Home | History | Annotate | Download | only in i18n
      1 /*
      2  **********************************************************************
      3  *   Copyright (C) 1999-2012, 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 "utypeinfo.h"  // for 'typeid' to work
     12 
     13 #include "unicode/utypes.h"
     14 
     15 #if !UCONFIG_NO_TRANSLITERATION
     16 
     17 #include "unicode/putil.h"
     18 #include "unicode/translit.h"
     19 #include "unicode/locid.h"
     20 #include "unicode/msgfmt.h"
     21 #include "unicode/rep.h"
     22 #include "unicode/resbund.h"
     23 #include "unicode/unifilt.h"
     24 #include "unicode/uniset.h"
     25 #include "unicode/uscript.h"
     26 #include "unicode/strenum.h"
     27 #include "unicode/utf16.h"
     28 #include "cpdtrans.h"
     29 #include "nultrans.h"
     30 #include "rbt_data.h"
     31 #include "rbt_pars.h"
     32 #include "rbt.h"
     33 #include "transreg.h"
     34 #include "name2uni.h"
     35 #include "nortrans.h"
     36 #include "remtrans.h"
     37 #include "titletrn.h"
     38 #include "tolowtrn.h"
     39 #include "toupptrn.h"
     40 #include "uni2name.h"
     41 #include "brktrans.h"
     42 #include "esctrn.h"
     43 #include "unesctrn.h"
     44 #include "tridpars.h"
     45 #include "anytrans.h"
     46 #include "util.h"
     47 #include "hash.h"
     48 #include "mutex.h"
     49 #include "ucln_in.h"
     50 #include "uassert.h"
     51 #include "cmemory.h"
     52 #include "cstring.h"
     53 #include "uinvchar.h"
     54 
     55 static const UChar TARGET_SEP  = 0x002D; /*-*/
     56 static const UChar ID_DELIM    = 0x003B; /*;*/
     57 static const UChar VARIANT_SEP = 0x002F; // '/'
     58 
     59 /**
     60  * Prefix for resource bundle key for the display name for a
     61  * transliterator.  The ID is appended to this to form the key.
     62  * The resource bundle value should be a String.
     63  */
     64 static const char RB_DISPLAY_NAME_PREFIX[] = "%Translit%%";
     65 
     66 /**
     67  * Prefix for resource bundle key for the display name for a
     68  * transliterator SCRIPT.  The ID is appended to this to form the key.
     69  * The resource bundle value should be a String.
     70  */
     71 static const char RB_SCRIPT_DISPLAY_NAME_PREFIX[] = "%Translit%";
     72 
     73 /**
     74  * Resource bundle key for display name pattern.
     75  * The resource bundle value should be a String forming a
     76  * MessageFormat pattern, e.g.:
     77  * "{0,choice,0#|1#{1} Transliterator|2#{1} to {2} Transliterator}".
     78  */
     79 static const char RB_DISPLAY_NAME_PATTERN[] = "TransliteratorNamePattern";
     80 
     81 /**
     82  * Resource bundle key for the list of RuleBasedTransliterator IDs.
     83  * The resource bundle value should be a String[] with each element
     84  * being a valid ID.  The ID will be appended to RB_RULE_BASED_PREFIX
     85  * to obtain the class name in which the RB_RULE key will be sought.
     86  */
     87 static const char RB_RULE_BASED_IDS[] = "RuleBasedTransliteratorIDs";
     88 
     89 /**
     90  * The mutex controlling access to registry object.
     91  */
     92 static UMutex registryMutex = U_MUTEX_INITIALIZER;
     93 
     94 /**
     95  * System transliterator registry; non-null when initialized.
     96  */
     97 static icu::TransliteratorRegistry* registry = 0;
     98 
     99 // Macro to check/initialize the registry. ONLY USE WITHIN
    100 // MUTEX. Avoids function call when registry is initialized.
    101 #define HAVE_REGISTRY(status) (registry!=0 || initializeRegistry(status))
    102 
    103 U_NAMESPACE_BEGIN
    104 
    105 UOBJECT_DEFINE_ABSTRACT_RTTI_IMPLEMENTATION(Transliterator)
    106 
    107 /**
    108  * Return TRUE if the given UTransPosition is valid for text of
    109  * the given length.
    110  */
    111 static inline UBool positionIsValid(UTransPosition& index, int32_t len) {
    112     return !(index.contextStart < 0 ||
    113              index.start < index.contextStart ||
    114              index.limit < index.start ||
    115              index.contextLimit < index.limit ||
    116              len < index.contextLimit);
    117 }
    118 
    119 /**
    120  * Default constructor.
    121  * @param theID the string identifier for this transliterator
    122  * @param theFilter the filter.  Any character for which
    123  * <tt>filter.contains()</tt> returns <tt>FALSE</tt> will not be
    124  * altered by this transliterator.  If <tt>filter</tt> is
    125  * <tt>null</tt> then no filtering is applied.
    126  */
    127 Transliterator::Transliterator(const UnicodeString& theID,
    128                                UnicodeFilter* adoptedFilter) :
    129     UObject(), ID(theID), filter(adoptedFilter),
    130     maximumContextLength(0)
    131 {
    132     // NUL-terminate the ID string, which is a non-aliased copy.
    133     ID.append((UChar)0);
    134     ID.truncate(ID.length()-1);
    135 }
    136 
    137 /**
    138  * Destructor.
    139  */
    140 Transliterator::~Transliterator() {
    141     if (filter) {
    142         delete filter;
    143     }
    144 }
    145 
    146 /**
    147  * Copy constructor.
    148  */
    149 Transliterator::Transliterator(const Transliterator& other) :
    150     UObject(other), ID(other.ID), filter(0),
    151     maximumContextLength(other.maximumContextLength)
    152 {
    153     // NUL-terminate the ID string, which is a non-aliased copy.
    154     ID.append((UChar)0);
    155     ID.truncate(ID.length()-1);
    156 
    157     if (other.filter != 0) {
    158         // We own the filter, so we must have our own copy
    159         filter = (UnicodeFilter*) other.filter->clone();
    160     }
    161 }
    162 
    163 Transliterator* Transliterator::clone() const {
    164     return NULL;
    165 }
    166 
    167 /**
    168  * Assignment operator.
    169  */
    170 Transliterator& Transliterator::operator=(const Transliterator& other) {
    171     ID = other.ID;
    172     // NUL-terminate the ID string
    173     ID.getTerminatedBuffer();
    174 
    175     maximumContextLength = other.maximumContextLength;
    176     adoptFilter((other.filter == 0) ? 0 : (UnicodeFilter*) other.filter->clone());
    177     return *this;
    178 }
    179 
    180 /**
    181  * Transliterates a segment of a string.  <code>Transliterator</code> API.
    182  * @param text the string to be transliterated
    183  * @param start the beginning index, inclusive; <code>0 <= start
    184  * <= limit</code>.
    185  * @param limit the ending index, exclusive; <code>start <= limit
    186  * <= text.length()</code>.
    187  * @return the new limit index, or -1
    188  */
    189 int32_t Transliterator::transliterate(Replaceable& text,
    190                                       int32_t start, int32_t limit) const {
    191     if (start < 0 ||
    192         limit < start ||
    193         text.length() < limit) {
    194         return -1;
    195     }
    196 
    197     UTransPosition offsets;
    198     offsets.contextStart= start;
    199     offsets.contextLimit = limit;
    200     offsets.start = start;
    201     offsets.limit = limit;
    202     filteredTransliterate(text, offsets, FALSE, TRUE);
    203     return offsets.limit;
    204 }
    205 
    206 /**
    207  * Transliterates an entire string in place. Convenience method.
    208  * @param text the string to be transliterated
    209  */
    210 void Transliterator::transliterate(Replaceable& text) const {
    211     transliterate(text, 0, text.length());
    212 }
    213 
    214 /**
    215  * Transliterates the portion of the text buffer that can be
    216  * transliterated unambiguosly after new text has been inserted,
    217  * typically as a result of a keyboard event.  The new text in
    218  * <code>insertion</code> will be inserted into <code>text</code>
    219  * at <code>index.contextLimit</code>, advancing
    220  * <code>index.contextLimit</code> by <code>insertion.length()</code>.
    221  * Then the transliterator will try to transliterate characters of
    222  * <code>text</code> between <code>index.start</code> and
    223  * <code>index.contextLimit</code>.  Characters before
    224  * <code>index.start</code> will not be changed.
    225  *
    226  * <p>Upon return, values in <code>index</code> will be updated.
