* The UnicodeSet class is not designed to be subclassed. * *
* UnicodeSet supports two APIs. The first is the operand
* API that allows the caller to modify the value of a UnicodeSet
* object. It conforms to Java 2's java.util.Set interface,
* although UnicodeSet does not actually implement that interface.
* All methods of Set are supported, with the modification that
* they take a character range or single character instead of an
* Object, and they take a UnicodeSet instead of a
* Collection. The operand API may be thought of in terms of
* boolean logic: a boolean OR is implemented by add, a boolean AND
* is implemented by retain, a boolean XOR is implemented by
* complement taking an argument, and a boolean NOT is implemented
* by complement with no argument. In terms of traditional set
* theory function names, add is a union, retain is an
* intersection, remove is an asymmetric difference, and
* complement with no argument is a set complement with respect to
* the superset range MIN_VALUE-MAX_VALUE
*
*
* The second API is the applyPattern()/toPattern()
* API from the java.text.Format-derived classes. Unlike the
* methods that add characters, add categories, and control the logic of the
* set, the method applyPattern() sets all attributes of a
* UnicodeSet at once, based on a string pattern.
*
*
* Pattern syntax *
* * Patterns are accepted by the constructors and theapplyPattern()
* methods and returned by the toPattern() method. These patterns
* follow a syntax similar to that employed by version 8 regular expression
* character classes. Here are some simple examples:
*
* ** * Any character may be preceded by a backslash in order to remove any special * meaning. White space characters, as defined by the Unicode * Pattern_White_Space property, are ignored, unless they are escaped. * **
** ** []No characters ** ** [a]The character 'a' ** ** [ae]The characters 'a' and 'e' ** ** [a-e]The characters 'a' through 'e' inclusive, in Unicode code * point order ** ** [\\u4E01]The character U+4E01 ** ** [a{ab}{ac}]The character 'a' and the multicharacter strings "ab" and * "ac" ** ** [\p{Lu}]All characters in the general category Uppercase Letter *
* Property patterns specify a set of characters having a certain property as * defined by the Unicode standard. Both the POSIX-like "[:Lu:]" and the * Perl-like syntax "\p{Lu}" are recognized. For a complete list of supported * property patterns, see the User's Guide for UnicodeSet at * * http://www.icu-project.org/userguide/unicodeSet.html. Actual * determination of property data is defined by the underlying Unicode database * as implemented by UCharacter. * *
* Patterns specify individual characters, ranges of characters, and Unicode * property sets. When elements are concatenated, they specify their union. To * complement a set, place a '^' immediately after the opening '['. Property * patterns are inverted by modifying their delimiters; "[:^foo]" and "\P{foo}". * In any other location, '^' has no special meaning. * *
* Ranges are indicated by placing two a '-' between two characters, as in * "a-z". This specifies the range of all characters from the left to the right, * in Unicode order. If the left character is greater than or equal to the right * character it is a syntax error. If a '-' occurs as the first character after * the opening '[' or '[^', or if it occurs as the last character before the * closing ']', then it is taken as a literal. Thus "[a\\-b]", "[-ab]", and * "[ab-]" all indicate the same set of three characters, 'a', 'b', and '-'. * *
* Sets may be intersected using the {@literal '&'} operator or the asymmetric * set difference may be taken using the '-' operator, for example, * "{@code [[:L:]&[\\u0000-\\u0FFF]]}" indicates the set of all Unicode letters * with values less than 4096. Operators ({@literal '&'} and '|') have equal * precedence and bind left-to-right. Thus "[[:L:]-[a-z]-[\\u0100-\\u01FF]]" is * equivalent to "[[[:L:]-[a-z]]-[\\u0100-\\u01FF]]". This only really matters * for difference; intersection is commutative. * *
[a]
* | The set containing 'a' * |
[a-z]
* | The set containing 'a' through 'z' and all letters in between, in Unicode * order * |
[^a-z]
* | The set containing all characters but 'a' through 'z', that is, U+0000 * through 'a'-1 and 'z'+1 through U+10FFFF * |
[[pat1][pat2]]
* | The union of sets specified by pat1 and pat2 * |
[[pat1]&[pat2]]
* | The intersection of sets specified by pat1 and pat2 * |
[[pat1]-[pat2]]
* | The asymmetric difference of sets specified by pat1 and * pat2 * |
[:Lu:] or \p{Lu}
* | The set of characters having the specified Unicode property; in this * case, Unicode uppercase letters * |
[:^Lu:] or \P{Lu}
* | The set of characters not having the given Unicode property * |
