• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  **********************************************************************
3  *   Copyright (C) 1999-2011, 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 "patternprops.h"
37 #include "util.h"
38 #include "cmemory.h"
39 #include "uprops.h"
40 #include "putilimp.h"
41 
42 // Operators
43 #define VARIABLE_DEF_OP ((UChar)0x003D) /*=*/
44 #define FORWARD_RULE_OP ((UChar)0x003E) /*>*/
45 #define REVERSE_RULE_OP ((UChar)0x003C) /*<*/
46 #define FWDREV_RULE_OP  ((UChar)0x007E) /*~*/ // internal rep of <> op
47 
48 // Other special characters
49 #define QUOTE             ((UChar)0x0027) /*'*/
50 #define ESCAPE            ((UChar)0x005C) /*\*/
51 #define END_OF_RULE       ((UChar)0x003B) /*;*/
52 #define RULE_COMMENT_CHAR ((UChar)0x0023) /*#*/
53 
54 #define SEGMENT_OPEN       ((UChar)0x0028) /*(*/
55 #define SEGMENT_CLOSE      ((UChar)0x0029) /*)*/
56 #define CONTEXT_ANTE       ((UChar)0x007B) /*{*/
57 #define CONTEXT_POST       ((UChar)0x007D) /*}*/
58 #define CURSOR_POS         ((UChar)0x007C) /*|*/
59 #define CURSOR_OFFSET      ((UChar)0x0040) /*@*/
60 #define ANCHOR_START       ((UChar)0x005E) /*^*/
61 #define KLEENE_STAR        ((UChar)0x002A) /***/
62 #define ONE_OR_MORE        ((UChar)0x002B) /*+*/
63 #define ZERO_OR_ONE        ((UChar)0x003F) /*?*/
64 
65 #define DOT                ((UChar)46)     /*.*/
66 
67 static const UChar DOT_SET[] = { // "[^[:Zp:][:Zl:]\r\n$]";
68     91, 94, 91, 58, 90, 112, 58, 93, 91, 58, 90,
69     108, 58, 93, 92, 114, 92, 110, 36, 93, 0
70 };
71 
72 // A function is denoted &Source-Target/Variant(text)
73 #define FUNCTION           ((UChar)38)     /*&*/
74 
75 // Aliases for some of the syntax characters. These are provided so
76 // transliteration rules can be expressed in XML without clashing with
77 // XML syntax characters '<', '>', and '&'.
78 #define ALT_REVERSE_RULE_OP ((UChar)0x2190) // Left Arrow
79 #define ALT_FORWARD_RULE_OP ((UChar)0x2192) // Right Arrow
80 #define ALT_FWDREV_RULE_OP  ((UChar)0x2194) // Left Right Arrow
81 #define ALT_FUNCTION        ((UChar)0x2206) // Increment (~Greek Capital Delta)
82 
83 // Special characters disallowed at the top level
84 static const UChar ILLEGAL_TOP[] = {41,0}; // ")"
85 
86 // Special characters disallowed within a segment
87 static const UChar ILLEGAL_SEG[] = {123,125,124,64,0}; // "{}|@"
88 
89 // Special characters disallowed within a function argument
90 static const UChar ILLEGAL_FUNC[] = {94,40,46,42,43,63,123,125,124,64,0}; // "^(.*+?{}|@"
91 
92 // By definition, the ANCHOR_END special character is a
93 // trailing SymbolTable.SYMBOL_REF character.
94 // private static final char ANCHOR_END       = '$';
95 
96 static const UChar gOPERATORS[] = { // "=><"
97     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
98     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
99     0
100 };
101 
102 static const UChar HALF_ENDERS[] = { // "=><;"
103     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
104     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
105     END_OF_RULE,
106     0
107 };
108 
109 // These are also used in Transliterator::toRules()
110 static const int32_t ID_TOKEN_LEN = 2;
111 static const UChar   ID_TOKEN[]   = { 0x3A, 0x3A }; // ':', ':'
112 
113 /*
114 commented out until we do real ::BEGIN/::END functionality
115 static const int32_t BEGIN_TOKEN_LEN = 5;
116 static const UChar BEGIN_TOKEN[] = { 0x42, 0x45, 0x47, 0x49, 0x4e }; // 'BEGIN'
117 
118 static const int32_t END_TOKEN_LEN = 3;
119 static const UChar END_TOKEN[] = { 0x45, 0x4e, 0x44 }; // 'END'
120 */
121 
122 U_NAMESPACE_BEGIN
123 
124 //----------------------------------------------------------------------
125 // BEGIN ParseData
126 //----------------------------------------------------------------------
127 
128 /**
129  * This class implements the SymbolTable interface.  It is used
130  * during parsing to give UnicodeSet access to variables that
131  * have been defined so far.  Note that it uses variablesVector,
132  * _not_ data.setVariables.