    227  * <code>index.contextStart</code> will be advanced to the first
    228  * character that future calls to this method will read.
    229  * <code>index.start</code> and <code>index.contextLimit</code> will
    230  * be adjusted to delimit the range of text that future calls to
    231  * this method may change.
    232  *
    233  * <p>Typical usage of this method begins with an initial call
    234  * with <code>index.contextStart</code> and <code>index.contextLimit</code>
    235  * set to indicate the portion of <code>text</code> to be
    236  * transliterated, and <code>index.start == index.contextStart</code>.
    237  * Thereafter, <code>index</code> can be used without
    238  * modification in future calls, provided that all changes to
    239  * <code>text</code> are made via this method.
    240  *
    241  * <p>This method assumes that future calls may be made that will
    242  * insert new text into the buffer.  As a result, it only performs
    243  * unambiguous transliterations.  After the last call to this
    244  * method, there may be untransliterated text that is waiting for
    245  * more input to resolve an ambiguity.  In order to perform these
    246  * pending transliterations, clients should call {@link
    247  * #finishKeyboardTransliteration} after the last call to this
    248  * method has been made.
    249  *
    250  * @param text the buffer holding transliterated and untransliterated text
    251  * @param index an array of three integers.
    252  *
    253  * <ul><li><code>index.contextStart</code>: the beginning index,
    254  * inclusive; <code>0 <= index.contextStart <= index.contextLimit</code>.
    255  *
    256  * <li><code>index.contextLimit</code>: the ending index, exclusive;
    257  * <code>index.contextStart <= index.contextLimit <= text.length()</code>.
    258  * <code>insertion</code> is inserted at
    259  * <code>index.contextLimit</code>.
    260  *
    261  * <li><code>index.start</code>: the next character to be
    262  * considered for transliteration; <code>index.contextStart <=
    263  * index.start <= index.contextLimit</code>.  Characters before
    264  * <code>index.start</code> will not be changed by future calls
    265  * to this method.</ul>
    266  *
    267  * @param insertion text to be inserted and possibly
    268  * transliterated into the translation buffer at
    269  * <code>index.contextLimit</code>.  If <code>null</code> then no text
    270  * is inserted.
    271  * @see #START
    272  * @see #LIMIT
    273  * @see #CURSOR
    274  * @see #handleTransliterate
    275  * @exception IllegalArgumentException if <code>index</code>
    276  * is invalid
    277  */
    278 void Transliterator::transliterate(Replaceable& text,
    279                                    UTransPosition& index,
    280                                    const UnicodeString& insertion,
    281                                    UErrorCode &status) const {
    282     _transliterate(text, index, &insertion, status);
    283 }
    284 
    285 /**
    286  * Transliterates the portion of the text buffer that can be
    287  * transliterated unambiguosly after a new character has been
    288  * inserted, typically as a result of a keyboard event.  This is a
    289  * convenience method; see {@link
    290  * #transliterate(Replaceable, int[], String)} for details.
    291  * @param text the buffer holding transliterated and
    292  * untransliterated text
    293  * @param index an array of three integers.  See {@link
    294  * #transliterate(Replaceable, int[], String)}.
    295  * @param insertion text to be inserted and possibly
    296  * transliterated into the translation buffer at
    297  * <code>index.contextLimit</code>.
    298  * @see #transliterate(Replaceable, int[], String)
    299  */
    300 void Transliterator::transliterate(Replaceable& text,
    301                                    UTransPosition& index,
    302                                    UChar32 insertion,
    303                                    UErrorCode& status) const {
    304     UnicodeString str(insertion);
    305     _transliterate(text, index, &str, status);
    306 }
    307 
    308 /**
    309  * Transliterates the portion of the text buffer that can be
    310  * transliterated unambiguosly.  This is a convenience method; see
    311  * {@link #transliterate(Replaceable, int[], String)} for
    312  * details.
    313  * @param text the buffer holding transliterated and
    314  * untransliterated text
    315  * @param index an array of three integers.  See {@link
    316  * #transliterate(Replaceable, int[], String)}.
    317  * @see #transliterate(Replaceable, int[], String)
    318  */
    319 void Transliterator::transliterate(Replaceable& text,
    320                                    UTransPosition& index,
    321                                    UErrorCode& status) const {
    322     _transliterate(text, index, 0, status);
    323 }
    324 
    325 /**
    326  * Finishes any pending transliterations that were waiting for
    327  * more characters.  Clients should call this method as the last
    328  * call after a sequence of one or more calls to
    329  * <code>transliterate()</code>.
    330  * @param text the buffer holding transliterated and
    331  * untransliterated text.
    332  * @param index the array of indices previously passed to {@link
    333  * #transliterate}
    334  */
    335 void Transliterator::finishTransliteration(Replaceable& text,
    336                                            UTransPosition& index) const {
    337     if (!positionIsValid(index, text.length())) {
    338         return;
    339     }
    340 
    341     filteredTransliterate(text, index, FALSE, TRUE);
    342 }
    343 
    344 /**
    345  * This internal method does keyboard transliteration.  If the
    346  * 'insertion' is non-null then we append it to 'text' before
    347  * proceeding.  This method calls through to the pure virtual
    348  * framework method handleTransliterate() to do the actual
    349  * work.
    350  */
    351 void Transliterator::_transliterate(Replaceable& text,
    352                                     UTransPosition& index,
    353                                     const UnicodeString* insertion,
    354                                     UErrorCode &status) const {
    355     if (U_FAILURE(status)) {
    356         return;
    357     }
    358 
    359     if (!positionIsValid(index, text.length())) {
    360         status = U_ILLEGAL_ARGUMENT_ERROR;
    361         return;
    362     }
    363 
    364 //    int32_t originalStart = index.contextStart;
    365     if (insertion != 0) {
    366         text.handleReplaceBetween(index.limit, index.limit, *insertion);
    367         index.limit += insertion->length();
    368         index.contextLimit += insertion->length();
    369     }
    370 
    371     if (index.limit > 0 &&
    372         U16_IS_LEAD(text.charAt(index.limit - 1))) {
    373         // Oops, there is a dangling lead surrogate in the buffer.
    374         // This will break most transliterators, since they will
    375         // assume it is part of a pair.  Don't transliterate until
    376         // more text comes in.
    377         return;
    378     }
    379 
    380     filteredTransliterate(text, index, TRUE, TRUE);
    381 
    382 #if 0
    383     // TODO
    384     // I CAN'T DO what I'm attempting below now that the Kleene star
    385     // operator is supported.  For example, in the rule
    386 
    387     //   ([:Lu:]+) { x } > $1;
    388 
    389     // what is the maximum context length?  getMaximumContextLength()
    390     // will return 1, but this is just the length of the ante context
    391     // part of the pattern string -- 1 character, which is a standin
    392     // for a Quantifier, which contains a StringMatcher, which
    393     // contains a UnicodeSet.
    394 
    395     // There is a complicated way to make this work again, and that's
    396     // to add a "maximum left context" protocol into the
    397     // UnicodeMatcher hierarchy.  At present I'm not convinced this is
    398     // worth it.
    399 
    400     // ---
    401 
    402     // The purpose of the code below is to keep the context small
    403     // while doing incremental transliteration.  When part of the left
    404     // context (between contextStart and start) is no longer needed,
    405     // we try to advance contextStart past that portion.  We use the
    406     // maximum context length to do so.
    407     int32_t newCS = index.start;
    408     int32_t n = getMaximumContextLength();
    409     while (newCS > originalStart && n-- > 0) {
    410         --newCS;
    411         newCS -= U16_LENGTH(text.char32At(newCS)) - 1;
    412     }
    413     index.contextStart = uprv_max(newCS, originalStart);
    414 #endif
    415 }
    416 
    417 /**
    418  * This method breaks up the input text into runs of unfiltered
    419  * characters.  It passes each such run to
    420  * <subclass>.handleTransliterate().  Subclasses that can handle the
    421  * filter logic more efficiently themselves may override this method.
    422  *
    423  * All transliteration calls in this class go through this method.
    424  */
    425 void Transliterator::filteredTransliterate(Replaceable& text,
    426                                            UTransPosition& index,
    427                                            UBool incremental,
    428                                            UBool rollback) const {
    429     // Short circuit path for transliterators with no filter in
    430     // non-incremental mode.
    431     if (filter == 0 && !rollback) {
    432         handleTransliterate(text, index, incremental);
    433         return;
    434     }
    435 
    436     //----------------------------------------------------------------------
    437     // This method processes text in two groupings:
    438     //
    439     // RUNS -- A run is a contiguous group of characters which are contained
    440     // in the filter for this transliterator (filter.contains(ch) == TRUE).
