2 *******************************************************************************
\r
3 * Copyright (C) 1996-2010, International Business Machines Corporation and *
\r
4 * others. All Rights Reserved. *
\r
5 *******************************************************************************
\r
7 package com.ibm.icu.text;
\r
9 import java.io.IOException;
\r
10 import java.text.ParseException;
\r
11 import java.util.Arrays;
\r
12 import java.util.Enumeration;
\r
13 import java.util.Hashtable;
\r
14 import java.util.Vector;
\r
16 import com.ibm.icu.impl.IntTrieBuilder;
\r
17 import com.ibm.icu.impl.Norm2AllModes;
\r
18 import com.ibm.icu.impl.Normalizer2Impl;
\r
19 import com.ibm.icu.impl.TrieBuilder;
\r
20 import com.ibm.icu.impl.TrieIterator;
\r
21 import com.ibm.icu.impl.UCharacterProperty;
\r
22 import com.ibm.icu.impl.Utility;
\r
23 import com.ibm.icu.lang.UCharacter;
\r
24 import com.ibm.icu.lang.UCharacterCategory;
\r
25 import com.ibm.icu.util.RangeValueIterator;
\r
26 import com.ibm.icu.util.VersionInfo;
\r
29 * Class for building a collator from a list of collation rules. This class is
\r
30 * uses CollationRuleParser
\r
32 * @author Syn Wee Quek
\r
33 * @since release 2.2, June 11 2002
\r
35 final class CollationParsedRuleBuilder {
\r
36 // package private constructors ------------------------------------------
\r
43 * @exception ParseException
\r
44 * thrown when argument rules have an invalid syntax
\r
46 CollationParsedRuleBuilder(String rules) throws ParseException {
\r
47 m_parser_ = new CollationRuleParser(rules);
\r
48 m_parser_.assembleTokenList();
\r
49 m_utilColEIter_ = RuleBasedCollator.UCA_
\r
50 .getCollationElementIterator("");
\r
53 // package private inner classes -----------------------------------------
\r
56 * Inverse UCA wrapper
\r
58 static class InverseUCA {
\r
59 // package private constructor ---------------------------------------
\r
64 // package private data member ---------------------------------------
\r
67 * Array list of characters
\r
71 * Array list of continuation characters
\r
73 char m_continuations_[];
\r
76 * UCA version of inverse UCA table
\r
78 VersionInfo m_UCA_version_;
\r
80 // package private method --------------------------------------------
\r
83 * Returns the previous inverse ces of the argument ces
\r
88 * continuation ce to test
\r
90 * collation strength
\r
92 * an array to store the return results previous inverse ce
\r
93 * and previous inverse continuation ce
\r
94 * @return result of the inverse ce
\r
96 final int getInversePrevCE(int ce, int contce, int strength,
\r
98 int result = findInverseCE(ce, contce);
\r
101 prevresult[0] = CollationElementIterator.NULLORDER;
\r
105 ce &= STRENGTH_MASK_[strength];
\r
106 contce &= STRENGTH_MASK_[strength];
\r
108 prevresult[0] = ce;
\r
109 prevresult[1] = contce;
\r
111 while ((prevresult[0] & STRENGTH_MASK_[strength]) == ce
\r
112 && (prevresult[1] & STRENGTH_MASK_[strength]) == contce
\r
114 // this condition should prevent falling off the edge of the
\r
116 // here, we end up in a singularity - zero
\r
117 prevresult[0] = m_table_[3 * (--result)];
\r
118 prevresult[1] = m_table_[3 * result + 1];
\r
123 final int getCEStrengthDifference(int CE, int contCE, int prevCE,
\r
125 int strength = Collator.TERTIARY;
\r
126 while (((prevCE & STRENGTH_MASK_[strength]) != (CE & STRENGTH_MASK_[strength]) || (prevContCE & STRENGTH_MASK_[strength]) != (contCE & STRENGTH_MASK_[strength]))
\r
127 && (strength != 0)) {
\r
133 private int compareCEs(int source0, int source1, int target0,
\r
135 int s1 = source0, s2, t1 = target0, t2;
\r
136 if (RuleBasedCollator.isContinuation(source1)) {
\r
141 if (RuleBasedCollator.isContinuation(target1)) {
\r
148 if (s1 == t1 && s2 == t2) {
\r
151 s = (s1 & 0xFFFF0000) | ((s2 & 0xFFFF0000) >>> 16);
\r
152 t = (t1 & 0xFFFF0000) | ((t2 & 0xFFFF0000) >>> 16);
\r
154 s = (s1 & 0x0000FF00) | (s2 & 0x0000FF00) >> 8;
\r
155 t = (t1 & 0x0000FF00) | (t2 & 0x0000FF00) >> 8;
\r
157 s = (s1 & 0x000000FF) << 8 | (s2 & 0x000000FF);
\r
158 t = (t1 & 0x000000FF) << 8 | (t2 & 0x000000FF);
\r
159 return Utility.compareUnsigned(s, t);
\r
161 return Utility.compareUnsigned(s, t);
\r
164 return Utility.compareUnsigned(s, t);
\r
169 * Finding the inverse CE of the argument CEs
\r
175 * @return inverse CE
\r
177 int findInverseCE(int ce, int contce) {
\r
179 int top = m_table_.length / 3;
\r
182 while (bottom < top - 1) {
\r
183 result = (top + bottom) >> 1;
\r
184 int first = m_table_[3 * result];
\r
185 int second = m_table_[3 * result + 1];
\r
186 int comparison = compareCEs(first, second, ce, contce);
\r
187 if (comparison > 0) {
\r
189 } else if (comparison < 0) {
\r
200 * Getting gap offsets in the inverse UCA
\r
202 * @param listheader
\r
203 * parsed token lists
\r
204 * @exception Exception
\r
205 * thrown when error occurs while finding the collation
\r
208 void getInverseGapPositions(
\r
209 CollationRuleParser.TokenListHeader listheader)
\r
211 // reset all the gaps
\r
212 CollationRuleParser.Token token = listheader.m_first_;
\r
213 int tokenstrength = token.m_strength_;
\r
215 for (int i = 0; i < 3; i++) {
\r
216 listheader.m_gapsHi_[3 * i] = 0;
\r
217 listheader.m_gapsHi_[3 * i + 1] = 0;
\r
218 listheader.m_gapsHi_[3 * i + 2] = 0;
\r
219 listheader.m_gapsLo_[3 * i] = 0;
\r
220 listheader.m_gapsLo_[3 * i + 1] = 0;
\r
221 listheader.m_gapsLo_[3 * i + 2] = 0;
\r
222 listheader.m_numStr_[i] = 0;
\r
223 listheader.m_fStrToken_[i] = null;
\r
224 listheader.m_lStrToken_[i] = null;
\r
225 listheader.m_pos_[i] = -1;
\r
228 if ((listheader.m_baseCE_ >>> 24) >= RuleBasedCollator.UCA_CONSTANTS_.PRIMARY_IMPLICIT_MIN_
\r
229 && (listheader.m_baseCE_ >>> 24) <= RuleBasedCollator.UCA_CONSTANTS_.PRIMARY_IMPLICIT_MAX_) {
\r
231 listheader.m_pos_[0] = 0;
\r
232 int t1 = listheader.m_baseCE_;
\r
233 int t2 = listheader.m_baseContCE_;
\r
234 listheader.m_gapsLo_[0] = mergeCE(t1, t2, Collator.PRIMARY);
\r
235 listheader.m_gapsLo_[1] = mergeCE(t1, t2, Collator.SECONDARY);
\r
236 listheader.m_gapsLo_[2] = mergeCE(t1, t2, Collator.TERTIARY);
\r
237 int primaryCE = t1 & RuleBasedCollator.CE_PRIMARY_MASK_
\r
238 | (t2 & RuleBasedCollator.CE_PRIMARY_MASK_) >>> 16;
\r
239 primaryCE = RuleBasedCollator.impCEGen_
\r
240 .getImplicitFromRaw(RuleBasedCollator.impCEGen_
\r
241 .getRawFromImplicit(primaryCE) + 1);
\r
243 t1 = primaryCE & RuleBasedCollator.CE_PRIMARY_MASK_ | 0x0505;
\r
244 t2 = (primaryCE << 16) & RuleBasedCollator.CE_PRIMARY_MASK_
\r
245 | RuleBasedCollator.CE_CONTINUATION_MARKER_;
\r
247 // if (listheader.m_baseCE_ < 0xEF000000) {
\r
248 // // first implicits have three byte primaries, with a gap of
\r
249 // // one so we esentially need to add 2 to the top byte in
\r
250 // // listheader.m_baseContCE_
\r
251 // t2 += 0x02000000;
\r
254 // // second implicits have four byte primaries, with a gap of
\r
255 // // IMPLICIT_LAST2_MULTIPLIER_
\r
256 // // Now, this guy is not really accessible here, so until we
\r
257 // // find a better way to pass it around, assume that the gap
\r
259 // t2 += 0x00020000;
\r
261 listheader.m_gapsHi_[0] = mergeCE(t1, t2, Collator.PRIMARY);
\r
262 listheader.m_gapsHi_[1] = mergeCE(t1, t2, Collator.SECONDARY);
\r
263 listheader.m_gapsHi_[2] = mergeCE(t1, t2, Collator.TERTIARY);
\r
264 } else if (listheader.m_indirect_ == true
\r
265 && listheader.m_nextCE_ != 0) {
\r
266 listheader.m_pos_[0] = 0;
\r
267 int t1 = listheader.m_baseCE_;
\r
268 int t2 = listheader.m_baseContCE_;
\r
269 listheader.m_gapsLo_[0] = mergeCE(t1, t2, Collator.PRIMARY);
\r
270 listheader.m_gapsLo_[1] = mergeCE(t1, t2, Collator.SECONDARY);
\r
271 listheader.m_gapsLo_[2] = mergeCE(t1, t2, Collator.TERTIARY);
\r
272 t1 = listheader.m_nextCE_;
\r
273 t2 = listheader.m_nextContCE_;
\r
274 listheader.m_gapsHi_[0] = mergeCE(t1, t2, Collator.PRIMARY);
\r
275 listheader.m_gapsHi_[1] = mergeCE(t1, t2, Collator.SECONDARY);
\r
276 listheader.m_gapsHi_[2] = mergeCE(t1, t2, Collator.TERTIARY);
\r
279 if (tokenstrength < CE_BASIC_STRENGTH_LIMIT_) {
\r
280 listheader.m_pos_[tokenstrength] = getInverseNext(
\r
281 listheader, tokenstrength);
\r
282 if (listheader.m_pos_[tokenstrength] >= 0) {
\r
283 listheader.m_fStrToken_[tokenstrength] = token;
\r
285 // The CE must be implicit, since it's not in the
\r
288 throw new Exception("Internal program error");
\r
292 while (token != null && token.m_strength_ >= tokenstrength) {
\r
293 if (tokenstrength < CE_BASIC_STRENGTH_LIMIT_) {
\r
294 listheader.m_lStrToken_[tokenstrength] = token;
\r
296 token = token.m_next_;
\r
298 if (tokenstrength < CE_BASIC_STRENGTH_LIMIT_ - 1) {
\r
299 // check if previous interval is the same and merge the
\r
300 // intervals if it is so
\r
301 if (listheader.m_pos_[tokenstrength] == listheader.m_pos_[tokenstrength + 1]) {
\r
302 listheader.m_fStrToken_[tokenstrength] = listheader.m_fStrToken_[tokenstrength + 1];
\r
303 listheader.m_fStrToken_[tokenstrength + 1] = null;
\r
304 listheader.m_lStrToken_[tokenstrength + 1] = null;
\r
305 listheader.m_pos_[tokenstrength + 1] = -1;
\r
308 if (token != null) {
\r
309 tokenstrength = token.m_strength_;
\r
314 for (int st = 0; st < 3; st++) {
\r
315 int pos = listheader.m_pos_[st];
\r
317 int t1 = m_table_[3 * pos];
\r
318 int t2 = m_table_[3 * pos + 1];
\r
319 listheader.m_gapsHi_[3 * st] = mergeCE(t1, t2,
\r
321 listheader.m_gapsHi_[3 * st + 1] = mergeCE(t1, t2,
\r
322 Collator.SECONDARY);
\r
323 listheader.m_gapsHi_[3 * st + 2] = (t1 & 0x3f) << 24
\r
324 | (t2 & 0x3f) << 16;
\r
326 // t1 = m_table_[3 * pos];
\r
327 // t2 = m_table_[3 * pos + 1];
\r
328 t1 = listheader.m_baseCE_;
\r
329 t2 = listheader.m_baseContCE_;
\r
331 listheader.m_gapsLo_[3 * st] = mergeCE(t1, t2,
\r
333 listheader.m_gapsLo_[3 * st + 1] = mergeCE(t1, t2,
\r
334 Collator.SECONDARY);
\r
335 listheader.m_gapsLo_[3 * st + 2] = (t1 & 0x3f) << 24
\r
336 | (t2 & 0x3f) << 16;
\r
343 * Gets the next CE in the inverse table
\r
345 * @param listheader
\r
346 * token list header
\r
348 * collation strength
\r
351 private final int getInverseNext(
\r
352 CollationRuleParser.TokenListHeader listheader, int strength) {
\r
353 int ce = listheader.m_baseCE_;
\r
354 int secondce = listheader.m_baseContCE_;
\r
355 int result = findInverseCE(ce, secondce);
\r
361 ce &= STRENGTH_MASK_[strength];
\r
362 secondce &= STRENGTH_MASK_[strength];
\r
365 int nextcontce = secondce;
\r
367 while ((nextce & STRENGTH_MASK_[strength]) == ce
\r
368 && (nextcontce & STRENGTH_MASK_[strength]) == secondce) {
\r
369 nextce = m_table_[3 * (++result)];
\r
370 nextcontce = m_table_[3 * result + 1];
\r
373 listheader.m_nextCE_ = nextce;
\r
374 listheader.m_nextContCE_ = nextcontce;
\r
380 // package private data members ------------------------------------------
\r
383 * Inverse UCA, instantiate only when required
\r
385 static final InverseUCA INVERSE_UCA_;
\r
388 * UCA and Inverse UCA version do not match
\r
390 private static final String INV_UCA_VERSION_MISMATCH_ = "UCA versions of UCA and inverse UCA should match";
\r
393 * UCA and Inverse UCA version do not match
\r
395 private static final String UCA_NOT_INSTANTIATED_ = "UCA is not instantiated!";
\r
398 * Initializing the inverse UCA
\r
401 InverseUCA temp = null;
\r
403 temp = CollatorReader.getInverseUCA();
\r
404 } catch (IOException e) {
\r
407 * try { String invdat = "/com/ibm/icu/impl/data/invuca.icu";
\r
409 * CollationParsedRuleBuilder.class.getResourceAsStream(invdat);
\r
410 * BufferedInputStream b = new BufferedInputStream(i, 110000);
\r
411 * INVERSE_UCA_ = CollatorReader.readInverseUCA(b); b.close();
\r
412 * i.close(); } catch (Exception e) { e.printStackTrace(); throw new
\r
413 * RuntimeException(e.getMessage()); }
\r
416 if (temp != null && RuleBasedCollator.UCA_ != null) {
\r
417 if (!temp.m_UCA_version_
\r
418 .equals(RuleBasedCollator.UCA_.m_UCA_version_)) {
\r
419 throw new RuntimeException(INV_UCA_VERSION_MISMATCH_);
\r
422 throw new RuntimeException(UCA_NOT_INSTANTIATED_);
\r
425 INVERSE_UCA_ = temp;
\r
428 // package private methods -----------------------------------------------
\r
431 * Parse and sets the collation rules in the argument collator
\r
435 * @exception Exception
\r
436 * thrown when internal program error occurs
\r
438 void setRules(RuleBasedCollator collator) throws Exception {
\r
439 if (m_parser_.m_resultLength_ > 0 || m_parser_.m_removeSet_ != null) {
\r
440 // we have a set of rules, let's make something of it
\r
441 assembleTailoringTable(collator);
\r
442 } else { // no rules, but no error either must be only options
\r
443 // We will init the collator from UCA
\r
444 collator.setWithUCATables();
\r
446 // And set only the options
\r
447 m_parser_.setDefaultOptionsInCollator(collator);
\r
450 private void copyRangeFromUCA(BuildTable t, int start, int end) {
\r
452 for (u = start; u <= end; u++) {
\r
453 // if ((CE = ucmpe32_get(t.m_mapping, u)) == UCOL_NOT_FOUND
\r
454 int CE = t.m_mapping_.getValue(u);
\r
455 if (CE == CE_NOT_FOUND_
\r
456 // this test is for contractions that are missing the starting
\r
457 // element. Looks like latin-1 should be done before
\r
458 // assembling the table, even if it results in more false
\r
459 // closure elements
\r
460 || (isContractionTableElement(CE) && getCE(
\r
461 t.m_contractions_, CE, 0) == CE_NOT_FOUND_)) {
\r
462 // m_utilElement_.m_uchars_ = str.toString();
\r
463 m_utilElement_.m_uchars_ = UCharacter.toString(u);
\r
464 m_utilElement_.m_cPoints_ = m_utilElement_.m_uchars_;
\r
465 m_utilElement_.m_prefix_ = 0;
\r
466 m_utilElement_.m_CELength_ = 0;
\r
467 m_utilElement_.m_prefixChars_ = null;
\r
468 m_utilColEIter_.setText(m_utilElement_.m_uchars_);
\r
469 while (CE != CollationElementIterator.NULLORDER) {
\r
470 CE = m_utilColEIter_.next();
\r
471 if (CE != CollationElementIterator.NULLORDER) {
\r
472 m_utilElement_.m_CEs_[m_utilElement_.m_CELength_++] = CE;
\r
475 addAnElement(t, m_utilElement_);
\r
481 * 2. Eliminate the negative lists by doing the following for each non-null
\r
482 * negative list: o if previousCE(baseCE, strongestN) != some ListHeader X's
\r
483 * baseCE, create new ListHeader X o reverse the list, add to the end of X's
\r
484 * positive list. Reset the strength of the first item you add, based on the
\r
485 * stronger strength levels of the two lists.
\r
487 * 3. For each ListHeader with a non-null positive list: o Find all
\r
488 * character strings with CEs between the baseCE and the next/previous CE,
\r
489 * at the strength of the first token. Add these to the tailoring. ? That
\r
490 * is, if UCA has ... x <<< X << x' <<< X' < y ..., and the tailoring has &
\r
491 * x < z... ? Then we change the tailoring to & x <<< X << x' <<< X' < z ...
\r
493 * It is possible that this part should be done even while constructing list
\r
494 * The problem is that it is unknown what is going to be the strongest
\r
495 * weight. So we might as well do it here o Allocate CEs for each token in
\r
496 * the list, based on the total number N of the largest level difference,
\r
497 * and the gap G between baseCE and nextCE at that level. The relation *
\r
498 * between the last item and nextCE is the same as the strongest strength. o
\r
499 * Example: baseCE < a << b <<< q << c < d < e * nextCE(X,1) ? There are 3
\r
500 * primary items: a, d, e. Fit them into the primary gap. Then fit b and c
\r
501 * into the secondary gap between a and d, then fit q into the tertiary gap
\r
502 * between b and c. o Example: baseCE << b <<< q << c * nextCE(X,2) ? There
\r
503 * are 2 secondary items: b, c. Fit them into the secondary gap. Then fit q
\r
504 * into the tertiary gap between b and c. o When incrementing primary
\r
505 * values, we will not cross high byte boundaries except where there is only
\r
506 * a single-byte primary. That is to ensure that the script reordering will
\r
507 * continue to work.
