• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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/uobject.h"
16 #include "unicode/parseerr.h"
17 #include "unicode/parsepos.h"
18 #include "unicode/putil.h"
19 #include "unicode/uchar.h"
20 #include "unicode/ustring.h"
21 #include "unicode/uniset.h"
22 #include "cstring.h"
23 #include "funcrepl.h"
24 #include "hash.h"
25 #include "quant.h"
26 #include "rbt.h"
27 #include "rbt_data.h"
28 #include "rbt_pars.h"
29 #include "rbt_rule.h"
30 #include "strmatch.h"
31 #include "strrepl.h"
32 #include "unicode/symtable.h"
33 #include "tridpars.h"
34 #include "uvector.h"
35 #include "hash.h"
36 #include "util.h"
37 #include "cmemory.h"
38 #include "uprops.h"
39 #include "putilimp.h"
40 
41 // Operators
42 #define VARIABLE_DEF_OP ((UChar)0x003D) /*=*/
43 #define FORWARD_RULE_OP ((UChar)0x003E) /*>*/
44 #define REVERSE_RULE_OP ((UChar)0x003C) /*<*/
45 #define FWDREV_RULE_OP  ((UChar)0x007E) /*~*/ // internal rep of <> op
46 
47 // Other special characters
48 #define QUOTE             ((UChar)0x0027) /*'*/
49 #define ESCAPE            ((UChar)0x005C) /*\*/
50 #define END_OF_RULE       ((UChar)0x003B) /*;*/
51 #define RULE_COMMENT_CHAR ((UChar)0x0023) /*#*/
52 
53 #define SEGMENT_OPEN       ((UChar)0x0028) /*(*/
54 #define SEGMENT_CLOSE      ((UChar)0x0029) /*)*/
55 #define CONTEXT_ANTE       ((UChar)0x007B) /*{*/
56 #define CONTEXT_POST       ((UChar)0x007D) /*}*/
57 #define CURSOR_POS         ((UChar)0x007C) /*|*/
58 #define CURSOR_OFFSET      ((UChar)0x0040) /*@*/
59 #define ANCHOR_START       ((UChar)0x005E) /*^*/
60 #define KLEENE_STAR        ((UChar)0x002A) /***/
61 #define ONE_OR_MORE        ((UChar)0x002B) /*+*/
62 #define ZERO_OR_ONE        ((UChar)0x003F) /*?*/
63 
64 #define DOT                ((UChar)46)     /*.*/
65 
66 static const UChar DOT_SET[] = { // "[^[:Zp:][:Zl:]\r\n$]";
67     91, 94, 91, 58, 90, 112, 58, 93, 91, 58, 90,
68     108, 58, 93, 92, 114, 92, 110, 36, 93, 0
69 };
70 
71 // A function is denoted &Source-Target/Variant(text)
72 #define FUNCTION           ((UChar)38)     /*&*/
73 
74 // Aliases for some of the syntax characters. These are provided so
75 // transliteration rules can be expressed in XML without clashing with
76 // XML syntax characters '<', '>', and '&'.
77 #define ALT_REVERSE_RULE_OP ((UChar)0x2190) // Left Arrow
78 #define ALT_FORWARD_RULE_OP ((UChar)0x2192) // Right Arrow
79 #define ALT_FWDREV_RULE_OP  ((UChar)0x2194) // Left Right Arrow
80 #define ALT_FUNCTION        ((UChar)0x2206) // Increment (~Greek Capital Delta)
81 
82 // Special characters disallowed at the top level
83 static const UChar ILLEGAL_TOP[] = {41,0}; // ")"
84 
85 // Special characters disallowed within a segment
86 static const UChar ILLEGAL_SEG[] = {123,125,124,64,0}; // "{}|@"
87 
88 // Special characters disallowed within a function argument
89 static const UChar ILLEGAL_FUNC[] = {94,40,46,42,43,63,123,125,124,64,0}; // "^(.*+?{}|@"
90 
91 // By definition, the ANCHOR_END special character is a
92 // trailing SymbolTable.SYMBOL_REF character.
93 // private static final char ANCHOR_END       = '$';
94 
95 static const UChar gOPERATORS[] = { // "=><"
96     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
97     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
98     0
99 };
100 
101 static const UChar HALF_ENDERS[] = { // "=><;"
102     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
103     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
104     END_OF_RULE,
105     0
106 };
107 
108 // These are also used in Transliterator::toRules()
109 static const int32_t ID_TOKEN_LEN = 2;
110 static const UChar   ID_TOKEN[]   = { 0x3A, 0x3A }; // ':', ':'
111 
112 /*
113 commented out until we do real ::BEGIN/::END functionality
114 static const int32_t BEGIN_TOKEN_LEN = 5;
115 static const UChar BEGIN_TOKEN[] = { 0x42, 0x45, 0x47, 0x49, 0x4e }; // 'BEGIN'
116 
117 static const int32_t END_TOKEN_LEN = 3;
118 static const UChar END_TOKEN[] = { 0x45, 0x4e, 0x44 }; // 'END'
119 */
120 
121 U_NAMESPACE_BEGIN
122 
123 //----------------------------------------------------------------------
124 // BEGIN ParseData
125 //----------------------------------------------------------------------
126 
127 /**
128  * This class implements the SymbolTable interface.  It is used
129  * during parsing to give UnicodeSet access to variables that
130  * have been defined so far.  Note that it uses variablesVector,
131  * _not_ data.setVariables.
132  */
133 class ParseData : public UMemory, public SymbolTable {
134 public:
135     const TransliterationRuleData* data; // alias
136 
137     const UVector* variablesVector; // alias
138 
139     const Hashtable* variableNames; // alias
140 
141     ParseData(const TransliterationRuleData* data = 0,
142               const UVector* variablesVector = 0,
143               const Hashtable* variableNames = 0);
144 
145     virtual const UnicodeString* lookup(const UnicodeString& s) const;
146 
147     virtual const UnicodeFunctor* lookupMatcher(UChar32 ch) const;
148 
149     virtual UnicodeString parseReference(const UnicodeString& text,
150                                          ParsePosition& pos, int32_t limit) const;
151     /**
152      * Return true if the given character is a matcher standin or a plain
153      * character (non standin).
154      */
155     UBool isMatcher(UChar32 ch);
156 
157     /**
158      * Return true if the given character is a replacer standin or a plain
159      * character (non standin).
160      */
161     UBool isReplacer(UChar32 ch);
162 
163 private:
164     ParseData(const ParseData &other); // forbid copying of this class
165     ParseData &operator=(const ParseData &other); // forbid copying of this class
166 };
167 
ParseData(const TransliterationRuleData * d,const UVector * sets,const Hashtable * vNames)168 ParseData::ParseData(const TransliterationRuleData* d,
169                      const UVector* sets,
170                      const Hashtable* vNames) :
171     data(d), variablesVector(sets), variableNames(vNames) {}
172 
173 /**
174  * Implement SymbolTable API.
175  */
lookup(const UnicodeString & name) const176 const UnicodeString* ParseData::lookup(const UnicodeString& name) const {
177     return (const UnicodeString*) variableNames->get(name);
178 }
179 
180 /**
181  * Implement SymbolTable API.
182  */
lookupMatcher(UChar32 ch) const183 const UnicodeFunctor* ParseData::lookupMatcher(UChar32 ch) const {
184     // Note that we cannot use data.lookupSet() because the
185     // set array has not been constructed yet.
186     const UnicodeFunctor* set = NULL;
187     int32_t i = ch - data->variablesBase;
188     if (i >= 0 && i < variablesVector->size()) {
189         int32_t i = ch - data->variablesBase;
190         set = (i < variablesVector->size()) ?
191             (UnicodeFunctor*) variablesVector->elementAt(i) : 0;
192     }
193     return set;
194 }
195 
196 /**
197  * Implement SymbolTable API.  Parse out a symbol reference
198  * name.
