organicmaps/icu
13
Fork 0
mirror of https://github.com/unicode-org/icu.git synced 2025-04-13 08:53:20 +00:00
Code Issues 1 Projects Releases Wiki Activity Actions 3

ICU-3729 port [before] fix

Browse source 
X-SVN-Rev: 15889
This commit is contained in:
Vladimir Weinstein 2004-06-16 18:19:51 +00:00
parent 3b290bedcb
commit 152733ca80
4 changed files with 94 additions and 20 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
icu4j/src/com/ibm/icu/dev/test/collator/CollationMiscTest.java
View file

@ -2085,9 +2085,9 @@ public class CollationMiscTest extends TestFmwk {
"xAx"
};
/* TODO: port builder fixes to before */
/*genericRulesStarter(rules, test);*/
genericRulesStarter(rules, test);
genericLocaleStarter(new Locale("zh"), test);
/*genericRulesStarter(rules, test2);*/
genericRulesStarter(rules, test2);
genericLocaleStarter(new Locale("zh"), test2);
}

37
icu4j/src/com/ibm/icu/text/CollationParsedRuleBuilder.java
View file

@ -110,12 +110,25 @@ final class CollationParsedRuleBuilder
&& result > 0) {
// this condition should prevent falling off the edge of the
// world
// here, we end up in a singularity - zero
prevresult[0] = m_table_[3 * (-- result)];
prevresult[1] = m_table_[3 * result + 1];
// here, we end up in a singularity - zero
prevresult[0] = m_table_[3 * (-- result)];
prevresult[1] = m_table_[3 * result + 1];
}
return result;
}
final int getCEStrengthDifference(int CE, int contCE,
int prevCE, int prevContCE) {
int strength = Collator.TERTIARY;
while(
((prevCE & STRENGTH_MASK_[strength]) != (CE & STRENGTH_MASK_[strength])
|| (prevContCE & STRENGTH_MASK_[strength]) != (contCE & STRENGTH_MASK_[strength]))
&& (strength != 0)) {
strength--;
}
return strength;
}
return result;
}
/**
* Finding the inverse CE of the argument CEs
@ -299,9 +312,12 @@ final class CollationParsedRuleBuilder
Collator.SECONDARY);
listheader.m_gapsHi_[3 * st + 2] = (t1 & 0x3f) << 24
| (t2 & 0x3f) << 16;
pos --;
t1 = m_table_[3 * pos];
t2 = m_table_[3 * pos + 1];
//pos --;
//t1 = m_table_[3 * pos];
//t2 = m_table_[3 * pos + 1];
t1 = listheader.m_baseCE_;
t2 = listheader.m_baseContCE_;
listheader.m_gapsLo_[3 * st] = mergeCE(t1, t2,
Collator.PRIMARY);
listheader.m_gapsLo_[3 * st + 1] = mergeCE(t1, t2,
@ -1500,9 +1516,10 @@ final class CollationParsedRuleBuilder
if (Utility.compareUnsigned(low,
RuleBasedCollator.COMMON_BOTTOM_2_ << 24) < 0) {
g.m_rangesLength_ = allocateWeights(
RuleBasedCollator.COMMON_BOTTOM_2_ << 24,
RuleBasedCollator.BYTE_UNSHIFTED_MIN_ << 24,
high, count, maxbyte, g.m_ranges_);
g.m_current_ = RuleBasedCollator.COMMON_BOTTOM_2_ << 24;
g.m_current_ = nextWeight(g);
//g.m_current_ = RuleBasedCollator.COMMON_BOTTOM_2_ << 24;
return g.m_current_;
}
}