\r
510 * the rule based collator to update
\r
511 * @exception Exception
\r
512 * thrown when internal program error occurs
\r
514 void assembleTailoringTable(RuleBasedCollator collator) throws Exception {
\r
516 for (int i = 0; i < m_parser_.m_resultLength_; i++) {
\r
517 // now we need to generate the CEs
\r
518 // We stuff the initial value in the buffers, and increase the
\r
519 // appropriate buffer according to strength
\r
520 if (m_parser_.m_listHeader_[i].m_first_ != null) {
\r
521 // if there are any elements
\r
522 // due to the way parser works, subsequent tailorings
\r
523 // may remove all the elements from a sequence, therefore
\r
524 // leaving an empty tailoring sequence.
\r
525 initBuffers(m_parser_.m_listHeader_[i]);
\r
529 if (m_parser_.m_variableTop_ != null) {
\r
530 // stuff the variable top value
\r
531 m_parser_.m_options_.m_variableTopValue_ = m_parser_.m_variableTop_.m_CE_[0] >>> 16;
\r
532 // remove it from the list
\r
533 if (m_parser_.m_variableTop_.m_listHeader_.m_first_ == m_parser_.m_variableTop_) { // first
\r
536 m_parser_.m_variableTop_.m_listHeader_.m_first_ = m_parser_.m_variableTop_.m_next_;
\r
538 if (m_parser_.m_variableTop_.m_listHeader_.m_last_ == m_parser_.m_variableTop_) {
\r
540 m_parser_.m_variableTop_.m_listHeader_.m_last_ = m_parser_.m_variableTop_.m_previous_;
\r
542 if (m_parser_.m_variableTop_.m_next_ != null) {
\r
543 m_parser_.m_variableTop_.m_next_.m_previous_ = m_parser_.m_variableTop_.m_previous_;
\r
545 if (m_parser_.m_variableTop_.m_previous_ != null) {
\r
546 m_parser_.m_variableTop_.m_previous_.m_next_ = m_parser_.m_variableTop_.m_next_;
\r
550 BuildTable t = new BuildTable(m_parser_);
\r
552 // After this, we have assigned CE values to all regular CEs now we
\r
553 // will go through list once more and resolve expansions, make
\r
554 // UCAElements structs and add them to table
\r
555 for (int i = 0; i < m_parser_.m_resultLength_; i++) {
\r
556 // now we need to generate the CEs
\r
557 // We stuff the initial value in the buffers, and increase the
\r
558 // appropriate buffer according to strength */
\r
559 createElements(t, m_parser_.m_listHeader_[i]);
\r
562 m_utilElement_.clear();
\r
563 StringBuilder str = new StringBuilder();
\r
565 // add latin-1 stuff
\r
566 copyRangeFromUCA(t, 0, 0xFF);
\r
568 // add stuff for copying
\r
569 if (m_parser_.m_copySet_ != null) {
\r
571 for (i = 0; i < m_parser_.m_copySet_.getRangeCount(); i++) {
\r
572 copyRangeFromUCA(t, m_parser_.m_copySet_.getRangeStart(i),
\r
573 m_parser_.m_copySet_.getRangeEnd(i));
\r
577 // copy contractions from the UCA - this is felt mostly for cyrillic
\r
578 char conts[] = RuleBasedCollator.UCA_CONTRACTIONS_;
\r
580 while (conts[offset] != 0) {
\r
581 // tailoredCE = ucmpe32_get(t.m_mapping, *conts);
\r
582 int tailoredCE = t.m_mapping_.getValue(conts[offset]);
\r
583 Elements prefixElm = null;
\r
584 if (tailoredCE != CE_NOT_FOUND_) {
\r
585 boolean needToAdd = true;
\r
586 if (isContractionTableElement(tailoredCE)) {
\r
587 if (isTailored(t.m_contractions_, tailoredCE, conts,
\r
588 offset + 1) == true) {
\r
592 if (!needToAdd && isPrefix(tailoredCE)
\r
593 && conts[offset + 1] == 0) {
\r
594 // pre-context character in UCA
\r
595 // The format for pre-context character is
\r
596 // conts[0]: baseCP conts[1]:0 conts[2]:pre-context CP
\r
597 Elements elm = new Elements();
\r
598 elm.m_cPoints_ = m_utilElement_.m_uchars_;
\r
599 elm.m_CELength_ = 0;
\r
600 elm.m_uchars_ = UCharacter.toString(conts[offset]);
\r
601 elm.m_prefixChars_ = UCharacter.toString(conts[offset + 2]);
\r
602 elm.m_prefix_ = 0; // TODO(claireho) : confirm!
\r
603 prefixElm = t.m_prefixLookup_.get(elm);
\r
604 if ((prefixElm == null)
\r
605 || (prefixElm.m_prefixChars_.charAt(0) != conts[offset + 2])) {
\r
609 if (m_parser_.m_removeSet_ != null
\r
610 && m_parser_.m_removeSet_.contains(conts[offset])) {
\r
614 if (needToAdd == true) {
\r
615 // we need to add if this contraction is not tailored.
\r
616 if (conts[offset + 1] != 0) { // not precontext
\r
617 m_utilElement_.m_prefix_ = 0;
\r
618 m_utilElement_.m_prefixChars_ = null;
\r
619 m_utilElement_.m_cPoints_ = m_utilElement_.m_uchars_;
\r
620 str.delete(0, str.length());
\r
621 str.append(conts[offset]);
\r
622 str.append(conts[offset + 1]);
\r
623 if (conts[offset + 2] != 0) {
\r
624 str.append(conts[offset + 2]);
\r
626 m_utilElement_.m_uchars_ = str.toString();
\r
627 m_utilElement_.m_CELength_ = 0;
\r
628 m_utilColEIter_.setText(m_utilElement_.m_uchars_);
\r
629 } else { // add a pre-context element
\r
631 str.delete(0, str.length()); // clean up
\r
632 m_utilElement_.m_cPoints_ = UCharacter
\r
633 .toString(conts[offset]);
\r
634 m_utilElement_.m_CELength_ = 0;
\r
635 m_utilElement_.m_uchars_ = UCharacter
\r
636 .toString(conts[offset]);
\r
637 m_utilElement_.m_prefixChars_ = UCharacter
\r
638 .toString(conts[offset + 2]);
\r
639 if (prefixElm == null) {
\r
640 m_utilElement_.m_prefix_ = 0;
\r
641 } else { // TODO (claireho): confirm!
\r
642 m_utilElement_.m_prefix_ = m_utilElement_.m_prefix_;
\r
643 // m_utilElement_.m_prefix_= prefixElm.m_prefix_;
\r
645 m_utilColEIter_.setText(m_utilElement_.m_prefixChars_);
\r
646 while (m_utilColEIter_.next() != CollationElementIterator.NULLORDER) {
\r
647 // count number of keys for pre-context char.
\r
650 str.append(conts[offset + 2]);
\r
651 str.append(conts[offset]);
\r
652 m_utilColEIter_.setText(str.toString());
\r
653 // Skip the keys for prefix character, then copy the
\r
655 while ((preKeyLen-- > 0)
\r
656 && m_utilColEIter_.next() != CollationElementIterator.NULLORDER) {
\r
662 int CE = m_utilColEIter_.next();
\r
663 if (CE != CollationElementIterator.NULLORDER) {
\r
664 m_utilElement_.m_CEs_[m_utilElement_.m_CELength_++] = CE;
\r
669 addAnElement(t, m_utilElement_);
\r
671 } else if (m_parser_.m_removeSet_ != null
\r
672 && m_parser_.m_removeSet_.contains(conts[offset])) {
\r
673 copyRangeFromUCA(t, conts[offset], conts[offset]);
\r
679 // Add completely ignorable elements
\r
680 processUCACompleteIgnorables(t);
\r
682 // canonical closure
\r
683 canonicalClosure(t);
\r
685 // still need to produce compatibility closure
\r
686 assembleTable(t, collator);
\r
689 // private inner classes -------------------------------------------------
\r
691 @SuppressWarnings("unused")
\r
692 private static class CEGenerator {
\r
693 // package private data members --------------------------------------
\r
695 WeightRange m_ranges_[];
\r
696 int m_rangesLength_;
\r
703 int m_fLow_; // forbidden Low
\r
704 int m_fHigh_; // forbidden High
\r
706 // package private constructor ---------------------------------------
\r
709 m_ranges_ = new WeightRange[7];
\r
710 for (int i = 6; i >= 0; i--) {
\r
711 m_ranges_[i] = new WeightRange();
\r
716 private static class WeightRange implements Comparable<WeightRange> {
\r
717 // public methods ----------------------------------------------------
\r
720 * Compares this object with target
\r
722 * @param target object to compare with
\r
723 * @return 0 if equals, 1 if this is > target, -1 otherwise
\r
725 public int compareTo(WeightRange target) {
\r
726 if (this == target) {
\r
729 int tstart = target.m_start_;
\r
730 if (m_start_ == tstart) {
\r
733 if (m_start_ > tstart) {
\r
742 public void clear() {
\r
751 // package private data members --------------------------------------
\r
760 // package private constructor ---------------------------------------
\r
767 * Copy constructor. Cloneable is troublesome, needs to check for
\r
773 WeightRange(WeightRange source) {
\r
774 m_start_ = source.m_start_;
\r
775 m_end_ = source.m_end_;
\r
776 m_length_ = source.m_length_;
\r
777 m_count_ = source.m_count_;
\r
778 m_length2_ = source.m_length2_;
\r
779 m_count2_ = source.m_count2_;
\r
783 private static class MaxJamoExpansionTable {
\r
784 // package private data members --------------------------------------
\r
786 Vector<Integer> m_endExpansionCE_;
\r
787 // vector of booleans
\r
788 Vector<Boolean> m_isV_;
\r
793 // package private constructor ---------------------------------------
\r
795 MaxJamoExpansionTable() {
\r
796 m_endExpansionCE_ = new Vector<Integer>();
\r
797 m_isV_ = new Vector<Boolean>();
\r
798 m_endExpansionCE_.add(new Integer(0));
\r
799 m_isV_.add(Boolean.FALSE);
\r
805 MaxJamoExpansionTable(MaxJamoExpansionTable table) {
\r
806 m_endExpansionCE_ = new Vector<Integer>(table.m_endExpansionCE_);
\r
807 m_isV_ = new Vector<Boolean>(table.m_isV_);
\r
808 m_maxLSize_ = table.m_maxLSize_;
\r
809 m_maxVSize_ = table.m_maxVSize_;
\r
810 m_maxTSize_ = table.m_maxTSize_;
\r
814 private static class MaxExpansionTable {
\r
815 // package private constructor --------------------------------------
\r
817 MaxExpansionTable() {
\r
818 m_endExpansionCE_ = new Vector<Integer>();
\r
819 m_expansionCESize_ = new Vector<Byte>();
\r
820 m_endExpansionCE_.add(new Integer(0));
\r
821 m_expansionCESize_.add(new Byte((byte) 0));
\r
824 MaxExpansionTable(MaxExpansionTable table) {
\r
825 m_endExpansionCE_ = new Vector<Integer>(table.m_endExpansionCE_);
\r
826 m_expansionCESize_ = new Vector<Byte>(table.m_expansionCESize_);
\r
829 // package private data member --------------------------------------
\r
831 Vector<Integer> m_endExpansionCE_;
\r
832 Vector<Byte> m_expansionCESize_;
\r
835 private static class BasicContractionTable {
\r
836 // package private constructors -------------------------------------
\r
838 BasicContractionTable() {
\r
839 m_CEs_ = new Vector<Integer>();
\r
840 m_codePoints_ = new StringBuilder();
\r
843 // package private data members -------------------------------------
\r
845 StringBuilder m_codePoints_;
\r
846 Vector<Integer> m_CEs_;
\r
849 private static class ContractionTable {
\r
850 // package private constructor --------------------------------------
\r
853 * Builds a contraction table
\r
857 ContractionTable(IntTrieBuilder mapping) {
\r
858 m_mapping_ = mapping;
\r
859 m_elements_ = new Vector<BasicContractionTable>();
\r
860 m_CEs_ = new Vector<Integer>();
\r
861 m_codePoints_ = new StringBuilder();
\r
862 m_offsets_ = new Vector<Integer>();
\r
863 m_currentTag_ = CE_NOT_FOUND_TAG_;
\r
867 * Copies a contraction table. Not all data will be copied into their
\r
872 ContractionTable(ContractionTable table) {
\r
873 m_mapping_ = table.m_mapping_;
\r
874 m_elements_ = new Vector<BasicContractionTable>(table.m_elements_);
\r
875 m_codePoints_ = new StringBuilder(table.m_codePoints_);
\r
876 m_CEs_ = new Vector<Integer>(table.m_CEs_);
\r
877 m_offsets_ = new Vector<Integer>(table.m_offsets_);
\r
878 m_currentTag_ = table.m_currentTag_;
\r
881 // package private data members ------------------------------------
\r
884 * Vector of BasicContractionTable
\r
886 Vector<BasicContractionTable> m_elements_;
\r
887 IntTrieBuilder m_mapping_;
\r
888 StringBuilder m_codePoints_;
\r
889 Vector<Integer> m_CEs_;
\r
890 Vector<Integer> m_offsets_;
\r
895 * Private class for combining mark table. The table is indexed by the class
\r
898 @SuppressWarnings("unused")
\r
899 private static class CombinClassTable {
\r
901 * accumulated numbers of combining marks.
\r
903 int[] index = new int[256];
\r
906 * code point array for combining marks.
\r
916 CombinClassTable() {
\r
924 * Copy the combining mark table from ccc and index in compact way.
\r
927 * : code point array
\r
931 * : index of combining classes(0-255)
\r
933 void generate(char[] cps, int numOfCM, int[] ccIndex) {
\r
936 cPoints = new char[numOfCM];
\r
937 for (int i = 0; i < 256; i++) {
\r
938 for (int j = 0; j < ccIndex[i]; j++) {
\r
939 cPoints[count++] = cps[(i << 8) + j];
\r
947 * Get first CM(combining mark) with the combining class value cClass.
\r
950 * : combining class value.
\r
951 * @return combining mark codepoint or 0 if no combining make with class
\r
954 char GetFirstCM(int cClass) {
\r
956 if (cPoints == null || cClass == 0
\r
957 || index[cClass] == index[cClass - 1]) {
\r
961 return cPoints[index[cClass - 1]];
\r
965 * Get next CM(combining mark) with the combining class value cClass.
\r
966 * Return combining mark codepoint or 0 if no next CM.
\r
969 if (cPoints == null
\r
970 || index[curClass] == (index[curClass - 1] + pos)) {
\r
973 return cPoints[index[curClass - 1] + (pos++)];
\r
976 // private data members
\r
981 private static final class BuildTable implements TrieBuilder.DataManipulate {
\r
982 // package private methods ------------------------------------------
\r
985 * For construction of the Trie tables. Has to be labeled public
\r
987 * @param cp The value of the code point.
\r
988 * @param offset The value of the offset.
\r
989 * @return data offset or 0
\r
991 public int getFoldedValue(int cp, int offset) {
\r
992 int limit = cp + 0x400;
\r
993 while (cp < limit) {
\r
994 int value = m_mapping_.getValue(cp);
\r
995 boolean inBlockZero = m_mapping_.isInZeroBlock(cp);
\r
996 int tag = getCETag(value);
\r
997 if (inBlockZero == true) {
\r
998 cp += TrieBuilder.DATA_BLOCK_LENGTH;
\r
999 } else if (!(isSpecial(value) && (tag == CE_IMPLICIT_TAG_ || tag == CE_NOT_FOUND_TAG_))) {
\r
1000 // These are values that are starting in either UCA
\r
1001 // (IMPLICIT_TAG) or in the tailorings (NOT_FOUND_TAG).
\r
1002 // Presence of these tags means that there is nothing in
\r
1003 // this position and that it should be skipped.
