icuSources/i18n/rbt.cpp

   1 /*
   2 **********************************************************************
   3 *   Copyright (C) 1999-2008, International Business Machines
   4 *   Corporation and others.  All Rights Reserved.
   5 **********************************************************************
   6 *   Date        Name        Description
   7 *   11/17/99    aliu        Creation.
   8 **********************************************************************
   9 */
  10
  11 #include "unicode/utypes.h"
  12
  13 #if !UCONFIG_NO_TRANSLITERATION
  14
  15 #include "unicode/rep.h"
  16 #include "unicode/uniset.h"
  17 #include "rbt_pars.h"
  18 #include "rbt_data.h"
  19 #include "rbt_rule.h"
  20 #include "rbt.h"
  21 #include "umutex.h"
  22
  23 U_NAMESPACE_BEGIN
  24
  25 UOBJECT_DEFINE_RTTI_IMPLEMENTATION(RuleBasedTransliterator)
  26
  27 static UMTX  transliteratorDataMutex = NULL;
  28 static Replaceable *gLockedText = NULL;
  29
  30 void RuleBasedTransliterator::_construct(const UnicodeString& rules,
  31                                          UTransDirection direction,
  32                                          UParseError& parseError,
  33                                          UErrorCode& status) {
  34     fData = 0;
  35     isDataOwned = TRUE;
  36     if (U_FAILURE(status)) {
  37         return;
  38     }
  39
  40     TransliteratorParser parser(status);
  41     parser.parse(rules, direction, parseError, status);
  42     if (U_FAILURE(status)) {
  43         return;
  44     }
  45
  46     if (parser.idBlockVector.size() != 0 ||
  47         parser.compoundFilter != NULL ||
  48         parser.dataVector.size() == 0) {
  49         status = U_INVALID_RBT_SYNTAX; // ::ID blocks disallowed in RBT
  50         return;
  51     }
  52
  53     fData = (TransliterationRuleData*)parser.dataVector.orphanElementAt(0);
  54     setMaximumContextLength(fData->ruleSet.getMaximumContextLength());
  55 }
  56
  57 /**
  58  * Constructs a new transliterator from the given rules.
  59  * @param id            the id for the transliterator.
  60  * @param rules         rules, separated by ';'
  61  * @param direction     either FORWARD or REVERSE.
  62  * @param adoptedFilter the filter for this transliterator.
  63  * @param parseError    Struct to recieve information on position
  64  *                      of error if an error is encountered
  65  * @param status        Output param set to success/failure code.
  66  * @exception IllegalArgumentException if rules are malformed
  67  * or direction is invalid.
  68  */
  69 RuleBasedTransliterator::RuleBasedTransliterator(
  70                             const UnicodeString& id,
  71                             const UnicodeString& rules,
  72                             UTransDirection direction,
  73                             UnicodeFilter* adoptedFilter,
  74                             UParseError& parseError,
  75                             UErrorCode& status) :
  76     Transliterator(id, adoptedFilter) {
  77     _construct(rules, direction,parseError,status);
  78 }
  79
  80 /**
  81  * Constructs a new transliterator from the given rules.
  82  * @param id            the id for the transliterator.
  83  * @param rules         rules, separated by ';'
  84  * @param direction     either FORWARD or REVERSE.
  85  * @param adoptedFilter the filter for this transliterator.
  86  * @param status        Output param set to success/failure code.
  87  * @exception IllegalArgumentException if rules are malformed
  88  * or direction is invalid.
  89  */
  90 /*RuleBasedTransliterator::RuleBasedTransliterator(
  91                             const UnicodeString& id,
  92                             const UnicodeString& rules,
  93                             UTransDirection direction,
  94                             UnicodeFilter* adoptedFilter,
  95                             UErrorCode& status) :
  96     Transliterator(id, adoptedFilter) {
  97     UParseError parseError;
  98     _construct(rules, direction,parseError, status);
  99 }*/
 100
 101 /**
 102  * Covenience constructor with no filter.
