michael@0: /*
michael@0: *******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 2011, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: *******************************************************************************
michael@0: *   file name:  uniset_closure.cpp
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2011may30
michael@0: *   created by: Markus W. Scherer
michael@0: *
michael@0: *   UnicodeSet::closeOver() and related methods moved here from uniset_props.cpp
michael@0: *   to simplify dependencies.
michael@0: *   In particular, this depends on the BreakIterator, but the BreakIterator
michael@0: *   code also builds UnicodeSets from patterns and needs uniset_props.
michael@0: */
michael@0: 
michael@0: #include "unicode/brkiter.h"
michael@0: #include "unicode/locid.h"
michael@0: #include "unicode/parsepos.h"
michael@0: #include "unicode/uniset.h"
michael@0: #include "cmemory.h"
michael@0: #include "ruleiter.h"
michael@0: #include "ucase.h"
michael@0: #include "util.h"
michael@0: #include "uvector.h"
michael@0: 
michael@0: // initial storage. Must be >= 0
michael@0: // *** same as in uniset.cpp ! ***
michael@0: #define START_EXTRA 16
michael@0: 
michael@0: U_NAMESPACE_BEGIN
michael@0: 
michael@0: // TODO memory debugging provided inside uniset.cpp
michael@0: // could be made available here but probably obsolete with use of modern
michael@0: // memory leak checker tools
michael@0: #define _dbgct(me)
michael@0: 
michael@0: //----------------------------------------------------------------
michael@0: // Constructors &c
michael@0: //----------------------------------------------------------------
michael@0: 
michael@0: UnicodeSet::UnicodeSet(const UnicodeString& pattern,
michael@0:                        uint32_t options,
michael@0:                        const SymbolTable* symbols,
michael@0:                        UErrorCode& status) :
michael@0:     len(0), capacity(START_EXTRA), list(0), bmpSet(0), buffer(0),
michael@0:     bufferCapacity(0), patLen(0), pat(NULL), strings(NULL), stringSpan(NULL),
michael@0:     fFlags(0)
michael@0: {
michael@0:     if(U_SUCCESS(status)){
michael@0:         list = (UChar32*) uprv_malloc(sizeof(UChar32) * capacity);
michael@0:         /* test for NULL */
michael@0:         if(list == NULL) {
michael@0:             status = U_MEMORY_ALLOCATION_ERROR;  
michael@0:         }else{
michael@0:             allocateStrings(status);
michael@0:             applyPattern(pattern, options, symbols, status);
michael@0:         }
michael@0:     }
michael@0:     _dbgct(this);
michael@0: }
michael@0: 
michael@0: UnicodeSet::UnicodeSet(const UnicodeString& pattern, ParsePosition& pos,
michael@0:                        uint32_t options,
michael@0:                        const SymbolTable* symbols,
michael@0:                        UErrorCode& status) :
michael@0:     len(0), capacity(START_EXTRA), list(0), bmpSet(0), buffer(0),
michael@0:     bufferCapacity(0), patLen(0), pat(NULL), strings(NULL), stringSpan(NULL),
michael@0:     fFlags(0)
michael@0: {
michael@0:     if(U_SUCCESS(status)){
michael@0:         list = (UChar32*) uprv_malloc(sizeof(UChar32) * capacity);
michael@0:         /* test for NULL */
michael@0:         if(list == NULL) {
michael@0:             status = U_MEMORY_ALLOCATION_ERROR;   
michael@0:         }else{
michael@0:             allocateStrings(status);
michael@0:             applyPattern(pattern, pos, options, symbols, status);
michael@0:         }
michael@0:     }
michael@0:     _dbgct(this);
michael@0: }
michael@0: 
michael@0: //----------------------------------------------------------------
michael@0: // Public API
michael@0: //----------------------------------------------------------------
michael@0: 
michael@0: UnicodeSet& UnicodeSet::applyPattern(const UnicodeString& pattern,
michael@0:                                      uint32_t options,
michael@0:                                      const SymbolTable* symbols,
michael@0:                                      UErrorCode& status) {
michael@0:     ParsePosition pos(0);
michael@0:     applyPattern(pattern, pos, options, symbols, status);
michael@0:     if (U_FAILURE(status)) return *this;
michael@0: 
michael@0:     int32_t i = pos.getIndex();
michael@0: 
michael@0:     if (options & USET_IGNORE_SPACE) {
michael@0:         // Skip over trailing whitespace
michael@0:         ICU_Utility::skipWhitespace(pattern, i, TRUE);
michael@0:     }
michael@0: 
michael@0:     if (i != pattern.length()) {
michael@0:         status = U_ILLEGAL_ARGUMENT_ERROR;
michael@0:     }
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: UnicodeSet& UnicodeSet::applyPattern(const UnicodeString& pattern,
michael@0:                               ParsePosition& pos,
michael@0:                               uint32_t options,
michael@0:                               const SymbolTable* symbols,
michael@0:                               UErrorCode& status) {
michael@0:     if (U_FAILURE(status)) {
michael@0:         return *this;
michael@0:     }
michael@0:     if (isFrozen()) {
michael@0:         status = U_NO_WRITE_PERMISSION;
michael@0:         return *this;
michael@0:     }
michael@0:     // Need to build the pattern in a temporary string because
michael@0:     // _applyPattern calls add() etc., which set pat to empty.