\r
1004 return RuleBasedCollator.CE_SPECIAL_FLAG_
\r
1005 | (CE_SURROGATE_TAG_ << 24) | offset;
\r
1013 // package private constructor --------------------------------------
\r
1018 BuildTable(CollationRuleParser parser) {
\r
1019 m_collator_ = new RuleBasedCollator();
\r
1020 m_collator_.setWithUCAData();
\r
1021 MaxExpansionTable maxet = new MaxExpansionTable();
\r
1022 MaxJamoExpansionTable maxjet = new MaxJamoExpansionTable();
\r
1023 m_options_ = parser.m_options_;
\r
1024 m_expansions_ = new Vector<Integer>();
\r
1025 // Do your own mallocs for the structure, array and have linear
\r
1027 int trieinitialvalue = RuleBasedCollator.CE_SPECIAL_FLAG_
\r
1028 | (CE_NOT_FOUND_TAG_ << 24);
\r
1029 // temporary fix for jb3822, 0x100000 -> 30000
\r
1030 m_mapping_ = new IntTrieBuilder(null, 0x30000, trieinitialvalue,
\r
1031 trieinitialvalue, true);
\r
1032 m_prefixLookup_ = new Hashtable<Elements, Elements>();
\r
1033 // uhash_open(prefixLookupHash, prefixLookupComp);
\r
1034 m_contractions_ = new ContractionTable(m_mapping_);
\r
1035 // copy UCA's maxexpansion and merge as we go along
\r
1036 m_maxExpansions_ = maxet;
\r
1037 // adding an extra initial value for easier manipulation
\r
1038 for (int i = 0; i < RuleBasedCollator.UCA_.m_expansionEndCE_.length; i++) {
\r
1039 maxet.m_endExpansionCE_.add(new Integer(
\r
1040 RuleBasedCollator.UCA_.m_expansionEndCE_[i]));
\r
1041 maxet.m_expansionCESize_.add(new Byte(
\r
1042 RuleBasedCollator.UCA_.m_expansionEndCEMaxSize_[i]));
\r
1044 m_maxJamoExpansions_ = maxjet;
\r
1046 m_unsafeCP_ = new byte[UNSAFECP_TABLE_SIZE_];
\r
1047 m_contrEndCP_ = new byte[UNSAFECP_TABLE_SIZE_];
\r
1048 Arrays.fill(m_unsafeCP_, (byte) 0);
\r
1049 Arrays.fill(m_contrEndCP_, (byte) 0);
\r
1053 * Duplicating a BuildTable. Not all data will be duplicated into their
\r
1059 BuildTable(BuildTable table) {
\r
1060 m_collator_ = table.m_collator_;
\r
1061 m_mapping_ = new IntTrieBuilder(table.m_mapping_);
\r
1062 m_expansions_ = new Vector<Integer>(table.m_expansions_);
\r
1063 m_contractions_ = new ContractionTable(table.m_contractions_);
\r
1064 m_contractions_.m_mapping_ = m_mapping_;
\r
1065 m_options_ = table.m_options_;
\r
1066 m_maxExpansions_ = new MaxExpansionTable(table.m_maxExpansions_);
\r
1067 m_maxJamoExpansions_ = new MaxJamoExpansionTable(
\r
1068 table.m_maxJamoExpansions_);
\r
1069 m_unsafeCP_ = new byte[table.m_unsafeCP_.length];
\r
1070 System.arraycopy(table.m_unsafeCP_, 0, m_unsafeCP_, 0,
\r
1071 m_unsafeCP_.length);
\r
1072 m_contrEndCP_ = new byte[table.m_contrEndCP_.length];
\r
1073 System.arraycopy(table.m_contrEndCP_, 0, m_contrEndCP_, 0,
\r
1074 m_contrEndCP_.length);
\r
1077 // package private data members -------------------------------------
\r
1079 RuleBasedCollator m_collator_;
\r
1080 IntTrieBuilder m_mapping_;
\r
1081 Vector<Integer> m_expansions_;
\r
1082 ContractionTable m_contractions_;
\r
1083 // UCATableHeader image;
\r
1084 CollationRuleParser.OptionSet m_options_;
\r
1085 MaxExpansionTable m_maxExpansions_;
\r
1086 MaxJamoExpansionTable m_maxJamoExpansions_;
\r
1087 byte m_unsafeCP_[];
\r
1088 byte m_contrEndCP_[];
\r
1089 Hashtable<Elements, Elements> m_prefixLookup_;
\r
1090 CombinClassTable cmLookup = null;
\r
1093 private static class Elements {
\r
1094 // package private data members -------------------------------------
\r
1096 String m_prefixChars_;
\r
1102 String m_cPoints_;
\r
1104 * Offset to the working string
\r
1106 int m_cPointsOffset_;
\r
1108 * These are collation elements - there could be more than one - in case
\r
1114 * This is the value element maps in original table
\r
1117 int m_sizePrim_[];
\r
1120 boolean m_variableTop_;
\r
1121 boolean m_caseBit_;
\r
1123 // package private constructors -------------------------------------
\r
1126 * Package private constructor
\r
1129 m_sizePrim_ = new int[128];
\r
1130 m_sizeSec_ = new int[128];
\r
1131 m_sizeTer_ = new int[128];
\r
1132 m_CEs_ = new int[256];
\r
1137 * Package private constructor
\r
1139 Elements(Elements element) {
\r
1140 m_prefixChars_ = element.m_prefixChars_;
\r
1141 m_prefix_ = element.m_prefix_;
\r
1142 m_uchars_ = element.m_uchars_;
\r
1143 m_cPoints_ = element.m_cPoints_;
\r
1144 m_cPointsOffset_ = element.m_cPointsOffset_;
\r
1145 m_CEs_ = element.m_CEs_;
\r
1146 m_CELength_ = element.m_CELength_;
\r
1147 m_mapCE_ = element.m_mapCE_;
\r
1148 m_sizePrim_ = element.m_sizePrim_;
\r
1149 m_sizeSec_ = element.m_sizeSec_;
\r
1150 m_sizeTer_ = element.m_sizeTer_;
\r
1151 m_variableTop_ = element.m_variableTop_;
\r
1152 m_caseBit_ = element.m_caseBit_;
\r
1155 // package private methods -------------------------------------------
\r
1158 * Initializing the elements
\r
1160 public void clear() {
\r
1161 m_prefixChars_ = null;
\r
1164 m_cPoints_ = null;
\r
1165 m_cPointsOffset_ = 0;
\r
1168 Arrays.fill(m_sizePrim_, 0);
\r
1169 Arrays.fill(m_sizeSec_, 0);
\r
1170 Arrays.fill(m_sizeTer_, 0);
\r
1171 m_variableTop_ = false;
\r
1172 m_caseBit_ = false;
\r
1176 * Hashcode calculation for token
\r
1178 * @return the hashcode
\r
1180 public int hashCode() {
\r
1181 String str = m_cPoints_.substring(m_cPointsOffset_);
\r
1182 return str.hashCode();
\r
1186 * Equals calculation
\r
1188 * @param target Object to compare
\r
1189 * @return true if target is the same as this object
\r
1191 public boolean equals(Object target) {
\r
1192 if (target == this) {
\r
1195 if (target instanceof Elements) {
\r
1196 Elements t = (Elements) target;
\r
1197 int size = m_cPoints_.length() - m_cPointsOffset_;
\r
1198 if (size == t.m_cPoints_.length() - t.m_cPointsOffset_) {
\r
1199 return t.m_cPoints_.regionMatches(t.m_cPointsOffset_,
\r
1200 m_cPoints_, m_cPointsOffset_, size);
\r
1207 // private data member ---------------------------------------------------
\r
1210 * Maximum strength used in CE building
\r
1212 private static final int CE_BASIC_STRENGTH_LIMIT_ = 3;
\r
1214 * Maximum collation strength
\r
1216 private static final int CE_STRENGTH_LIMIT_ = 16;
\r
1218 * Strength mask array, used in inverse UCA
\r
1220 private static final int STRENGTH_MASK_[] = { 0xFFFF0000, 0xFFFFFF00,
\r
1223 * CE tag for not found
\r
1225 private static final int CE_NOT_FOUND_ = 0xF0000000;
\r
1227 * CE tag for not found
\r
1229 private static final int CE_NOT_FOUND_TAG_ = 0;
\r
1231 * This code point results in an expansion
\r
1233 private static final int CE_EXPANSION_TAG_ = 1;
\r
1235 * Start of a contraction
\r
1237 private static final int CE_CONTRACTION_TAG_ = 2;
\r
1239 * Thai character - do the reordering
\r
1241 // private static final int CE_THAI_TAG_ = 3;
\r
1243 * Charset processing, not yet implemented
\r
1245 // private static final int CE_CHARSET_TAG_ = 4;
\r
1247 * Lead surrogate that is tailored and doesn't start a contraction
\r
1249 private static final int CE_SURROGATE_TAG_ = 5;
\r
1253 // private static final int CE_HANGUL_SYLLABLE_TAG_ = 6;
\r
1257 // private static final int CE_LEAD_SURROGATE_TAG_ = 7;
\r
1261 // private static final int CE_TRAIL_SURROGATE_TAG_ = 8;
\r
1263 * 0x3400-0x4DB5, 0x4E00-0x9FA5, 0xF900-0xFA2D
\r
1265 // private static final int CE_CJK_IMPLICIT_TAG_ = 9;
\r
1266 private static final int CE_IMPLICIT_TAG_ = 10;
\r
1267 private static final int CE_SPEC_PROC_TAG_ = 11;
\r
1269 * This is a three byte primary with starting secondaries and tertiaries. It
\r
1270 * fits in a single 32 bit CE and is used instead of expansion to save space
\r
1271 * without affecting the performance (hopefully)
\r
1273 private static final int CE_LONG_PRIMARY_TAG_ = 12;
\r
1275 * Unsafe UChar hash table table size. Size is 32 bytes for 1 bit for each
\r
1276 * latin 1 char + some power of two for hashing the rest of the chars. Size
\r
1279 private static final int UNSAFECP_TABLE_SIZE_ = 1056;
\r
1281 * Mask value down to "some power of two" -1. Number of bits, not num of
\r
1284 private static final int UNSAFECP_TABLE_MASK_ = 0x1fff;
\r
1288 private static final int UPPER_CASE_ = 0x80;
\r
1289 private static final int MIXED_CASE_ = 0x40;
\r
1290 private static final int LOWER_CASE_ = 0x00;
\r
1292 * Initial table size
\r
1294 // private static final int INIT_TABLE_SIZE_ = 1028;
\r
1296 * Header size, copied from ICU4C, to be changed when that value changes
\r
1298 // private static final int HEADER_SIZE_ = 0xC4;
\r
1300 * Contraction table new element indicator
\r
1302 private static final int CONTRACTION_TABLE_NEW_ELEMENT_ = 0xFFFFFF;
\r
1304 * Parser for the rules
\r
1306 private CollationRuleParser m_parser_;
\r
1308 * Utility UCA collation element iterator
\r
1310 private CollationElementIterator m_utilColEIter_;
\r
1312 * Utility data members
\r
1314 private CEGenerator m_utilGens_[] = { new CEGenerator(), new CEGenerator(),
\r
1315 new CEGenerator() };
\r
1316 private int m_utilCEBuffer_[] = new int[CE_BASIC_STRENGTH_LIMIT_];
\r
1317 private int m_utilIntBuffer_[] = new int[CE_STRENGTH_LIMIT_];
\r
1318 private Elements m_utilElement_ = new Elements();
\r
1319 private Elements m_utilElement2_ = new Elements();
\r
1320 private CollationRuleParser.Token m_utilToken_ = new CollationRuleParser.Token();
\r
1321 private int m_utilCountBuffer_[] = new int[6];
\r
1322 private long m_utilLongBuffer_[] = new long[5];
\r
1323 private WeightRange m_utilLowerWeightRange_[] = { new WeightRange(),
\r
1324 new WeightRange(), new WeightRange(), new WeightRange(),
\r
1325 new WeightRange() };
\r
1326 private WeightRange m_utilUpperWeightRange_[] = { new WeightRange(),
\r
1327 new WeightRange(), new WeightRange(), new WeightRange(),
\r
1328 new WeightRange() };
\r
1329 private WeightRange m_utilWeightRange_ = new WeightRange();
\r
1330 private final Normalizer2Impl m_nfcImpl_ = Norm2AllModes.getNFCInstance().impl;
\r
1331 private CanonicalIterator m_utilCanIter_ = new CanonicalIterator("");
\r
1332 private StringBuilder m_utilStringBuffer_ = new StringBuilder("");
\r
1333 // Flag indicating a combining marks table is required or not.
\r
1334 private static boolean buildCMTabFlag = false;
\r
1336 // private methods -------------------------------------------------------
\r
1339 * @param listheader
\r
1340 * parsed rule tokens
\r
1341 * @exception Exception
\r
1342 * thrown when internal error occurs
\r
1344 private void initBuffers(CollationRuleParser.TokenListHeader listheader)
\r
1345 throws Exception {
\r
1346 CollationRuleParser.Token token = listheader.m_last_;
\r
1347 Arrays.fill(m_utilIntBuffer_, 0, CE_STRENGTH_LIMIT_, 0);
\r
1349 token.m_toInsert_ = 1;
\r
1350 m_utilIntBuffer_[token.m_strength_] = 1;
\r
1351 while (token.m_previous_ != null) {
\r
1352 if (token.m_previous_.m_strength_ < token.m_strength_) {
\r
1354 m_utilIntBuffer_[token.m_strength_] = 0;
\r
1355 m_utilIntBuffer_[token.m_previous_.m_strength_]++;
\r
1356 } else if (token.m_previous_.m_strength_ > token.m_strength_) {
\r
1358 m_utilIntBuffer_[token.m_previous_.m_strength_] = 1;
\r
1360 m_utilIntBuffer_[token.m_strength_]++;
\r
1362 token = token.m_previous_;
\r
1363 token.m_toInsert_ = m_utilIntBuffer_[token.m_strength_];
\r
1366 token.m_toInsert_ = m_utilIntBuffer_[token.m_strength_];
\r
1367 INVERSE_UCA_.getInverseGapPositions(listheader);
\r
1369 token = listheader.m_first_;
\r
1370 int fstrength = Collator.IDENTICAL;
\r
1371 int initstrength = Collator.IDENTICAL;
\r
1373 m_utilCEBuffer_[Collator.PRIMARY] = mergeCE(listheader.m_baseCE_,
\r
1374 listheader.m_baseContCE_, Collator.PRIMARY);
\r
1375 m_utilCEBuffer_[Collator.SECONDARY] = mergeCE(listheader.m_baseCE_,
\r
1376 listheader.m_baseContCE_, Collator.SECONDARY);
\r
1377 m_utilCEBuffer_[Collator.TERTIARY] = mergeCE(listheader.m_baseCE_,
\r
1378 listheader.m_baseContCE_, Collator.TERTIARY);
\r
1379 while (token != null) {
\r
1380 fstrength = token.m_strength_;
\r
1381 if (fstrength < initstrength) {
\r
1382 initstrength = fstrength;
\r
1383 if (listheader.m_pos_[fstrength] == -1) {
\r
1384 while (listheader.m_pos_[fstrength] == -1 && fstrength > 0) {
\r
1387 if (listheader.m_pos_[fstrength] == -1) {
\r
1388 throw new Exception("Internal program error");
\r
1391 if (initstrength == Collator.TERTIARY) {
\r
1392 // starting with tertiary
\r
1393 m_utilCEBuffer_[Collator.PRIMARY] = listheader.m_gapsLo_[fstrength * 3];
\r
1394 m_utilCEBuffer_[Collator.SECONDARY] = listheader.m_gapsLo_[fstrength * 3 + 1];
\r
1395 m_utilCEBuffer_[Collator.TERTIARY] = getCEGenerator(
\r
1396 m_utilGens_[Collator.TERTIARY],
\r
1397 listheader.m_gapsLo_, listheader.m_gapsHi_, token,
\r
1399 } else if (initstrength == Collator.SECONDARY) {
\r
1401 m_utilCEBuffer_[Collator.PRIMARY] = listheader.m_gapsLo_[fstrength * 3];
\r
1402 m_utilCEBuffer_[Collator.SECONDARY] = getCEGenerator(
\r
1403 m_utilGens_[Collator.SECONDARY],
\r
1404 listheader.m_gapsLo_, listheader.m_gapsHi_, token,
\r
1406 m_utilCEBuffer_[Collator.TERTIARY] = getSimpleCEGenerator(
\r
1407 m_utilGens_[Collator.TERTIARY], token,
\r
1408 Collator.TERTIARY);
\r
1411 m_utilCEBuffer_[Collator.PRIMARY] = getCEGenerator(
\r
1412 m_utilGens_[Collator.PRIMARY],
\r
1413 listheader.m_gapsLo_, listheader.m_gapsHi_, token,
\r
1415 m_utilCEBuffer_[Collator.SECONDARY] = getSimpleCEGenerator(
\r
1416 m_utilGens_[Collator.SECONDARY], token,
\r
1417 Collator.SECONDARY);
\r
1418 m_utilCEBuffer_[Collator.TERTIARY] = getSimpleCEGenerator(
\r
1419 m_utilGens_[Collator.TERTIARY], token,
\r
1420 Collator.TERTIARY);
\r
1423 if (token.m_strength_ == Collator.TERTIARY) {
\r
1424 m_utilCEBuffer_[Collator.TERTIARY] = getNextGenerated(m_utilGens_[Collator.TERTIARY]);
\r
1425 } else if (token.m_strength_ == Collator.SECONDARY) {
\r
1426 m_utilCEBuffer_[Collator.SECONDARY] = getNextGenerated(m_utilGens_[Collator.SECONDARY]);
\r
1427 m_utilCEBuffer_[Collator.TERTIARY] = getSimpleCEGenerator(
\r
1428 m_utilGens_[Collator.TERTIARY], token,
\r
1429 Collator.TERTIARY);
\r
1430 } else if (token.m_strength_ == Collator.PRIMARY) {
\r
1431 m_utilCEBuffer_[Collator.PRIMARY] = getNextGenerated(m_utilGens_[Collator.PRIMARY]);
\r
1432 m_utilCEBuffer_[Collator.SECONDARY] = getSimpleCEGenerator(
\r
1433 m_utilGens_[Collator.SECONDARY], token,
\r
1434 Collator.SECONDARY);
\r
1435 m_utilCEBuffer_[Collator.TERTIARY] = getSimpleCEGenerator(
\r
1436 m_utilGens_[Collator.TERTIARY], token,
\r
1437 Collator.TERTIARY);
\r
1440 doCE(m_utilCEBuffer_, token);
\r
1441 token = token.m_next_;
\r
1446 * Get the next generated ce
\r
1450 * @return next generated ce
\r
1452 private int getNextGenerated(CEGenerator g) {
\r
1453 g.m_current_ = nextWeight(g);
\r
1454 return g.m_current_;
\r
1463 * @return ce generator
\r
1464 * @exception Exception
\r
1465 * thrown when internal error occurs
\r
1467 private int getSimpleCEGenerator(CEGenerator g,
\r
1468 CollationRuleParser.Token token, int strength) throws Exception {
\r
1469 int high, low, count = 1;
\r
1470 int maxbyte = (strength == Collator.TERTIARY) ? 0x3F : 0xFF;
\r
1472 if (strength == Collator.SECONDARY) {
\r
1473 low = RuleBasedCollator.COMMON_TOP_2_ << 24;
\r
1474 high = 0xFFFFFFFF;
\r
1475 count = 0xFF - RuleBasedCollator.COMMON_TOP_2_;
\r
1477 low = RuleBasedCollator.BYTE_COMMON_ << 24; // 0x05000000;
\r
1478 high = 0x40000000;
\r
1479 count = 0x40 - RuleBasedCollator.BYTE_COMMON_;
\r
1482 if (token.m_next_ != null && token.m_next_.m_strength_ == strength) {
\r
1483 count = token.m_next_.m_toInsert_;
\r
1486 g.m_rangesLength_ = allocateWeights(low, high, count, maxbyte,
\r
1488 g.m_current_ = RuleBasedCollator.BYTE_COMMON_ << 24;
\r
1490 if (g.m_rangesLength_ == 0) {
\r
1491 throw new Exception("Internal program error");
\r
1493 return g.m_current_;
\r
1497 * Combines 2 ce into one with respect to the argument strength
\r
1505 * @return combined ce
\r
1507 private static int mergeCE(int ce1, int ce2, int strength) {
\r
1508 int mask = RuleBasedCollator.CE_TERTIARY_MASK_;
\r
1509 if (strength == Collator.SECONDARY) {
\r
1510 mask = RuleBasedCollator.CE_SECONDARY_MASK_;
\r
1511 } else if (strength == Collator.PRIMARY) {
\r
1512 mask = RuleBasedCollator.CE_PRIMARY_MASK_;
\r
1516 switch (strength) {
\r
1517 case Collator.PRIMARY:
\r
1518 return ce1 | ce2 >>> 16;
\r
1519 case Collator.SECONDARY:
\r
1520 return ce1 << 16 | ce2 << 8;
\r
1522 return ce1 << 24 | ce2 << 16;
\r
1535 * @param fstrength
\r
1536 * @exception Exception
\r
1537 * thrown when internal error occurs
\r
1539 private int getCEGenerator(CEGenerator g, int lows[], int highs[],
\r
1540 CollationRuleParser.Token token, int fstrength) throws Exception {
\r
1541 int strength = token.m_strength_;
\r
1542 int low = lows[fstrength * 3 + strength];
\r
1543 int high = highs[fstrength * 3 + strength];
\r
1545 if (strength == Collator.TERTIARY) {
\r
1547 } else if (strength == Collator.PRIMARY) {
\r
1553 int count = token.m_toInsert_;
\r
1555 if (Utility.compareUnsigned(low, high) >= 0
\r
1556 && strength > Collator.PRIMARY) {
\r
1560 if (lows[fstrength * 3 + s] != highs[fstrength * 3 + s]) {
\r
1561 if (strength == Collator.SECONDARY) {
\r
1562 if (low < (RuleBasedCollator.COMMON_TOP_2_ << 24)) {
\r
1563 // Override if low range is less than
\r
1564 // UCOL_COMMON_TOP2.
\r
1565 low = RuleBasedCollator.COMMON_TOP_2_ << 24;
\r
1567 high = 0xFFFFFFFF;
\r
1569 if (low < RuleBasedCollator.COMMON_BOTTOM_3 << 24) {
\r
1570 // Override if low range is less than
\r
1571 // UCOL_COMMON_BOT3.