* Warning: you cannot add an empty string ("") to a UnicodeSet. *
* ** Formal syntax *
* ****
** ** pattern :=* ('[' '^'? item* ']') | * property* ** item :=* char | (char '-' char) | pattern-expr
** ** pattern-expr :=* pattern | pattern-expr pattern | * pattern-expr op pattern
** ** op :=* '&' | '-'
** ** special :=* '[' | ']' | '-'
** ** char :=any character that is not *specialany character
* | ('\\')
* | ('\u' hex hex hex hex)
** ** hex :=any character for which * *Character.digit(c, 16)returns a non-negative * result* ** property :=a Unicode property set pattern *
**
** *Legend: * **
** ** a := b* * amay be replaced byb* ** a?* zero or one instance of *a
** ** a** one or more instances of *a
** ** a | b* either *aorb
** ** 'a'* the literal string between the quotes *
* To iterate over contents of UnicodeSet, the following are available: *
* To replace, count elements, or delete spans, see
* {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
*
* @author Alan Liu
* @stable ICU 2.0
*/
public class UnicodeSet {
private static final int LOW = 0x000000; // LOW <= all valid values. ZERO for codepoints
private static final int HIGH = 0x110000; // HIGH > all valid values. 10000 for code units.
// 110000 for codepoints
/**
* Minimum value that can be stored in a UnicodeSet.
*
* @stable ICU 2.0
*/
public static final int MIN_VALUE = LOW;
/**
* Maximum value that can be stored in a UnicodeSet.
*
* @stable ICU 2.0
*/
public static final int MAX_VALUE = HIGH - 1;
private int len; // length used; list may be longer to minimize reallocs
private int[] list; // MUST be terminated with HIGH
private int[] rangeList; // internal buffer
private int[] buffer; // internal buffer
// NOTE: normally the field should be of type SortedSet; but that is missing a
// public clone!!
// is not private so that UnicodeSetIterator can get access
TreeSet
* To replace, count elements, or delete spans, see
* {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
*
* @param s The string to be spanned
* @param spanCondition The span condition
* @return the length of the span
* @stable ICU 4.4
*/
public int span(CharSequence s, SpanCondition spanCondition) {
return span(s, 0, spanCondition);
}
/**
* Span a string using this UnicodeSet. If the start index is less than 0, span
* will start from 0. If the start index is greater than the string length, span
* returns the string length.
*
* To replace, count elements, or delete spans, see
* {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
*
* @param s The string to be spanned
* @param start The start index that the span begins
* @param spanCondition The span condition
* @return the string index which ends the span (i.e. exclusive)
* @stable ICU 4.4
*/
public int span(CharSequence s, int start, SpanCondition spanCondition) {
int end = s.length();
if (start < 0) {
start = 0;
} else if (start >= end) {
return end;
}
if (bmpSet != null) {
// Frozen set without strings, or no string is relevant for span().
return bmpSet.span(s, start, spanCondition, null);
}
if (stringSpan != null) {
return stringSpan.span(s, start, spanCondition);
} else if (!strings.isEmpty()) {
int which = spanCondition == SpanCondition.NOT_CONTAINED ? UnicodeSetStringSpan.FWD_UTF16_NOT_CONTAINED
: UnicodeSetStringSpan.FWD_UTF16_CONTAINED;
UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList
* To replace, count elements, or delete spans, see
* {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
*
* @param outCount An output-only object (must not be null) for returning the
* count.
* @return the limit (exclusive end) of the span
*/
public int spanAndCount(CharSequence s, int start, SpanCondition spanCondition, OutputInt outCount) {
if (outCount == null) {
throw new IllegalArgumentException("outCount must not be null");
}
int end = s.length();
if (start < 0) {
start = 0;
} else if (start >= end) {
return end;
}
if (stringSpan != null) {
// We might also have bmpSet != null,
// but fully-contained strings are relevant for counting elements.
return stringSpan.spanAndCount(s, start, spanCondition, outCount);
} else if (bmpSet != null) {
return bmpSet.span(s, start, spanCondition, outCount);
} else if (!strings.isEmpty()) {
int which = spanCondition == SpanCondition.NOT_CONTAINED ? UnicodeSetStringSpan.FWD_UTF16_NOT_CONTAINED
: UnicodeSetStringSpan.FWD_UTF16_CONTAINED;
which |= UnicodeSetStringSpan.WITH_COUNT;
UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList
* To replace, count elements, or delete spans, see
* {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
*
* @param s The string to be spanned
* @param fromIndex The index of the char (exclusive) that the string should
* be spanned backwards
* @param spanCondition The span condition
* @return The string index which starts the span (i.e. inclusive).