    441     // Text outside of runs may appear as context but it is not modified.
    442     // The start and limit Position values are narrowed to each run.
    443     //
    444     // PASSES (incremental only) -- To make incremental mode work correctly,
    445     // each run is broken up into n passes, where n is the length (in code
    446     // points) of the run.  Each pass contains the first n characters.  If a
    447     // pass is completely transliterated, it is committed, and further passes
    448     // include characters after the committed text.  If a pass is blocked,
    449     // and does not transliterate completely, then this method rolls back
    450     // the changes made during the pass, extends the pass by one code point,
    451     // and tries again.
    452     //----------------------------------------------------------------------
    453 
    454     // globalLimit is the limit value for the entire operation.  We
    455     // set index.limit to the end of each unfiltered run before
    456     // calling handleTransliterate(), so we need to maintain the real
    457     // value of index.limit here.  After each transliteration, we
    458     // update globalLimit for insertions or deletions that have
    459     // happened.
    460     int32_t globalLimit = index.limit;
    461 
    462     // If there is a non-null filter, then break the input text up.  Say the
    463     // input text has the form:
    464     //   xxxabcxxdefxx
    465     // where 'x' represents a filtered character (filter.contains('x') ==
    466     // false).  Then we break this up into:
    467     //   xxxabc xxdef xx
    468     // Each pass through the loop consumes a run of filtered
    469     // characters (which are ignored) and a subsequent run of
    470     // unfiltered characters (which are transliterated).
    471 
    472     for (;;) {
    473 
    474         if (filter != NULL) {
    475             // Narrow the range to be transliterated to the first segment
    476             // of unfiltered characters at or after index.start.
    477 
    478             // Advance past filtered chars
    479             UChar32 c;
    480             while (index.start < globalLimit &&
    481                    !filter->contains(c=text.char32At(index.start))) {
    482                 index.start += U16_LENGTH(c);
    483             }
    484 
    485             // Find the end of this run of unfiltered chars
    486             index.limit = index.start;
    487             while (index.limit < globalLimit &&
    488                    filter->contains(c=text.char32At(index.limit))) {
    489                 index.limit += U16_LENGTH(c);
    490             }
    491         }
    492 
    493         // Check to see if the unfiltered run is empty.  This only
    494         // happens at the end of the string when all the remaining
    495         // characters are filtered.
    496         if (index.limit == index.start) {
    497             // assert(index.start == globalLimit);
    498             break;
    499         }
    500 
    501         // Is this run incremental?  If there is additional
    502         // filtered text (if limit < globalLimit) then we pass in
    503         // an incremental value of FALSE to force the subclass to
    504         // complete the transliteration for this run.
    505         UBool isIncrementalRun =
    506             (index.limit < globalLimit ? FALSE : incremental);
    507 
    508         int32_t delta;
    509 
    510         // Implement rollback.  To understand the need for rollback,
    511         // consider the following transliterator:
    512         //
    513         //  "t" is "a > A;"
    514         //  "u" is "A > b;"
    515         //  "v" is a compound of "t; NFD; u" with a filter [:Ll:]
    516         //
    517         // Now apply "c" to the input text "a".  The result is "b".  But if
    518         // the transliteration is done incrementally, then the NFD holds
    519         // things up after "t" has already transformed "a" to "A".  When
    520         // finishTransliterate() is called, "A" is _not_ processed because
    521         // it gets excluded by the [:Ll:] filter, and the end result is "A"
    522         // -- incorrect.  The problem is that the filter is applied to a
    523         // partially-transliterated result, when we only want it to apply to
    524         // input text.  Although this example hinges on a compound
    525         // transliterator containing NFD and a specific filter, it can
    526         // actually happen with any transliterator which may do a partial
    527         // transformation in incremental mode into characters outside its
    528         // filter.
    529         //
    530         // To handle this, when in incremental mode we supply characters to
    531         // handleTransliterate() in several passes.  Each pass adds one more
    532         // input character to the input text.  That is, for input "ABCD", we
    533         // first try "A", then "AB", then "ABC", and finally "ABCD".  If at
    534         // any point we block (upon return, start < limit) then we roll
    535         // back.  If at any point we complete the run (upon return start ==
    536         // limit) then we commit that run.
    537 
    538         if (rollback && isIncrementalRun) {
    539 
    540             int32_t runStart = index.start;
    541             int32_t runLimit = index.limit;
    542             int32_t runLength =  runLimit - runStart;
    543 
    544             // Make a rollback copy at the end of the string
    545             int32_t rollbackOrigin = text.length();
    546             text.copy(runStart, runLimit, rollbackOrigin);
    547 
    548             // Variables reflecting the commitment of completely
    549             // transliterated text.  passStart is the runStart, advanced
    550             // past committed text.  rollbackStart is the rollbackOrigin,
    551             // advanced past rollback text that corresponds to committed
    552             // text.
    553             int32_t passStart = runStart;
    554             int32_t rollbackStart = rollbackOrigin;
    555 
    556             // The limit for each pass; we advance by one code point with
    557             // each iteration.
    558             int32_t passLimit = index.start;
    559 
    560             // Total length, in 16-bit code units, of uncommitted text.
    561             // This is the length to be rolled back.
    562             int32_t uncommittedLength = 0;
    563 
    564             // Total delta (change in length) for all passes
    565             int32_t totalDelta = 0;
    566 
    567             // PASS MAIN LOOP -- Start with a single character, and extend
    568             // the text by one character at a time.  Roll back partial
    569             // transliterations and commit complete transliterations.
    570             for (;;) {
    571                 // Length of additional code point, either one or two
    572                 int32_t charLength = U16_LENGTH(text.char32At(passLimit));
    573                 passLimit += charLength;
    574                 if (passLimit > runLimit) {
    575                     break;
    576                 }
    577                 uncommittedLength += charLength;
    578 
    579                 index.limit = passLimit;
    580 
    581                 // Delegate to subclass for actual transliteration.  Upon
    582                 // return, start will be updated to point after the
    583                 // transliterated text, and limit and contextLimit will be
    584                 // adjusted for length changes.
    585                 handleTransliterate(text, index, TRUE);
    586 
    587                 delta = index.limit - passLimit; // change in length
    588 
    589                 // We failed to completely transliterate this pass.
    590                 // Roll back the text.  Indices remain unchanged; reset
    591                 // them where necessary.
    592                 if (index.start != index.limit) {
    593                     // Find the rollbackStart, adjusted for length changes
    594                     // and the deletion of partially transliterated text.
    595                     int32_t rs = rollbackStart + delta - (index.limit - passStart);
    596 
    597                     // Delete the partially transliterated text
    598                     text.handleReplaceBetween(passStart, index.limit, UnicodeString());
    599 
    600                     // Copy the rollback text back
    601                     text.copy(rs, rs + uncommittedLength, passStart);
    602 
    603                     // Restore indices to their original values
    604                     index.start = passStart;
    605                     index.limit = passLimit;
    606                     index.contextLimit -= delta;
    607                 }
    608 
    609                 // We did completely transliterate this pass.  Update the
    610                 // commit indices to record how far we got.  Adjust indices
    611                 // for length change.
    612                 else {
    613                     // Move the pass indices past the committed text.
    614                     passStart = passLimit = index.start;
    615 
    616                     // Adjust the rollbackStart for length changes and move
    617                     // it past the committed text.  All characters we've
    618                     // processed to this point are committed now, so zero
    619                     // out the uncommittedLength.
    620                     rollbackStart += delta + uncommittedLength;
    621                     uncommittedLength = 0;
    622 
    623                     // Adjust indices for length changes.
    624                     runLimit += delta;
    625                     totalDelta += delta;
    626                 }
    627             }
    628 
    629             // Adjust overall limit and rollbackOrigin for insertions and
    630             // deletions.  Don't need to worry about contextLimit because
    631             // handleTransliterate() maintains that.
    632             rollbackOrigin += totalDelta;
    633             globalLimit += totalDelta;
    634 
    635             // Delete the rollback copy
    636             text.handleReplaceBetween(rollbackOrigin, rollbackOrigin + runLength, UnicodeString());
    637 
    638             // Move start past committed text
    639             index.start = passStart;
    640         }
    641 
    642         else {
    643             // Delegate to subclass for actual transliteration.
    644             int32_t limit = index.limit;
    645             handleTransliterate(text, index, isIncrementalRun);
    646             delta = index.limit - limit; // change in length
    647 
    648             // In a properly written transliterator, start == limit after
    649             // handleTransliterate() returns when incremental is false.