\r
1572 low = RuleBasedCollator.COMMON_BOTTOM_3 << 24;
\r
1574 high = 0x40000000;
\r
1579 throw new Exception("Internal program error");
\r
1586 if (strength == Collator.SECONDARY) { // similar as simple
\r
1587 if (Utility.compareUnsigned(low,
\r
1588 RuleBasedCollator.COMMON_BOTTOM_2_ << 24) >= 0
\r
1589 && Utility.compareUnsigned(low,
\r
1590 RuleBasedCollator.COMMON_TOP_2_ << 24) < 0) {
\r
1591 low = RuleBasedCollator.COMMON_TOP_2_ << 24;
\r
1593 if (Utility.compareUnsigned(high,
\r
1594 RuleBasedCollator.COMMON_BOTTOM_2_ << 24) > 0
\r
1595 && Utility.compareUnsigned(high,
\r
1596 RuleBasedCollator.COMMON_TOP_2_ << 24) < 0) {
\r
1597 high = RuleBasedCollator.COMMON_TOP_2_ << 24;
\r
1599 if (Utility.compareUnsigned(low,
\r
1600 RuleBasedCollator.COMMON_BOTTOM_2_ << 24) < 0) {
\r
1601 g.m_rangesLength_ = allocateWeights(
\r
1602 RuleBasedCollator.BYTE_UNSHIFTED_MIN_ << 24, high,
\r
1603 count, maxbyte, g.m_ranges_);
\r
1604 g.m_current_ = nextWeight(g);
\r
1605 // g.m_current_ = RuleBasedCollator.COMMON_BOTTOM_2_ << 24;
\r
1606 return g.m_current_;
\r
1610 g.m_rangesLength_ = allocateWeights(low, high, count, maxbyte,
\r
1612 if (g.m_rangesLength_ == 0) {
\r
1613 throw new Exception("Internal program error");
\r
1615 g.m_current_ = nextWeight(g);
\r
1616 return g.m_current_;
\r
1621 * list of collation elements parts
\r
1624 * @exception Exception
\r
1625 * thrown when forming case bits for expansions fails
\r
1627 private void doCE(int ceparts[], CollationRuleParser.Token token)
\r
1628 throws Exception {
\r
1629 // this one makes the table and stuff
\r
1630 // int noofbytes[] = new int[3];
\r
1631 for (int i = 0; i < 3; i++) {
\r
1632 // noofbytes[i] = countBytes(ceparts[i]);
\r
1633 m_utilIntBuffer_[i] = countBytes(ceparts[i]);
\r
1636 // Here we have to pack CEs from parts
\r
1640 while ((cei << 1) < m_utilIntBuffer_[0] || cei < m_utilIntBuffer_[1]
\r
1641 || cei < m_utilIntBuffer_[2]) {
\r
1643 value = RuleBasedCollator.CE_CONTINUATION_MARKER_;
\r
1648 if ((cei << 1) < m_utilIntBuffer_[0]) {
\r
1649 value |= ((ceparts[0] >> (32 - ((cei + 1) << 4))) & 0xFFFF) << 16;
\r
1651 if (cei < m_utilIntBuffer_[1]) {
\r
1652 value |= ((ceparts[1] >> (32 - ((cei + 1) << 3))) & 0xFF) << 8;
\r
1655 if (cei < m_utilIntBuffer_[2]) {
\r
1656 value |= ((ceparts[2] >> (32 - ((cei + 1) << 3))) & 0x3F);
\r
1658 token.m_CE_[cei] = value;
\r
1661 if (cei == 0) { // totally ignorable
\r
1662 token.m_CELength_ = 1;
\r
1663 token.m_CE_[0] = 0;
\r
1664 } else { // there is at least something
\r
1665 token.m_CELength_ = cei;
\r
1668 // Case bits handling for expansion
\r
1669 if (token.m_CE_[0] != 0) { // case bits should be set only for
\r
1671 int startoftokenrule = token.m_source_ & 0xFF;
\r
1672 if ((token.m_source_ >>> 24) > 1) {
\r
1674 int length = token.m_source_ >>> 24;
\r
1675 String tokenstr = token.m_rules_.substring(startoftokenrule,
\r
1676 startoftokenrule + length);
\r
1677 token.m_CE_[0] |= getCaseBits(tokenstr);
\r
1679 // Copy it from the UCA
\r
1680 int caseCE = getFirstCE(token.m_rules_.charAt(startoftokenrule));
\r
1681 token.m_CE_[0] |= (caseCE & 0xC0);
\r
1687 * Count the number of non-zero bytes used in the ce
\r
1690 * @return number of non-zero bytes used in ce
\r
1692 private static final int countBytes(int ce) {
\r
1693 int mask = 0xFFFFFFFF;
\r
1695 while (mask != 0) {
\r
1696 if ((ce & mask) != 0) {
\r
1705 * We are ready to create collation elements
\r
1708 * build table to insert
\r
1710 * rule token list header
\r
1712 private void createElements(BuildTable t,
\r
1713 CollationRuleParser.TokenListHeader lh) {
\r
1714 CollationRuleParser.Token tok = lh.m_first_;
\r
1715 m_utilElement_.clear();
\r
1716 while (tok != null) {
\r
1717 // first, check if there are any expansions
\r
1718 // if there are expansions, we need to do a little bit more
\r
1719 // processing since parts of expansion can be tailored, while
\r
1721 if (tok.m_expansion_ != 0) {
\r
1722 int len = tok.m_expansion_ >>> 24;
\r
1723 int currentSequenceLen = len;
\r
1724 int expOffset = tok.m_expansion_ & 0x00FFFFFF;
\r
1725 m_utilToken_.m_source_ = currentSequenceLen | expOffset;
\r
1726 m_utilToken_.m_rules_ = m_parser_.m_source_;
\r
1729 currentSequenceLen = len;
\r
1730 while (currentSequenceLen > 0) {
\r
1731 m_utilToken_.m_source_ = (currentSequenceLen << 24)
\r
1733 CollationRuleParser.Token expt = m_parser_.m_hashTable_.get(m_utilToken_);
\r
1735 && expt.m_strength_ != CollationRuleParser.TOKEN_RESET_) {
\r
1736 // expansion is tailored
\r
1737 int noOfCEsToCopy = expt.m_CELength_;
\r
1738 for (int j = 0; j < noOfCEsToCopy; j++) {
\r
1739 tok.m_expCE_[tok.m_expCELength_ + j] = expt.m_CE_[j];
\r
1741 tok.m_expCELength_ += noOfCEsToCopy;
\r
1742 // never try to add codepoints and CEs.
\r
1743 // For some odd reason, it won't work.
\r
1744 expOffset += currentSequenceLen; // noOfCEsToCopy;
\r
1745 len -= currentSequenceLen; // noOfCEsToCopy;
\r
1748 currentSequenceLen--;
\r
1751 if (currentSequenceLen == 0) {
\r
1752 // couldn't find any tailored subsequence, will have to
\r
1753 // get one from UCA. first, get the UChars from the
\r
1754 // rules then pick CEs out until there is no more and
\r
1755 // stuff them into expansion
\r
1756 m_utilColEIter_.setText(m_parser_.m_source_.substring(
\r
1757 expOffset, expOffset + 1));
\r
1759 int order = m_utilColEIter_.next();
\r
1760 if (order == CollationElementIterator.NULLORDER) {
\r
1763 tok.m_expCE_[tok.m_expCELength_++] = order;
\r
1770 tok.m_expCELength_ = 0;
\r
1773 // set the ucaelement with obtained values
\r
1774 m_utilElement_.m_CELength_ = tok.m_CELength_ + tok.m_expCELength_;
\r
1777 System.arraycopy(tok.m_CE_, 0, m_utilElement_.m_CEs_, 0,
\r
1779 System.arraycopy(tok.m_expCE_, 0, m_utilElement_.m_CEs_,
\r
1780 tok.m_CELength_, tok.m_expCELength_);
\r
1783 // We kept prefix and source kind of together, as it is a kind of a
\r
1785 // However, now we have to slice the prefix off the main thing -
\r
1786 m_utilElement_.m_prefix_ = 0;// el.m_prefixChars_;
\r
1787 m_utilElement_.m_cPointsOffset_ = 0; // el.m_uchars_;
\r
1788 if (tok.m_prefix_ != 0) {
\r
1789 // we will just copy the prefix here, and adjust accordingly in
\r
1790 // the addPrefix function in ucol_elm. The reason is that we
\r
1791 // need to add both composed AND decomposed elements to the
\r
1793 int size = tok.m_prefix_ >> 24;
\r
1794 int offset = tok.m_prefix_ & 0x00FFFFFF;
\r
1795 m_utilElement_.m_prefixChars_ = m_parser_.m_source_.substring(
\r
1796 offset, offset + size);
\r
1797 size = (tok.m_source_ >> 24) - (tok.m_prefix_ >> 24);
\r
1798 offset = (tok.m_source_ & 0x00FFFFFF) + (tok.m_prefix_ >> 24);
\r
1799 m_utilElement_.m_uchars_ = m_parser_.m_source_.substring(
\r
1800 offset, offset + size);
\r
1802 m_utilElement_.m_prefixChars_ = null;
\r
1803 int offset = tok.m_source_ & 0x00FFFFFF;
\r
1804 int size = tok.m_source_ >>> 24;
\r
1805 m_utilElement_.m_uchars_ = m_parser_.m_source_.substring(
\r
1806 offset, offset + size);
\r
1808 m_utilElement_.m_cPoints_ = m_utilElement_.m_uchars_;
\r
1810 boolean containCombinMarks = false;
\r
1811 for (int i = 0; i < m_utilElement_.m_cPoints_.length()
\r
1812 - m_utilElement_.m_cPointsOffset_; i++) {
\r
1813 if (isJamo(m_utilElement_.m_cPoints_.charAt(i))) {
\r
1814 t.m_collator_.m_isJamoSpecial_ = true;
\r
1817 if (!buildCMTabFlag) {
\r
1818 // check combining class
\r
1819 int fcd = m_nfcImpl_.getFCD16FromSingleLead(m_utilElement_.m_cPoints_.charAt(i)); // TODO: review for handling supplementary characters
\r
1820 if ((fcd & 0xff) == 0) {
\r
1821 // reset flag when current char is not combining mark.
\r
1822 containCombinMarks = false;
\r
1824 containCombinMarks = true;
\r
1829 if (!buildCMTabFlag && containCombinMarks) {
\r
1830 buildCMTabFlag = true;
\r
1834 * // Case bits handling m_utilElement_.m_CEs_[0] &= 0xFFFFFF3F; //
\r
1835 * Clean the case bits field if (m_utilElement_.m_cPoints_.length()
\r
1836 * - m_utilElement_.m_cPointsOffset_ > 1) { // Do it manually
\r
1837 * m_utilElement_.m_CEs_[0] |=
\r
1838 * getCaseBits(m_utilElement_.m_cPoints_); } else { // Copy it from
\r
1839 * the UCA int caseCE =
\r
1840 * getFirstCE(m_utilElement_.m_cPoints_.charAt(0));
\r
1841 * m_utilElement_.m_CEs_[0] |= (caseCE & 0xC0); }
\r
1843 // and then, add it
\r
1844 addAnElement(t, m_utilElement_);
\r
1845 tok = tok.m_next_;
\r
1850 * Testing if the string argument has case
\r
1854 * @return the case for this char array
\r
1855 * @exception Exception
\r
1856 * thrown when internal program error occurs
\r
1858 private final int getCaseBits(String src) throws Exception {
\r
1861 src = Normalizer.decompose(src, true);
\r
1862 m_utilColEIter_.setText(src);
\r
1863 for (int i = 0; i < src.length(); i++) {
\r
1864 m_utilColEIter_.setText(src.substring(i, i + 1));
\r
1865 int order = m_utilColEIter_.next();
\r
1866 if (RuleBasedCollator.isContinuation(order)) {
\r
1867 throw new Exception("Internal program error");
\r
1869 if ((order & RuleBasedCollator.CE_CASE_BIT_MASK_) == UPPER_CASE_) {
\r
1872 char ch = src.charAt(i);
\r
1873 if (UCharacter.isLowerCase(ch)) {
\r
1876 if (toSmallKana(ch) == ch && toLargeKana(ch) != ch) {
\r
1883 if (uCount != 0 && lCount != 0) {
\r
1884 return MIXED_CASE_;
\r
1885 } else if (uCount != 0) {
\r
1886 return UPPER_CASE_;
\r
1888 return LOWER_CASE_;
\r
1893 * Converts a char to the uppercase Kana
\r
1896 * character to convert
\r
1897 * @return the converted Kana character
\r
1899 private static final char toLargeKana(char ch) {
\r
1900 if (0x3042 < ch && ch < 0x30ef) { // Kana range
\r
1901 switch (ch - 0x3000) {
\r
1934 * Converts a char to the lowercase Kana
\r
1937 * character to convert
\r
1938 * @return the converted Kana character
\r
1940 private static final char toSmallKana(char ch) {
\r
1941 if (0x3042 < ch && ch < 0x30ef) { // Kana range
\r
1942 switch (ch - 0x3000) {
\r
1975 * This should be connected to special Jamo handling.
\r
1977 private int getFirstCE(char ch) {
\r
1978 m_utilColEIter_.setText(UCharacter.toString(ch));
\r
1979 return m_utilColEIter_.next();
\r
1983 * This adds a read element, while testing for existence
\r
1990 private int addAnElement(BuildTable t, Elements element) {
\r
1991 Vector<Integer> expansions = t.m_expansions_;
\r
1992 element.m_mapCE_ = 0;
\r
1994 if (element.m_CELength_ == 1) {
\r
1995 element.m_mapCE_ = element.m_CEs_[0];
\r
1998 // unfortunately, it looks like we have to look for a long primary
\r
1999 // here since in canonical closure we are going to hit some long
\r
2000 // primaries from the first phase, and they will come back as
\r
2001 // continuations/expansions destroying the effect of the previous
\r
2002 // opitimization. A long primary is a three byte primary with
\r
2003 // starting secondaries and tertiaries. It can appear in long runs
\r
2004 // of only primary differences (like east Asian tailorings) also,
\r
2005 // it should not be an expansion, as expansions would break with
\r
2007 if (element.m_CELength_ == 2 // a two CE expansion
\r
2008 && RuleBasedCollator.isContinuation(element.m_CEs_[1])
\r
2009 && (element.m_CEs_[1] & (~(0xFF << 24 | RuleBasedCollator.CE_CONTINUATION_MARKER_))) == 0 // that
\r
2015 && (((element.m_CEs_[0] >> 8) & 0xFF) == RuleBasedCollator.BYTE_COMMON_)
\r
2016 // a common secondary
\r
2017 && ((element.m_CEs_[0] & 0xFF) == RuleBasedCollator.BYTE_COMMON_) // and
\r
2022 element.m_mapCE_ = RuleBasedCollator.CE_SPECIAL_FLAG_
\r
2023 // a long primary special
\r
2024 | (CE_LONG_PRIMARY_TAG_ << 24)
\r
2025 // first and second byte of primary
\r
2026 | ((element.m_CEs_[0] >> 8) & 0xFFFF00)
\r
2027 // third byte of primary
\r
2028 | ((element.m_CEs_[1] >> 24) & 0xFF);
\r
2030 // omitting expansion offset in builder
\r
2031 // (HEADER_SIZE_ >> 2)
\r
2032 int expansion = RuleBasedCollator.CE_SPECIAL_FLAG_
\r
2033 | (CE_EXPANSION_TAG_ << RuleBasedCollator.CE_TAG_SHIFT_)
\r
2034 | (addExpansion(expansions, element.m_CEs_[0]) << 4)
\r
2037 for (int i = 1; i < element.m_CELength_; i++) {
\r
2038 addExpansion(expansions, element.m_CEs_[i]);
\r
2040 if (element.m_CELength_ <= 0xF) {
\r
2041 expansion |= element.m_CELength_;
\r
2043 addExpansion(expansions, 0);
\r
2045 element.m_mapCE_ = expansion;
\r
2046 setMaxExpansion(element.m_CEs_[element.m_CELength_ - 1],
\r
2047 (byte) element.m_CELength_, t.m_maxExpansions_);
\r
2048 if (isJamo(element.m_cPoints_.charAt(0))) {
\r
2049 t.m_collator_.m_isJamoSpecial_ = true;
\r
2050 setMaxJamoExpansion(element.m_cPoints_.charAt(0),
\r
2051 element.m_CEs_[element.m_CELength_ - 1],
\r
2052 (byte) element.m_CELength_, t.m_maxJamoExpansions_);
\r
2057 // We treat digits differently - they are "uber special" and should be
\r
2058 // processed differently if numeric collation is on.
\r
2060 if ((element.m_uchars_.length() == 2)
\r
2061 && UTF16.isLeadSurrogate(element.m_uchars_.charAt(0))) {
\r
2062 uniChar = UCharacterProperty.getRawSupplementary(element.m_uchars_
\r
2063 .charAt(0), element.m_uchars_.charAt(1));
\r
2064 } else if (element.m_uchars_.length() == 1) {
\r
2065 uniChar = element.m_uchars_.charAt(0);
\r
2068 // Here, we either have one normal CE OR mapCE is set. Therefore, we
\r
2069 // stuff only one element to the expansion buffer. When we encounter a
\r
2070 // digit and we don't do numeric collation, we will just pick the CE
\r
2071 // we have and break out of case (see ucol.cpp ucol_prv_getSpecialCE
\r
2072 // && ucol_prv_getSpecialPrevCE). If we picked a special, further
\r
2073 // processing will occur. If it's a simple CE, we'll return due
\r
2074 // to how the loop is constructed.
\r
2075 if (uniChar != 0 && UCharacter.isDigit(uniChar)) {
\r
2076 // prepare the element
\r
2077 int expansion = RuleBasedCollator.CE_SPECIAL_FLAG_
\r
2078 | (CollationElementIterator.CE_DIGIT_TAG_ << RuleBasedCollator.CE_TAG_SHIFT_)
\r
2080 if (element.m_mapCE_ != 0) {
\r
2081 // if there is an expansion, we'll pick it here
\r
2082 expansion |= (addExpansion(expansions, element.m_mapCE_) << 4);
\r
2084 expansion |= (addExpansion(expansions, element.m_CEs_[0]) << 4);
\r
2086 element.m_mapCE_ = expansion;
\r
2089 // here we want to add the prefix structure.
\r
2090 // I will try to process it as a reverse contraction, if possible.
\r
2091 // prefix buffer is already reversed.
\r
2093 if (element.m_prefixChars_ != null
\r
2094 && element.m_prefixChars_.length() - element.m_prefix_ > 0) {
\r
2095 // We keep the seen prefix starter elements in a hashtable we need
\r
2096 // it to be able to distinguish between the simple codepoints and
\r
2097 // prefix starters. Also, we need to use it for canonical closure.
\r
2098 m_utilElement2_.m_caseBit_ = element.m_caseBit_;
\r
2099 m_utilElement2_.m_CELength_ = element.m_CELength_;
\r
2100 m_utilElement2_.m_CEs_ = element.m_CEs_;
\r
2101 m_utilElement2_.m_mapCE_ = element.m_mapCE_;
\r
2102 // m_utilElement2_.m_prefixChars_ = element.m_prefixChars_;
\r
2103 m_utilElement2_.m_sizePrim_ = element.m_sizePrim_;
\r
2104 m_utilElement2_.m_sizeSec_ = element.m_sizeSec_;
\r
2105 m_utilElement2_.m_sizeTer_ = element.m_sizeTer_;
\r
2106 m_utilElement2_.m_variableTop_ = element.m_variableTop_;
\r
2107 m_utilElement2_.m_prefix_ = element.m_prefix_;
\r
2108 m_utilElement2_.m_prefixChars_ = Normalizer.compose(
\r
2109 element.m_prefixChars_, false);
\r
2110 m_utilElement2_.m_uchars_ = element.m_uchars_;
\r
2111 m_utilElement2_.m_cPoints_ = element.m_cPoints_;
\r
2112 m_utilElement2_.m_cPointsOffset_ = 0;
\r
2114 if (t.m_prefixLookup_ != null) {
\r
2115 Elements uCE = t.m_prefixLookup_.get(element);
\r
2116 if (uCE != null) {
\r
2117 // there is already a set of code points here
\r
2118 element.m_mapCE_ = addPrefix(t, uCE.m_mapCE_, element);
\r
2119 } else { // no code points, so this spot is clean
\r
2120 element.m_mapCE_ = addPrefix(t, CE_NOT_FOUND_, element);
\r
2121 uCE = new Elements(element);
\r
2122 uCE.m_cPoints_ = uCE.m_uchars_;
\r
2123 t.m_prefixLookup_.put(uCE, uCE);
\r
2125 if (m_utilElement2_.m_prefixChars_.length() != element.m_prefixChars_
\r
2127 - element.m_prefix_
\r
2128 || !m_utilElement2_.m_prefixChars_.regionMatches(0,
\r
2129 element.m_prefixChars_, element.m_prefix_,
\r
2130 m_utilElement2_.m_prefixChars_.length())) {
\r
2132 m_utilElement2_.m_mapCE_ = addPrefix(t, element.m_mapCE_,
\r
2138 // We need to use the canonical iterator here
\r
2139 // the way we do it is to generate the canonically equivalent strings
\r
2140 // for the contraction and then add the sequences that pass FCD check
\r
2141 if (element.m_cPoints_.length() - element.m_cPointsOffset_ > 1
\r
2142 && !(element.m_cPoints_.length() - element.m_cPointsOffset_ == 2
\r
2143 && UTF16.isLeadSurrogate(element.m_cPoints_.charAt(0)) && UTF16
\r
2144 .isTrailSurrogate(element.m_cPoints_.charAt(1)))) {
\r
2145 // this is a contraction, we should check whether a composed form
\r
2146 // should also be included
\r
2147 m_utilCanIter_.setSource(element.m_cPoints_);
\r
2148 String source = m_utilCanIter_.next();
\r
2149 while (source != null && source.length() > 0) {
\r
2150 if (Normalizer.quickCheck(source, Normalizer.FCD, 0) != Normalizer.NO) {
\r
2151 element.m_uchars_ = source;
\r
2152 element.m_cPoints_ = element.m_uchars_;
\r
2153 finalizeAddition(t, element);
\r
2155 source = m_utilCanIter_.next();
\r
2158 return element.m_mapCE_;
\r
2160 return finalizeAddition(t, element);
\r
2165 * Adds an expansion ce to the expansion vector
\r
2167 * @param expansions
\r
2168 * vector to add to
\r
2170 * of the expansion
\r
2171 * @return the current position of the new element
\r
2173 private static final int addExpansion(Vector<Integer> expansions, int value) {
\r
2174 expansions.add(new Integer(value));
\r
2175 return expansions.size() - 1;
\r
2179 * Looks for the maximum length of all expansion sequences ending with the
\r
2180 * same collation element. The size required for maxexpansion and maxsize is
\r
2181 * returned if the arrays are too small.