* @stable ICU 4.4
*/
public int spanBack(CharSequence s, int fromIndex, SpanCondition spanCondition) {
if (fromIndex <= 0) {
return 0;
}
if (fromIndex > s.length()) {
fromIndex = s.length();
}
if (bmpSet != null) {
// Frozen set without strings, or no string is relevant for spanBack().
return bmpSet.spanBack(s, fromIndex, spanCondition);
}
if (stringSpan != null) {
return stringSpan.spanBack(s, fromIndex, spanCondition);
} else if (!strings.isEmpty()) {
int which = (spanCondition == SpanCondition.NOT_CONTAINED) ? UnicodeSetStringSpan.BACK_UTF16_NOT_CONTAINED
: UnicodeSetStringSpan.BACK_UTF16_CONTAINED;
UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList
* The functionality is straightforward for sets with only single code points,
* without strings (which is the common case):
*
* Note: Unpaired surrogates are treated like surrogate code points. Similarly,
* set strings match only on code point boundaries, never in the middle of a
* surrogate pair.
*
* @stable ICU 4.4
*/
public enum SpanCondition {
/**
* Continues a span() while there is no set element at the current position.
* Increments by one code point at a time. Stops before the first set element
* (character or string). (For code points only, this is like while
* contains(current)==false).
*
* When span() returns, the substring between where it started and the position
* it returned consists only of characters that are not in the set, and none of
* its strings overlap with the span.
*
* @stable ICU 4.4
*/
NOT_CONTAINED,
/**
* Spans the longest substring that is a concatenation of set elements
* (characters or strings). (For characters only, this is like while
* contains(current)==true).
*
* When span() returns, the substring between where it started and the position
* it returned consists only of set elements (characters or strings) that are in
* the set.
*
* If a set contains strings, then the span will be the longest substring for
* which there exists at least one non-overlapping concatenation of set elements
* (characters or strings). This is equivalent to a POSIX regular expression for
*
* When span() returns, the substring between where it started and the position
* it returned consists only of set elements (characters or strings) that are in
* the set.
*
* If a set only contains single characters, then this is the same as CONTAINED.
*
* If a set contains strings, then the span will be the longest substring with a
* match at each position with the longest single set element (character or
* string).
*
* Use this span condition together with other longest-match algorithms, such as
* ICU converters (ucnv_getUnicodeSet()).
*
* @stable ICU 4.4
*/
SIMPLE,
}
}
end >
* start then an empty set is created.
*
* @param start first character, inclusive, of range
* @param end last character, inclusive, of range
* @stable ICU 2.0
*/
public UnicodeSet(int start, int end) {
this();
complement(start, end);
}
/**
* Constructs a set from the given pattern. See the class description for the
* syntax of the pattern language. Whitespace is ignored.
*
* @param pattern a string specifying what characters are in the set
* @exception java.lang.IllegalArgumentException if the pattern contains a
* syntax error.
* @stable ICU 2.0
*/
public UnicodeSet(String pattern) {
this();
applyPattern(pattern, null);
}
/**
* Make this object represent the same set as other.
*
* @param other a UnicodeSet whose value will be copied to this
* object
* @stable ICU 2.0
*/
public UnicodeSet set(UnicodeSet other) {
checkFrozen();
list = other.list.clone();
len = other.len;
strings = new TreeSet
* Warning: you cannot add an empty string ("") to a UnicodeSet.
*
* @param s the source string
* @return this object, for chaining
* @stable ICU 2.0
*/
public final UnicodeSet add(CharSequence s) {
checkFrozen();
int cp = getSingleCP(s);
if (cp < 0) {
strings.add(s.toString());
} else {
add_unchecked(cp, cp);
}
return this;
}
/**
* Utility for getting code point from single code point CharSequence. See the
* public UTF16.getSingleCodePoint()
*
* @return a code point IF the string consists of a single one. otherwise
* returns -1.