    650             // Catch cases where the subclass doesn't do this, and throw
    651             // an exception.  (Just pinning start to limit is a bad idea,
    652             // because what's probably happening is that the subclass
    653             // isn't transliterating all the way to the end, and it should
    654             // in non-incremental mode.)
    655             if (!incremental && index.start != index.limit) {
    656                 // We can't throw an exception, so just fudge things
    657                 index.start = index.limit;
    658             }
    659 
    660             // Adjust overall limit for insertions/deletions.  Don't need
    661             // to worry about contextLimit because handleTransliterate()
    662             // maintains that.
    663             globalLimit += delta;
    664         }
    665 
    666         if (filter == NULL || isIncrementalRun) {
    667             break;
    668         }
    669 
    670         // If we did completely transliterate this
    671         // run, then repeat with the next unfiltered run.
    672     }
    673 
    674     // Start is valid where it is.  Limit needs to be put back where
    675     // it was, modulo adjustments for deletions/insertions.
    676     index.limit = globalLimit;
    677 }
    678 
    679 void Transliterator::filteredTransliterate(Replaceable& text,
    680                                            UTransPosition& index,
    681                                            UBool incremental) const {
    682     filteredTransliterate(text, index, incremental, FALSE);
    683 }
    684 
    685 /**
    686  * Method for subclasses to use to set the maximum context length.
    687  * @see #getMaximumContextLength
    688  */
    689 void Transliterator::setMaximumContextLength(int32_t maxContextLength) {
    690     maximumContextLength = maxContextLength;
    691 }
    692 
    693 /**
    694  * Returns a programmatic identifier for this transliterator.
    695  * If this identifier is passed to <code>getInstance()</code>, it
    696  * will return this object, if it has been registered.
    697  * @see #registerInstance
    698  * @see #getAvailableIDs
    699  */
    700 const UnicodeString& Transliterator::getID(void) const {
    701     return ID;
    702 }
    703 
    704 /**
    705  * Returns a name for this transliterator that is appropriate for
    706  * display to the user in the default locale.  See {@link
    707  * #getDisplayName(Locale)} for details.
    708  */
    709 UnicodeString& U_EXPORT2 Transliterator::getDisplayName(const UnicodeString& ID,
    710                                               UnicodeString& result) {
    711     return getDisplayName(ID, Locale::getDefault(), result);
    712 }
    713 
    714 /**
    715  * Returns a name for this transliterator that is appropriate for
    716  * display to the user in the given locale.  This name is taken
    717  * from the locale resource data in the standard manner of the
    718  * <code>java.text</code> package.
    719  *
    720  * <p>If no localized names exist in the system resource bundles,
    721  * a name is synthesized using a localized
    722  * <code>MessageFormat</code> pattern from the resource data.  The
    723  * arguments to this pattern are an integer followed by one or two
    724  * strings.  The integer is the number of strings, either 1 or 2.
    725  * The strings are formed by splitting the ID for this
    726  * transliterator at the first TARGET_SEP.  If there is no TARGET_SEP, then the
    727  * entire ID forms the only string.
    728  * @param inLocale the Locale in which the display name should be
    729  * localized.
    730  * @see java.text.MessageFormat
    731  */
    732 UnicodeString& U_EXPORT2 Transliterator::getDisplayName(const UnicodeString& id,
    733                                               const Locale& inLocale,
    734                                               UnicodeString& result) {
    735     UErrorCode status = U_ZERO_ERROR;
    736 
    737     ResourceBundle bundle(U_ICUDATA_TRANSLIT, inLocale, status);
    738 
    739     // Suspend checking status until later...
    740 
    741     result.truncate(0);
    742 
    743     // Normalize the ID
    744     UnicodeString source, target, variant;
    745     UBool sawSource;
    746     TransliteratorIDParser::IDtoSTV(id, source, target, variant, sawSource);
    747     if (target.length() < 1) {
    748         // No target; malformed id
    749         return result;
    750     }
    751     if (variant.length() > 0) { // Change "Foo" to "/Foo"
    752         variant.insert(0, VARIANT_SEP);
    753     }
    754     UnicodeString ID(source);
    755     ID.append(TARGET_SEP).append(target).append(variant);
    756 
    757     // build the char* key
    758     if (uprv_isInvariantUString(ID.getBuffer(), ID.length())) {
    759         char key[200];
    760         uprv_strcpy(key, RB_DISPLAY_NAME_PREFIX);
    761         int32_t length=(int32_t)uprv_strlen(RB_DISPLAY_NAME_PREFIX);
    762         ID.extract(0, (int32_t)(sizeof(key)-length), key+length, (int32_t)(sizeof(key)-length), US_INV);
    763 
    764         // Try to retrieve a UnicodeString from the bundle.
    765         UnicodeString resString = bundle.getStringEx(key, status);
    766 
    767         if (U_SUCCESS(status) && resString.length() != 0) {
    768             return result = resString; // [sic] assign & return
    769         }
    770 
    771 #if !UCONFIG_NO_FORMATTING
    772         // We have failed to get a name from the locale data.  This is
    773         // typical, since most transliterators will not have localized
    774         // name data.  The next step is to retrieve the MessageFormat
    775         // pattern from the locale data and to use it to synthesize the
    776         // name from the ID.
    777 
    778         status = U_ZERO_ERROR;
    779         resString = bundle.getStringEx(RB_DISPLAY_NAME_PATTERN, status);
    780 
    781         if (U_SUCCESS(status) && resString.length() != 0) {
    782             MessageFormat msg(resString, inLocale, status);
    783             // Suspend checking status until later...
    784 
    785             // We pass either 2 or 3 Formattable objects to msg.
    786             Formattable args[3];
    787             int32_t nargs;
    788             args[0].setLong(2); // # of args to follow
    789             args[1].setString(source);
    790             args[2].setString(target);
    791             nargs = 3;
    792 
    793             // Use display names for the scripts, if they exist
    794             UnicodeString s;
    795             length=(int32_t)uprv_strlen(RB_SCRIPT_DISPLAY_NAME_PREFIX);
    796             for (int j=1; j<=2; ++j) {
    797                 status = U_ZERO_ERROR;
    798                 uprv_strcpy(key, RB_SCRIPT_DISPLAY_NAME_PREFIX);
    799                 args[j].getString(s);
    800                 if (uprv_isInvariantUString(s.getBuffer(), s.length())) {
    801                     s.extract(0, sizeof(key)-length-1, key+length, (int32_t)sizeof(key)-length-1, US_INV);
    802 
    803                     resString = bundle.getStringEx(key, status);
    804 
    805                     if (U_SUCCESS(status)) {
    806                         args[j] = resString;
    807                     }
    808                 }
    809             }
    810 
    811             status = U_ZERO_ERROR;
    812             FieldPosition pos; // ignored by msg
    813             msg.format(args, nargs, result, pos, status);
    814             if (U_SUCCESS(status)) {
    815                 result.append(variant);
    816                 return result;
    817             }
    818         }
    819 #endif
    820     }
    821 
    822     // We should not reach this point unless there is something
    823     // wrong with the build or the RB_DISPLAY_NAME_PATTERN has
    824     // been deleted from the root RB_LOCALE_ELEMENTS resource.
    825     result = ID;
    826     return result;
    827 }
    828 
    829 /**
    830  * Returns the filter used by this transliterator, or <tt>null</tt>
    831  * if this transliterator uses no filter.  Caller musn't delete
    832  * the result!
    833  */
    834 const UnicodeFilter* Transliterator::getFilter(void) const {
    835     return filter;
    836 }
    837 
    838 /**
    839  * Returns the filter used by this transliterator, or
    840  * <tt>NULL</tt> if this transliterator uses no filter.  The
    841  * caller must eventually delete the result.  After this call,
    842  * this transliterator's filter is set to <tt>NULL</tt>.
    843  */
    844 UnicodeFilter* Transliterator::orphanFilter(void) {
    845     UnicodeFilter *result = filter;
    846     filter = NULL;
    847     return result;
    848 }
    849 
    850 /**
    851  * Changes the filter used by this transliterator.  If the filter
    852  * is set to <tt>null</tt> then no filtering will occur.
    853  *
    854  * <p>Callers must take care if a transliterator is in use by
    855  * multiple threads.  The filter should not be changed by one
    856  * thread while another thread may be transliterating.