\r
2183 * @param endexpansion
\r
2184 * the last expansion collation element to be added
\r
2185 * @param expansionsize
\r
2186 * size of the expansion
\r
2187 * @param maxexpansion
\r
2188 * data structure to store the maximum expansion data.
\r
2189 * @returns size of the maxexpansion and maxsize used.
\r
2191 private static int setMaxExpansion(int endexpansion, byte expansionsize,
\r
2192 MaxExpansionTable maxexpansion) {
\r
2194 int limit = maxexpansion.m_endExpansionCE_.size();
\r
2195 long unsigned = (long) endexpansion;
\r
2196 unsigned &= 0xFFFFFFFFl;
\r
2198 // using binary search to determine if last expansion element is
\r
2199 // already in the array
\r
2201 while (start < limit - 1) {
\r
2202 int mid = start + ((limit - start) >> 1);
\r
2203 long unsignedce = (maxexpansion.m_endExpansionCE_
\r
2204 .get(mid)).intValue();
\r
2205 unsignedce &= 0xFFFFFFFFl;
\r
2206 if (unsigned <= unsignedce) {
\r
2213 if ((maxexpansion.m_endExpansionCE_.get(start)).intValue() == endexpansion) {
\r
2215 } else if ((maxexpansion.m_endExpansionCE_.get(limit))
\r
2216 .intValue() == endexpansion) {
\r
2219 if (result > -1) {
\r
2220 // found the ce in expansion, we'll just modify the size if it
\r
2222 Object currentsize = maxexpansion.m_expansionCESize_.get(result);
\r
2223 if (((Byte) currentsize).byteValue() < expansionsize) {
\r
2224 maxexpansion.m_expansionCESize_.set(result, new Byte(
\r
2228 // we'll need to squeeze the value into the array. initial
\r
2229 // implementation. shifting the subarray down by 1
\r
2230 maxexpansion.m_endExpansionCE_.insertElementAt(new Integer(
\r
2231 endexpansion), start + 1);
\r
2232 maxexpansion.m_expansionCESize_.insertElementAt(new Byte(
\r
2233 expansionsize), start + 1);
\r
2235 return maxexpansion.m_endExpansionCE_.size();
\r
2239 * Sets the maximum length of all jamo expansion sequences ending with the
\r
2240 * same collation element. The size required for maxexpansion and maxsize is
\r
2241 * returned if the arrays are too small.
\r
2244 * the jamo codepoint
\r
2245 * @param endexpansion
\r
2246 * the last expansion collation element to be added
\r
2247 * @param expansionsize
\r
2248 * size of the expansion
\r
2249 * @param maxexpansion
\r
2250 * data structure to store the maximum expansion data.
\r
2251 * @returns size of the maxexpansion and maxsize used.
\r
2253 private static int setMaxJamoExpansion(char ch, int endexpansion,
\r
2254 byte expansionsize, MaxJamoExpansionTable maxexpansion) {
\r
2255 boolean isV = true;
\r
2256 if (ch >= 0x1100 && ch <= 0x1112) {
\r
2257 // determines L for Jamo, doesn't need to store this since it is
\r
2258 // never at the end of a expansion
\r
2259 if (maxexpansion.m_maxLSize_ < expansionsize) {
\r
2260 maxexpansion.m_maxLSize_ = expansionsize;
\r
2262 return maxexpansion.m_endExpansionCE_.size();
\r
2265 if (ch >= 0x1161 && ch <= 0x1175) {
\r
2266 // determines V for Jamo
\r
2267 if (maxexpansion.m_maxVSize_ < expansionsize) {
\r
2268 maxexpansion.m_maxVSize_ = expansionsize;
\r
2272 if (ch >= 0x11A8 && ch <= 0x11C2) {
\r
2274 // determines T for Jamo
\r
2275 if (maxexpansion.m_maxTSize_ < expansionsize) {
\r
2276 maxexpansion.m_maxTSize_ = expansionsize;
\r
2280 int pos = maxexpansion.m_endExpansionCE_.size();
\r
2283 if ((maxexpansion.m_endExpansionCE_.get(pos)).intValue() == endexpansion) {
\r
2284 return maxexpansion.m_endExpansionCE_.size();
\r
2287 maxexpansion.m_endExpansionCE_.add(new Integer(endexpansion));
\r
2288 maxexpansion.m_isV_.add(isV ? Boolean.TRUE : Boolean.FALSE);
\r
2290 return maxexpansion.m_endExpansionCE_.size();
\r
2294 * Adds a prefix to the table
\r
2297 * build table to update
\r
2299 * collation element to add
\r
2301 * rule element to add
\r
2302 * @return modified ce
\r
2304 private int addPrefix(BuildTable t, int CE, Elements element) {
\r
2305 // currently the longest prefix we're supporting in Japanese is two
\r
2306 // characters long. Although this table could quite easily mimic
\r
2307 // complete contraction stuff there is no good reason to make a general
\r
2308 // solution, as it would require some error prone messing.
\r
2309 ContractionTable contractions = t.m_contractions_;
\r
2310 String oldCP = element.m_cPoints_;
\r
2311 int oldCPOffset = element.m_cPointsOffset_;
\r
2313 contractions.m_currentTag_ = CE_SPEC_PROC_TAG_;
\r
2314 // here, we will normalize & add prefix to the table.
\r
2315 int size = element.m_prefixChars_.length() - element.m_prefix_;
\r
2316 for (int j = 1; j < size; j++) {
\r
2317 // First add NFD prefix chars to unsafe CP hash table
\r
2318 // Unless it is a trail surrogate, which is handled algoritmically
\r
2319 // and shouldn't take up space in the table.
\r
2320 char ch = element.m_prefixChars_.charAt(j + element.m_prefix_);
\r
2321 if (!UTF16.isTrailSurrogate(ch)) {
\r
2322 unsafeCPSet(t.m_unsafeCP_, ch);
\r
2326 // StringBuffer reversed = new StringBuffer();
\r
2327 m_utilStringBuffer_.delete(0, m_utilStringBuffer_.length());
\r
2328 for (int j = 0; j < size; j++) {
\r
2329 // prefixes are going to be looked up backwards
\r
2330 // therefore, we will promptly reverse the prefix buffer...
\r
2331 int offset = element.m_prefixChars_.length() - j - 1;
\r
2332 m_utilStringBuffer_.append(element.m_prefixChars_.charAt(offset));
\r
2334 element.m_prefixChars_ = m_utilStringBuffer_.toString();
\r
2335 element.m_prefix_ = 0;
\r
2337 // the first codepoint is also unsafe, as it forms a 'contraction' with
\r
2339 if (!UTF16.isTrailSurrogate(element.m_cPoints_.charAt(0))) {
\r
2340 unsafeCPSet(t.m_unsafeCP_, element.m_cPoints_.charAt(0));
\r
2343 element.m_cPoints_ = element.m_prefixChars_;
\r
2344 element.m_cPointsOffset_ = element.m_prefix_;
\r
2346 // Add the last char of the contraction to the contraction-end hash
\r
2347 // table. unless it is a trail surrogate, which is handled
\r
2348 // algorithmically and shouldn't be in the table
\r
2349 if (!UTF16.isTrailSurrogate(element.m_cPoints_
\r
2350 .charAt(element.m_cPoints_.length() - 1))) {
\r
2351 ContrEndCPSet(t.m_contrEndCP_, element.m_cPoints_
\r
2352 .charAt(element.m_cPoints_.length() - 1));
\r
2354 // First we need to check if contractions starts with a surrogate
\r
2355 // int cp = UTF16.charAt(element.m_cPoints_, element.m_cPointsOffset_);
\r
2357 // If there are any Jamos in the contraction, we should turn on special
\r
2358 // processing for Jamos
\r
2359 if (isJamo(element.m_prefixChars_.charAt(element.m_prefix_))) {
\r
2360 t.m_collator_.m_isJamoSpecial_ = true;
\r
2362 // then we need to deal with it
\r
2363 // we could aready have something in table - or we might not
\r
2364 if (!isPrefix(CE)) {
\r
2365 // if it wasn't contraction, we wouldn't end up here
\r
2366 int firstContractionOffset = addContraction(contractions,
\r
2367 CONTRACTION_TABLE_NEW_ELEMENT_, (char) 0, CE);
\r
2368 int newCE = processContraction(contractions, element, CE_NOT_FOUND_);
\r
2369 addContraction(contractions, firstContractionOffset,
\r
2370 element.m_prefixChars_.charAt(element.m_prefix_), newCE);
\r
2371 addContraction(contractions, firstContractionOffset, (char) 0xFFFF,
\r
2373 CE = constructSpecialCE(CE_SPEC_PROC_TAG_, firstContractionOffset);
\r
2375 // we are adding to existing contraction
\r
2376 // there were already some elements in the table, so we need to add
\r
2377 // a new contraction
\r
2378 // Two things can happen here: either the codepoint is already in
\r
2379 // the table, or it is not
\r
2380 char ch = element.m_prefixChars_.charAt(element.m_prefix_);
\r
2381 int position = findCP(contractions, CE, ch);
\r
2382 if (position > 0) {
\r
2383 // if it is we just continue down the chain
\r
2384 int eCE = getCE(contractions, CE, position);
\r
2385 int newCE = processContraction(contractions, element, eCE);
\r
2386 setContraction(contractions, CE, position, ch, newCE);
\r
2388 // if it isn't, we will have to create a new sequence
\r
2389 processContraction(contractions, element, CE_NOT_FOUND_);
\r
2390 insertContraction(contractions, CE, ch, element.m_mapCE_);
\r
2394 element.m_cPoints_ = oldCP;
\r
2395 element.m_cPointsOffset_ = oldCPOffset;
\r
2401 * Checks if the argument ce is a contraction
\r
2404 * collation element
\r
2405 * @return true if argument ce is a contraction
\r
2407 private static final boolean isContraction(int CE) {
\r
2408 return isSpecial(CE) && (getCETag(CE) == CE_CONTRACTION_TAG_);
\r
2412 * Checks if the argument ce has a prefix
\r
2415 * collation element
\r
2416 * @return true if argument ce has a prefix
\r
2418 private static final boolean isPrefix(int CE) {
\r
2419 return isSpecial(CE) && (getCETag(CE) == CE_SPEC_PROC_TAG_);
\r
2423 * Checks if the argument ce is special
\r
2426 * collation element
\r
2427 * @return true if argument ce is special
\r
2429 private static final boolean isSpecial(int CE) {
\r
2430 return (CE & RuleBasedCollator.CE_SPECIAL_FLAG_) == 0xF0000000;
\r
2434 * Checks if the argument ce has a prefix
\r
2437 * collation element
\r
2438 * @return true if argument ce has a prefix
\r
2440 private static final int getCETag(int CE) {
\r
2441 return (CE & RuleBasedCollator.CE_TAG_MASK_) >>> RuleBasedCollator.CE_TAG_SHIFT_;
\r
2445 * Gets the ce at position in contraction table
\r
2448 * contraction table
\r
2450 * offset to the contraction table
\r
2453 private static final int getCE(ContractionTable table, int element,
\r
2455 element &= 0xFFFFFF;
\r
2456 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2458 if (tbl == null) {
\r
2459 return CE_NOT_FOUND_;
\r
2461 if (position > tbl.m_CEs_.size() || position == -1) {
\r
2462 return CE_NOT_FOUND_;
\r
2464 return tbl.m_CEs_.get(position).intValue();
\r
2469 * Sets the unsafe character
\r
2474 * character to be added
\r
2476 private static final void unsafeCPSet(byte table[], char c) {
\r
2478 if (hash >= (UNSAFECP_TABLE_SIZE_ << 3)) {
\r
2479 if (hash >= 0xd800 && hash <= 0xf8ff) {
\r
2480 // Part of a surrogate, or in private use area.
\r
2481 // These don't go in the table
\r
2484 hash = (hash & UNSAFECP_TABLE_MASK_) + 256;
\r
2486 table[hash >> 3] |= (1 << (hash & 7));
\r
2490 * Sets the contraction end character
\r
2493 * contraction end table
\r
2495 * character to be added
\r
2497 private static final void ContrEndCPSet(byte table[], char c) {
\r
2499 if (hash >= (UNSAFECP_TABLE_SIZE_ << 3)) {
\r
2500 hash = (hash & UNSAFECP_TABLE_MASK_) + 256;
\r
2502 table[hash >> 3] |= (1 << (hash & 7));
\r
2506 * Adds more contractions in table. If element is non existant, it creates
\r
2507 * on. Returns element handle
\r
2510 * contraction table
\r
2512 * offset to the contraction table
\r
2513 * @param codePoint
\r
2514 * codepoint to add
\r
2516 * @return collation element
\r
2518 private static int addContraction(ContractionTable table, int element,
\r
2519 char codePoint, int value) {
\r
2520 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2521 if (tbl == null) {
\r
2522 tbl = addAContractionElement(table);
\r
2523 element = table.m_elements_.size() - 1;
\r
2526 tbl.m_CEs_.add(new Integer(value));
\r
2527 tbl.m_codePoints_.append(codePoint);
\r
2528 return constructSpecialCE(table.m_currentTag_, element);
\r
2532 * Adds a contraction element to the table
\r
2535 * contraction table to update
\r
2536 * @return contraction
\r
2538 private static BasicContractionTable addAContractionElement(
\r
2539 ContractionTable table) {
\r
2540 BasicContractionTable result = new BasicContractionTable();
\r
2541 table.m_elements_.add(result);
\r
2546 * Constructs a special ce
\r
2551 * collation element
\r
2552 * @return a contraction ce
\r
2554 private static final int constructSpecialCE(int tag, int CE) {
\r
2555 return RuleBasedCollator.CE_SPECIAL_FLAG_
\r
2556 | (tag << RuleBasedCollator.CE_TAG_SHIFT_) | (CE & 0xFFFFFF);
\r
2560 * Sets and inserts the element that has a contraction
\r
2562 * @param contractions
\r
2563 * contraction table
\r
2565 * contracting element
\r
2566 * @param existingCE
\r
2567 * @return contraction ce
\r
2569 private static int processContraction(ContractionTable contractions,
\r
2570 Elements element, int existingCE) {
\r
2571 int firstContractionOffset = 0;
\r
2572 // end of recursion
\r
2573 if (element.m_cPoints_.length() - element.m_cPointsOffset_ == 1) {
\r
2574 if (isContractionTableElement(existingCE)
\r
2575 && getCETag(existingCE) == contractions.m_currentTag_) {
\r
2576 changeContraction(contractions, existingCE, (char) 0,
\r
2577 element.m_mapCE_);
\r
2578 changeContraction(contractions, existingCE, (char) 0xFFFF,
\r
2579 element.m_mapCE_);
\r
2580 return existingCE;
\r
2582 // can't do just that. existingCe might be a contraction,
\r
2583 // meaning that we need to do another step
\r
2584 return element.m_mapCE_;
\r
2588 // this recursion currently feeds on the only element we have...
\r
2589 // We will have to copy it in order to accomodate for both backward
\r
2590 // and forward cycles
\r
2591 // we encountered either an empty space or a non-contraction element
\r
2592 // this means we are constructing a new contraction sequence
\r
2593 element.m_cPointsOffset_++;
\r
2594 if (!isContractionTableElement(existingCE)) {
\r
2595 // if it wasn't contraction, we wouldn't end up here
\r
2596 firstContractionOffset = addContraction(contractions,
\r
2597 CONTRACTION_TABLE_NEW_ELEMENT_, (char) 0, existingCE);
\r
2598 int newCE = processContraction(contractions, element, CE_NOT_FOUND_);
\r
2599 addContraction(contractions, firstContractionOffset,
\r
2600 element.m_cPoints_.charAt(element.m_cPointsOffset_), newCE);
\r
2601 addContraction(contractions, firstContractionOffset, (char) 0xFFFF,
\r
2603 existingCE = constructSpecialCE(contractions.m_currentTag_,
\r
2604 firstContractionOffset);
\r
2606 // we are adding to existing contraction
\r
2607 // there were already some elements in the table, so we need to add
\r
2608 // a new contraction
\r
2609 // Two things can happen here: either the codepoint is already in
\r
2610 // the table, or it is not
\r
2611 int position = findCP(contractions, existingCE, element.m_cPoints_
\r
2612 .charAt(element.m_cPointsOffset_));
\r
2613 if (position > 0) {
\r
2614 // if it is we just continue down the chain
\r
2615 int eCE = getCE(contractions, existingCE, position);
\r
2616 int newCE = processContraction(contractions, element, eCE);
\r
2617 setContraction(contractions, existingCE, position,
\r
2618 element.m_cPoints_.charAt(element.m_cPointsOffset_),
\r
2621 // if it isn't, we will have to create a new sequence
\r
2622 int newCE = processContraction(contractions, element,
\r
2624 insertContraction(contractions, existingCE, element.m_cPoints_
\r
2625 .charAt(element.m_cPointsOffset_), newCE);
\r
2628 element.m_cPointsOffset_--;
\r
2629 return existingCE;
\r
2633 * Checks if CE belongs to the contraction table
\r
2636 * collation element to test
\r
2637 * @return true if CE belongs to the contraction table
\r
2639 private static final boolean isContractionTableElement(int CE) {
\r
2640 return isSpecial(CE)
\r
2641 && (getCETag(CE) == CE_CONTRACTION_TAG_ || getCETag(CE) == CE_SPEC_PROC_TAG_);
\r
2645 * Gets the codepoint
\r
2648 * contraction table
\r
2650 * offset to the contraction element in the table
\r
2651 * @param codePoint
\r
2652 * code point to look for
\r
2653 * @return the offset to the code point
\r
2655 private static int findCP(ContractionTable table, int element,
\r
2657 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2658 if (tbl == null) {
\r
2663 while (codePoint > tbl.m_codePoints_.charAt(position)) {
\r
2665 if (position > tbl.m_codePoints_.length()) {
\r
2669 if (codePoint == tbl.m_codePoints_.charAt(position)) {
\r
2677 * Gets the contraction element out of the contraction table
\r
2680 * contraction table
\r
2682 * to the element in the contraction table
\r
2683 * @return basic contraction element at offset in the contraction table
\r
2685 private static final BasicContractionTable getBasicContractionTable(
\r
2686 ContractionTable table, int offset) {
\r
2687 offset &= 0xFFFFFF;
\r
2688 if (offset == 0xFFFFFF) {
\r
2691 return table.m_elements_.get(offset);
\r
2695 * Changes the contraction element
\r
2698 * contraction table
\r
2700 * offset to the element in the contraction table
\r
2701 * @param codePoint
\r
2704 * new collation element
\r
2705 * @return basic contraction element at offset in the contraction table
\r
2707 private static final int changeContraction(ContractionTable table,
\r
2708 int element, char codePoint, int newCE) {
\r
2709 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2710 if (tbl == null) {
\r
2714 while (codePoint > tbl.m_codePoints_.charAt(position)) {
\r
2716 if (position > tbl.m_codePoints_.length()) {
\r
2717 return CE_NOT_FOUND_;
\r
2720 if (codePoint == tbl.m_codePoints_.charAt(position)) {
\r
2721 tbl.m_CEs_.set(position, new Integer(newCE));
\r
2722 return element & 0xFFFFFF;
\r
2724 return CE_NOT_FOUND_;
\r
2729 * Sets a part of contraction sequence in table. If element is non existant,
\r
2730 * it creates on. Returns element handle.