199  */
parseReference(const UnicodeString & text,ParsePosition & pos,int32_t limit) const200 UnicodeString ParseData::parseReference(const UnicodeString& text,
201                                         ParsePosition& pos, int32_t limit) const {
202     int32_t start = pos.getIndex();
203     int32_t i = start;
204     UnicodeString result;
205     while (i < limit) {
206         UChar c = text.charAt(i);
207         if ((i==start && !u_isIDStart(c)) || !u_isIDPart(c)) {
208             break;
209         }
210         ++i;
211     }
212     if (i == start) { // No valid name chars
213         return result; // Indicate failure with empty string
214     }
215     pos.setIndex(i);
216     text.extractBetween(start, i, result);
217     return result;
218 }
219 
isMatcher(UChar32 ch)220 UBool ParseData::isMatcher(UChar32 ch) {
221     // Note that we cannot use data.lookup() because the
222     // set array has not been constructed yet.
223     int32_t i = ch - data->variablesBase;
224     if (i >= 0 && i < variablesVector->size()) {
225         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
226         return f != NULL && f->toMatcher() != NULL;
227     }
228     return TRUE;
229 }
230 
231 /**
232  * Return true if the given character is a replacer standin or a plain
233  * character (non standin).
234  */
isReplacer(UChar32 ch)235 UBool ParseData::isReplacer(UChar32 ch) {
236     // Note that we cannot use data.lookup() because the
237     // set array has not been constructed yet.
238     int i = ch - data->variablesBase;
239     if (i >= 0 && i < variablesVector->size()) {
240         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
241         return f != NULL && f->toReplacer() != NULL;
242     }
243     return TRUE;
244 }
245 
246 //----------------------------------------------------------------------
247 // BEGIN RuleHalf
248 //----------------------------------------------------------------------
249 
250 /**
251  * A class representing one side of a rule.  This class knows how to
252  * parse half of a rule.  It is tightly coupled to the method
253  * RuleBasedTransliterator.Parser.parseRule().
254  */
255 class RuleHalf : public UMemory {
256 
257 public:
258 
259     UnicodeString text;
260 
261     int32_t cursor; // position of cursor in text
262     int32_t ante;   // position of ante context marker '{' in text
263     int32_t post;   // position of post context marker '}' in text
264 
265     // Record the offset to the cursor either to the left or to the
266     // right of the key.  This is indicated by characters on the output
267     // side that allow the cursor to be positioned arbitrarily within
268     // the matching text.  For example, abc{def} > | @@@ xyz; changes
269     // def to xyz and moves the cursor to before abc.  Offset characters
270     // must be at the start or end, and they cannot move the cursor past
271     // the ante- or postcontext text.  Placeholders are only valid in
272     // output text.  The length of the ante and post context is
273     // determined at runtime, because of supplementals and quantifiers.
274     int32_t cursorOffset; // only nonzero on output side
275 
276     // Position of first CURSOR_OFFSET on _right_.  This will be -1
277     // for |@, -2 for |@@, etc., and 1 for @|, 2 for @@|, etc.
278     int32_t cursorOffsetPos;
279 
280     UBool anchorStart;
281     UBool anchorEnd;
282 
283     /**
284      * The segment number from 1..n of the next '(' we see
285      * during parsing; 1-based.
286      */
287     int32_t nextSegmentNumber;
288 
289     TransliteratorParser& parser;
290 
291     //--------------------------------------------------
292     // Methods
293 
294     RuleHalf(TransliteratorParser& parser);
295     ~RuleHalf();
296 
297     int32_t parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status);
298 
299     int32_t parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
300                          UnicodeString& buf,
301                          const UnicodeString& illegal,
302                          UBool isSegment,
303                          UErrorCode& status);
304 
305     /**
306      * Remove context.
307      */
308     void removeContext();
309 
310     /**
311      * Return true if this half looks like valid output, that is, does not
312      * contain quantifiers or other special input-only elements.
313      */
314     UBool isValidOutput(TransliteratorParser& parser);
315 
316     /**
317      * Return true if this half looks like valid input, that is, does not
318      * contain functions or other special output-only elements.
319      */
320     UBool isValidInput(TransliteratorParser& parser);
321 
syntaxError(UErrorCode code,const UnicodeString & rule,int32_t start,UErrorCode & status)322     int syntaxError(UErrorCode code,
323                     const UnicodeString& rule,
324                     int32_t start,
325                     UErrorCode& status) {
326         return parser.syntaxError(code, rule, start, status);
327     }
328 
329 private:
330     // Disallowed methods; no impl.
331     RuleHalf(const RuleHalf&);
332     RuleHalf& operator=(const RuleHalf&);
333 };
334 
RuleHalf(TransliteratorParser & p)335 RuleHalf::RuleHalf(TransliteratorParser& p) :
336     parser(p)
337 {
338     cursor = -1;
339     ante = -1;
340     post = -1;
341     cursorOffset = 0;
342     cursorOffsetPos = 0;
343     anchorStart = anchorEnd = FALSE;
344     nextSegmentNumber = 1;
345 }
346 
~RuleHalf()347 RuleHalf::~RuleHalf() {
348 }
349 
350 /**
351  * Parse one side of a rule, stopping at either the limit,
352  * the END_OF_RULE character, or an operator.
353  * @return the index after the terminating character, or
354  * if limit was reached, limit
355  */
parse(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)356 int32_t RuleHalf::parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
357     int32_t start = pos;
358     text.truncate(0);
359     pos = parseSection(rule, pos, limit, text, ILLEGAL_TOP, FALSE, status);
360 
361     if (cursorOffset > 0 && cursor != cursorOffsetPos) {
362         return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
363     }
364 
365     return pos;
366 }
367 
368 /**
369  * Parse a section of one side of a rule, stopping at either
370  * the limit, the END_OF_RULE character, an operator, or a
371  * segment close character.  This method parses both a
372  * top-level rule half and a segment within such a rule half.
373  * It calls itself recursively to parse segments and nested
374  * segments.
375  * @param buf buffer into which to accumulate the rule pattern
376  * characters, either literal characters from the rule or
377  * standins for UnicodeMatcher objects including segments.
378  * @param illegal the set of special characters that is illegal during
379  * this parse.
380  * @param isSegment if true, then we've already seen a '(' and
381  * pos on entry points right after it.  Accumulate everything
382  * up to the closing ')', put it in a segment matcher object,
383  * generate a standin for it, and add the standin to buf.  As
384  * a side effect, update the segments vector with a reference
385  * to the segment matcher.  This works recursively for nested
386  * segments.  If isSegment is false, just accumulate
387  * characters into buf.
388  * @return the index after the terminating character, or
389  * if limit was reached, limit
390  */
parseSection(const UnicodeString & rule,int32_t pos,int32_t limit,UnicodeString & buf,const UnicodeString & illegal,UBool isSegment,UErrorCode & status)391 int32_t RuleHalf::parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
392                                UnicodeString& buf,
393                                const UnicodeString& illegal,
394                                UBool isSegment, UErrorCode& status) {
395     int32_t start = pos;
396     ParsePosition pp;
397     UnicodeString scratch;
398     UBool done = FALSE;
399     int32_t quoteStart = -1; // Most recent 'single quoted string'
400     int32_t quoteLimit = -1;
401     int32_t varStart = -1; // Most recent $variableReference
402     int32_t varLimit = -1;
403     int32_t bufStart = buf.length();
404 
405     while (pos < limit && !done) {
406         // Since all syntax characters are in the BMP, fetching
407         // 16-bit code units suffices here.
408         UChar c = rule.charAt(pos++);
409         if (uprv_isRuleWhiteSpace(c)) {
410             // Ignore whitespace.  Note that this is not Unicode
411             // spaces, but Java spaces -- a subset, representing
412             // whitespace likely to be seen in code.