michael@0:     UnicodeString rebuiltPat;
michael@0:     RuleCharacterIterator chars(pattern, symbols, pos);
michael@0:     applyPattern(chars, symbols, rebuiltPat, options, &UnicodeSet::closeOver, status);
michael@0:     if (U_FAILURE(status)) return *this;
michael@0:     if (chars.inVariable()) {
michael@0:         // syntaxError(chars, "Extra chars in variable value");
michael@0:         status = U_MALFORMED_SET;
michael@0:         return *this;
michael@0:     }
michael@0:     setPattern(rebuiltPat);
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: // USetAdder implementation
michael@0: // Does not use uset.h to reduce code dependencies
michael@0: static void U_CALLCONV
michael@0: _set_add(USet *set, UChar32 c) {
michael@0:     ((UnicodeSet *)set)->add(c);
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: _set_addRange(USet *set, UChar32 start, UChar32 end) {
michael@0:     ((UnicodeSet *)set)->add(start, end);
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: _set_addString(USet *set, const UChar *str, int32_t length) {
michael@0:     ((UnicodeSet *)set)->add(UnicodeString((UBool)(length<0), str, length));
michael@0: }
michael@0: 
michael@0: //----------------------------------------------------------------
michael@0: // Case folding API
michael@0: //----------------------------------------------------------------
michael@0: 
michael@0: // add the result of a full case mapping to the set
michael@0: // use str as a temporary string to avoid constructing one
michael@0: static inline void
michael@0: addCaseMapping(UnicodeSet &set, int32_t result, const UChar *full, UnicodeString &str) {
michael@0:     if(result >= 0) {
michael@0:         if(result > UCASE_MAX_STRING_LENGTH) {
michael@0:             // add a single-code point case mapping
michael@0:             set.add(result);
michael@0:         } else {
michael@0:             // add a string case mapping from full with length result
michael@0:             str.setTo((UBool)FALSE, full, result);
michael@0:             set.add(str);
michael@0:         }
michael@0:     }
michael@0:     // result < 0: the code point mapped to itself, no need to add it
michael@0:     // see ucase.h
michael@0: }
michael@0: 
michael@0: UnicodeSet& UnicodeSet::closeOver(int32_t attribute) {
michael@0:     if (isFrozen() || isBogus()) {
michael@0:         return *this;
michael@0:     }
michael@0:     if (attribute & (USET_CASE_INSENSITIVE | USET_ADD_CASE_MAPPINGS)) {
michael@0:         const UCaseProps *csp = ucase_getSingleton();
michael@0:         {
michael@0:             UnicodeSet foldSet(*this);
michael@0:             UnicodeString str;
michael@0:             USetAdder sa = {
michael@0:                 foldSet.toUSet(),
michael@0:                 _set_add,
michael@0:                 _set_addRange,
michael@0:                 _set_addString,
michael@0:                 NULL, // don't need remove()
michael@0:                 NULL // don't need removeRange()
michael@0:             };
michael@0: 
michael@0:             // start with input set to guarantee inclusion
michael@0:             // USET_CASE: remove strings because the strings will actually be reduced (folded);
michael@0:             //            therefore, start with no strings and add only those needed
michael@0:             if (attribute & USET_CASE_INSENSITIVE) {
michael@0:                 foldSet.strings->removeAllElements();
michael@0:             }
michael@0: 
michael@0:             int32_t n = getRangeCount();
michael@0:             UChar32 result;
michael@0:             const UChar *full;
michael@0:             int32_t locCache = 0;
michael@0: 
michael@0:             for (int32_t i=0; i<n; ++i) {