* @param s to test
*/
private static int getSingleCP(CharSequence s) {
if (s.length() < 1) {
throw new IllegalArgumentException("Can't use zero-length strings in UnicodeSet");
}
if (s.length() > 2)
return -1;
if (s.length() == 1)
return s.charAt(0);
// at this point, len = 2
int cp = UTF16.charAt(s, 0);
if (cp > 0xFFFF) { // is surrogate pair
return cp;
}
return -1;
}
/**
* Complements the specified range in this set. Any character in the range will
* be removed if it is in this set, or will be added if it is not in this set.
* If {@code end > start} then an empty range is complemented, leaving the set
* unchanged.
*
* @param start first character, inclusive, of range to be removed from this
* set.
* @param end last character, inclusive, of range to be removed from this set.
* @stable ICU 2.0
*/
public UnicodeSet complement(int start, int end) {
checkFrozen();
if (start < MIN_VALUE || start > MAX_VALUE) {
throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(start, 6));
}
if (end < MIN_VALUE || end > MAX_VALUE) {
throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(end, 6));
}
if (start <= end) {
xor(range(start, end), 2, 0);
}
return this;
}
/**
* Returns true if this set contains the given character.
*
* @param c character to be checked for containment
* @return true if the test condition is met
* @stable ICU 2.0
*/
public boolean contains(int c) {
if (c < MIN_VALUE || c > MAX_VALUE) {
throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(c, 6));
}
if (bmpSet != null) {
return bmpSet.contains(c);
}
if (stringSpan != null) {
return stringSpan.contains(c);
}
/*
* // Set i to the index of the start item greater than ch // We know we will
* terminate without length test! int i = -1; while (true) { if (c < list[++i])
* break; }
*/
int i = findCodePoint(c);
return ((i & 1) != 0); // return true if odd
}
/**
* Returns the smallest value i such that c < list[i]. Caller must ensure that c
* is a legal value or this method will enter an infinite loop. This method
* performs a binary search.
*
* @param c a character in the range MIN_VALUE..MAX_VALUE inclusive
* @return the smallest integer i in the range 0..len-1, inclusive, such that c
* < list[i]
*/
private final int findCodePoint(int c) {
/*
* Examples: findCodePoint(c) set list[] c=0 1 3 4 7 8 === ==============
* =========== [] [110000] 0 0 0 0 0 0 [\u0000-\u0003] [0, 4, 110000] 1 1 1 2 2
* 2 [\u0004-\u0007] [4, 8, 110000] 0 0 0 1 1 2 [:all:] [0, 110000] 1 1 1 1 1 1
*/
// Return the smallest i such that c < list[i]. Assume
// list[len - 1] == HIGH and that c is legal (0..HIGH-1).
if (c < list[0])
return 0;
// High runner test. c is often after the last range, so an
// initial check for this condition pays off.
if (len >= 2 && c >= list[len - 2])
return len - 1;
int lo = 0;
int hi = len - 1;
// invariant: c >= list[lo]
// invariant: c < list[hi]
for (;;) {
int i = (lo + hi) >>> 1;
if (i == lo)
return hi;
if (c < list[i]) {
hi = i;
} else {
lo = i;
}
}
}
/**
* Retains only the elements in this set that are contained in the specified
* set. In other words, removes from this set all of its elements that are not
* contained in the specified set. This operation effectively modifies this set
* so that its value is the intersection of the two sets.
*
* @param c set that defines which elements this set will retain.
* @stable ICU 2.0
*/
public UnicodeSet retainAll(UnicodeSet c) {
checkFrozen();
retain(c.list, c.len, 0);
strings.retainAll(c.strings);
return this;
}
/**
* Removes all of the elements from this set. This set will be empty after this
* call returns.
*
* @stable ICU 2.0
*/
public UnicodeSet clear() {
checkFrozen();
list[0] = HIGH;
len = 1;
strings.clear();
return this;
}
/**
* Iteration method that returns the number of ranges contained in this set.
*
* @see #getRangeStart
* @see #getRangeEnd
* @stable ICU 2.0
*/
public int getRangeCount() {
return len / 2;
}
/**
* Iteration method that returns the first character in the specified range of
* this set.