413             continue;
414         }
415         if (u_strchr(HALF_ENDERS, c) != NULL) {
416             if (isSegment) {
417                 // Unclosed segment
418                 return syntaxError(U_UNCLOSED_SEGMENT, rule, start, status);
419             }
420             break;
421         }
422         if (anchorEnd) {
423             // Text after a presumed end anchor is a syntax err
424             return syntaxError(U_MALFORMED_VARIABLE_REFERENCE, rule, start, status);
425         }
426         if (UnicodeSet::resemblesPattern(rule, pos-1)) {
427             pp.setIndex(pos-1); // Backup to opening '['
428             buf.append(parser.parseSet(rule, pp, status));
429             if (U_FAILURE(status)) {
430                 return syntaxError(U_MALFORMED_SET, rule, start, status);
431             }
432             pos = pp.getIndex();
433             continue;
434         }
435         // Handle escapes
436         if (c == ESCAPE) {
437             if (pos == limit) {
438                 return syntaxError(U_TRAILING_BACKSLASH, rule, start, status);
439             }
440             UChar32 escaped = rule.unescapeAt(pos); // pos is already past '\\'
441             if (escaped == (UChar32) -1) {
442                 return syntaxError(U_MALFORMED_UNICODE_ESCAPE, rule, start, status);
443             }
444             if (!parser.checkVariableRange(escaped)) {
445                 return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
446             }
447             buf.append(escaped);
448             continue;
449         }
450         // Handle quoted matter
451         if (c == QUOTE) {
452             int32_t iq = rule.indexOf(QUOTE, pos);
453             if (iq == pos) {
454                 buf.append(c); // Parse [''] outside quotes as [']
455                 ++pos;
456             } else {
457                 /* This loop picks up a run of quoted text of the
458                  * form 'aaaa' each time through.  If this run
459                  * hasn't really ended ('aaaa''bbbb') then it keeps
460                  * looping, each time adding on a new run.  When it
461                  * reaches the final quote it breaks.
462                  */
463                 quoteStart = buf.length();
464                 for (;;) {
465                     if (iq < 0) {
466                         return syntaxError(U_UNTERMINATED_QUOTE, rule, start, status);
467                     }
468                     scratch.truncate(0);
469                     rule.extractBetween(pos, iq, scratch);
470                     buf.append(scratch);
471                     pos = iq+1;
472                     if (pos < limit && rule.charAt(pos) == QUOTE) {
473                         // Parse [''] inside quotes as [']
474                         iq = rule.indexOf(QUOTE, pos+1);
475                         // Continue looping
476                     } else {
477                         break;
478                     }
479                 }
480                 quoteLimit = buf.length();
481 
482                 for (iq=quoteStart; iq<quoteLimit; ++iq) {
483                     if (!parser.checkVariableRange(buf.charAt(iq))) {
484                         return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
485                     }
486                 }
487             }
488             continue;
489         }
490 
491         if (!parser.checkVariableRange(c)) {
492             return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
493         }
494 
495         if (illegal.indexOf(c) >= 0) {
496             syntaxError(U_ILLEGAL_CHARACTER, rule, start, status);
497         }
498 
499         switch (c) {
500 
501         //------------------------------------------------------
502         // Elements allowed within and out of segments
503         //------------------------------------------------------
504         case ANCHOR_START:
505             if (buf.length() == 0 && !anchorStart) {
506                 anchorStart = TRUE;
507             } else {
508               return syntaxError(U_MISPLACED_ANCHOR_START,
509                                  rule, start, status);
510             }
511           break;
512         case SEGMENT_OPEN:
513             {
514                 // bufSegStart is the offset in buf to the first
515                 // character of the segment we are parsing.
516                 int32_t bufSegStart = buf.length();
517 
518                 // Record segment number now, since nextSegmentNumber
519                 // will be incremented during the call to parseSection
520                 // if there are nested segments.
521                 int32_t segmentNumber = nextSegmentNumber++; // 1-based
522 
523                 // Parse the segment
524                 pos = parseSection(rule, pos, limit, buf, ILLEGAL_SEG, TRUE, status);
525 
526                 // After parsing a segment, the relevant characters are
527                 // in buf, starting at offset bufSegStart.  Extract them
528                 // into a string matcher, and replace them with a
529                 // standin for that matcher.
530                 StringMatcher* m =
531                     new StringMatcher(buf, bufSegStart, buf.length(),
532                                       segmentNumber, *parser.curData);
533                 if (m == NULL) {
534                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
535                 }
536 
537                 // Record and associate object and segment number
538                 parser.setSegmentObject(segmentNumber, m, status);
539                 buf.truncate(bufSegStart);
540                 buf.append(parser.getSegmentStandin(segmentNumber, status));
541             }
542             break;
543         case FUNCTION:
544         case ALT_FUNCTION:
545             {
546                 int32_t iref = pos;
547                 TransliteratorIDParser::SingleID* single =
548                     TransliteratorIDParser::parseFilterID(rule, iref);
549                 // The next character MUST be a segment open
550                 if (single == NULL ||
551                     !ICU_Utility::parseChar(rule, iref, SEGMENT_OPEN)) {
552                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
553                 }
554 
555                 Transliterator *t = single->createInstance();
556                 delete single;
557                 if (t == NULL) {
558                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
559                 }
560 
561                 // bufSegStart is the offset in buf to the first
562                 // character of the segment we are parsing.
563                 int32_t bufSegStart = buf.length();
564 
565                 // Parse the segment
566                 pos = parseSection(rule, iref, limit, buf, ILLEGAL_FUNC, TRUE, status);
567 
568                 // After parsing a segment, the relevant characters are
569                 // in buf, starting at offset bufSegStart.
570                 UnicodeString output;
571                 buf.extractBetween(bufSegStart, buf.length(), output);
572                 FunctionReplacer *r =
573                     new FunctionReplacer(t, new StringReplacer(output, parser.curData));
574                 if (r == NULL) {
575                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
576                 }
577 
578                 // Replace the buffer contents with a stand-in
579                 buf.truncate(bufSegStart);
580                 buf.append(parser.generateStandInFor(r, status));
581             }
582             break;
583         case SymbolTable::SYMBOL_REF:
584             // Handle variable references and segment references "$1" .. "$9"
585             {
586                 // A variable reference must be followed immediately
587                 // by a Unicode identifier start and zero or more
588                 // Unicode identifier part characters, or by a digit
589                 // 1..9 if it is a segment reference.
590                 if (pos == limit) {
591                     // A variable ref character at the end acts as
592                     // an anchor to the context limit, as in perl.
593                     anchorEnd = TRUE;
594                     break;
595                 }
596                 // Parse "$1" "$2" .. "$9" .. (no upper limit)
597                 c = rule.charAt(pos);
598                 int32_t r = u_digit(c, 10);
599                 if (r >= 1 && r <= 9) {
600                     r = ICU_Utility::parseNumber(rule, pos, 10);
601                     if (r < 0) {
602                         return syntaxError(U_UNDEFINED_SEGMENT_REFERENCE,
603                                            rule, start, status);
604                     }
605                     buf.append(parser.getSegmentStandin(r, status));
606                 } else {
607                     pp.setIndex(pos);
608                     UnicodeString name = parser.parseData->
609                                     parseReference(rule, pp, limit);
610                     if (name.length() == 0) {
611                         // This means the '$' was not followed by a
612                         // valid name.  Try to interpret it as an
613                         // end anchor then.  If this also doesn't work
614                         // (if we see a following character) then signal
615                         // an error.
616                         anchorEnd = TRUE;
617                         break;
618                     }
619                     pos = pp.getIndex();
620                     // If this is a variable definition statement,
621                     // then the LHS variable will be undefined.  In
622                     // that case appendVariableDef() will append the
623                     // special placeholder char variableLimit-1.
624                     varStart = buf.length();
625                     parser.appendVariableDef(name, buf, status);
626                     varLimit = buf.length();
627                 }
628             }
629             break;
630         case DOT:
631             buf.append(parser.getDotStandIn(status));
632             break;
633         case KLEENE_STAR:
634         case ONE_OR_MORE:
635         case ZERO_OR_ONE:
636             // Quantifiers.  We handle single characters, quoted strings,
637             // variable references, and segments.