 103  */
 104 /*RuleBasedTransliterator::RuleBasedTransliterator(
 105                             const UnicodeString& id,
 106                             const UnicodeString& rules,
 107                             UTransDirection direction,
 108                             UErrorCode& status) :
 109     Transliterator(id, 0) {
 110     UParseError parseError;
 111     _construct(rules, direction,parseError, status);
 112 }*/
 113
 114 /**
 115  * Covenience constructor with no filter and FORWARD direction.
 116  */
 117 /*RuleBasedTransliterator::RuleBasedTransliterator(
 118                             const UnicodeString& id,
 119                             const UnicodeString& rules,
 120                             UErrorCode& status) :
 121     Transliterator(id, 0) {
 122     UParseError parseError;
 123     _construct(rules, UTRANS_FORWARD, parseError, status);
 124 }*/
 125
 126 /**
 127  * Covenience constructor with FORWARD direction.
 128  */
 129 /*RuleBasedTransliterator::RuleBasedTransliterator(
 130                             const UnicodeString& id,
 131                             const UnicodeString& rules,
 132                             UnicodeFilter* adoptedFilter,
 133                             UErrorCode& status) :
 134     Transliterator(id, adoptedFilter) {
 135     UParseError parseError;
 136     _construct(rules, UTRANS_FORWARD,parseError, status);
 137 }*/
 138
 139 RuleBasedTransliterator::RuleBasedTransliterator(const UnicodeString& id,
 140                                  const TransliterationRuleData* theData,
 141                                  UnicodeFilter* adoptedFilter) :
 142     Transliterator(id, adoptedFilter),
 143     fData((TransliterationRuleData*)theData), // cast away const
 144     isDataOwned(FALSE) {
 145     setMaximumContextLength(fData->ruleSet.getMaximumContextLength());
 146 }
 147
 148 /**
 149  * Internal constructor.
 150  */
 151 RuleBasedTransliterator::RuleBasedTransliterator(const UnicodeString& id,
 152                                                  TransliterationRuleData* theData,
 153                                                  UBool isDataAdopted) :
 154     Transliterator(id, 0),
 155     fData(theData),
 156     isDataOwned(isDataAdopted) {
 157     setMaximumContextLength(fData->ruleSet.getMaximumContextLength());
 158 }
 159
 160 /**
 161  * Copy constructor.
 162  */
 163 RuleBasedTransliterator::RuleBasedTransliterator(
 164         const RuleBasedTransliterator& other) :
 165     Transliterator(other), fData(other.fData),
 166     isDataOwned(other.isDataOwned) {
 167
 168     // The data object may or may not be owned.  If it is not owned we
 169     // share it; it is invariant.  If it is owned, it's still
 170     // invariant, but we need to copy it to prevent double-deletion.
 171     // If this becomes a performance issue (if people do a lot of RBT
 172     // copying -- unlikely) we can reference count the data object.
 173
 174     // Only do a deep copy if this is owned data, that is, data that
 175     // will be later deleted.  System transliterators contain
 176     // non-owned data.
 177     if (isDataOwned) {
 178         fData = new TransliterationRuleData(*other.fData);
 179     }
 180 }
 181
 182 /**
 183  * Destructor.
 184  */
 185 RuleBasedTransliterator::~RuleBasedTransliterator() {
 186     // Delete the data object only if we own it.
 187     if (isDataOwned) {
 188         delete fData;
 189     }
 190 }
 191
 192 Transliterator* // Covariant return NOT ALLOWED (for portability)
 193 RuleBasedTransliterator::clone(void) const {
 194     return new RuleBasedTransliterator(*this);
 195 }
 196
 197 /**
 198  * Implements {@link Transliterator#handleTransliterate}.