*
* @exception ArrayIndexOutOfBoundsException if index is outside the range
* 0..getRangeCount()-1
* @see #getRangeCount
* @see #getRangeEnd
* @stable ICU 2.0
*/
public int getRangeStart(int index) {
return list[index * 2];
}
/**
* Iteration method that returns the last character in the specified range of
* this set.
*
* @exception ArrayIndexOutOfBoundsException if index is outside the range
* 0..getRangeCount()-1
* @see #getRangeStart
* @see #getRangeEnd
* @stable ICU 2.0
*/
public int getRangeEnd(int index) {
return (list[index * 2 + 1] - 1);
}
// ----------------------------------------------------------------
// Implementation: Pattern parsing
// ----------------------------------------------------------------
/**
* Parses the given pattern, starting at the given position. The character at
* pattern.charAt(pos.getIndex()) must be '[', or the parse fails. Parsing
* continues until the corresponding closing ']'. If a syntax error is
* encountered between the opening and closing brace, the parse fails. Upon
* return from a successful parse, the ParsePosition is updated to point to the
* character following the closing ']', and an inversion list for the parsed
* pattern is returned. This method calls itself recursively to parse embedded
* subpatterns.
*
* @param pattern the string containing the pattern to be parsed. The portion of
* the string from pos.getIndex(), which must be a '[', to the
* corresponding closing ']', is parsed.
* @param pos upon entry, the position at which to being parsing. The
* character at pattern.charAt(pos.getIndex()) must be a '['.
* Upon return from a successful parse, pos.getIndex() is either
* the character after the closing ']' of the parsed pattern, or
* pattern.length() if the closing ']' is the last character of
* the pattern string.
* @return an inversion list for the parsed substring of pattern
* @exception java.lang.IllegalArgumentException if the parse fails.
*/
private UnicodeSet applyPattern(String pattern, ParsePosition pos) {
if ("[:age=3.2:]".equals(pattern)) {
checkFrozen();
VersionInfo version = VersionInfo.getInstance("3.2");
applyFilter(new VersionFilter(version), UCharacterProperty.SRC_PROPSVEC);
} else {
throw new IllegalStateException("UnicodeSet.applyPattern(unexpected pattern " + pattern + ")");
}
return this;
}
// ----------------------------------------------------------------
// Implementation: Utility methods
// ----------------------------------------------------------------
private void ensureCapacity(int newLen) {
if (newLen <= list.length)
return;
int[] temp = new int[newLen + GROW_EXTRA];
System.arraycopy(list, 0, temp, 0, len);
list = temp;
}
private void ensureBufferCapacity(int newLen) {
if (buffer != null && newLen <= buffer.length)
return;
buffer = new int[newLen + GROW_EXTRA];
}
/**
* Assumes start <= end.
*/
private int[] range(int start, int end) {
if (rangeList == null) {
rangeList = new int[] { start, end + 1, HIGH };
} else {
rangeList[0] = start;
rangeList[1] = end + 1;
}
return rangeList;
}
// ----------------------------------------------------------------
// Implementation: Fundamental operations
// ----------------------------------------------------------------
// polarity = 0, 3 is normal: x xor y
// polarity = 1, 2: x xor ~y == x === y
private UnicodeSet xor(int[] other, int otherLen, int polarity) {
ensureBufferCapacity(len + otherLen);
int i = 0, j = 0, k = 0;
int a = list[i++];
int b;
if (polarity == 1 || polarity == 2) {
b = LOW;
if (other[j] == LOW) { // skip base if already LOW
++j;
b = other[j];
}
} else {
b = other[j++];
}
// simplest of all the routines
// sort the values, discarding identicals!
while (true) {
if (a < b) {
buffer[k++] = a;
a = list[i++];
} else if (b < a) {
buffer[k++] = b;
b = other[j++];
} else if (a != HIGH) { // at this point, a == b
// discard both values!
a = list[i++];
b = other[j++];
} else { // DONE!
buffer[k++] = HIGH;
len = k;
break;
}
}
// swap list and buffer
int[] temp = list;
list = buffer;
buffer = temp;
return this;
}
// polarity = 0 is normal: x union y
// polarity = 2: x union ~y
// polarity = 1: ~x union y
// polarity = 3: ~x union ~y
private UnicodeSet add(int[] other, int otherLen, int polarity) {
ensureBufferCapacity(len + otherLen);
int i = 0, j = 0, k = 0;
int a = list[i++];
int b = other[j++];
// change from xor is that we have to check overlapping pairs
// polarity bit 1 means a is second, bit 2 means b is.