638             //  a+      matches  aaa
639             //  'foo'+  matches  foofoofoo
640             //  $v+     matches  xyxyxy if $v == xy
641             //  (seg)+  matches  segsegseg
642             {
643                 if (isSegment && buf.length() == bufStart) {
644                     // The */+ immediately follows '('
645                     return syntaxError(U_MISPLACED_QUANTIFIER, rule, start, status);
646                 }
647 
648                 int32_t qstart, qlimit;
649                 // The */+ follows an isolated character or quote
650                 // or variable reference
651                 if (buf.length() == quoteLimit) {
652                     // The */+ follows a 'quoted string'
653                     qstart = quoteStart;
654                     qlimit = quoteLimit;
655                 } else if (buf.length() == varLimit) {
656                     // The */+ follows a $variableReference
657                     qstart = varStart;
658                     qlimit = varLimit;
659                 } else {
660                     // The */+ follows a single character, possibly
661                     // a segment standin
662                     qstart = buf.length() - 1;
663                     qlimit = qstart + 1;
664                 }
665 
666                 UnicodeFunctor *m =
667                     new StringMatcher(buf, qstart, qlimit, 0, *parser.curData);
668                 if (m == NULL) {
669                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
670                 }
671                 int32_t min = 0;
672                 int32_t max = Quantifier::MAX;
673                 switch (c) {
674                 case ONE_OR_MORE:
675                     min = 1;
676                     break;
677                 case ZERO_OR_ONE:
678                     min = 0;
679                     max = 1;
680                     break;
681                 // case KLEENE_STAR:
682                 //    do nothing -- min, max already set
683                 }
684                 m = new Quantifier(m, min, max);
685                 if (m == NULL) {
686                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
687                 }
688                 buf.truncate(qstart);
689                 buf.append(parser.generateStandInFor(m, status));
690             }
691             break;
692 
693         //------------------------------------------------------
694         // Elements allowed ONLY WITHIN segments
695         //------------------------------------------------------
696         case SEGMENT_CLOSE:
697             // assert(isSegment);
698             // We're done parsing a segment.
699             done = TRUE;
700             break;
701 
702         //------------------------------------------------------
703         // Elements allowed ONLY OUTSIDE segments
704         //------------------------------------------------------
705         case CONTEXT_ANTE:
706             if (ante >= 0) {
707                 return syntaxError(U_MULTIPLE_ANTE_CONTEXTS, rule, start, status);
708             }
709             ante = buf.length();
710             break;
711         case CONTEXT_POST:
712             if (post >= 0) {
713                 return syntaxError(U_MULTIPLE_POST_CONTEXTS, rule, start, status);
714             }
715             post = buf.length();
716             break;
717         case CURSOR_POS:
718             if (cursor >= 0) {
719                 return syntaxError(U_MULTIPLE_CURSORS, rule, start, status);
720             }
721             cursor = buf.length();
722             break;
723         case CURSOR_OFFSET:
724             if (cursorOffset < 0) {
725                 if (buf.length() > 0) {
726                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
727                 }
728                 --cursorOffset;
729             } else if (cursorOffset > 0) {
730                 if (buf.length() != cursorOffsetPos || cursor >= 0) {
731                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
732                 }
733                 ++cursorOffset;
734             } else {
735                 if (cursor == 0 && buf.length() == 0) {
736                     cursorOffset = -1;
737                 } else if (cursor < 0) {
738                     cursorOffsetPos = buf.length();
739                     cursorOffset = 1;
740                 } else {
741                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
742                 }
743             }
744             break;
745 
746 
747         //------------------------------------------------------
748         // Non-special characters
749         //------------------------------------------------------
750         default:
751             // Disallow unquoted characters other than [0-9A-Za-z]
752             // in the printable ASCII range.  These characters are
753             // reserved for possible future use.
754             if (c >= 0x0021 && c <= 0x007E &&
755                 !((c >= 0x0030/*'0'*/ && c <= 0x0039/*'9'*/) ||
756                   (c >= 0x0041/*'A'*/ && c <= 0x005A/*'Z'*/) ||
757                   (c >= 0x0061/*'a'*/ && c <= 0x007A/*'z'*/))) {
758                 return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
759             }
760             buf.append(c);
761             break;
762         }
763     }
764 
765     return pos;
766 }
767 
768 /**
769  * Remove context.
770  */
removeContext()771 void RuleHalf::removeContext() {
772     //text = text.substring(ante < 0 ? 0 : ante,
773     //                      post < 0 ? text.length() : post);
774     if (post >= 0) {
775         text.remove(post);
776     }
777     if (ante >= 0) {
778         text.removeBetween(0, ante);
779     }
780     ante = post = -1;
781     anchorStart = anchorEnd = FALSE;
782 }
783 
784 /**
785  * Return true if this half looks like valid output, that is, does not
786  * contain quantifiers or other special input-only elements.
787  */
isValidOutput(TransliteratorParser & transParser)788 UBool RuleHalf::isValidOutput(TransliteratorParser& transParser) {
789     for (int32_t i=0; i<text.length(); ) {
790         UChar32 c = text.char32At(i);
791         i += UTF_CHAR_LENGTH(c);
792         if (!transParser.parseData->isReplacer(c)) {
793             return FALSE;
794         }
795     }
796     return TRUE;
797 }
798 
799 /**
800  * Return true if this half looks like valid input, that is, does not
801  * contain functions or other special output-only elements.
802  */
isValidInput(TransliteratorParser & transParser)803 UBool RuleHalf::isValidInput(TransliteratorParser& transParser) {
804     for (int32_t i=0; i<text.length(); ) {
805         UChar32 c = text.char32At(i);
806         i += UTF_CHAR_LENGTH(c);
807         if (!transParser.parseData->isMatcher(c)) {
808             return FALSE;
809         }
810     }
811     return TRUE;
812 }
813 
814 //----------------------------------------------------------------------
815 // PUBLIC API
816 //----------------------------------------------------------------------
817 
818 /**
819  * Constructor.
820  */
TransliteratorParser(UErrorCode & statusReturn)821 TransliteratorParser::TransliteratorParser(UErrorCode &statusReturn) :
822 dataVector(statusReturn),
823 idBlockVector(statusReturn),
824 variablesVector(statusReturn),
825 segmentObjects(statusReturn)
826 {
827     idBlockVector.setDeleter(uhash_deleteUnicodeString);
828     curData = NULL;
829     compoundFilter = NULL;
830     parseData = NULL;
831     variableNames.setValueDeleter(uhash_deleteUnicodeString);
832 }
833 
834 /**
835  * Destructor.
836  */
~TransliteratorParser()837 TransliteratorParser::~TransliteratorParser() {
838     while (!dataVector.isEmpty())
839         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
840     delete compoundFilter;
841     delete parseData;
842     while (!variablesVector.isEmpty())
843         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
844 }
845 
846 void
parse(const UnicodeString & rules,UTransDirection transDirection,UParseError & pe,UErrorCode & ec)847 TransliteratorParser::parse(const UnicodeString& rules,
848                             UTransDirection transDirection,
849                             UParseError& pe,
850                             UErrorCode& ec) {
851     if (U_SUCCESS(ec)) {
852         parseRules(rules, transDirection, ec);
853         pe = parseError;
854     }
855 }
856 
857 /**
858  * Return the compound filter parsed by parse().  Caller owns result.
859  */
orphanCompoundFilter()860 UnicodeSet* TransliteratorParser::orphanCompoundFilter() {
861     UnicodeSet* f = compoundFilter;
862     compoundFilter = NULL;
863     return f;
864 }
865 
866 //----------------------------------------------------------------------
867 // Private implementation
868 //----------------------------------------------------------------------
869 
870 /**
871  * Parse the given string as a sequence of rules, separated by newline
872  * characters ('\n'), and cause this object to implement those rules.  Any
873  * previous rules are discarded.  Typically this method is called exactly
874  * once, during construction.
875  * @exception IllegalArgumentException if there is a syntax error in the
876  * rules
877  */
parseRules(const UnicodeString & rule,UTransDirection theDirection,UErrorCode & status)878 void TransliteratorParser::parseRules(const UnicodeString& rule,
879                                       UTransDirection theDirection,
880                                       UErrorCode& status)
881 {
882     // Clear error struct
883     uprv_memset(&parseError, 0, sizeof(parseError));
884     parseError.line = parseError.offset = -1;
885 
886     UBool parsingIDs = TRUE;
887     int32_t ruleCount = 0;
888 
889     while (!dataVector.isEmpty()) {
890         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
891     }
892     if (U_FAILURE(status)) {
893         return;
894     }
895 
896     idBlockVector.removeAllElements();
897     curData = NULL;
898     direction = theDirection;
899     ruleCount = 0;
900 
901     delete compoundFilter;
902     compoundFilter = NULL;
903 
904     while (!variablesVector.isEmpty()) {
905         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
906     }
907     variableNames.removeAll();
908     parseData = new ParseData(0, &variablesVector, &variableNames);
909     if (parseData == NULL) {
910         status = U_MEMORY_ALLOCATION_ERROR;
911         return;
912     }
913 
914     dotStandIn = (UChar) -1;
915 
916     UnicodeString *tempstr = NULL; // used for memory allocation error checking
917     UnicodeString str; // scratch
918     UnicodeString idBlockResult;
919     int32_t pos = 0;
920     int32_t limit = rule.length();
921 
922     // The compound filter offset is an index into idBlockResult.