 199  */
 200 void
 201 RuleBasedTransliterator::handleTransliterate(Replaceable& text, UTransPosition& index,
 202                                              UBool isIncremental) const {
 203     /* We keep contextStart and contextLimit fixed the entire time,
 204      * relative to the text -- contextLimit may move numerically if
 205      * text is inserted or removed.  The start offset moves toward
 206      * limit, with replacements happening under it.
 207      *
 208      * Example: rules 1. ab>x|y
 209      *                2. yc>z
 210      *
 211      * |eabcd   begin - no match, advance start
 212      * e|abcd   match rule 1 - change text & adjust start
 213      * ex|ycd   match rule 2 - change text & adjust start
 214      * exz|d    no match, advance start
 215      * exzd|    done
 216      */
 217
 218     /* A rule like
 219      *   a>b|a
 220      * creates an infinite loop. To prevent that, we put an arbitrary
 221      * limit on the number of iterations that we take, one that is
 222      * high enough that any reasonable rules are ok, but low enough to
 223      * prevent a server from hanging.  The limit is 16 times the
 224      * number of characters n, unless n is so large that 16n exceeds a
 225      * uint32_t.
 226      */
 227     uint32_t loopCount = 0;
 228     uint32_t loopLimit = index.limit - index.start;
 229     if (loopLimit >= 0x10000000) {
 230         loopLimit = 0xFFFFFFFF;
 231     } else {
 232         loopLimit <<= 4;
 233     }
 234
 235     // Transliterator locking.  Rule-based Transliterators are not thread safe; concurrent
 236     //   operations must be prevented.
 237     // A Complication: compound transliterators can result in recursive entries to this
 238     //   function, sometimes with different "This" objects, always with the same text.
 239     //   Double-locking must be prevented in these cases.
 240     //
 241
 242     // If the transliteration data is exclusively owned by this transliterator object,
 243     //   we don't need to do any locking.  No sharing between transliterators is possible,
 244     //   so no concurrent access from multiple threads is possible.
 245     UBool    lockedMutexAtThisLevel = FALSE;
 246     if (isDataOwned == FALSE) {
 247         // Test whether this request is operating on the same text string as some
 248         //   some other transliteration that is still in progress and holding the
 249         //   transliteration mutex.  If so, do not lock the transliteration
 250         //    mutex again.
 251         UBool needToLock;
 252         UMTX_CHECK(NULL, (&text != gLockedText), needToLock);
 253         if (needToLock) {
 254             umtx_lock(&transliteratorDataMutex);
 255             gLockedText = &text;
 256             lockedMutexAtThisLevel = TRUE;
 257         }
 258     }
 259
 260     // Check to make sure we don't dereference a null pointer.
 261     if (fData != NULL) {
 262             while (index.start < index.limit &&
 263                    loopCount <= loopLimit &&
 264                    fData->ruleSet.transliterate(text, index, isIncremental)) {
 265                 ++loopCount;
 266             }
 267     }
 268     if (lockedMutexAtThisLevel) {
 269         gLockedText = NULL;
 270         umtx_unlock(&transliteratorDataMutex);
 271     }
 272 }
 273
 274 UnicodeString& RuleBasedTransliterator::toRules(UnicodeString& rulesSource,
 275                                                 UBool escapeUnprintable) const {
 276     return fData->ruleSet.toRules(rulesSource, escapeUnprintable);
 277 }
 278
 279 /**
 280  * Implement Transliterator framework
 281  */
 282 void RuleBasedTransliterator::handleGetSourceSet(UnicodeSet& result) const {
 283     fData->ruleSet.getSourceTargetSet(result, FALSE);
 284 }
 285
 286 /**
 287  * Override Transliterator framework
 288  */
 289 UnicodeSet& RuleBasedTransliterator::getTargetSet(UnicodeSet& result) const {
 290     return fData->ruleSet.getSourceTargetSet(result, TRUE);
 291 }
 292
 293 U_NAMESPACE_END
 294
 295 #endif /* #if !UCONFIG_NO_TRANSLITERATION */