main: while (true) {
switch (polarity) {
case 0: // both first; take lower if unequal
if (a < b) { // take a
// Back up over overlapping ranges in buffer[]
if (k > 0 && a <= buffer[k - 1]) {
// Pick latter end value in buffer[] vs. list[]
a = max(list[i], buffer[--k]);
} else {
// No overlap
buffer[k++] = a;
a = list[i];
}
i++; // Common if/else code factored out
polarity ^= 1;
} else if (b < a) { // take b
if (k > 0 && b <= buffer[k - 1]) {
b = max(other[j], buffer[--k]);
} else {
buffer[k++] = b;
b = other[j];
}
j++;
polarity ^= 2;
} else { // a == b, take a, drop b
if (a == HIGH)
break main;
// This is symmetrical; it doesn't matter if
// we backtrack with a or b. - liu
if (k > 0 && a <= buffer[k - 1]) {
a = max(list[i], buffer[--k]);
} else {
// No overlap
buffer[k++] = a;
a = list[i];
}
i++;
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
case 3: // both second; take higher if unequal, and drop other
if (b <= a) { // take a
if (a == HIGH)
break main;
buffer[k++] = a;
} else { // take b
if (b == HIGH)
break main;
buffer[k++] = b;
}
a = list[i++];
polarity ^= 1; // factored common code
b = other[j++];
polarity ^= 2;
break;
case 1: // a second, b first; if b < a, overlap
if (a < b) { // no overlap, take a
buffer[k++] = a;
a = list[i++];
polarity ^= 1;
} else if (b < a) { // OVERLAP, drop b
b = other[j++];
polarity ^= 2;
} else { // a == b, drop both!
if (a == HIGH)
break main;
a = list[i++];
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
case 2: // a first, b second; if a < b, overlap
if (b < a) { // no overlap, take b
buffer[k++] = b;
b = other[j++];
polarity ^= 2;
} else if (a < b) { // OVERLAP, drop a
a = list[i++];
polarity ^= 1;
} else { // a == b, drop both!
if (a == HIGH)
break main;
a = list[i++];
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
}
}
buffer[k++] = HIGH; // terminate
len = k;
// swap list and buffer
int[] temp = list;
list = buffer;
buffer = temp;
return this;
}
// polarity = 0 is normal: x intersect y
// polarity = 2: x intersect ~y == set-minus
// polarity = 1: ~x intersect y
// polarity = 3: ~x intersect ~y
private UnicodeSet retain(int[] other, int otherLen, int polarity) {
ensureBufferCapacity(len + otherLen);
int i = 0, j = 0, k = 0;
int a = list[i++];
int b = other[j++];
// change from xor is that we have to check overlapping pairs
// polarity bit 1 means a is second, bit 2 means b is.
main: while (true) {
switch (polarity) {
case 0: // both first; drop the smaller
if (a < b) { // drop a
a = list[i++];
polarity ^= 1;
} else if (b < a) { // drop b
b = other[j++];
polarity ^= 2;
} else { // a == b, take one, drop other
if (a == HIGH)
break main;
buffer[k++] = a;
a = list[i++];
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
case 3: // both second; take lower if unequal
if (a < b) { // take a
buffer[k++] = a;
a = list[i++];
polarity ^= 1;
} else if (b < a) { // take b
buffer[k++] = b;
b = other[j++];
polarity ^= 2;
} else { // a == b, take one, drop other
if (a == HIGH)
break main;
buffer[k++] = a;
a = list[i++];
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
case 1: // a second, b first;
if (a < b) { // NO OVERLAP, drop a
a = list[i++];
polarity ^= 1;
} else if (b < a) { // OVERLAP, take b
buffer[k++] = b;
b = other[j++];
polarity ^= 2;
} else { // a == b, drop both!
if (a == HIGH)
break main;
a = list[i++];
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
case 2: // a first, b second; if a < b, overlap
if (b < a) { // no overlap, drop b
b = other[j++];
polarity ^= 2;
} else if (a < b) { // OVERLAP, take a
buffer[k++] = a;
a = list[i++];
polarity ^= 1;
} else { // a == b, drop both!