923     // If it is 0, then the compound filter occurred at the start,
924     // and it is the offset to the _start_ of the compound filter
925     // pattern.  Otherwise it is the offset to the _limit_ of the
926     // compound filter pattern within idBlockResult.
927     compoundFilter = NULL;
928     int32_t compoundFilterOffset = -1;
929 
930     while (pos < limit && U_SUCCESS(status)) {
931         UChar c = rule.charAt(pos++);
932         if (uprv_isRuleWhiteSpace(c)) {
933             // Ignore leading whitespace.
934             continue;
935         }
936         // Skip lines starting with the comment character
937         if (c == RULE_COMMENT_CHAR) {
938             pos = rule.indexOf((UChar)0x000A /*\n*/, pos) + 1;
939             if (pos == 0) {
940                 break; // No "\n" found; rest of rule is a commnet
941             }
942             continue; // Either fall out or restart with next line
943         }
944 
945         // skip empty rules
946         if (c == END_OF_RULE)
947             continue;
948 
949         // keep track of how many rules we've seen
950         ++ruleCount;
951 
952         // We've found the start of a rule or ID.  c is its first
953         // character, and pos points past c.
954         --pos;
955         // Look for an ID token.  Must have at least ID_TOKEN_LEN + 1
956         // chars left.
957         if ((pos + ID_TOKEN_LEN + 1) <= limit &&
958                 rule.compare(pos, ID_TOKEN_LEN, ID_TOKEN) == 0) {
959             pos += ID_TOKEN_LEN;
960             c = rule.charAt(pos);
961             while (uprv_isRuleWhiteSpace(c) && pos < limit) {
962                 ++pos;
963                 c = rule.charAt(pos);
964             }
965 
966             int32_t p = pos;
967 
968             if (!parsingIDs) {
969                 if (curData != NULL) {
970                     if (direction == UTRANS_FORWARD)
971                         dataVector.addElement(curData, status);
972                     else
973                         dataVector.insertElementAt(curData, 0, status);
974                     curData = NULL;
975                 }
976                 parsingIDs = TRUE;
977             }
978 
979             TransliteratorIDParser::SingleID* id =
980                 TransliteratorIDParser::parseSingleID(rule, p, direction, status);
981             if (p != pos && ICU_Utility::parseChar(rule, p, END_OF_RULE)) {
982                 // Successful ::ID parse.
983 
984                 if (direction == UTRANS_FORWARD) {
985                     idBlockResult.append(id->canonID).append(END_OF_RULE);
986                 } else {
987                     idBlockResult.insert(0, END_OF_RULE);
988                     idBlockResult.insert(0, id->canonID);
989                 }
990 
991             } else {
992                 // Couldn't parse an ID.  Try to parse a global filter
993                 int32_t withParens = -1;
994                 UnicodeSet* f = TransliteratorIDParser::parseGlobalFilter(rule, p, direction, withParens, NULL);
995                 if (f != NULL) {
996                     if (ICU_Utility::parseChar(rule, p, END_OF_RULE)
997                         && (direction == UTRANS_FORWARD) == (withParens == 0))
998                     {
999                         if (compoundFilter != NULL) {
1000                             // Multiple compound filters
1001                             syntaxError(U_MULTIPLE_COMPOUND_FILTERS, rule, pos, status);
1002                             delete f;
1003                         } else {
1004                             compoundFilter = f;
1005                             compoundFilterOffset = ruleCount;
1006                         }
1007                     } else {
1008                         delete f;
1009                     }
1010                 } else {
1011                     // Invalid ::id
1012                     // Can be parsed as neither an ID nor a global filter
1013                     syntaxError(U_INVALID_ID, rule, pos, status);
1014                 }
1015             }
1016             delete id;
1017             pos = p;
1018         } else {
1019             if (parsingIDs) {
1020                 tempstr = new UnicodeString(idBlockResult);
1021                 // NULL pointer check
1022                 if (tempstr == NULL) {
1023                     status = U_MEMORY_ALLOCATION_ERROR;
1024                     return;
1025                 }
1026                 if (direction == UTRANS_FORWARD)
1027                     idBlockVector.addElement(tempstr, status);
1028                 else
1029                     idBlockVector.insertElementAt(tempstr, 0, status);
1030                 idBlockResult.remove();
1031                 parsingIDs = FALSE;
1032                 curData = new TransliterationRuleData(status);
1033                 // NULL pointer check
1034                 if (curData == NULL) {
1035                     status = U_MEMORY_ALLOCATION_ERROR;
1036                     return;
1037                 }
1038                 parseData->data = curData;
1039 
1040                 // By default, rules use part of the private use area
1041                 // E000..F8FF for variables and other stand-ins.  Currently
1042                 // the range F000..F8FF is typically sufficient.  The 'use
1043                 // variable range' pragma allows rule sets to modify this.
1044                 setVariableRange(0xF000, 0xF8FF, status);
1045             }
1046 
1047             if (resemblesPragma(rule, pos, limit)) {
1048                 int32_t ppp = parsePragma(rule, pos, limit, status);
1049                 if (ppp < 0) {
1050                     syntaxError(U_MALFORMED_PRAGMA, rule, pos, status);
1051                 }
1052                 pos = ppp;
1053             // Parse a rule
1054             } else {
1055                 pos = parseRule(rule, pos, limit, status);
1056             }
1057         }
1058     }
1059 
1060     if (parsingIDs && idBlockResult.length() > 0) {
1061         tempstr = new UnicodeString(idBlockResult);
1062         // NULL pointer check
1063         if (tempstr == NULL) {
1064             status = U_MEMORY_ALLOCATION_ERROR;
1065             return;
1066         }
1067         if (direction == UTRANS_FORWARD)
1068             idBlockVector.addElement(tempstr, status);
1069         else
1070             idBlockVector.insertElementAt(tempstr, 0, status);
1071     }
1072     else if (!parsingIDs && curData != NULL) {
1073         if (direction == UTRANS_FORWARD)
1074             dataVector.addElement(curData, status);
1075         else
1076             dataVector.insertElementAt(curData, 0, status);
1077     }
1078 
1079     if (U_SUCCESS(status)) {
1080         // Convert the set vector to an array
1081         int32_t i, dataVectorSize = dataVector.size();
1082         for (i = 0; i < dataVectorSize; i++) {
1083             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1084             data->variablesLength = variablesVector.size();
1085             if (data->variablesLength == 0) {
1086                 data->variables = 0;
1087             } else {
1088                 data->variables = (UnicodeFunctor**)uprv_malloc(data->variablesLength * sizeof(UnicodeFunctor*));
1089                 // NULL pointer check
1090                 if (data->variables == NULL) {
1091                     status = U_MEMORY_ALLOCATION_ERROR;
1092                     return;
1093                 }
1094                 data->variablesAreOwned = (i == 0);
1095             }
1096 
1097             for (int32_t j = 0; j < data->variablesLength; j++) {
1098                 data->variables[j] =
1099                     ((UnicodeSet*)variablesVector.elementAt(j));
1100             }
1101 
1102             data->variableNames.removeAll();
1103             int32_t pos = -1;
1104             const UHashElement* he = variableNames.nextElement(pos);
1105             while (he != NULL) {
1106                 UnicodeString* tempus = (UnicodeString*)(((UnicodeString*)(he->value.pointer))->clone());
1107                 if (tempus == NULL) {
1108                     status = U_MEMORY_ALLOCATION_ERROR;
1109                     return;
1110                 }
1111                 data->variableNames.put(*((UnicodeString*)(he->key.pointer)),
1112                     tempus, status);
1113                 he = variableNames.nextElement(pos);
1114             }
1115         }
1116         variablesVector.removeAllElements();   // keeps them from getting deleted when we succeed
1117 
1118         // Index the rules
1119         if (compoundFilter != NULL) {
1120             if ((direction == UTRANS_FORWARD && compoundFilterOffset != 1) ||
1121                 (direction == UTRANS_REVERSE && compoundFilterOffset != ruleCount)) {
1122                 status = U_MISPLACED_COMPOUND_FILTER;
1123             }
1124         }
1125 
1126         for (i = 0; i < dataVectorSize; i++) {
1127             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1128             data->ruleSet.freeze(parseError, status);
1129         }
1130         if (idBlockVector.size() == 1 && ((UnicodeString*)idBlockVector.elementAt(0))->isEmpty()) {
1131             idBlockVector.removeElementAt(0);
1132         }
1133     }
1134 }
1135 
1136 /**
1137  * Set the variable range to [start, end] (inclusive).