\r
2733 * contraction table
\r
2735 * offset to the contraction table
\r
2737 * @param codePoint
\r
2738 * contraction character
\r
2741 * @return new contraction ce
\r
2743 private static final int setContraction(ContractionTable table,
\r
2744 int element, int offset, char codePoint, int value) {
\r
2745 element &= 0xFFFFFF;
\r
2746 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2747 if (tbl == null) {
\r
2748 tbl = addAContractionElement(table);
\r
2749 element = table.m_elements_.size() - 1;
\r
2752 tbl.m_CEs_.set(offset, new Integer(value));
\r
2753 tbl.m_codePoints_.setCharAt(offset, codePoint);
\r
2754 return constructSpecialCE(table.m_currentTag_, element);
\r
2758 * Inserts a part of contraction sequence in table. Sequences behind the
\r
2759 * offset are moved back. If element is non existent, it creates on.
\r
2764 * offset to the table contraction
\r
2765 * @param codePoint
\r
2768 * collation element value
\r
2769 * @return contraction collation element
\r
2771 private static final int insertContraction(ContractionTable table,
\r
2772 int element, char codePoint, int value) {
\r
2773 element &= 0xFFFFFF;
\r
2774 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2775 if (tbl == null) {
\r
2776 tbl = addAContractionElement(table);
\r
2777 element = table.m_elements_.size() - 1;
\r
2781 while (tbl.m_codePoints_.charAt(offset) < codePoint
\r
2782 && offset < tbl.m_codePoints_.length()) {
\r
2786 tbl.m_CEs_.insertElementAt(new Integer(value), offset);
\r
2787 tbl.m_codePoints_.insert(offset, codePoint);
\r
2789 return constructSpecialCE(table.m_currentTag_, element);
\r
2793 * Finalize addition
\r
2800 private final static int finalizeAddition(BuildTable t, Elements element) {
\r
2801 int CE = CE_NOT_FOUND_;
\r
2802 // This should add a completely ignorable element to the
\r
2803 // unsafe table, so that backward iteration will skip
\r
2804 // over it when treating contractions.
\r
2805 if (element.m_mapCE_ == 0) {
\r
2806 for (int i = 0; i < element.m_cPoints_.length(); i++) {
\r
2807 char ch = element.m_cPoints_.charAt(i);
\r
2808 if (!UTF16.isTrailSurrogate(ch)) {
\r
2809 unsafeCPSet(t.m_unsafeCP_, ch);
\r
2814 if (element.m_cPoints_.length() - element.m_cPointsOffset_ > 1) {
\r
2815 // we're adding a contraction
\r
2816 int cp = UTF16.charAt(element.m_cPoints_, element.m_cPointsOffset_);
\r
2817 CE = t.m_mapping_.getValue(cp);
\r
2818 CE = addContraction(t, CE, element);
\r
2821 CE = t.m_mapping_.getValue(element.m_cPoints_
\r
2822 .charAt(element.m_cPointsOffset_));
\r
2824 if (CE != CE_NOT_FOUND_) {
\r
2825 if (isContractionTableElement(CE)) {
\r
2826 // adding a non contraction element (thai, expansion,
\r
2827 // single) to already existing contraction
\r
2828 if (!isPrefix(element.m_mapCE_)) {
\r
2829 // we cannot reenter prefix elements - as we are going
\r
2830 // to create a dead loop
\r
2831 // Only expansions and regular CEs can go here...
\r
2832 // Contractions will never happen in this place
\r
2833 setContraction(t.m_contractions_, CE, 0, (char) 0,
\r
2834 element.m_mapCE_);
\r
2835 // This loop has to change the CE at the end of
\r
2836 // contraction REDO!
\r
2837 changeLastCE(t.m_contractions_, CE, element.m_mapCE_);
\r
2841 .setValue(element.m_cPoints_
\r
2842 .charAt(element.m_cPointsOffset_),
\r
2843 element.m_mapCE_);
\r
2844 if (element.m_prefixChars_ != null
\r
2845 && element.m_prefixChars_.length() > 0
\r
2846 && getCETag(CE) != CE_IMPLICIT_TAG_) {
\r
2847 // Add CE for standalone precontext char.
\r
2848 Elements origElem = new Elements();
\r
2849 origElem.m_prefixChars_ = null;
\r
2850 origElem.m_uchars_ = element.m_cPoints_;
\r
2851 origElem.m_cPoints_ = origElem.m_uchars_;
\r
2852 origElem.m_CEs_[0] = CE;
\r
2853 origElem.m_mapCE_ = CE;
\r
2854 origElem.m_CELength_ = 1;
\r
2855 finalizeAddition(t, origElem);
\r
2859 t.m_mapping_.setValue(element.m_cPoints_
\r
2860 .charAt(element.m_cPointsOffset_), element.m_mapCE_);
\r
2867 * Note regarding surrogate handling: We are interested only in the single
\r
2868 * or leading surrogates in a contraction. If a surrogate is somewhere else
\r
2869 * in the contraction, it is going to be handled as a pair of code units, as
\r
2870 * it doesn't affect the performance AND handling surrogates specially would
\r
2871 * complicate code way too much.
\r
2873 private static int addContraction(BuildTable t, int CE, Elements element) {
\r
2874 ContractionTable contractions = t.m_contractions_;
\r
2875 contractions.m_currentTag_ = CE_CONTRACTION_TAG_;
\r
2877 // First we need to check if contractions starts with a surrogate
\r
2878 int cp = UTF16.charAt(element.m_cPoints_, 0);
\r
2880 if (UCharacter.isSupplementary(cp)) {
\r
2883 if (cpsize < element.m_cPoints_.length()) {
\r
2884 // This is a real contraction, if there are other characters after
\r
2886 int size = element.m_cPoints_.length() - element.m_cPointsOffset_;
\r
2887 for (int j = 1; j < size; j++) {
\r
2888 // First add contraction chars to unsafe CP hash table
\r
2889 // Unless it is a trail surrogate, which is handled
\r
2890 // algoritmically and shouldn't take up space in the table.
\r
2891 if (!UTF16.isTrailSurrogate(element.m_cPoints_
\r
2892 .charAt(element.m_cPointsOffset_ + j))) {
\r
2893 unsafeCPSet(t.m_unsafeCP_, element.m_cPoints_
\r
2894 .charAt(element.m_cPointsOffset_ + j));
\r
2897 // Add the last char of the contraction to the contraction-end
\r
2898 // hash table. unless it is a trail surrogate, which is handled
\r
2899 // algorithmically and shouldn't be in the table
\r
2900 if (!UTF16.isTrailSurrogate(element.m_cPoints_
\r
2901 .charAt(element.m_cPoints_.length() - 1))) {
\r
2902 ContrEndCPSet(t.m_contrEndCP_, element.m_cPoints_
\r
2903 .charAt(element.m_cPoints_.length() - 1));
\r
2906 // If there are any Jamos in the contraction, we should turn on
\r
2907 // special processing for Jamos
\r
2908 if (isJamo(element.m_cPoints_.charAt(element.m_cPointsOffset_))) {
\r
2909 t.m_collator_.m_isJamoSpecial_ = true;
\r
2911 // then we need to deal with it
\r
2912 // we could aready have something in table - or we might not
\r
2913 element.m_cPointsOffset_ += cpsize;
\r
2914 if (!isContraction(CE)) {
\r
2915 // if it wasn't contraction, we wouldn't end up here
\r
2916 int firstContractionOffset = addContraction(contractions,
\r
2917 CONTRACTION_TABLE_NEW_ELEMENT_, (char) 0, CE);
\r
2918 int newCE = processContraction(contractions, element,
\r
2920 addContraction(contractions, firstContractionOffset,
\r
2921 element.m_cPoints_.charAt(element.m_cPointsOffset_),
\r
2923 addContraction(contractions, firstContractionOffset,
\r
2924 (char) 0xFFFF, CE);
\r
2925 CE = constructSpecialCE(CE_CONTRACTION_TAG_,
\r
2926 firstContractionOffset);
\r
2928 // we are adding to existing contraction
\r
2929 // there were already some elements in the table, so we need to
\r
2930 // add a new contraction
\r
2931 // Two things can happen here: either the codepoint is already
\r
2932 // in the table, or it is not
\r
2933 int position = findCP(contractions, CE, element.m_cPoints_
\r
2934 .charAt(element.m_cPointsOffset_));
\r
2935 if (position > 0) {
\r
2936 // if it is we just continue down the chain
\r
2937 int eCE = getCE(contractions, CE, position);
\r
2938 int newCE = processContraction(contractions, element, eCE);
\r
2943 element.m_cPoints_.charAt(element.m_cPointsOffset_),
\r
2946 // if it isn't, we will have to create a new sequence
\r
2947 int newCE = processContraction(contractions, element,
\r
2949 insertContraction(contractions, CE, element.m_cPoints_
\r
2950 .charAt(element.m_cPointsOffset_), newCE);
\r
2953 element.m_cPointsOffset_ -= cpsize;
\r
2954 t.m_mapping_.setValue(cp, CE);
\r
2955 } else if (!isContraction(CE)) {
\r
2956 // this is just a surrogate, and there is no contraction
\r
2957 t.m_mapping_.setValue(cp, element.m_mapCE_);
\r
2959 // fill out the first stage of the contraction with the surrogate
\r
2961 changeContraction(contractions, CE, (char) 0, element.m_mapCE_);
\r
2962 changeContraction(contractions, CE, (char) 0xFFFF, element.m_mapCE_);
\r
2968 * this is for adding non contractions
\r
2971 * contraction table
\r
2973 * offset to the contraction table
\r
2975 * collation element value
\r
2976 * @return new collation element
\r
2978 private static final int changeLastCE(ContractionTable table, int element,
\r
2980 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
2981 if (tbl == null) {
\r
2985 tbl.m_CEs_.set(tbl.m_CEs_.size() - 1, new Integer(value));
\r
2986 return constructSpecialCE(table.m_currentTag_, element & 0xFFFFFF);
\r
2990 * Given a set of ranges calculated by allocWeights(), iterate through the
\r
2991 * weights. Sets the next weight in cegenerator.m_current_.
\r
2993 * @param cegenerator
\r
2994 * object that contains ranges weight range array and its
\r
2996 * @return the next weight
\r
2998 private static int nextWeight(CEGenerator cegenerator) {
\r
2999 if (cegenerator.m_rangesLength_ > 0) {
\r
3000 // get maxByte from the .count field
\r
3001 int maxByte = cegenerator.m_ranges_[0].m_count_;
\r
3002 // get the next weight
\r
3003 int weight = cegenerator.m_ranges_[0].m_start_;
\r
3004 if (weight == cegenerator.m_ranges_[0].m_end_) {
\r
3005 // this range is finished, remove it and move the following
\r
3007 cegenerator.m_rangesLength_--;
\r
3008 if (cegenerator.m_rangesLength_ > 0) {
\r
3009 System.arraycopy(cegenerator.m_ranges_, 1,
\r
3010 cegenerator.m_ranges_, 0,
\r
3011 cegenerator.m_rangesLength_);
\r
3012 cegenerator.m_ranges_[0].m_count_ = maxByte;
\r
3013 // keep maxByte in ranges[0]
\r
3016 // increment the weight for the next value
\r
3017 cegenerator.m_ranges_[0].m_start_ = incWeight(weight,
\r
3018 cegenerator.m_ranges_[0].m_length2_, maxByte);
\r
3026 * Increment the collation weight
\r
3032 * @return new incremented weight
\r
3034 private static final int incWeight(int weight, int length, int maxByte) {
\r
3036 int b = getWeightByte(weight, length);
\r
3037 if (b < maxByte) {
\r
3038 return setWeightByte(weight, length, b + 1);
\r
3040 // roll over, set this byte to BYTE_FIRST_TAILORED_ and
\r
3041 // increment the previous one
\r
3042 weight = setWeightByte(weight, length,
\r
3043 RuleBasedCollator.BYTE_FIRST_TAILORED_);
\r
3050 * Gets the weight byte
\r
3056 private static final int getWeightByte(int weight, int index) {
\r
3057 return (weight >> ((4 - index) << 3)) & 0xff;
\r
3061 * Set the weight byte in table
\r
3068 private static final int setWeightByte(int weight, int index, int b) {
\r
3070 // 0xffffffff except a 00 "hole" for the index-th byte
\r
3071 int mask = 0xffffffff >>> index;
\r
3072 index = 32 - index;
\r
3073 mask |= 0xffffff00 << index;
\r
3074 return (weight & mask) | (b << index);
\r
3078 * Call getWeightRanges and then determine heuristically which ranges to use
\r
3079 * for a given number of weights between (excluding) two limits
\r
3081 * @param lowerLimit
\r
3082 * @param upperLimit
\r
3088 private int allocateWeights(int lowerLimit, int upperLimit, int n,
\r
3089 int maxByte, WeightRange ranges[]) {
\r
3090 // number of usable byte values 3..maxByte
\r
3091 int countBytes = maxByte - RuleBasedCollator.BYTE_FIRST_TAILORED_ + 1;
\r
3092 // [0] unused, [5] to make index checks unnecessary, m_utilCountBuffer_
\r
3093 // countBytes to the power of index, m_utilLongBuffer_ for unsignedness
\r
3094 // gcc requires explicit initialization
\r
3095 m_utilLongBuffer_[0] = 1;
\r
3096 m_utilLongBuffer_[1] = countBytes;
\r
3097 m_utilLongBuffer_[2] = m_utilLongBuffer_[1] * countBytes;
\r
3098 m_utilLongBuffer_[3] = m_utilLongBuffer_[2] * countBytes;
\r
3099 m_utilLongBuffer_[4] = m_utilLongBuffer_[3] * countBytes;
\r
3100 int rangeCount = getWeightRanges(lowerLimit, upperLimit, maxByte,
\r
3101 countBytes, ranges);
\r
3102 if (rangeCount <= 0) {
\r
3105 // what is the maximum number of weights with these ranges?
\r
3106 long maxCount = 0;
\r
3107 for (int i = 0; i < rangeCount; ++i) {
\r
3108 maxCount += (long) ranges[i].m_count_
\r
3109 * m_utilLongBuffer_[4 - ranges[i].m_length_];
\r
3111 if (maxCount < n) {
\r
3114 // set the length2 and count2 fields
\r
3115 for (int i = 0; i < rangeCount; ++i) {
\r
3116 ranges[i].m_length2_ = ranges[i].m_length_;
\r
3117 ranges[i].m_count2_ = ranges[i].m_count_;
\r
3119 // try until we find suitably large ranges
\r
3121 // get the smallest number of bytes in a range
\r
3122 int minLength = ranges[0].m_length2_;
\r
3123 // sum up the number of elements that fit into ranges of each byte
\r
3125 Arrays.fill(m_utilCountBuffer_, 0);
\r
3126 for (int i = 0; i < rangeCount; ++i) {
\r
3127 m_utilCountBuffer_[ranges[i].m_length2_] += ranges[i].m_count2_;
\r
3129 // now try to allocate n elements in the available short ranges
\r
3130 if (n <= m_utilCountBuffer_[minLength]
\r
3131 + m_utilCountBuffer_[minLength + 1]) {
\r
3132 // trivial cases, use the first few ranges
\r
3136 maxCount += ranges[rangeCount].m_count2_;
\r
3138 } while (n > maxCount);
\r
3140 } else if (n <= ranges[0].m_count2_ * countBytes) {
\r
3141 // easy case, just make this one range large enough by
\r
3142 // lengthening it once more, possibly split it
\r
3144 // calculate how to split the range between maxLength-1
\r
3145 // (count1) and maxLength (count2)
\r
3146 long power_1 = m_utilLongBuffer_[minLength
\r
3147 - ranges[0].m_length_];
\r
3148 long power = power_1 * countBytes;
\r
3149 int count2 = (int) ((n + power - 1) / power);
\r
3150 int count1 = ranges[0].m_count_ - count2;
\r
3151 // split the range
\r
3153 // lengthen the entire range to maxLength
\r
3154 lengthenRange(ranges, 0, maxByte, countBytes);
\r
3156 // really split the range
\r
3157 // create a new range with the end and initial and current
\r
3158 // length of the old one
\r
3160 ranges[1].m_end_ = ranges[0].m_end_;
\r
3161 ranges[1].m_length_ = ranges[0].m_length_;
\r
3162 ranges[1].m_length2_ = minLength;
\r
3163 // set the end of the first range according to count1
\r
3164 int i = ranges[0].m_length_;
\r
3165 int b = getWeightByte(ranges[0].m_start_, i) + count1 - 1;
\r
3166 // ranges[0].count and count1 may be >countBytes from
\r
3167 // merging adjacent ranges; b > maxByte is possible
\r
3168 if (b <= maxByte) {
\r
3169 ranges[0].m_end_ = setWeightByte(ranges[0].m_start_, i,
\r
3172 ranges[0].m_end_ = setWeightByte(incWeight(
\r
3173 ranges[0].m_start_, i - 1, maxByte), i, b
\r
3176 // set the bytes in the end weight at length + 1..length2
\r
3178 b = (maxByte << 24) | (maxByte << 16) | (maxByte << 8)
\r
3179 | maxByte; // this used to be 0xffffffff
\r
3180 ranges[0].m_end_ = truncateWeight(ranges[0].m_end_, i)
\r
3181 | (b >>> (i << 3)) & (b << ((4 - minLength) << 3));
\r
3182 // set the start of the second range to immediately follow
\r
3183 // the end of the first one
\r
3184 ranges[1].m_start_ = incWeight(ranges[0].m_end_, minLength,
\r
3186 // set the count values (informational)
\r
3187 ranges[0].m_count_ = count1;
\r
3188 ranges[1].m_count_ = count2;
\r
3190 ranges[0].m_count2_ = (int) (count1 * power_1);
\r
3191 // will be *countBytes when lengthened
\r
3192 ranges[1].m_count2_ = (int) (count2 * power_1);
\r
3194 // lengthen the second range to maxLength
\r
3195 lengthenRange(ranges, 1, maxByte, countBytes);
\r
3199 // no good match, lengthen all minLength ranges and iterate
\r
3200 for (int i = 0; ranges[i].m_length2_ == minLength; ++i) {
\r
3201 lengthenRange(ranges, i, maxByte, countBytes);
\r
3205 if (rangeCount > 1) {
\r
3206 // sort the ranges by weight values
\r
3207 Arrays.sort(ranges, 0, rangeCount);
\r
3210 // set maxByte in ranges[0] for ucol_nextWeight()
\r
3211 ranges[0].m_count_ = maxByte;
\r
3213 return rangeCount;
\r
3217 * Updates the range length
\r
3220 * weight range array
\r
3222 * to weight range array
\r
3224 * @param countBytes
\r
3225 * @return new length
\r
3227 private static final int lengthenRange(WeightRange range[], int offset,
\r
3228 int maxByte, int countBytes) {
\r
3229 int length = range[offset].m_length2_ + 1;
\r
3230 range[offset].m_start_ = setWeightTrail(range[offset].m_start_, length,
\r
3231 RuleBasedCollator.BYTE_FIRST_TAILORED_);
\r
3232 range[offset].m_end_ = setWeightTrail(range[offset].m_end_, length,
\r
3234 range[offset].m_count2_ *= countBytes;
\r
3235 range[offset].m_length2_ = length;
\r
3245 * @return new weight
\r
3247 private static final int setWeightTrail(int weight, int length, int trail) {
\r
3248 length = (4 - length) << 3;
\r
3249 return (weight & (0xffffff00 << length)) | (trail << length);
\r
3253 * take two CE weights and calculate the possible ranges of weights between
\r
3254 * the two limits, excluding them for weights with up to 4 bytes there are
\r
3255 * up to 2*4-1=7 ranges
\r
3257 * @param lowerLimit
\r
3258 * @param upperLimit
\r
3260 * @param countBytes
\r
3262 * @return weight ranges
\r
3264 private int getWeightRanges(int lowerLimit, int upperLimit, int maxByte,
\r
3265 int countBytes, WeightRange ranges[]) {
\r
3266 // assume that both lowerLimit & upperLimit are not 0
\r
3267 // get the lengths of the limits
\r
3268 int lowerLength = lengthOfWeight(lowerLimit);
\r
3269 int upperLength = lengthOfWeight(upperLimit);
\r
3270 if (Utility.compareUnsigned(lowerLimit, upperLimit) >= 0) {
\r
3273 // check that neither is a prefix of the other
\r
3274 if (lowerLength < upperLength) {
\r
3275 if (lowerLimit == truncateWeight(upperLimit, lowerLength)) {
\r
3279 // if the upper limit is a prefix of the lower limit then the earlier
\r
3280 // test lowerLimit >= upperLimit has caught it
\r
3281 // reset local variables
\r
3282 // With the limit lengths of 1..4, there are up to 7 ranges for
\r
3284 // range minimum length
\r
3292 // We are now going to calculate up to 7 ranges.