    857  */
    858 void Transliterator::adoptFilter(UnicodeFilter* filterToAdopt) {
    859     delete filter;
    860     filter = filterToAdopt;
    861 }
    862 
    863 /**
    864  * Returns this transliterator's inverse.  See the class
    865  * documentation for details.  This implementation simply inverts
    866  * the two entities in the ID and attempts to retrieve the
    867  * resulting transliterator.  That is, if <code>getID()</code>
    868  * returns "A-B", then this method will return the result of
    869  * <code>getInstance("B-A")</code>, or <code>null</code> if that
    870  * call fails.
    871  *
    872  * <p>This method does not take filtering into account.  The
    873  * returned transliterator will have no filter.
    874  *
    875  * <p>Subclasses with knowledge of their inverse may wish to
    876  * override this method.
    877  *
    878  * @return a transliterator that is an inverse, not necessarily
    879  * exact, of this transliterator, or <code>null</code> if no such
    880  * transliterator is registered.
    881  * @see #registerInstance
    882  */
    883 Transliterator* Transliterator::createInverse(UErrorCode& status) const {
    884     UParseError parseError;
    885     return Transliterator::createInstance(ID, UTRANS_REVERSE,parseError,status);
    886 }
    887 
    888 Transliterator* U_EXPORT2
    889 Transliterator::createInstance(const UnicodeString& ID,
    890                                 UTransDirection dir,
    891                                 UErrorCode& status)
    892 {
    893     UParseError parseError;
    894     return createInstance(ID, dir, parseError, status);
    895 }
    896 
    897 /**
    898  * Returns a <code>Transliterator</code> object given its ID.
    899  * The ID must be either a system transliterator ID or a ID registered
    900  * using <code>registerInstance()</code>.
    901  *
    902  * @param ID a valid ID, as enumerated by <code>getAvailableIDs()</code>
    903  * @return A <code>Transliterator</code> object with the given ID
    904  * @see #registerInstance
    905  * @see #getAvailableIDs
    906  * @see #getID
    907  */
    908 Transliterator* U_EXPORT2
    909 Transliterator::createInstance(const UnicodeString& ID,
    910                                 UTransDirection dir,
    911                                 UParseError& parseError,
    912                                 UErrorCode& status)
    913 {
    914     if (U_FAILURE(status)) {
    915         return 0;
    916     }
    917 
    918     UnicodeString canonID;
    919     UVector list(status);
    920     if (U_FAILURE(status)) {
    921         return NULL;
    922     }
    923 
    924     UnicodeSet* globalFilter;
    925     // TODO add code for parseError...currently unused, but
    926     // later may be used by parsing code...
    927     if (!TransliteratorIDParser::parseCompoundID(ID, dir, canonID, list, globalFilter)) {
    928         status = U_INVALID_ID;
    929         return NULL;
    930     }
    931 
    932     TransliteratorIDParser::instantiateList(list, status);
    933     if (U_FAILURE(status)) {
    934         return NULL;
    935     }
    936 
    937     U_ASSERT(list.size() > 0);
    938     Transliterator* t = NULL;
    939 
    940     if (list.size() > 1 || canonID.indexOf(ID_DELIM) >= 0) {
    941         // [NOTE: If it's a compoundID, we instantiate a CompoundTransliterator even if it only
    942         // has one child transliterator.  This is so that toRules() will return the right thing
    943         // (without any inactive ID), but our main ID still comes out correct.  That is, if we
    944         // instantiate "(Lower);Latin-Greek;", we want the rules to come out as "::Latin-Greek;"
    945         // even though the ID is "(Lower);Latin-Greek;".
    946         t = new CompoundTransliterator(list, parseError, status);
    947     }
    948     else {
    949         t = (Transliterator*)list.elementAt(0);
    950     }
    951     // Check null pointer
    952     if (t != NULL) {
    953         t->setID(canonID);
    954         if (globalFilter != NULL) {
    955             t->adoptFilter(globalFilter);
    956         }
    957     }
    958     else if (U_SUCCESS(status)) {
    959         status = U_MEMORY_ALLOCATION_ERROR;
    960     }
    961     return t;
    962 }
    963 
    964 /**
    965  * Create a transliterator from a basic ID.  This is an ID
    966  * containing only the forward direction source, target, and
    967  * variant.
    968  * @param id a basic ID of the form S-T or S-T/V.
    969  * @return a newly created Transliterator or null if the ID is
    970  * invalid.
    971  */
    972 Transliterator* Transliterator::createBasicInstance(const UnicodeString& id,
    973                                                     const UnicodeString* canon) {
    974     UParseError pe;
    975     UErrorCode ec = U_ZERO_ERROR;
    976     TransliteratorAlias* alias = 0;
    977     Transliterator* t = 0;
    978 
    979     umtx_lock(&registryMutex);
    980     if (HAVE_REGISTRY(ec)) {
    981         t = registry->get(id, alias, ec);
    982     }
    983     umtx_unlock(&registryMutex);
    984 
    985     if (U_FAILURE(ec)) {
    986         delete t;
    987         delete alias;
    988         return 0;
    989     }
    990 
    991     // We may have not gotten a transliterator:  Because we can't
    992     // instantiate a transliterator from inside TransliteratorRegistry::
    993     // get() (that would deadlock), we sometimes pass back an alias.  This
    994     // contains the data we need to finish the instantiation outside the
    995     // registry mutex.  The alias may, in turn, generate another alias, so
    996     // we handle aliases in a loop.  The max times through the loop is two.
    997     // [alan]
    998     while (alias != 0) {
    999         U_ASSERT(t==0);
   1000         // Rule-based aliases are handled with TransliteratorAlias::
   1001         // parse(), followed by TransliteratorRegistry::reget().
   1002         // Other aliases are handled with TransliteratorAlias::create().
   1003         if (alias->isRuleBased()) {
   1004             // Step 1. parse
   1005             TransliteratorParser parser(ec);
   1006             alias->parse(parser, pe, ec);
   1007             delete alias;
   1008             alias = 0;
   1009 
   1010             // Step 2. reget
   1011             umtx_lock(&registryMutex);
   1012             if (HAVE_REGISTRY(ec)) {
   1013                 t = registry->reget(id, parser, alias, ec);
   1014             }
   1015             umtx_unlock(&registryMutex);
   1016 
   1017             // Step 3. Loop back around!
   1018         } else {
   1019             t = alias->create(pe, ec);
   1020             delete alias;
   1021             alias = 0;
   1022             break;
   1023         }
   1024         if (U_FAILURE(ec)) {
   1025             delete t;
   1026             delete alias;
   1027             t = NULL;
   1028             break;
   1029         }
   1030     }
   1031 
   1032     if (t != NULL && canon != NULL) {
   1033         t->setID(*canon);
   1034     }
   1035 
   1036     return t;
   1037 }
   1038 
   1039 /**
   1040  * Returns a <code>Transliterator</code> object constructed from
   1041  * the given rule string.  This will be a RuleBasedTransliterator,
   1042  * if the rule string contains only rules, or a
   1043  * CompoundTransliterator, if it contains ID blocks, or a
   1044  * NullTransliterator, if it contains ID blocks which parse as
   1045  * empty for the given direction.
   1046  */
   1047 Transliterator* U_EXPORT2
   1048 Transliterator::createFromRules(const UnicodeString& ID,
   1049                                 const UnicodeString& rules,
   1050                                 UTransDirection dir,
   1051                                 UParseError& parseError,
   1052                                 UErrorCode& status)
   1053 {
   1054     Transliterator* t = NULL;
   1055 
   1056     TransliteratorParser parser(status);
   1057     parser.parse(rules, dir, parseError, status);
   1058 
   1059     if (U_FAILURE(status)) {
   1060         return 0;
   1061     }
   1062 
   1063     // NOTE: The logic here matches that in TransliteratorRegistry.
   1064     if (parser.idBlockVector.size() == 0 && parser.dataVector.size() == 0) {
   1065         t = new NullTransliterator();
   1066     }
   1067     else if (parser.idBlockVector.size() == 0 && parser.dataVector.size() == 1) {
   1068         t = new RuleBasedTransliterator(ID, (TransliterationRuleData*)parser.dataVector.orphanElementAt(0), TRUE);
   1069     }
   1070     else if (parser.idBlockVector.size() == 1 && parser.dataVector.size() == 0) {
   1071         // idBlock, no data -- this is an alias.  The ID has
   1072         // been munged from reverse into forward mode, if
   1073         // necessary, so instantiate the ID in the forward
   1074         // direction.