1138  */
setVariableRange(int32_t start,int32_t end,UErrorCode & status)1139 void TransliteratorParser::setVariableRange(int32_t start, int32_t end, UErrorCode& status) {
1140     if (start > end || start < 0 || end > 0xFFFF) {
1141         status = U_MALFORMED_PRAGMA;
1142         return;
1143     }
1144 
1145     curData->variablesBase = (UChar) start;
1146     if (dataVector.size() == 0) {
1147         variableNext = (UChar) start;
1148         variableLimit = (UChar) (end + 1);
1149     }
1150 }
1151 
1152 /**
1153  * Assert that the given character is NOT within the variable range.
1154  * If it is, return FALSE.  This is neccesary to ensure that the
1155  * variable range does not overlap characters used in a rule.
1156  */
checkVariableRange(UChar32 ch) const1157 UBool TransliteratorParser::checkVariableRange(UChar32 ch) const {
1158     return !(ch >= curData->variablesBase && ch < variableLimit);
1159 }
1160 
1161 /**
1162  * Set the maximum backup to 'backup', in response to a pragma
1163  * statement.
1164  */
pragmaMaximumBackup(int32_t)1165 void TransliteratorParser::pragmaMaximumBackup(int32_t /*backup*/) {
1166     //TODO Finish
1167 }
1168 
1169 /**
1170  * Begin normalizing all rules using the given mode, in response
1171  * to a pragma statement.
1172  */
pragmaNormalizeRules(UNormalizationMode)1173 void TransliteratorParser::pragmaNormalizeRules(UNormalizationMode /*mode*/) {
1174     //TODO Finish
1175 }
1176 
1177 static const UChar PRAGMA_USE[] = {0x75,0x73,0x65,0x20,0}; // "use "
1178 
1179 static const UChar PRAGMA_VARIABLE_RANGE[] = {0x7E,0x76,0x61,0x72,0x69,0x61,0x62,0x6C,0x65,0x20,0x72,0x61,0x6E,0x67,0x65,0x20,0x23,0x20,0x23,0x7E,0x3B,0}; // "~variable range # #~;"
1180 
1181 static const UChar PRAGMA_MAXIMUM_BACKUP[] = {0x7E,0x6D,0x61,0x78,0x69,0x6D,0x75,0x6D,0x20,0x62,0x61,0x63,0x6B,0x75,0x70,0x20,0x23,0x7E,0x3B,0}; // "~maximum backup #~;"
1182 
1183 static const UChar PRAGMA_NFD_RULES[] = {0x7E,0x6E,0x66,0x64,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfd rules~;"
1184 
1185 static const UChar PRAGMA_NFC_RULES[] = {0x7E,0x6E,0x66,0x63,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfc rules~;"
1186 
1187 /**
1188  * Return true if the given rule looks like a pragma.
1189  * @param pos offset to the first non-whitespace character
1190  * of the rule.
1191  * @param limit pointer past the last character of the rule.
1192  */
resemblesPragma(const UnicodeString & rule,int32_t pos,int32_t limit)1193 UBool TransliteratorParser::resemblesPragma(const UnicodeString& rule, int32_t pos, int32_t limit) {
1194     // Must start with /use\s/i
1195     return ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_USE, NULL) >= 0;
1196 }
1197 
1198 /**
1199  * Parse a pragma.  This method assumes resemblesPragma() has
1200  * already returned true.
1201  * @param pos offset to the first non-whitespace character
1202  * of the rule.
1203  * @param limit pointer past the last character of the rule.
1204  * @return the position index after the final ';' of the pragma,
1205  * or -1 on failure.
1206  */
parsePragma(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1207 int32_t TransliteratorParser::parsePragma(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1208     int32_t array[2];
1209 
1210     // resemblesPragma() has already returned true, so we
1211     // know that pos points to /use\s/i; we can skip 4 characters
1212     // immediately
1213     pos += 4;
1214 
1215     // Here are the pragmas we recognize:
1216     // use variable range 0xE000 0xEFFF;
1217     // use maximum backup 16;
1218     // use nfd rules;
1219     // use nfc rules;
1220     int p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_VARIABLE_RANGE, array);
1221     if (p >= 0) {
1222         setVariableRange(array[0], array[1], status);
1223         return p;
1224     }
1225 
1226     p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_MAXIMUM_BACKUP, array);
1227     if (p >= 0) {
1228         pragmaMaximumBackup(array[0]);
1229         return p;
1230     }
1231 
1232     p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_NFD_RULES, NULL);
1233     if (p >= 0) {
1234         pragmaNormalizeRules(UNORM_NFD);
1235         return p;
1236     }
1237 
1238     p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_NFC_RULES, NULL);
1239     if (p >= 0) {
1240         pragmaNormalizeRules(UNORM_NFC);
1241         return p;
1242     }
1243 
1244     // Syntax error: unable to parse pragma
1245     return -1;
1246 }
1247 
1248 /**
1249  * MAIN PARSER.  Parse the next rule in the given rule string, starting
1250  * at pos.  Return the index after the last character parsed.  Do not
1251  * parse characters at or after limit.
1252  *
1253  * Important:  The character at pos must be a non-whitespace character
1254  * that is not the comment character.
1255  *
1256  * This method handles quoting, escaping, and whitespace removal.  It
1257  * parses the end-of-rule character.  It recognizes context and cursor
1258  * indicators.  Once it does a lexical breakdown of the rule at pos, it
1259  * creates a rule object and adds it to our rule list.
1260  */
parseRule(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1261 int32_t TransliteratorParser::parseRule(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1262     // Locate the left side, operator, and right side
1263     int32_t start = pos;
1264     UChar op = 0;
1265     int32_t i;
1266 
1267     // Set up segments data
1268     segmentStandins.truncate(0);
1269     segmentObjects.removeAllElements();
1270 
1271     // Use pointers to automatics to make swapping possible.
1272     RuleHalf _left(*this), _right(*this);
1273     RuleHalf* left = &_left;
1274     RuleHalf* right = &_right;
1275 
1276     undefinedVariableName.remove();
1277     pos = left->parse(rule, pos, limit, status);
1278     if (U_FAILURE(status)) {
1279         return start;
1280     }
1281 
1282     if (pos == limit || u_strchr(gOPERATORS, (op = rule.charAt(--pos))) == NULL) {
1283         return syntaxError(U_MISSING_OPERATOR, rule, start, status);
1284     }
1285     ++pos;
1286 
1287     // Found an operator char.  Check for forward-reverse operator.
1288     if (op == REVERSE_RULE_OP &&
1289         (pos < limit && rule.charAt(pos) == FORWARD_RULE_OP)) {
1290         ++pos;
1291         op = FWDREV_RULE_OP;
1292     }
1293 
1294     // Translate alternate op characters.
1295     switch (op) {
1296     case ALT_FORWARD_RULE_OP:
1297         op = FORWARD_RULE_OP;
1298         break;
1299     case ALT_REVERSE_RULE_OP:
1300         op = REVERSE_RULE_OP;
1301         break;
1302     case ALT_FWDREV_RULE_OP:
1303         op = FWDREV_RULE_OP;
1304         break;
1305     }
1306 
1307     pos = right->parse(rule, pos, limit, status);
1308     if (U_FAILURE(status)) {
1309         return start;
1310     }
1311 
1312     if (pos < limit) {
1313         if (rule.charAt(--pos) == END_OF_RULE) {
1314             ++pos;
1315         } else {
1316             // RuleHalf parser must have terminated at an operator
1317             return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
1318         }
1319     }
1320 
1321     if (op == VARIABLE_DEF_OP) {
1322         // LHS is the name.  RHS is a single character, either a literal
1323         // or a set (already parsed).  If RHS is longer than one
1324         // character, it is either a multi-character string, or multiple
1325         // sets, or a mixture of chars and sets -- syntax error.