\r
3293 // Some of them will typically overlap, so we will then have to merge
\r
3294 // and eliminate ranges.
\r
3296 // We have to clean cruft from previous invocations
\r
3297 // before doing anything. C++ already does that
\r
3298 for (int length = 0; length < 5; length++) {
\r
3299 m_utilLowerWeightRange_[length].clear();
\r
3300 m_utilUpperWeightRange_[length].clear();
\r
3302 m_utilWeightRange_.clear();
\r
3304 int weight = lowerLimit;
\r
3305 for (int length = lowerLength; length >= 2; --length) {
\r
3306 m_utilLowerWeightRange_[length].clear();
\r
3307 int trail = getWeightByte(weight, length);
\r
3308 if (trail < maxByte) {
\r
3309 m_utilLowerWeightRange_[length].m_start_ = incWeightTrail(
\r
3311 m_utilLowerWeightRange_[length].m_end_ = setWeightTrail(weight,
\r
3313 m_utilLowerWeightRange_[length].m_length_ = length;
\r
3314 m_utilLowerWeightRange_[length].m_count_ = maxByte - trail;
\r
3316 weight = truncateWeight(weight, length - 1);
\r
3318 m_utilWeightRange_.m_start_ = incWeightTrail(weight, 1);
\r
3320 weight = upperLimit;
\r
3321 // [0] and [1] are not used - this simplifies indexing,
\r
3322 // m_utilUpperWeightRange_
\r
3324 for (int length = upperLength; length >= 2; length--) {
\r
3325 int trail = getWeightByte(weight, length);
\r
3326 if (trail > RuleBasedCollator.BYTE_FIRST_TAILORED_) {
\r
3327 m_utilUpperWeightRange_[length].m_start_ = setWeightTrail(
\r
3328 weight, length, RuleBasedCollator.BYTE_FIRST_TAILORED_);
\r
3329 m_utilUpperWeightRange_[length].m_end_ = decWeightTrail(weight,
\r
3331 m_utilUpperWeightRange_[length].m_length_ = length;
\r
3332 m_utilUpperWeightRange_[length].m_count_ = trail
\r
3333 - RuleBasedCollator.BYTE_FIRST_TAILORED_;
\r
3335 weight = truncateWeight(weight, length - 1);
\r
3337 m_utilWeightRange_.m_end_ = decWeightTrail(weight, 1);
\r
3339 // set the middle range
\r
3340 m_utilWeightRange_.m_length_ = 1;
\r
3341 if (Utility.compareUnsigned(m_utilWeightRange_.m_end_,
\r
3342 m_utilWeightRange_.m_start_) >= 0) {
\r
3343 // if (m_utilWeightRange_.m_end_ >= m_utilWeightRange_.m_start_) {
\r
3344 m_utilWeightRange_.m_count_ = ((m_utilWeightRange_.m_end_ - m_utilWeightRange_.m_start_) >>> 24) + 1;
\r
3346 // eliminate overlaps
\r
3347 // remove the middle range
\r
3348 m_utilWeightRange_.m_count_ = 0;
\r
3349 // reduce or remove the lower ranges that go beyond upperLimit
\r
3350 for (int length = 4; length >= 2; --length) {
\r
3351 if (m_utilLowerWeightRange_[length].m_count_ > 0
\r
3352 && m_utilUpperWeightRange_[length].m_count_ > 0) {
\r
3353 int start = m_utilUpperWeightRange_[length].m_start_;
\r
3354 int end = m_utilLowerWeightRange_[length].m_end_;
\r
3356 || incWeight(end, length, maxByte) == start) {
\r
3357 // lower and upper ranges collide or are directly
\r
3358 // adjacent: merge these two and remove all shorter
\r
3360 start = m_utilLowerWeightRange_[length].m_start_;
\r
3361 end = m_utilLowerWeightRange_[length].m_end_ = m_utilUpperWeightRange_[length].m_end_;
\r
3362 // merging directly adjacent ranges needs to subtract
\r
3363 // the 0/1 gaps in between;
\r
3364 // it may result in a range with count>countBytes
\r
3365 m_utilLowerWeightRange_[length].m_count_ = getWeightByte(
\r
3367 - getWeightByte(start, length)
\r
3370 * (getWeightByte(end, length - 1) - getWeightByte(
\r
3371 start, length - 1));
\r
3372 m_utilUpperWeightRange_[length].m_count_ = 0;
\r
3373 while (--length >= 2) {
\r
3374 m_utilLowerWeightRange_[length].m_count_ = m_utilUpperWeightRange_[length].m_count_ = 0;
\r
3382 // copy the ranges, shortest first, into the result array
\r
3383 int rangeCount = 0;
\r
3384 if (m_utilWeightRange_.m_count_ > 0) {
\r
3385 ranges[0] = new WeightRange(m_utilWeightRange_);
\r
3388 for (int length = 2; length <= 4; ++length) {
\r
3389 // copy upper first so that later the middle range is more likely
\r
3390 // the first one to use
\r
3391 if (m_utilUpperWeightRange_[length].m_count_ > 0) {
\r
3392 ranges[rangeCount] = new WeightRange(
\r
3393 m_utilUpperWeightRange_[length]);
\r
3396 if (m_utilLowerWeightRange_[length].m_count_ > 0) {
\r
3397 ranges[rangeCount] = new WeightRange(
\r
3398 m_utilLowerWeightRange_[length]);
\r
3402 return rangeCount;
\r
3406 * Truncates the weight with length
\r
3410 * @return truncated weight
\r
3412 private static final int truncateWeight(int weight, int length) {
\r
3413 return weight & (0xffffffff << ((4 - length) << 3));
\r
3417 * Length of the weight
\r
3420 * @return length of the weight
\r
3422 private static final int lengthOfWeight(int weight) {
\r
3423 if ((weight & 0xffffff) == 0) {
\r
3425 } else if ((weight & 0xffff) == 0) {
\r
3427 } else if ((weight & 0xff) == 0) {
\r
3434 * Increment the weight trail
\r
3438 * @return new weight
\r
3440 private static final int incWeightTrail(int weight, int length) {
\r
3441 return weight + (1 << ((4 - length) << 3));
\r
3445 * Decrement the weight trail
\r
3449 * @return new weight
\r
3451 private static int decWeightTrail(int weight, int length) {
\r
3452 return weight - (1 << ((4 - length) << 3));
\r
3456 * Gets the codepoint
\r
3459 * contraction table
\r
3460 * @param codePoint
\r
3461 * code point to look for
\r
3462 * @return the offset to the code point
\r
3464 private static int findCP(BasicContractionTable tbl, char codePoint) {
\r
3466 while (codePoint > tbl.m_codePoints_.charAt(position)) {
\r
3468 if (position > tbl.m_codePoints_.length()) {
\r
3472 if (codePoint == tbl.m_codePoints_.charAt(position)) {
\r
3480 * Finds a contraction ce
\r
3487 private static int findCE(ContractionTable table, int element, char ch) {
\r
3488 if (table == null) {
\r
3489 return CE_NOT_FOUND_;
\r
3491 BasicContractionTable tbl = getBasicContractionTable(table, element);
\r
3492 if (tbl == null) {
\r
3493 return CE_NOT_FOUND_;
\r
3495 int position = findCP(tbl, ch);
\r
3496 if (position > tbl.m_CEs_.size() || position < 0) {
\r
3497 return CE_NOT_FOUND_;
\r
3499 return tbl.m_CEs_.get(position).intValue();
\r
3503 * Checks if the string is tailored in the contraction
\r
3506 * contraction table
\r
3509 * character array to check
\r
3512 * @return true if it is tailored
\r
3514 private static boolean isTailored(ContractionTable table, int element,
\r
3515 char array[], int offset) {
\r
3516 while (array[offset] != 0) {
\r
3517 element = findCE(table, element, array[offset]);
\r
3518 if (element == CE_NOT_FOUND_) {
\r
3521 if (!isContractionTableElement(element)) {
\r
3526 if (getCE(table, element, 0) != CE_NOT_FOUND_) {
\r
3534 * Assemble RuleBasedCollator
\r
3541 private void assembleTable(BuildTable t, RuleBasedCollator collator) {
\r
3542 IntTrieBuilder mapping = t.m_mapping_;
\r
3543 Vector<Integer> expansions = t.m_expansions_;
\r
3544 ContractionTable contractions = t.m_contractions_;
\r
3545 MaxExpansionTable maxexpansion = t.m_maxExpansions_;
\r
3547 // contraction offset has to be in since we are building on the
\r
3548 // UCA contractions
\r
3549 // int beforeContractions = (HEADER_SIZE_
\r
3550 // + paddedsize(expansions.size() << 2)) >>> 1;
\r
3551 collator.m_contractionOffset_ = 0;
\r
3552 int contractionsSize = constructTable(contractions);
\r
3554 // the following operation depends on the trie data. Therefore, we have
\r
3555 // to do it before the trie is compacted
\r
3556 // sets jamo expansions
\r
3557 getMaxExpansionJamo(mapping, maxexpansion, t.m_maxJamoExpansions_,
\r
3558 collator.m_isJamoSpecial_);
\r
3560 // TODO: LATIN1 array is now in the utrie - it should be removed from
\r
3561 // the calculation
\r
3562 setAttributes(collator, t.m_options_);
\r
3563 // copy expansions
\r
3564 int size = expansions.size();
\r
3565 collator.m_expansion_ = new int[size];
\r
3566 for (int i = 0; i < size; i++) {
\r
3567 collator.m_expansion_[i] = expansions.get(i).intValue();
\r
3569 // contractions block
\r
3570 if (contractionsSize != 0) {
\r
3571 // copy contraction index
\r
3572 collator.m_contractionIndex_ = new char[contractionsSize];
\r
3573 contractions.m_codePoints_.getChars(0, contractionsSize,
\r
3574 collator.m_contractionIndex_, 0);
\r
3575 // copy contraction collation elements
\r
3576 collator.m_contractionCE_ = new int[contractionsSize];
\r
3577 for (int i = 0; i < contractionsSize; i++) {
\r
3578 collator.m_contractionCE_[i] = contractions.m_CEs_.get(i).intValue();
\r
3581 // copy mapping table
\r
3582 collator.m_trie_ = mapping.serialize(t,
\r
3583 RuleBasedCollator.DataManipulate.getInstance());
\r
3584 // copy max expansion table
\r
3585 // not copying the first element which is a dummy
\r
3586 // to be in synch with icu4c's builder, we continue to use the
\r
3587 // expansion offset
\r
3588 // omitting expansion offset in builder
\r
3589 collator.m_expansionOffset_ = 0;
\r
3590 size = maxexpansion.m_endExpansionCE_.size();
\r
3591 collator.m_expansionEndCE_ = new int[size - 1];
\r
3592 for (int i = 1; i < size; i++) {
\r
3593 collator.m_expansionEndCE_[i - 1] = maxexpansion.m_endExpansionCE_
\r
3594 .get(i).intValue();
\r
3596 collator.m_expansionEndCEMaxSize_ = new byte[size - 1];
\r
3597 for (int i = 1; i < size; i++) {
\r
3598 collator.m_expansionEndCEMaxSize_[i - 1] = maxexpansion.m_expansionCESize_
\r
3599 .get(i).byteValue();
\r
3601 // Unsafe chars table. Finish it off, then copy it.
\r
3602 unsafeCPAddCCNZ(t);
\r
3603 // Or in unsafebits from UCA, making a combined table.
\r
3604 for (int i = 0; i < UNSAFECP_TABLE_SIZE_; i++) {
\r
3605 t.m_unsafeCP_[i] |= RuleBasedCollator.UCA_.m_unsafe_[i];
\r
3607 collator.m_unsafe_ = t.m_unsafeCP_;
\r
3609 // Finish building Contraction Ending chars hash table and then copy it
\r
3611 // Or in unsafebits from UCA, making a combined table
\r
3612 for (int i = 0; i < UNSAFECP_TABLE_SIZE_; i++) {
\r
3613 t.m_contrEndCP_[i] |= RuleBasedCollator.UCA_.m_contractionEnd_[i];
\r
3615 collator.m_contractionEnd_ = t.m_contrEndCP_;
\r
3619 * Sets this collator to use the all options and tables in UCA.
\r
3622 * which attribute is to be set
\r
3626 private static final void setAttributes(RuleBasedCollator collator,
\r
3627 CollationRuleParser.OptionSet option) {
\r
3628 collator.latinOneFailed_ = true;
\r
3629 collator.m_caseFirst_ = option.m_caseFirst_;
\r
3630 collator.setDecomposition(option.m_decomposition_);
\r
3632 .setAlternateHandlingShifted(option.m_isAlternateHandlingShifted_);
\r
3633 collator.setCaseLevel(option.m_isCaseLevel_);
\r
3634 collator.setFrenchCollation(option.m_isFrenchCollation_);
\r
3635 collator.m_isHiragana4_ = option.m_isHiragana4_;
\r
3636 collator.setStrength(option.m_strength_);
\r
3637 collator.m_variableTopValue_ = option.m_variableTopValue_;
\r
3638 collator.latinOneFailed_ = false;
\r
3642 * Constructing the contraction table
\r
3645 * contraction table
\r
3648 private int constructTable(ContractionTable table) {
\r
3649 // See how much memory we need
\r
3650 int tsize = table.m_elements_.size();
\r
3654 table.m_offsets_.clear();
\r
3656 for (int i = 0; i < tsize; i++) {
\r
3657 table.m_offsets_.add(new Integer(position));
\r
3658 position += table.m_elements_.get(i).m_CEs_
\r
3661 table.m_CEs_.clear();
\r
3662 table.m_codePoints_.delete(0, table.m_codePoints_.length());
\r
3663 // Now stuff the things in
\r
3664 StringBuilder cpPointer = table.m_codePoints_;
\r
3665 Vector<Integer> CEPointer = table.m_CEs_;
\r
3666 for (int i = 0; i < tsize; i++) {
\r
3667 BasicContractionTable bct = table.m_elements_.get(i);
\r
3668 int size = bct.m_CEs_.size();
\r
3671 int offset = CEPointer.size();
\r
3672 CEPointer.add(bct.m_CEs_.get(0));
\r
3673 for (int j = 1; j < size; j++) {
\r
3674 char ch = bct.m_codePoints_.charAt(j);
\r
3675 char cc = (char) (UCharacter.getCombiningClass(ch) & 0xFF);
\r
3682 cpPointer.append(ch);
\r
3683 CEPointer.add(bct.m_CEs_.get(j));
\r
3685 cpPointer.insert(offset,
\r
3686 (char) (((ccMin == ccMax) ? 1 : 0 << 8) | ccMax));
\r
3687 for (int j = 0; j < size; j++) {
\r
3688 if (isContractionTableElement(CEPointer.get(offset + j).intValue())) {
\r
3689 int ce = CEPointer.get(offset + j).intValue();
\r
3690 CEPointer.set(offset + j,
\r
3691 new Integer(constructSpecialCE(getCETag(ce),
\r
3692 table.m_offsets_.get(getContractionOffset(ce))
\r
3698 for (int i = 0; i <= 0x10FFFF; i++) {
\r
3699 int CE = table.m_mapping_.getValue(i);
\r
3700 if (isContractionTableElement(CE)) {
\r
3701 CE = constructSpecialCE(getCETag(CE),
\r
3702 table.m_offsets_.get(getContractionOffset(CE)).intValue());
\r
3703 table.m_mapping_.setValue(i, CE);
\r
3710 * Get contraction offset
\r
3713 * collation element
\r
3714 * @return contraction offset
\r
3716 private static final int getContractionOffset(int ce) {
\r
3717 return ce & 0xFFFFFF;
\r
3721 * Gets the maximum Jamo expansion
\r
3725 * @param maxexpansion
\r
3726 * maximum expansion table
\r
3727 * @param maxjamoexpansion
\r
3728 * maximum jamo expansion table
\r
3729 * @param jamospecial
\r
3730 * is jamo special?