   1075         if (parser.compoundFilter != NULL) {
   1076             UnicodeString filterPattern;
   1077             parser.compoundFilter->toPattern(filterPattern, FALSE);
   1078             t = createInstance(filterPattern + UnicodeString(ID_DELIM)
   1079                     + *((UnicodeString*)parser.idBlockVector.elementAt(0)), UTRANS_FORWARD, parseError, status);
   1080         }
   1081         else
   1082             t = createInstance(*((UnicodeString*)parser.idBlockVector.elementAt(0)), UTRANS_FORWARD, parseError, status);
   1083 
   1084 
   1085         if (t != NULL) {
   1086             t->setID(ID);
   1087         }
   1088     }
   1089     else {
   1090         UVector transliterators(status);
   1091         int32_t passNumber = 1;
   1092 
   1093         int32_t limit = parser.idBlockVector.size();
   1094         if (parser.dataVector.size() > limit)
   1095             limit = parser.dataVector.size();
   1096 
   1097         for (int32_t i = 0; i < limit; i++) {
   1098             if (i < parser.idBlockVector.size()) {
   1099                 UnicodeString* idBlock = (UnicodeString*)parser.idBlockVector.elementAt(i);
   1100                 if (!idBlock->isEmpty()) {
   1101                     Transliterator* temp = createInstance(*idBlock, UTRANS_FORWARD, parseError, status);
   1102                     if (temp != NULL && typeid(*temp) != typeid(NullTransliterator))
   1103                         transliterators.addElement(temp, status);
   1104                     else
   1105                         delete temp;
   1106                 }
   1107             }
   1108             if (!parser.dataVector.isEmpty()) {
   1109                 TransliterationRuleData* data = (TransliterationRuleData*)parser.dataVector.orphanElementAt(0);
   1110                 // TODO: Should passNumber be turned into a decimal-string representation (1 -> "1")?
   1111                 RuleBasedTransliterator* temprbt = new RuleBasedTransliterator(UnicodeString(CompoundTransliterator::PASS_STRING) + UnicodeString(passNumber++),
   1112                         data, TRUE);
   1113                 // Check if NULL before adding it to transliterators to avoid future usage of NULL pointer.
   1114                 if (temprbt == NULL) {
   1115                 	status = U_MEMORY_ALLOCATION_ERROR;
   1116                 	return t;
   1117                 }
   1118                 transliterators.addElement(temprbt, status);
   1119             }
   1120         }
   1121 
   1122         t = new CompoundTransliterator(transliterators, passNumber - 1, parseError, status);
   1123         // Null pointer check
   1124         if (t != NULL) {
   1125             t->setID(ID);
   1126             t->adoptFilter(parser.orphanCompoundFilter());
   1127         }
   1128     }
   1129     if (U_SUCCESS(status) && t == NULL) {
   1130         status = U_MEMORY_ALLOCATION_ERROR;
   1131     }
   1132     return t;
   1133 }
   1134 
   1135 UnicodeString& Transliterator::toRules(UnicodeString& rulesSource,
   1136                                        UBool escapeUnprintable) const {
   1137     // The base class implementation of toRules munges the ID into
   1138     // the correct format.  That is: foo => ::foo
   1139     if (escapeUnprintable) {
   1140         rulesSource.truncate(0);
   1141         UnicodeString id = getID();
   1142         for (int32_t i=0; i<id.length();) {
   1143             UChar32 c = id.char32At(i);
   1144             if (!ICU_Utility::escapeUnprintable(rulesSource, c)) {
   1145                 rulesSource.append(c);
   1146             }
   1147             i += U16_LENGTH(c);
   1148         }
   1149     } else {
   1150         rulesSource = getID();
   1151     }
   1152     // KEEP in sync with rbt_pars
   1153     rulesSource.insert(0, UNICODE_STRING_SIMPLE("::"));
   1154     rulesSource.append(ID_DELIM);
   1155     return rulesSource;
   1156 }
   1157 
   1158 int32_t Transliterator::countElements() const {
   1159     const CompoundTransliterator* ct = dynamic_cast<const CompoundTransliterator*>(this);
   1160     return ct != NULL ? ct->getCount() : 0;
   1161 }
   1162 
   1163 const Transliterator& Transliterator::getElement(int32_t index, UErrorCode& ec) const {
   1164     if (U_FAILURE(ec)) {
   1165         return *this;
   1166     }
   1167     const CompoundTransliterator* cpd = dynamic_cast<const CompoundTransliterator*>(this);
   1168     int32_t n = (cpd == NULL) ? 1 : cpd->getCount();
   1169     if (index < 0 || index >= n) {
   1170         ec = U_INDEX_OUTOFBOUNDS_ERROR;
   1171         return *this;
   1172     } else {
   1173         return (n == 1) ? *this : cpd->getTransliterator(index);
   1174     }
   1175 }
   1176 
   1177 UnicodeSet& Transliterator::getSourceSet(UnicodeSet& result) const {
   1178     handleGetSourceSet(result);
   1179     if (filter != NULL) {
   1180         UnicodeSet* filterSet = dynamic_cast<UnicodeSet*>(filter);
   1181         UBool deleteFilterSet = FALSE;
   1182         // Most, but not all filters will be UnicodeSets.  Optimize for
   1183         // the high-runner case.
   1184         if (filterSet == NULL) {
   1185             filterSet = new UnicodeSet();
   1186             // Check null pointer
   1187             if (filterSet == NULL) {
   1188                 return result;
   1189             }
   1190             deleteFilterSet = TRUE;
   1191             filter->addMatchSetTo(*filterSet);
   1192         }
   1193         result.retainAll(*filterSet);
   1194         if (deleteFilterSet) {
   1195             delete filterSet;
   1196         }
   1197     }
   1198     return result;
   1199 }
   1200 
   1201 void Transliterator::handleGetSourceSet(UnicodeSet& result) const {
   1202     result.clear();
   1203 }
   1204 
   1205 UnicodeSet& Transliterator::getTargetSet(UnicodeSet& result) const {
   1206     return result.clear();
   1207 }
   1208 
   1209 // For public consumption
   1210 void U_EXPORT2 Transliterator::registerFactory(const UnicodeString& id,
   1211                                      Transliterator::Factory factory,
   1212                                      Transliterator::Token context) {
   1213     Mutex lock(&registryMutex);
   1214     UErrorCode ec = U_ZERO_ERROR;
   1215     if (HAVE_REGISTRY(ec)) {
   1216         _registerFactory(id, factory, context);
   1217     }
   1218 }
   1219 
   1220 // To be called only by Transliterator subclasses that are called
   1221 // to register themselves by initializeRegistry().
   1222 void Transliterator::_registerFactory(const UnicodeString& id,
   1223                                       Transliterator::Factory factory,
   1224                                       Transliterator::Token context) {
   1225     UErrorCode ec = U_ZERO_ERROR;
   1226     registry->put(id, factory, context, TRUE, ec);
   1227 }
   1228 
   1229 // To be called only by Transliterator subclasses that are called
   1230 // to register themselves by initializeRegistry().
   1231 void Transliterator::_registerSpecialInverse(const UnicodeString& target,
   1232                                              const UnicodeString& inverseTarget,
   1233                                              UBool bidirectional) {
   1234     UErrorCode status = U_ZERO_ERROR;
   1235     TransliteratorIDParser::registerSpecialInverse(target, inverseTarget, bidirectional, status);
   1236 }
   1237 
   1238 /**
   1239  * Registers a instance <tt>obj</tt> of a subclass of
   1240  * <code>Transliterator</code> with the system.  This object must
   1241  * implement the <tt>clone()</tt> method.  When
   1242  * <tt>getInstance()</tt> is called with an ID string that is
   1243  * equal to <tt>obj.getID()</tt>, then <tt>obj.clone()</tt> is
   1244  * returned.
   1245  *
   1246  * @param obj an instance of subclass of
   1247  * <code>Transliterator</code> that defines <tt>clone()</tt>
   1248  * @see #getInstance
   1249  * @see #unregister
   1250  */
   1251 void U_EXPORT2 Transliterator::registerInstance(Transliterator* adoptedPrototype) {
   1252     Mutex lock(&registryMutex);
   1253     UErrorCode ec = U_ZERO_ERROR;
   1254     if (HAVE_REGISTRY(ec)) {
   1255         _registerInstance(adoptedPrototype);
   1256     }
   1257 }
   1258 
   1259 void Transliterator::_registerInstance(Transliterator* adoptedPrototype) {
   1260     UErrorCode ec = U_ZERO_ERROR;
   1261     registry->put(adoptedPrototype, TRUE, ec);
   1262 }
   1263 
   1264 void U_EXPORT2 Transliterator::registerAlias(const UnicodeString& aliasID,
   1265                                              const UnicodeString& realID) {
   1266     Mutex lock(&registryMutex);
   1267     UErrorCode ec = U_ZERO_ERROR;
   1268     if (HAVE_REGISTRY(ec)) {
   1269         _registerAlias(aliasID, realID);
   1270     }
   1271 }
   1272 
   1273 void Transliterator::_registerAlias(const UnicodeString& aliasID,
   1274                                     const UnicodeString& realID) {
   1275     UErrorCode ec = U_ZERO_ERROR;
   1276     registry->put(aliasID, realID, FALSE, TRUE, ec);
   1277 }
   1278 
   1279 /**
   1280  * Unregisters a transliterator or class.  This may be either
   1281  * a system transliterator or a user transliterator or class.