1326 
1327         // We expect to see a single undefined variable (the one being
1328         // defined).
1329         if (undefinedVariableName.length() == 0) {
1330             // "Missing '$' or duplicate definition"
1331             return syntaxError(U_BAD_VARIABLE_DEFINITION, rule, start, status);
1332         }
1333         if (left->text.length() != 1 || left->text.charAt(0) != variableLimit) {
1334             // "Malformed LHS"
1335             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1336         }
1337         if (left->anchorStart || left->anchorEnd ||
1338             right->anchorStart || right->anchorEnd) {
1339             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1340         }
1341         // We allow anything on the right, including an empty string.
1342         UnicodeString* value = new UnicodeString(right->text);
1343         // NULL pointer check
1344         if (value == NULL) {
1345             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1346         }
1347         variableNames.put(undefinedVariableName, value, status);
1348         ++variableLimit;
1349         return pos;
1350     }
1351 
1352     // If this is not a variable definition rule, we shouldn't have
1353     // any undefined variable names.
1354     if (undefinedVariableName.length() != 0) {
1355         return syntaxError(// "Undefined variable $" + undefinedVariableName,
1356                     U_UNDEFINED_VARIABLE,
1357                     rule, start, status);
1358     }
1359 
1360     // Verify segments
1361     if (segmentStandins.length() > segmentObjects.size()) {
1362         syntaxError(U_UNDEFINED_SEGMENT_REFERENCE, rule, start, status);
1363     }
1364     for (i=0; i<segmentStandins.length(); ++i) {
1365         if (segmentStandins.charAt(i) == 0) {
1366             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1367         }
1368     }
1369     for (i=0; i<segmentObjects.size(); ++i) {
1370         if (segmentObjects.elementAt(i) == NULL) {
1371             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1372         }
1373     }
1374 
1375     // If the direction we want doesn't match the rule
1376     // direction, do nothing.
1377     if (op != FWDREV_RULE_OP &&
1378         ((direction == UTRANS_FORWARD) != (op == FORWARD_RULE_OP))) {
1379         return pos;
1380     }
1381 
1382     // Transform the rule into a forward rule by swapping the
1383     // sides if necessary.
1384     if (direction == UTRANS_REVERSE) {
1385         left = &_right;
1386         right = &_left;
1387     }
1388 
1389     // Remove non-applicable elements in forward-reverse
1390     // rules.  Bidirectional rules ignore elements that do not
1391     // apply.
1392     if (op == FWDREV_RULE_OP) {
1393         right->removeContext();
1394         left->cursor = -1;
1395         left->cursorOffset = 0;
1396     }
1397 
1398     // Normalize context
1399     if (left->ante < 0) {
1400         left->ante = 0;
1401     }
1402     if (left->post < 0) {
1403         left->post = left->text.length();
1404     }
1405 
1406     // Context is only allowed on the input side.  Cursors are only
1407     // allowed on the output side.  Segment delimiters can only appear
1408     // on the left, and references on the right.  Cursor offset
1409     // cannot appear without an explicit cursor.  Cursor offset
1410     // cannot place the cursor outside the limits of the context.
1411     // Anchors are only allowed on the input side.
1412     if (right->ante >= 0 || right->post >= 0 || left->cursor >= 0 ||
1413         (right->cursorOffset != 0 && right->cursor < 0) ||
1414         // - The following two checks were used to ensure that the
1415         // - the cursor offset stayed within the ante- or postcontext.
1416         // - However, with the addition of quantifiers, we have to
1417         // - allow arbitrary cursor offsets and do runtime checking.
1418         //(right->cursorOffset > (left->text.length() - left->post)) ||
1419         //(-right->cursorOffset > left->ante) ||
1420         right->anchorStart || right->anchorEnd ||
1421         !left->isValidInput(*this) || !right->isValidOutput(*this) ||
1422         left->ante > left->post) {
1423 
1424         return syntaxError(U_MALFORMED_RULE, rule, start, status);
1425     }
1426 
1427     // Flatten segment objects vector to an array
1428     UnicodeFunctor** segmentsArray = NULL;
1429     if (segmentObjects.size() > 0) {
1430         segmentsArray = (UnicodeFunctor **)uprv_malloc(segmentObjects.size() * sizeof(UnicodeFunctor *));
1431         // Null pointer check
1432         if (segmentsArray == NULL) {
1433             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1434         }
1435         segmentObjects.toArray((void**) segmentsArray);
1436     }
1437     TransliterationRule* temptr = new TransliterationRule(
1438             left->text, left->ante, left->post,
1439             right->text, right->cursor, right->cursorOffset,
1440             segmentsArray,
1441             segmentObjects.size(),
1442             left->anchorStart, left->anchorEnd,
1443             curData,
1444             status);
1445     //Null pointer check
1446     if (temptr == NULL) {
1447         uprv_free(segmentsArray);
1448         return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1449     }
1450 
1451     curData->ruleSet.addRule(temptr, status);
1452 
1453     return pos;
1454 }
1455 
1456 /**
1457  * Called by main parser upon syntax error.  Search the rule string
1458  * for the probable end of the rule.  Of course, if the error is that
1459  * the end of rule marker is missing, then the rule end will not be found.
1460  * In any case the rule start will be correctly reported.
1461  * @param msg error description
1462  * @param rule pattern string
1463  * @param start position of first character of current rule
1464  */
syntaxError(UErrorCode parseErrorCode,const UnicodeString & rule,int32_t pos,UErrorCode & status)1465 int32_t TransliteratorParser::syntaxError(UErrorCode parseErrorCode,
1466                                           const UnicodeString& rule,
1467                                           int32_t pos,
1468                                           UErrorCode& status)
1469 {
1470     parseError.offset = pos;
1471     parseError.line = 0 ; /* we are not using line numbers */
1472 
1473     // for pre-context
1474     const int32_t LEN = U_PARSE_CONTEXT_LEN - 1;
1475     int32_t start = uprv_max(pos - LEN, 0);
1476     int32_t stop  = pos;
1477 
1478     rule.extract(start,stop-start,parseError.preContext);
1479     //null terminate the buffer
1480     parseError.preContext[stop-start] = 0;
1481 
1482     //for post-context
1483     start = pos;
1484     stop  = uprv_min(pos + LEN, rule.length());
1485 
1486     rule.extract(start,stop-start,parseError.postContext);
1487     //null terminate the buffer
1488     parseError.postContext[stop-start]= 0;
1489 
1490     status = (UErrorCode)parseErrorCode;
1491     return pos;
1492 
1493 }
1494 
1495 /**
1496  * Parse a UnicodeSet out, store it, and return the stand-in character
1497  * used to represent it.
1498  */
parseSet(const UnicodeString & rule,ParsePosition & pos,UErrorCode & status)1499 UChar TransliteratorParser::parseSet(const UnicodeString& rule,
1500                                           ParsePosition& pos,
1501                                           UErrorCode& status) {
1502     UnicodeSet* set = new UnicodeSet(rule, pos, USET_IGNORE_SPACE, parseData, status);
1503     // Null pointer check
1504     if (set == NULL) {
1505         status = U_MEMORY_ALLOCATION_ERROR;
1506         return (UChar)0x0000; // Return empty character with error.
1507     }
1508     set->compact();
1509     return generateStandInFor(set, status);
1510 }
1511 
1512 /**
1513  * Generate and return a stand-in for a new UnicodeFunctor.  Store
1514  * the matcher (adopt it).
1515  */
generateStandInFor(UnicodeFunctor * adopted,UErrorCode & status)1516 UChar TransliteratorParser::generateStandInFor(UnicodeFunctor* adopted, UErrorCode& status) {
1517     // assert(obj != null);
1518 
1519     // Look up previous stand-in, if any.  This is a short list
1520     // (typical n is 0, 1, or 2); linear search is optimal.