if (a == HIGH)
break main;
a = list[i++];
polarity ^= 1;
b = other[j++];
polarity ^= 2;
}
break;
}
}
buffer[k++] = HIGH; // terminate
len = k;
// swap list and buffer
int[] temp = list;
list = buffer;
buffer = temp;
return this;
}
private static final int max(int a, int b) {
return (a > b) ? a : b;
}
// ----------------------------------------------------------------
// Generic filter-based scanning code
// ----------------------------------------------------------------
private static interface Filter {
boolean contains(int codePoint);
}
private static final VersionInfo NO_VERSION = VersionInfo.getInstance(0, 0, 0, 0);
private static class VersionFilter implements Filter {
VersionInfo version;
VersionFilter(VersionInfo version) {
this.version = version;
}
public boolean contains(int ch) {
VersionInfo v = UCharacter.getAge(ch);
// Reference comparison ok; VersionInfo caches and reuses
// unique objects.
return v != NO_VERSION && v.compareTo(version) <= 0;
}
}
private static synchronized UnicodeSet getInclusions(int src) {
if (src != UCharacterProperty.SRC_PROPSVEC) {
throw new IllegalStateException("UnicodeSet.getInclusions(unknown src " + src + ")");
}
if (INCLUSION == null) {
UnicodeSet incl = new UnicodeSet();
UCharacterProperty.INSTANCE.upropsvec_addPropertyStarts(incl);
INCLUSION = incl;
}
return INCLUSION;
}
/**
* Generic filter-based scanning code for UCD property UnicodeSets.
*/
private UnicodeSet applyFilter(Filter filter, int src) {
// Logically, walk through all Unicode characters, noting the start
// and end of each range for which filter.contain(c) is
// true. Add each range to a set.
//
// To improve performance, use an inclusions set which
// encodes information about character ranges that are known
// to have identical properties.
// getInclusions(src) contains exactly the first characters of
// same-value ranges for the given properties "source".
clear();
int startHasProperty = -1;
UnicodeSet inclusions = getInclusions(src);
int limitRange = inclusions.getRangeCount();
for (int j = 0; j < limitRange; ++j) {
// get current range
int start = inclusions.getRangeStart(j);
int end = inclusions.getRangeEnd(j);
// for all the code points in the range, process
for (int ch = start; ch <= end; ++ch) {
// only add to the unicodeset on inflection points --
// where the hasProperty value changes to false
if (filter.contains(ch)) {
if (startHasProperty < 0) {
startHasProperty = ch;
}
} else if (startHasProperty >= 0) {
add_unchecked(startHasProperty, ch - 1);
startHasProperty = -1;
}
}
}
if (startHasProperty >= 0) {
add_unchecked(startHasProperty, 0x10FFFF);
}
return this;
}
/**
* Is this frozen, according to the Freezable interface?
*
* @return value
* @stable ICU 3.8
*/
public boolean isFrozen() {
return (bmpSet != null || stringSpan != null);
}
/**
* Freeze this class, according to the Freezable interface.
*
* @return this
* @stable ICU 4.4
*/
public UnicodeSet freeze() {
if (!isFrozen()) {
// Do most of what compact() does before freezing because
// compact() will not work when the set is frozen.
// Small modification: Don't shrink if the savings would be tiny (<=GROW_EXTRA).
// Delete buffer first to defragment memory less.
buffer = null;
if (list.length > (len + GROW_EXTRA)) {
// Make the capacity equal to len or 1.
// We don't want to realloc of 0 size.
int capacity = (len == 0) ? 1 : len;
int[] oldList = list;
list = new int[capacity];
for (int i = capacity; i-- > 0;) {
list[i] = oldList[i];
}
}
// Optimize contains() and span() and similar functions.
if (!strings.isEmpty()) {
stringSpan = new UnicodeSetStringSpan(this, new ArrayList
*
* When a set contains multi-code point strings, then these statements may not
* be true, depending on the strings in the set (for example, whether they
* overlap with each other) and the string that is processed. For a set with
* strings:
*
*
* Note: If it is important to get the same boundaries whether iterating forward
* or backward through a string, then either only span() should be used and the
* boundaries cached for backward operation, or an ICU BreakIterator could be
* used.
* (OR of each set element)*. (Java/ICU/Perl regex stops at the
* first match of an OR.)
*
* @stable ICU 4.4
*/
CONTAINED,
/**
* Continues a span() while there is a set element at the current position.
* Increments by the longest matching element at each position. (For characters
* only, this is like while contains(current)==true).
*