michael@0:                 UChar32 start = getRangeStart(i);
michael@0:                 UChar32 end   = getRangeEnd(i);
michael@0: 
michael@0:                 if (attribute & USET_CASE_INSENSITIVE) {
michael@0:                     // full case closure
michael@0:                     for (UChar32 cp=start; cp<=end; ++cp) {
michael@0:                         ucase_addCaseClosure(csp, cp, &sa);
michael@0:                     }
michael@0:                 } else {
michael@0:                     // add case mappings
michael@0:                     // (does not add long s for regular s, or Kelvin for k, for example)
michael@0:                     for (UChar32 cp=start; cp<=end; ++cp) {
michael@0:                         result = ucase_toFullLower(csp, cp, NULL, NULL, &full, "", &locCache);
michael@0:                         addCaseMapping(foldSet, result, full, str);
michael@0: 
michael@0:                         result = ucase_toFullTitle(csp, cp, NULL, NULL, &full, "", &locCache);
michael@0:                         addCaseMapping(foldSet, result, full, str);
michael@0: 
michael@0:                         result = ucase_toFullUpper(csp, cp, NULL, NULL, &full, "", &locCache);
michael@0:                         addCaseMapping(foldSet, result, full, str);
michael@0: 
michael@0:                         result = ucase_toFullFolding(csp, cp, &full, 0);
michael@0:                         addCaseMapping(foldSet, result, full, str);
michael@0:                     }
michael@0:                 }
michael@0:             }
michael@0:             if (strings != NULL && strings->size() > 0) {
michael@0:                 if (attribute & USET_CASE_INSENSITIVE) {
michael@0:                     for (int32_t j=0; j<strings->size(); ++j) {
michael@0:                         str = *(const UnicodeString *) strings->elementAt(j);
michael@0:                         str.foldCase();
michael@0:                         if(!ucase_addStringCaseClosure(csp, str.getBuffer(), str.length(), &sa)) {
michael@0:                             foldSet.add(str); // does not map to code points: add the folded string itself
michael@0:                         }
michael@0:                     }
michael@0:                 } else {
michael@0:                     Locale root("");
michael@0: #if !UCONFIG_NO_BREAK_ITERATION
michael@0:                     UErrorCode status = U_ZERO_ERROR;
michael@0:                     BreakIterator *bi = BreakIterator::createWordInstance(root, status);
michael@0:                     if (U_SUCCESS(status)) {
michael@0: #endif
michael@0:                         const UnicodeString *pStr;
michael@0: 
michael@0:                         for (int32_t j=0; j<strings->size(); ++j) {
michael@0:                             pStr = (const UnicodeString *) strings->elementAt(j);
michael@0:                             (str = *pStr).toLower(root);
michael@0:                             foldSet.add(str);
michael@0: #if !UCONFIG_NO_BREAK_ITERATION
michael@0:                             (str = *pStr).toTitle(bi, root);
michael@0:                             foldSet.add(str);
michael@0: #endif
michael@0:                             (str = *pStr).toUpper(root);
michael@0:                             foldSet.add(str);
michael@0:                             (str = *pStr).foldCase();
michael@0:                             foldSet.add(str);
michael@0:                         }
michael@0: #if !UCONFIG_NO_BREAK_ITERATION
michael@0:                     }
michael@0:                     delete bi;
michael@0: #endif
michael@0:                 }
michael@0:             }
michael@0:             *this = foldSet;
michael@0:         }
michael@0:     }
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: U_NAMESPACE_END