1521     for (int32_t i=0; i<variablesVector.size(); ++i) {
1522         if (variablesVector.elementAt(i) == adopted) { // [sic] pointer comparison
1523             return (UChar) (curData->variablesBase + i);
1524         }
1525     }
1526 
1527     if (variableNext >= variableLimit) {
1528         delete adopted;
1529         status = U_VARIABLE_RANGE_EXHAUSTED;
1530         return 0;
1531     }
1532     variablesVector.addElement(adopted, status);
1533     return variableNext++;
1534 }
1535 
1536 /**
1537  * Return the standin for segment seg (1-based).
1538  */
getSegmentStandin(int32_t seg,UErrorCode & status)1539 UChar TransliteratorParser::getSegmentStandin(int32_t seg, UErrorCode& status) {
1540     // Special character used to indicate an empty spot
1541     UChar empty = curData->variablesBase - 1;
1542     while (segmentStandins.length() < seg) {
1543         segmentStandins.append(empty);
1544     }
1545     UChar c = segmentStandins.charAt(seg-1);
1546     if (c == empty) {
1547         if (variableNext >= variableLimit) {
1548             status = U_VARIABLE_RANGE_EXHAUSTED;
1549             return 0;
1550         }
1551         c = variableNext++;
1552         // Set a placeholder in the master variables vector that will be
1553         // filled in later by setSegmentObject().  We know that we will get
1554         // called first because setSegmentObject() will call us.
1555         variablesVector.addElement((void*) NULL, status);
1556         segmentStandins.setCharAt(seg-1, c);
1557     }
1558     return c;
1559 }
1560 
1561 /**
1562  * Set the object for segment seg (1-based).
1563  */
setSegmentObject(int32_t seg,StringMatcher * adopted,UErrorCode & status)1564 void TransliteratorParser::setSegmentObject(int32_t seg, StringMatcher* adopted, UErrorCode& status) {
1565     // Since we call parseSection() recursively, nested
1566     // segments will result in segment i+1 getting parsed
1567     // and stored before segment i; be careful with the
1568     // vector handling here.
1569     if (segmentObjects.size() < seg) {
1570         segmentObjects.setSize(seg, status);
1571     }
1572     int32_t index = getSegmentStandin(seg, status) - curData->variablesBase;
1573     if (segmentObjects.elementAt(seg-1) != NULL ||
1574         variablesVector.elementAt(index) != NULL) {
1575         // should never happen
1576         status = U_INTERNAL_TRANSLITERATOR_ERROR;
1577         return;
1578     }
1579     segmentObjects.setElementAt(adopted, seg-1);
1580     variablesVector.setElementAt(adopted, index);
1581 }
1582 
1583 /**
1584  * Return the stand-in for the dot set.  It is allocated the first
1585  * time and reused thereafter.
1586  */
getDotStandIn(UErrorCode & status)1587 UChar TransliteratorParser::getDotStandIn(UErrorCode& status) {
1588     if (dotStandIn == (UChar) -1) {
1589         UnicodeSet* tempus = new UnicodeSet(DOT_SET, status);
1590         // Null pointer check.
1591         if (tempus == NULL) {
1592             status = U_MEMORY_ALLOCATION_ERROR;
1593             return (UChar)0x0000;
1594         }
1595         dotStandIn = generateStandInFor(tempus, status);
1596     }
1597     return dotStandIn;
1598 }
1599 
1600 /**
1601  * Append the value of the given variable name to the given
1602  * UnicodeString.
1603  */
appendVariableDef(const UnicodeString & name,UnicodeString & buf,UErrorCode & status)1604 void TransliteratorParser::appendVariableDef(const UnicodeString& name,
1605                                                   UnicodeString& buf,
1606                                                   UErrorCode& status) {
1607     const UnicodeString* s = (const UnicodeString*) variableNames.get(name);
1608     if (s == NULL) {
1609         // We allow one undefined variable so that variable definition
1610         // statements work.  For the first undefined variable we return
1611         // the special placeholder variableLimit-1, and save the variable
1612         // name.
1613         if (undefinedVariableName.length() == 0) {
1614             undefinedVariableName = name;
1615             if (variableNext >= variableLimit) {
1616                 // throw new RuntimeException("Private use variables exhausted");
1617                 status = U_ILLEGAL_ARGUMENT_ERROR;
1618                 return;
1619             }
1620             buf.append((UChar) --variableLimit);
1621         } else {
1622             //throw new IllegalArgumentException("Undefined variable $"
1623             //                                   + name);
1624             status = U_ILLEGAL_ARGUMENT_ERROR;
1625             return;
1626         }
1627     } else {
1628         buf.append(*s);
1629     }
1630 }
1631 
1632 /**
1633  * Glue method to get around access restrictions in C++.
1634  */
1635 /*Transliterator* TransliteratorParser::createBasicInstance(const UnicodeString& id, const UnicodeString* canonID) {
1636     return Transliterator::createBasicInstance(id, canonID);
1637 }*/
1638 
1639 U_NAMESPACE_END
1640 
1641 U_CAPI int32_t
utrans_stripRules(const UChar * source,int32_t sourceLen,UChar * target,UErrorCode * status)1642 utrans_stripRules(const UChar *source, int32_t sourceLen, UChar *target, UErrorCode *status) {
1643     U_NAMESPACE_USE
1644 
1645     //const UChar *sourceStart = source;
1646     const UChar *targetStart = target;
1647     const UChar *sourceLimit = source+sourceLen;
1648     UChar *targetLimit = target+sourceLen;
1649     UChar32 c = 0;
1650     UBool quoted = FALSE;
1651     int32_t index;
1652 
1653     uprv_memset(target, 0, sourceLen*U_SIZEOF_UCHAR);
1654 
1655     /* read the rules into the buffer */
1656     while (source < sourceLimit)
1657     {
1658         index=0;
1659         U16_NEXT_UNSAFE(source, index, c);
1660         source+=index;
1661         if(c == QUOTE) {
1662             quoted = (UBool)!quoted;
1663         }
1664         else if (!quoted) {
1665             if (c == RULE_COMMENT_CHAR) {
1666                 /* skip comments and all preceding spaces */
1667                 while (targetStart < target && *(target - 1) == 0x0020) {
1668                     target--;
1669                 }
1670                 do {
1671                     c = *(source++);
1672                 }
1673                 while (c != CR && c != LF);
1674             }
1675             else if (c == ESCAPE) {
1676                 UChar32   c2 = *source;
1677                 if (c2 == CR || c2 == LF) {
1678                     /* A backslash at the end of a line. */
1679                     /* Since we're stripping lines, ignore the backslash. */
1680                     source++;
1681                     continue;
1682                 }
1683                 if (c2 == 0x0075 && source+5 < sourceLimit) { /* \u seen. \U isn't unescaped. */
1684                     int32_t escapeOffset = 0;
1685                     UnicodeString escapedStr(source, 5);
1686                     c2 = escapedStr.unescapeAt(escapeOffset);
1687 
1688                     if (c2 == (UChar32)0xFFFFFFFF || escapeOffset == 0)
1689                     {
1690                         *status = U_PARSE_ERROR;
1691                         return 0;
1692                     }
1693                     if (!uprv_isRuleWhiteSpace(c2) && !u_iscntrl(c2) && !u_ispunct(c2)) {
1694                         /* It was escaped for a reason. Write what it was suppose to be. */
1695                         source+=5;
1696                         c = c2;
1697                     }
1698                 }
1699                 else if (c2 == QUOTE) {
1700                     /* \' seen. Make sure we don't do anything when we see it again. */
1701                     quoted = (UBool)!quoted;
1702                 }
1703             }
1704         }
1705         if (c == CR || c == LF)
1706         {
1707             /* ignore spaces carriage returns, and all leading spaces on the next line.
1708             * and line feed unless in the form \uXXXX
1709             */
1710             quoted = FALSE;
1711             while (source < sourceLimit) {
1712                 c = *(source);
1713                 if (c != CR && c != LF && c != 0x0020) {
1714                     break;
1715                 }
1716                 source++;
1717             }
1718             continue;
1719         }
1720 
1721         /* Append UChar * after dissembling if c > 0xffff*/
1722         index=0;
1723         U16_APPEND_UNSAFE(target, index, c);
1724         target+=index;
1725     }
1726     if (target < targetLimit) {
1727         *target = 0;
1728     }
1729     return (int32_t)(target-targetStart);
1730 }
1731 
1732 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
1733