71
icu4j/src/com/ibm/icu/text/CollationRuleParser.java
View file

@ -153,6 +153,7 @@ final class CollationRuleParser
Token m_previous_;
Token m_next_;
StringBuffer m_rules_;
char m_flags_;
// package private constructors ---------------------------------------
@ -883,6 +884,18 @@ final class CollationRuleParser
m_extraCurrent_ += size + m_parsedToken_.m_extensionLen_;
}
}
// if the previous token was a reset before, the strength of this
// token must match the strength of before. Otherwise we have an
// undefined situation.
// In other words, we currently have a cludge which we use to
// represent &a >> x. This is written as &[before 2]a << x.
if((lastToken.m_flags_ & TOKEN_BEFORE_) != 0) {
int beforeStrength = (lastToken.m_flags_ & TOKEN_BEFORE_) - 1;
if(beforeStrength != sourceToken.m_strength_) {
throwParseException(m_source_.toString(), m_current_);
}
}
}
else {
if (lastToken != null && lastStrength == TOKEN_RESET_) {
@ -1585,6 +1598,36 @@ final class CollationRuleParser
int invpos = CollationParsedRuleBuilder.INVERSE_UCA_.getInversePrevCE(
basece, basecontce,
strength, m_utilCEBuffer_);
// we got the previous CE. Now we need to see if the difference between
// the two CEs is really of the requested strength.
// if it's a bigger difference (we asked for secondary and got primary), we
// need to modify the CE.
if(CollationParsedRuleBuilder.INVERSE_UCA_.getCEStrengthDifference(basece, basecontce, m_utilCEBuffer_[0], m_utilCEBuffer_[1]) < strength) {
// adjust the strength
// now we are in the situation where our baseCE should actually be modified in
// order to get the CE in the right position.
if(strength == Collator.SECONDARY) {
m_utilCEBuffer_[0] = basece - 0x0200;
} else { // strength == UCOL_TERTIARY
m_utilCEBuffer_[0] = basece - 0x02;
}
if(RuleBasedCollator.isContinuation(basecontce)) {
if(strength == Collator.SECONDARY) {
m_utilCEBuffer_[1] = basecontce - 0x0200;
} else { // strength == UCOL_TERTIARY
m_utilCEBuffer_[1] = basecontce - 0x02;
}
}
}
/*
// the code below relies on getting a code point from the inverse table, in order to be
// able to merge the situations like &x < 9 &[before 1]a < d. This won't work:
// 1. There are many code points that have the same CE
// 2. The CE to codepoint table (things pointed to by CETable[3*invPos+2] are broken.
// Also, in case when there is no equivalent strength before an element, we have to actually
// construct one. For example, &[before 2]a << x won't result in x << a, because the element
// before a is a primary difference.
ch = CollationParsedRuleBuilder.INVERSE_UCA_.m_table_[3 * invpos
+ 2];
if ((ch & INVERSE_SIZE_MASK_) != 0) {
@ -1606,16 +1649,27 @@ final class CollationRuleParser
| m_parsedToken_.m_charsOffset_;
m_utilToken_.m_rules_ = m_source_;
sourcetoken = (Token)m_hashTable_.get(m_utilToken_);
*/
// here is how it should be. The situation such as &[before 1]a < x, should be
// resolved exactly as if we wrote &a > x.
// therefore, I don't really care if the UCA value before a has been changed.
// However, I do care if the strength between my element and the previous element
// is bigger then I wanted. So, if CE < baseCE and I wanted &[before 2], then i'll
// have to construct the base CE.
// if we found a tailored thing, we have to use the UCA value and
// construct a new reset token with constructed name
if (sourcetoken != null && sourcetoken.m_strength_ != TOKEN_RESET_) {
//if (sourcetoken != null && sourcetoken.m_strength_ != TOKEN_RESET_) {
// character to which we want to anchor is already tailored.
// We need to construct a new token which will be the anchor point
m_source_.setCharAt(m_extraCurrent_ - 1, '\uFFFE');
m_source_.append(ch);
m_extraCurrent_ ++;
m_parsedToken_.m_charsLen_ ++;
//m_source_.setCharAt(m_extraCurrent_ - 1, '\uFFFE');
//m_source_.append(ch);
//m_extraCurrent_ ++;
//m_parsedToken_.m_charsLen_ ++;
// grab before
m_parsedToken_.m_charsOffset_ -= 10;
m_parsedToken_.m_charsLen_ += 10;
m_listHeader_[m_resultLength_] = new TokenListHeader();
m_listHeader_[m_resultLength_].m_baseCE_
= m_utilCEBuffer_[0] & 0xFFFFFF3F;
@ -1633,7 +1687,7 @@ final class CollationRuleParser
m_listHeader_[m_resultLength_].m_indirect_ = false;
sourcetoken = new Token();
initAReset(-1, sourcetoken);
}
//}
}
return sourcetoken;
}
@ -1665,6 +1719,9 @@ final class CollationRuleParser
| m_parsedToken_.m_charsOffset_;
targetToken.m_expansion_ = m_parsedToken_.m_extensionLen_ << 24
| m_parsedToken_.m_extensionOffset_;
// keep the flags around so that we know about before
targetToken.m_flags_ = m_parsedToken_.m_flags_;
if (m_parsedToken_.m_prefixOffset_ != 0) {
throwParseException(m_rules_, m_parsedToken_.m_charsOffset_ - 1);
}

2
icu4j/src/com/ibm/icu/text/RuleBasedCollator.java
View file

@ -1922,7 +1922,7 @@ public final class RuleBasedCollator extends Collator
private static final byte BYTE_LEVEL_SEPARATOR_ = (byte)0x01;
private static final byte BYTE_SORTKEY_GLUE_ = (byte)0x02;
private static final byte BYTE_SHIFT_PREFIX_ = (byte)0x03;
private static final byte BYTE_UNSHIFTED_MIN_ = BYTE_SHIFT_PREFIX_;
/*private*/ static final byte BYTE_UNSHIFTED_MIN_ = BYTE_SHIFT_PREFIX_;
private static final byte BYTE_FIRST_UCA_ = BYTE_COMMON_;
static final byte CODAN_PLACEHOLDER = 0x24;
private static final byte BYTE_LAST_LATIN_PRIMARY_ = (byte)0x4C;