(source);
696             }
697         }
698         return new RuleBasedCollationKey(source, primResult.toString());
699     }
700 
701     /**
702      * Standard override; no change in semantics.
703      */
clone()704     public Object clone() {
705         // if we know we're not actually a subclass of RuleBasedCollator
706         // (this class really should have been made final), bypass
707         // Object.clone() and use our "copy constructor".  This is faster.
708         if (getClass() == RuleBasedCollator.class) {
709             return new RuleBasedCollator(this);
710         }
711         else {
712             RuleBasedCollator result = (RuleBasedCollator) super.clone();
713             result.primResult = null;
714             result.secResult = null;
715             result.terResult = null;
716             result.sourceCursor = null;
717             result.targetCursor = null;
718             return result;
719         }
720     }
721 
722     /**
723      * Compares the equality of two collation objects.
724      * @param obj the table-based collation object to be compared with this.
725      * @return true if the current table-based collation object is the same
726      * as the table-based collation object obj; false otherwise.
727      */
equals(Object obj)728     public boolean equals(Object obj) {
729         if (obj == null) return false;
730         if (!super.equals(obj)) return false;  // super does class check
731         RuleBasedCollator other = (RuleBasedCollator) obj;
732         // all other non-transient information is also contained in rules.
733         return (getRules().equals(other.getRules()));
734     }
735 
736     /**
737      * Generates the hash code for the table-based collation object
738      */
hashCode()739     public int hashCode() {
740         return getRules().hashCode();
741     }
742 
743     /**
744      * Allows CollationElementIterator access to the tables object
745      */
getTables()746     RBCollationTables getTables() {
747         return tables;
748     }
749 
750     // ==============================================================
751     // private
752     // ==============================================================
753 
754     static final int CHARINDEX = 0x70000000;  // need look up in .commit()
755     static final int EXPANDCHARINDEX = 0x7E000000; // Expand index follows
756     static final int CONTRACTCHARINDEX = 0x7F000000;  // contract indexes follow
757     static final int UNMAPPED = 0xFFFFFFFF;
758 
759     private static final int COLLATIONKEYOFFSET = 1;
760 
761     private RBCollationTables tables = null;
762 
763     // Internal objects that are cached across calls so that they don't have to
764     // be created/destroyed on every call to compare() and getCollationKey()
765     private StringBuffer primResult = null;
766     private StringBuffer secResult = null;
767     private StringBuffer terResult = null;
768     private CollationElementIterator sourceCursor = null;
769     private CollationElementIterator targetCursor = null;
770 }
771