\r
3732 private static void getMaxExpansionJamo(IntTrieBuilder mapping,
\r
3733 MaxExpansionTable maxexpansion,
\r
3734 MaxJamoExpansionTable maxjamoexpansion, boolean jamospecial) {
\r
3735 int VBASE = 0x1161;
\r
3736 int TBASE = 0x11A8;
\r
3739 int v = VBASE + VCOUNT - 1;
\r
3740 int t = TBASE + TCOUNT - 1;
\r
3742 while (v >= VBASE) {
\r
3743 int ce = mapping.getValue(v);
\r
3744 if ((ce & RuleBasedCollator.CE_SPECIAL_FLAG_) != RuleBasedCollator.CE_SPECIAL_FLAG_) {
\r
3745 setMaxExpansion(ce, (byte) 2, maxexpansion);
\r
3750 while (t >= TBASE) {
\r
3751 int ce = mapping.getValue(t);
\r
3752 if ((ce & RuleBasedCollator.CE_SPECIAL_FLAG_) != RuleBasedCollator.CE_SPECIAL_FLAG_) {
\r
3753 setMaxExpansion(ce, (byte) 3, maxexpansion);
\r
3757 // According to the docs, 99% of the time, the Jamo will not be special
\r
3758 if (jamospecial) {
\r
3759 // gets the max expansion in all unicode characters
\r
3760 int count = maxjamoexpansion.m_endExpansionCE_.size();
\r
3761 byte maxTSize = (byte) (maxjamoexpansion.m_maxLSize_
\r
3762 + maxjamoexpansion.m_maxVSize_ + maxjamoexpansion.m_maxTSize_);
\r
3763 byte maxVSize = (byte) (maxjamoexpansion.m_maxLSize_ + maxjamoexpansion.m_maxVSize_);
\r
3765 while (count > 0) {
\r
3767 if ((maxjamoexpansion.m_isV_.get(count))
\r
3768 .booleanValue() == true) {
\r
3770 (maxjamoexpansion.m_endExpansionCE_
\r
3771 .get(count)).intValue(), maxVSize,
\r
3775 (maxjamoexpansion.m_endExpansionCE_
\r
3776 .get(count)).intValue(), maxTSize,
\r
3784 * To the UnsafeCP hash table, add all chars with combining class != 0
\r
3789 private final void unsafeCPAddCCNZ(BuildTable t) {
\r
3790 boolean buildCMTable = (buildCMTabFlag & (t.cmLookup == null));
\r
3791 char[] cm = null; // combining mark array
\r
3792 int[] index = new int[256];
\r
3795 if (buildCMTable) {
\r
3796 cm = new char[0x10000];
\r
3798 for (char c = 0; c < 0xffff; c++) {
\r
3799 int fcd = m_nfcImpl_.getFCD16FromSingleLead(c); // TODO: review for handling supplementary characters
\r
3800 if (fcd >= 0x100 || // if the leading combining class(c) > 0 ||
\r
3801 (UTF16.isLeadSurrogate(c) && fcd != 0)) {
\r
3802 // c is a leading surrogate with some FCD data
\r
3803 unsafeCPSet(t.m_unsafeCP_, c);
\r
3804 if (buildCMTable && (fcd != 0)) {
\r
3805 int cc = (fcd & 0xff);
\r
3806 int pos = (cc << 8) + index[cc];
\r
3814 if (t.m_prefixLookup_ != null) {
\r
3815 Enumeration<Elements> els = t.m_prefixLookup_.elements();
\r
3816 while (els.hasMoreElements()) {
\r
3817 Elements e = els.nextElement();
\r
3818 // codepoints here are in the NFD form. We need to add the
\r
3819 // first code point of the NFC form to unsafe, because
\r
3820 // strcoll needs to backup over them.
\r
3821 // weiv: This is wrong! See the comment above.
\r
3822 // String decomp = Normalizer.decompose(e.m_cPoints_, true);
\r
3823 // unsafeCPSet(t.m_unsafeCP_, decomp.charAt(0));
\r
3825 String comp = Normalizer.compose(e.m_cPoints_, false);
\r
3826 unsafeCPSet(t.m_unsafeCP_, comp.charAt(0));
\r
3830 if (buildCMTable) {
\r
3831 t.cmLookup = new CombinClassTable();
\r
3832 t.cmLookup.generate(cm, count, index);
\r
3842 * RuleBasedCollator
\r
3844 * collation element iterator
\r
3851 private boolean enumCategoryRangeClosureCategory(BuildTable t,
\r
3852 RuleBasedCollator collator, CollationElementIterator colEl,
\r
3853 int start, int limit, int type) {
\r
3854 if (type != UCharacterCategory.UNASSIGNED
\r
3855 && type != UCharacterCategory.PRIVATE_USE) {
\r
3856 // if the range is assigned - we might ommit more categories later
\r
3858 for (int u32 = start; u32 < limit; u32++) {
\r
3859 String decomp = m_nfcImpl_.getDecomposition(u32);
\r
3860 if (decomp != null) {
\r
3861 String comp = UCharacter.toString(u32);
\r
3862 if (!collator.equals(comp, decomp)) {
\r
3863 m_utilElement_.m_cPoints_ = decomp;
\r
3864 m_utilElement_.m_prefix_ = 0;
\r
3865 Elements prefix = t.m_prefixLookup_.get(m_utilElement_);
\r
3866 if (prefix == null) {
\r
3867 m_utilElement_.m_cPoints_ = comp;
\r
3868 m_utilElement_.m_prefix_ = 0;
\r
3869 m_utilElement_.m_prefixChars_ = null;
\r
3870 colEl.setText(decomp);
\r
3871 int ce = colEl.next();
\r
3872 m_utilElement_.m_CELength_ = 0;
\r
3873 while (ce != CollationElementIterator.NULLORDER) {
\r
3874 m_utilElement_.m_CEs_[m_utilElement_.m_CELength_++] = ce;
\r
3875 ce = colEl.next();
\r
3878 m_utilElement_.m_cPoints_ = comp;
\r
3879 m_utilElement_.m_prefix_ = 0;
\r
3880 m_utilElement_.m_prefixChars_ = null;
\r
3881 m_utilElement_.m_CELength_ = 1;
\r
3882 m_utilElement_.m_CEs_[0] = prefix.m_mapCE_;
\r
3883 // This character uses a prefix. We have to add it
\r
3884 // to the unsafe table, as it decomposed form is
\r
3885 // already in. In Japanese, this happens for \u309e
\r
3887 // Since unsafeCPSet is static in ucol_elm, we are
\r
3888 // going to wrap it up in the unsafeCPAddCCNZ
\r
3891 addAnElement(t, m_utilElement_);
\r
3900 * Determine if a character is a Jamo
\r
3903 * character to test
\r
3904 * @return true if ch is a Jamo, false otherwise
\r
3906 private static final boolean isJamo(char ch) {
\r
3907 return (ch >= 0x1100 && ch <= 0x1112) || (ch >= 0x1175 && ch <= 0x1161)
\r
3908 || (ch >= 0x11A8 && ch <= 0x11C2);
\r
3912 * Produces canonical closure
\r
3914 private void canonicalClosure(BuildTable t) {
\r
3915 BuildTable temp = new BuildTable(t);
\r
3916 assembleTable(temp, temp.m_collator_);
\r
3917 // produce canonical closure
\r
3918 CollationElementIterator coleiter = temp.m_collator_
\r
3919 .getCollationElementIterator("");
\r
3920 RangeValueIterator typeiter = UCharacter.getTypeIterator();
\r
3921 RangeValueIterator.Element element = new RangeValueIterator.Element();
\r
3922 while (typeiter.next(element)) {
\r
3923 enumCategoryRangeClosureCategory(t, temp.m_collator_, coleiter,
\r
3924 element.start, element.limit, element.value);
\r
3927 t.cmLookup = temp.cmLookup;
\r
3928 temp.cmLookup = null;
\r
3930 for (int i = 0; i < m_parser_.m_resultLength_; i++) {
\r
3931 char baseChar, firstCM;
\r
3932 // now we need to generate the CEs
\r
3933 // We stuff the initial value in the buffers, and increase the
\r
3934 // appropriate buffer according to strength */
\r
3935 // createElements(t, m_parser_.m_listHeader_[i]);
\r
3936 CollationRuleParser.Token tok = m_parser_.m_listHeader_[i].m_first_;
\r
3937 m_utilElement_.clear();
\r
3938 while (tok != null) {
\r
3939 m_utilElement_.m_prefix_ = 0;// el.m_prefixChars_;
\r
3940 m_utilElement_.m_cPointsOffset_ = 0; // el.m_uchars_;
\r
3941 if (tok.m_prefix_ != 0) {
\r
3942 // we will just copy the prefix here, and adjust accordingly
\r
3944 // the addPrefix function in ucol_elm. The reason is that we
\r
3945 // need to add both composed AND decomposed elements to the
\r
3947 int size = tok.m_prefix_ >> 24;
\r
3948 int offset = tok.m_prefix_ & 0x00FFFFFF;
\r
3949 m_utilElement_.m_prefixChars_ = m_parser_.m_source_
\r
3950 .substring(offset, offset + size);
\r
3951 size = (tok.m_source_ >> 24) - (tok.m_prefix_ >> 24);
\r
3952 offset = (tok.m_source_ & 0x00FFFFFF)
\r
3953 + (tok.m_prefix_ >> 24);
\r
3954 m_utilElement_.m_uchars_ = m_parser_.m_source_.substring(
\r
3955 offset, offset + size);
\r
3957 m_utilElement_.m_prefixChars_ = null;
\r
3958 int offset = tok.m_source_ & 0x00FFFFFF;
\r
3959 int size = tok.m_source_ >>> 24;
\r
3960 m_utilElement_.m_uchars_ = m_parser_.m_source_.substring(
\r
3961 offset, offset + size);
\r
3963 m_utilElement_.m_cPoints_ = m_utilElement_.m_uchars_;
\r
3965 baseChar = firstCM = 0; // reset
\r
3966 for (int j = 0; j < m_utilElement_.m_cPoints_.length()
\r
3967 - m_utilElement_.m_cPointsOffset_; j++) {
\r
3969 int fcd = m_nfcImpl_.getFCD16FromSingleLead(m_utilElement_.m_cPoints_.charAt(j)); // TODO: review for handling supplementary characters
\r
3970 if ((fcd & 0xff) == 0) {
\r
3971 baseChar = m_utilElement_.m_cPoints_.charAt(j);
\r
3973 if ((baseChar != 0) && (firstCM == 0)) {
\r
3974 firstCM = m_utilElement_.m_cPoints_.charAt(j); // first
\r
3981 if ((baseChar != 0) && (firstCM != 0)) {
\r
3982 addTailCanonicalClosures(t, temp.m_collator_, coleiter,
\r
3983 baseChar, firstCM);
\r
3985 tok = tok.m_next_;
\r
3990 private void addTailCanonicalClosures(BuildTable t,
\r
3991 RuleBasedCollator m_collator, CollationElementIterator colEl,
\r
3992 char baseChar, char cMark) {
\r
3993 if (t.cmLookup == null) {
\r
3996 CombinClassTable cmLookup = t.cmLookup;
\r
3997 int[] index = cmLookup.index;
\r
3998 int cClass = m_nfcImpl_.getFCD16FromSingleLead(cMark) & 0xff; // TODO: review for handling supplementary characters
\r
4000 char[] precompCh = new char[256];
\r
4001 int[] precompClass = new int[256];
\r
4002 int precompLen = 0;
\r
4003 Elements element = new Elements();
\r
4006 maxIndex = index[cClass - 1];
\r
4008 for (int i = 0; i < maxIndex; i++) {
\r
4009 StringBuilder decompBuf = new StringBuilder();
\r
4010 decompBuf.append(baseChar).append(cmLookup.cPoints[i]);
\r
4011 String comp = Normalizer.compose(decompBuf.toString(), false);
\r
4012 if (comp.length() == 1) {
\r
4013 precompCh[precompLen] = comp.charAt(0);
\r
4014 precompClass[precompLen] = (m_nfcImpl_.getFCD16FromSingleLead(cmLookup.cPoints[i]) & 0xff); // TODO: review for handling supplementary characters
\r
4016 StringBuilder decomp = new StringBuilder();
\r
4017 for (int j = 0; j < m_utilElement_.m_cPoints_.length(); j++) {
\r
4018 if (m_utilElement_.m_cPoints_.charAt(j) == cMark) {
\r
4019 decomp.append(cmLookup.cPoints[i]);
\r
4021 decomp.append(m_utilElement_.m_cPoints_.charAt(j));
\r
4024 comp = Normalizer.compose(decomp.toString(), false);
\r
4025 StringBuilder buf = new StringBuilder(comp);
\r
4026 buf.append(cMark);
\r
4027 decomp.append(cMark);
\r
4028 comp = buf.toString();
\r
4030 element.m_cPoints_ = decomp.toString();
\r
4031 element.m_CELength_ = 0;
\r
4032 element.m_prefix_ = 0;
\r
4033 Elements prefix = t.m_prefixLookup_.get(element);
\r
4034 element.m_cPoints_ = comp;
\r
4035 element.m_uchars_ = comp;
\r
4037 if (prefix == null) {
\r
4038 element.m_prefix_ = 0;
\r
4039 element.m_prefixChars_ = null;
\r
4040 colEl.setText(decomp.toString());
\r
4041 int ce = colEl.next();
\r
4042 element.m_CELength_ = 0;
\r
4043 while (ce != CollationElementIterator.NULLORDER) {
\r
4044 element.m_CEs_[element.m_CELength_++] = ce;
\r
4045 ce = colEl.next();
\r
4048 element.m_cPoints_ = comp;
\r
4049 element.m_prefix_ = 0;
\r
4050 element.m_prefixChars_ = null;
\r
4051 element.m_CELength_ = 1;
\r
4052 element.m_CEs_[0] = prefix.m_mapCE_;
\r
4054 setMapCE(t, element);
\r
4055 finalizeAddition(t, element);
\r
4057 if (comp.length() > 2) {
\r
4058 // This is a fix for tailoring contractions with accented
\r
4059 // character at the end of contraction string.
\r
4060 addFCD4AccentedContractions(t, colEl, comp, element);
\r
4062 if (precompLen > 1) {
\r
4063 precompLen = addMultiCMontractions(t, colEl, element,
\r
4064 precompCh, precompClass, precompLen, cMark, i,
\r
4065 decomp.toString());
\r
4072 private void setMapCE(BuildTable t, Elements element) {
\r
4073 Vector<Integer> expansions = t.m_expansions_;
\r
4074 element.m_mapCE_ = 0;
\r
4076 if (element.m_CELength_ == 2 // a two CE expansion
\r
4077 && RuleBasedCollator.isContinuation(element.m_CEs_[1])
\r
4078 && (element.m_CEs_[1] & (~(0xFF << 24 | RuleBasedCollator.CE_CONTINUATION_MARKER_))) == 0 // that
\r
4084 && (((element.m_CEs_[0] >> 8) & 0xFF) == RuleBasedCollator.BYTE_COMMON_)
\r
4085 // a common secondary
\r
4086 && ((element.m_CEs_[0] & 0xFF) == RuleBasedCollator.BYTE_COMMON_)) { // and
\r
4091 element.m_mapCE_ = RuleBasedCollator.CE_SPECIAL_FLAG_
\r
4092 // a long primary special
\r
4093 | (CE_LONG_PRIMARY_TAG_ << 24)
\r
4094 // first and second byte of primary
\r
4095 | ((element.m_CEs_[0] >> 8) & 0xFFFF00)
\r
4096 // third byte of primary
\r
4097 | ((element.m_CEs_[1] >> 24) & 0xFF);
\r
4099 // omitting expansion offset in builder
\r
4100 // (HEADER_SIZE_ >> 2)
\r
4101 int expansion = RuleBasedCollator.CE_SPECIAL_FLAG_
\r
4102 | (CE_EXPANSION_TAG_ << RuleBasedCollator.CE_TAG_SHIFT_)
\r
4103 | (addExpansion(expansions, element.m_CEs_[0]) << 4)
\r
4106 for (int i = 1; i < element.m_CELength_; i++) {
\r
4107 addExpansion(expansions, element.m_CEs_[i]);
\r
4109 if (element.m_CELength_ <= 0xF) {
\r
4110 expansion |= element.m_CELength_;
\r
4112 addExpansion(expansions, 0);
\r
4114 element.m_mapCE_ = expansion;
\r
4115 setMaxExpansion(element.m_CEs_[element.m_CELength_ - 1],
\r
4116 (byte) element.m_CELength_, t.m_maxExpansions_);
\r
4120 private int addMultiCMontractions(BuildTable t,
\r
4121 CollationElementIterator colEl, Elements element, char[] precompCh,
\r
4122 int[] precompClass, int maxComp, char cMark, int cmPos,
\r
4125 CombinClassTable cmLookup = t.cmLookup;
\r
4126 char[] combiningMarks = { cMark };
\r
4127 int cMarkClass = UCharacter.getCombiningClass(cMark) & 0xFF;
\r
4128 String comMark = new String(combiningMarks);
\r
4129 int noOfPrecomposedChs = maxComp;
\r
4131 for (int j = 0; j < maxComp; j++) {
\r
4133 StringBuilder temp;
\r
4136 String newDecomp, comp;
\r
4138 if (count == 0) { // Decompose the saved precomposed char.
\r
4139 newDecomp = Normalizer.decompose(
\r
4140 new String(precompCh, j, 1), false);
\r
4141 temp = new StringBuilder(newDecomp);
\r
4142 temp.append(cmLookup.cPoints[cmPos]);
\r
4143 newDecomp = temp.toString();
\r
4145 temp = new StringBuilder(decomp);
\r
4146 temp.append(precompCh[j]);
\r
4147 newDecomp = temp.toString();
\r
4149 comp = Normalizer.compose(newDecomp, false);
\r
4150 if (comp.length() == 1) {
\r
4151 temp.append(cMark);
\r
4152 element.m_cPoints_ = temp.toString();
\r
4153 element.m_CELength_ = 0;
\r
4154 element.m_prefix_ = 0;
\r
4155 Elements prefix = t.m_prefixLookup_.get(element);
\r
4156 element.m_cPoints_ = comp + comMark;
\r
4157 if (prefix == null) {
\r
4158 element.m_prefix_ = 0;
\r
4159 element.m_prefixChars_ = null;
\r
4160 colEl.setText(temp.toString());
\r
4161 int ce = colEl.next();
\r
4162 element.m_CELength_ = 0;
\r
4163 while (ce != CollationElementIterator.NULLORDER) {
\r
4164 element.m_CEs_[element.m_CELength_++] = ce;
\r
4165 ce = colEl.next();
\r
4168 element.m_cPoints_ = comp;
\r
4169 element.m_prefix_ = 0;
\r
4170 element.m_prefixChars_ = null;
\r
4171 element.m_CELength_ = 1;
\r
4172 element.m_CEs_[0] = prefix.m_mapCE_;
\r
4174 setMapCE(t, element);
\r
4175 finalizeAddition(t, element);
\r
4176 precompCh[noOfPrecomposedChs] = comp.charAt(0);
\r
4177 precompClass[noOfPrecomposedChs] = cMarkClass;
\r
4178 noOfPrecomposedChs++;
\r
4180 } while (++count < 2 && (precompClass[j] == cMarkClass));
\r
4182 return noOfPrecomposedChs;
\r
4185 private void addFCD4AccentedContractions(BuildTable t,
\r
4186 CollationElementIterator colEl, String data, Elements element) {
\r
4187 String decomp = Normalizer.decompose(data, false);
\r
4188 String comp = Normalizer.compose(data, false);
\r
4190 element.m_cPoints_ = decomp;
\r
4191 element.m_CELength_ = 0;
\r
4192 element.m_prefix_ = 0;
\r
4193 Elements prefix = t.m_prefixLookup_.get(element);
\r
4194 if (prefix == null) {
\r
4195 element.m_cPoints_ = comp;
\r
4196 element.m_prefix_ = 0;
\r
4197 element.m_prefixChars_ = null;
\r
4198 element.m_CELength_ = 0;
\r
4199 colEl.setText(decomp);
\r
4200 int ce = colEl.next();
\r
4201 element.m_CELength_ = 0;
\r
4202 while (ce != CollationElementIterator.NULLORDER) {
\r
4203 element.m_CEs_[element.m_CELength_++] = ce;
\r
4204 ce = colEl.next();
\r
4206 addAnElement(t, element);
\r
4210 private void processUCACompleteIgnorables(BuildTable t) {
\r
4211 TrieIterator trieiterator = new TrieIterator(
\r
4212 RuleBasedCollator.UCA_.m_trie_);
\r
4213 RangeValueIterator.Element element = new RangeValueIterator.Element();
\r
4214 while (trieiterator.next(element)) {
\r
4215 int start = element.start;
\r
4216 int limit = element.limit;
\r
4217 if (element.value == 0) {
\r
4218 while (start < limit) {
\r
4219 int CE = t.m_mapping_.getValue(start);
\r
4220 if (CE == CE_NOT_FOUND_) {
\r
4221 m_utilElement_.m_prefix_ = 0;
\r
4222 m_utilElement_.m_uchars_ = UCharacter.toString(start);
\r
4223 m_utilElement_.m_cPoints_ = m_utilElement_.m_uchars_;
\r
4224 m_utilElement_.m_cPointsOffset_ = 0;
\r
4225 m_utilElement_.m_CELength_ = 1;
\r
4226 m_utilElement_.m_CEs_[0] = 0;
\r
4227 addAnElement(t, m_utilElement_);
\r
projects / Dictionary.git / blob