   1282  *
   1283  * @param ID the ID of the transliterator or class
   1284  * @see #registerInstance
   1285 
   1286  */
   1287 void U_EXPORT2 Transliterator::unregister(const UnicodeString& ID) {
   1288     Mutex lock(&registryMutex);
   1289     UErrorCode ec = U_ZERO_ERROR;
   1290     if (HAVE_REGISTRY(ec)) {
   1291         registry->remove(ID);
   1292     }
   1293 }
   1294 
   1295 /**
   1296  * == OBSOLETE - remove in ICU 3.4 ==
   1297  * Return the number of IDs currently registered with the system.
   1298  * To retrieve the actual IDs, call getAvailableID(i) with
   1299  * i from 0 to countAvailableIDs() - 1.
   1300  */
   1301 int32_t U_EXPORT2 Transliterator::countAvailableIDs(void) {
   1302     int32_t retVal = 0;
   1303     Mutex lock(&registryMutex);
   1304     UErrorCode ec = U_ZERO_ERROR;
   1305     if (HAVE_REGISTRY(ec)) {
   1306         retVal = registry->countAvailableIDs();
   1307     }
   1308     return retVal;
   1309 }
   1310 
   1311 /**
   1312  * == OBSOLETE - remove in ICU 3.4 ==
   1313  * Return the index-th available ID.  index must be between 0
   1314  * and countAvailableIDs() - 1, inclusive.  If index is out of
   1315  * range, the result of getAvailableID(0) is returned.
   1316  */
   1317 const UnicodeString& U_EXPORT2 Transliterator::getAvailableID(int32_t index) {
   1318     const UnicodeString* result = NULL;
   1319     umtx_lock(&registryMutex);
   1320     UErrorCode ec = U_ZERO_ERROR;
   1321     if (HAVE_REGISTRY(ec)) {
   1322         result = &registry->getAvailableID(index);
   1323     }
   1324     umtx_unlock(&registryMutex);
   1325     U_ASSERT(result != NULL); // fail if no registry
   1326     return *result;
   1327 }
   1328 
   1329 StringEnumeration* U_EXPORT2 Transliterator::getAvailableIDs(UErrorCode& ec) {
   1330     if (U_FAILURE(ec)) return NULL;
   1331     StringEnumeration* result = NULL;
   1332     umtx_lock(&registryMutex);
   1333     if (HAVE_REGISTRY(ec)) {
   1334         result = registry->getAvailableIDs();
   1335     }
   1336     umtx_unlock(&registryMutex);
   1337     if (result == NULL) {
   1338         ec = U_INTERNAL_TRANSLITERATOR_ERROR;
   1339     }
   1340     return result;
   1341 }
   1342 
   1343 int32_t U_EXPORT2 Transliterator::countAvailableSources(void) {
   1344     Mutex lock(&registryMutex);
   1345     UErrorCode ec = U_ZERO_ERROR;
   1346     return HAVE_REGISTRY(ec) ? _countAvailableSources() : 0;
   1347 }
   1348 
   1349 UnicodeString& U_EXPORT2 Transliterator::getAvailableSource(int32_t index,
   1350                                                   UnicodeString& result) {
   1351     Mutex lock(&registryMutex);
   1352     UErrorCode ec = U_ZERO_ERROR;
   1353     if (HAVE_REGISTRY(ec)) {
   1354         _getAvailableSource(index, result);
   1355     }
   1356     return result;
   1357 }
   1358 
   1359 int32_t U_EXPORT2 Transliterator::countAvailableTargets(const UnicodeString& source) {
   1360     Mutex lock(&registryMutex);
   1361     UErrorCode ec = U_ZERO_ERROR;
   1362     return HAVE_REGISTRY(ec) ? _countAvailableTargets(source) : 0;
   1363 }
   1364 
   1365 UnicodeString& U_EXPORT2 Transliterator::getAvailableTarget(int32_t index,
   1366                                                   const UnicodeString& source,
   1367                                                   UnicodeString& result) {
   1368     Mutex lock(&registryMutex);
   1369     UErrorCode ec = U_ZERO_ERROR;
   1370     if (HAVE_REGISTRY(ec)) {
   1371         _getAvailableTarget(index, source, result);
   1372     }
   1373     return result;
   1374 }
   1375 
   1376 int32_t U_EXPORT2 Transliterator::countAvailableVariants(const UnicodeString& source,
   1377                                                const UnicodeString& target) {
   1378     Mutex lock(&registryMutex);
   1379     UErrorCode ec = U_ZERO_ERROR;
   1380     return HAVE_REGISTRY(ec) ? _countAvailableVariants(source, target) : 0;
   1381 }
   1382 
   1383 UnicodeString& U_EXPORT2 Transliterator::getAvailableVariant(int32_t index,
   1384                                                    const UnicodeString& source,
   1385                                                    const UnicodeString& target,
   1386                                                    UnicodeString& result) {
   1387     Mutex lock(&registryMutex);
   1388     UErrorCode ec = U_ZERO_ERROR;
   1389     if (HAVE_REGISTRY(ec)) {
   1390         _getAvailableVariant(index, source, target, result);
   1391     }
   1392     return result;
   1393 }
   1394 
   1395 int32_t Transliterator::_countAvailableSources(void) {
   1396     return registry->countAvailableSources();
   1397 }
   1398 
   1399 UnicodeString& Transliterator::_getAvailableSource(int32_t index,
   1400                                                   UnicodeString& result) {
   1401     return registry->getAvailableSource(index, result);
   1402 }
   1403 
   1404 int32_t Transliterator::_countAvailableTargets(const UnicodeString& source) {
   1405     return registry->countAvailableTargets(source);
   1406 }
   1407 
   1408 UnicodeString& Transliterator::_getAvailableTarget(int32_t index,
   1409                                                   const UnicodeString& source,
   1410                                                   UnicodeString& result) {
   1411     return registry->getAvailableTarget(index, source, result);
   1412 }
   1413 
   1414 int32_t Transliterator::_countAvailableVariants(const UnicodeString& source,
   1415                                                const UnicodeString& target) {
   1416     return registry->countAvailableVariants(source, target);
   1417 }
   1418 
   1419 UnicodeString& Transliterator::_getAvailableVariant(int32_t index,
   1420                                                    const UnicodeString& source,
   1421                                                    const UnicodeString& target,
   1422                                                    UnicodeString& result) {
   1423     return registry->getAvailableVariant(index, source, target, result);
   1424 }
   1425 
   1426 #ifdef U_USE_DEPRECATED_TRANSLITERATOR_API
   1427 
   1428 /**
   1429  * Method for subclasses to use to obtain a character in the given
   1430  * string, with filtering.
   1431  * @deprecated the new architecture provides filtering at the top
   1432  * level.  This method will be removed Dec 31 2001.
   1433  */
   1434 UChar Transliterator::filteredCharAt(const Replaceable& text, int32_t i) const {
   1435     UChar c;
   1436     const UnicodeFilter* localFilter = getFilter();
   1437     return (localFilter == 0) ? text.charAt(i) :
   1438         (localFilter->contains(c = text.charAt(i)) ? c : (UChar)0xFFFE);
   1439 }
   1440 
   1441 #endif
   1442 
   1443 /**
   1444  * If the registry is initialized, return TRUE.  If not, initialize it
   1445  * and return TRUE.  If the registry cannot be initialized, return
   1446  * FALSE (rare).
   1447  *
   1448  * IMPORTANT: Upon entry, registryMutex must be LOCKED.  The entire
   1449  * initialization is done with the lock held.  There is NO REASON to
   1450  * unlock, since no other thread that is waiting on the registryMutex
   1451  * cannot itself proceed until the registry is initialized.
   1452  */
   1453 UBool Transliterator::initializeRegistry(UErrorCode &status) {
   1454     if (registry != 0) {
   1455         return TRUE;
   1456     }
   1457 
   1458     registry = new TransliteratorRegistry(status);
   1459     if (registry == 0 || U_FAILURE(status)) {
   1460         delete registry;
   1461         registry = 0;
   1462         return FALSE; // can't create registry, no recovery
   1463     }
   1464 
   1465     /* The following code parses the index table located in
   1466      * icu/data/translit/root.txt.  The index is an n x 4 table
   1467      * that follows this format:
   1468      *  <id>{
   1469      *      file{
   1470      *          resource{"<resource>"}
   1471      *          direction{"<direction>"}
   1472      *      }
   1473      *  }
   1474      *  <id>{
   1475      *      internal{
   1476      *          resource{"<resource>"}
   1477      *          direction{"<direction"}
   1478      *       }
   1479      *  }
   1480      *  <id>{
   1481      *      alias{"<getInstanceArg"}
   1482      *  }
   1483      * <id> is the ID of the system transliterator being defined.  These
   1484      * are public IDs enumerated by Transliterator.getAvailableIDs(),
   1485      * unless the second field is "internal".
   1486      *
   1487      * <resource> is a ResourceReader resource name.  Currently these refer
   1488      * to file names under com/ibm/text/resources.  This string is passed
   1489      * directly to ResourceReader, together with <encoding>.
   1490      *
   1491      * <direction> is either "FORWARD" or "REVERSE".
   1492      *
   1493      * <getInstanceArg> is a string to be passed directly to
   1494      * Transliterator.getInstance().  The returned Transliterator object
   1495      * then has its ID changed to <id> and is returned.
   1496      *
   1497      * The extra blank field on "alias" lines is to make the array square.
   1498      */
   1499     //static const char translit_index[] = "translit_index";
   1500 
   1501     UResourceBundle *bundle, *transIDs, *colBund;
   1502     bundle = ures_open(U_ICUDATA_TRANSLIT, NULL/*open default locale*/, &status);
   1503     transIDs = ures_getByKey(bundle, RB_RULE_BASED_IDS, 0, &status);
   1504 
   1505     int32_t row, maxRows;
   1506     if (U_SUCCESS(status)) {
   1507         maxRows = ures_getSize(transIDs);
   1508         for (row = 0; row < maxRows; row++) {
   1509             colBund = ures_getByIndex(transIDs, row, 0, &status);
   1510             if (U_SUCCESS(status)) {
   1511                 UnicodeString id(ures_getKey(colBund), -1, US_INV);
   1512                 UResourceBundle* res = ures_getNextResource(colBund, NULL, &status);
   1513                 const char* typeStr = ures_getKey(res);
   1514                 UChar type;
   1515                 u_charsToUChars(typeStr, &type, 1);
   1516 
   1517                 if (U_SUCCESS(status)) {
   1518                     int32_t len = 0;
   1519                     const UChar *resString;
   1520                     switch (type) {
   1521                     case 0x66: // 'f'
   1522                     case 0x69: // 'i'
   1523                         // 'file' or 'internal';
   1524                         // row[2]=resource, row[3]=direction
   1525                         {
   1526 
   1527                             resString = ures_getStringByKey(res, "resource", &len, &status);
   1528                             UBool visible = (type == 0x0066 /*f*/);
   1529                             UTransDirection dir =
   1530                                 (ures_getUnicodeStringByKey(res, "direction", &status).charAt(0) ==
   1531                                  0x0046 /*F*/) ?
   1532                                 UTRANS_FORWARD : UTRANS_REVERSE;
   1533                             registry->put(id, UnicodeString(TRUE, resString, len), dir, TRUE, visible, status);
   1534                         }
   1535                         break;
   1536                     case 0x61: // 'a'
   1537                         // 'alias'; row[2]=createInstance argument
   1538                         resString = ures_getString(res, &len, &status);
   1539                         registry->put(id, UnicodeString(TRUE, resString, len), TRUE, TRUE, status);
   1540                         break;
   1541                     }
   1542                 }
   1543                 ures_close(res);
   1544             }
   1545             ures_close(colBund);
   1546         }
   1547     }
   1548 
   1549     ures_close(transIDs);
   1550     ures_close(bundle);
   1551 
   1552     // Manually add prototypes that the system knows about to the
   1553     // cache.  This is how new non-rule-based transliterators are
   1554     // added to the system.
   1555 
   1556     // This is to allow for null pointer check
   1557     NullTransliterator* tempNullTranslit = new NullTransliterator();
   1558     LowercaseTransliterator* tempLowercaseTranslit = new LowercaseTransliterator();
   1559     UppercaseTransliterator* tempUppercaseTranslit = new UppercaseTransliterator();
   1560     TitlecaseTransliterator* tempTitlecaseTranslit = new TitlecaseTransliterator();
   1561     UnicodeNameTransliterator* tempUnicodeTranslit = new UnicodeNameTransliterator();
   1562     NameUnicodeTransliterator* tempNameUnicodeTranslit = new NameUnicodeTransliterator();
   1563 #if !UCONFIG_NO_BREAK_ITERATION
   1564      // TODO: could or should these transliterators be referenced polymorphically once constructed?
   1565      BreakTransliterator* tempBreakTranslit         = new BreakTransliterator();
   1566 #endif
   1567     // Check for null pointers
   1568     if (tempNullTranslit == NULL || tempLowercaseTranslit == NULL || tempUppercaseTranslit == NULL ||
   1569         tempTitlecaseTranslit == NULL || tempUnicodeTranslit == NULL ||
   1570 #if !UCONFIG_NO_BREAK_ITERATION
   1571         tempBreakTranslit == NULL ||
   1572 #endif
   1573         tempNameUnicodeTranslit == NULL )
   1574     {
   1575         delete tempNullTranslit;
   1576         delete tempLowercaseTranslit;
   1577         delete tempUppercaseTranslit;
   1578         delete tempTitlecaseTranslit;
   1579         delete tempUnicodeTranslit;
   1580         delete tempNameUnicodeTranslit;
   1581 #if !UCONFIG_NO_BREAK_ITERATION
   1582         delete tempBreakTranslit;
   1583 #endif
   1584         // Since there was an error, remove registry
   1585         delete registry;
   1586         registry = NULL;
   1587 
   1588         status = U_MEMORY_ALLOCATION_ERROR;
   1589         return 0;
   1590     }
   1591 
   1592     registry->put(tempNullTranslit, TRUE, status);
   1593     registry->put(tempLowercaseTranslit, TRUE, status);
   1594     registry->put(tempUppercaseTranslit, TRUE, status);
   1595     registry->put(tempTitlecaseTranslit, TRUE, status);
   1596     registry->put(tempUnicodeTranslit, TRUE, status);
   1597     registry->put(tempNameUnicodeTranslit, TRUE, status);
   1598 #if !UCONFIG_NO_BREAK_ITERATION
   1599     registry->put(tempBreakTranslit, FALSE, status);   // FALSE means invisible.
   1600 #endif
   1601 
   1602     RemoveTransliterator::registerIDs(); // Must be within mutex
   1603     EscapeTransliterator::registerIDs();
   1604     UnescapeTransliterator::registerIDs();
   1605     NormalizationTransliterator::registerIDs();
   1606     AnyTransliterator::registerIDs();
   1607 
   1608     _registerSpecialInverse(UNICODE_STRING_SIMPLE("Null"),
   1609                             UNICODE_STRING_SIMPLE("Null"), FALSE);
   1610     _registerSpecialInverse(UNICODE_STRING_SIMPLE("Upper"),
   1611                             UNICODE_STRING_SIMPLE("Lower"), TRUE);
   1612     _registerSpecialInverse(UNICODE_STRING_SIMPLE("Title"),
   1613                             UNICODE_STRING_SIMPLE("Lower"), FALSE);
   1614 
   1615     ucln_i18n_registerCleanup(UCLN_I18N_TRANSLITERATOR, utrans_transliterator_cleanup);
   1616 
   1617     return TRUE;
   1618 }
   1619 
   1620 U_NAMESPACE_END
   1621 
   1622 // Defined in ucln_in.h:
   1623 
   1624 /**
   1625  * Release all static memory held by transliterator.  This will
   1626  * necessarily invalidate any rule-based transliterators held by the
   1627  * user, because RBTs hold pointers to common data objects.
   1628  */
   1629 U_CFUNC UBool utrans_transliterator_cleanup(void) {
   1630     U_NAMESPACE_USE
   1631     TransliteratorIDParser::cleanup();
   1632     if (registry) {
   1633         delete registry;
   1634         registry = NULL;
   1635     }
   1636     return TRUE;
   1637 }
   1638 
   1639 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
   1640 
   1641 //eof
   1642