133  */
134 class ParseData : public UMemory, public SymbolTable {
135 public:
136     const TransliterationRuleData* data; // alias
137 
138     const UVector* variablesVector; // alias
139 
140     const Hashtable* variableNames; // alias
141 
142     ParseData(const TransliterationRuleData* data = 0,
143               const UVector* variablesVector = 0,
144               const Hashtable* variableNames = 0);
145 
146     virtual const UnicodeString* lookup(const UnicodeString& s) const;
147 
148     virtual const UnicodeFunctor* lookupMatcher(UChar32 ch) const;
149 
150     virtual UnicodeString parseReference(const UnicodeString& text,
151                                          ParsePosition& pos, int32_t limit) const;
152     /**
153      * Return true if the given character is a matcher standin or a plain
154      * character (non standin).
155      */
156     UBool isMatcher(UChar32 ch);
157 
158     /**
159      * Return true if the given character is a replacer standin or a plain
160      * character (non standin).
161      */
162     UBool isReplacer(UChar32 ch);
163 
164 private:
165     ParseData(const ParseData &other); // forbid copying of this class
166     ParseData &operator=(const ParseData &other); // forbid copying of this class
167 };
168 
ParseData(const TransliterationRuleData * d,const UVector * sets,const Hashtable * vNames)169 ParseData::ParseData(const TransliterationRuleData* d,
170                      const UVector* sets,
171                      const Hashtable* vNames) :
172     data(d), variablesVector(sets), variableNames(vNames) {}
173 
174 /**
175  * Implement SymbolTable API.
176  */
lookup(const UnicodeString & name) const177 const UnicodeString* ParseData::lookup(const UnicodeString& name) const {
178     return (const UnicodeString*) variableNames->get(name);
179 }
180 
181 /**
182  * Implement SymbolTable API.
183  */
lookupMatcher(UChar32 ch) const184 const UnicodeFunctor* ParseData::lookupMatcher(UChar32 ch) const {
185     // Note that we cannot use data.lookupSet() because the
186     // set array has not been constructed yet.
187     const UnicodeFunctor* set = NULL;
188     int32_t i = ch - data->variablesBase;
189     if (i >= 0 && i < variablesVector->size()) {
190         int32_t i = ch - data->variablesBase;
191         set = (i < variablesVector->size()) ?
192             (UnicodeFunctor*) variablesVector->elementAt(i) : 0;
193     }
194     return set;
195 }
196 
197 /**
198  * Implement SymbolTable API.  Parse out a symbol reference
199  * name.
200  */
parseReference(const UnicodeString & text,ParsePosition & pos,int32_t limit) const201 UnicodeString ParseData::parseReference(const UnicodeString& text,
202                                         ParsePosition& pos, int32_t limit) const {
203     int32_t start = pos.getIndex();
204     int32_t i = start;
205     UnicodeString result;
206     while (i < limit) {
207         UChar c = text.charAt(i);
208         if ((i==start && !u_isIDStart(c)) || !u_isIDPart(c)) {
209             break;
210         }
211         ++i;
212     }
213     if (i == start) { // No valid name chars
214         return result; // Indicate failure with empty string
215     }
216     pos.setIndex(i);
217     text.extractBetween(start, i, result);
218     return result;
219 }
220 
isMatcher(UChar32 ch)221 UBool ParseData::isMatcher(UChar32 ch) {
222     // Note that we cannot use data.lookup() because the
223     // set array has not been constructed yet.
224     int32_t i = ch - data->variablesBase;
225     if (i >= 0 && i < variablesVector->size()) {
226         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
227         return f != NULL && f->toMatcher() != NULL;
228     }
229     return TRUE;
230 }
231 
232 /**
233  * Return true if the given character is a replacer standin or a plain
234  * character (non standin).
235  */
isReplacer(UChar32 ch)236 UBool ParseData::isReplacer(UChar32 ch) {
237     // Note that we cannot use data.lookup() because the
238     // set array has not been constructed yet.
239     int i = ch - data->variablesBase;
240     if (i >= 0 && i < variablesVector->size()) {
241         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
242         return f != NULL && f->toReplacer() != NULL;
243     }
244     return TRUE;
245 }
246 
247 //----------------------------------------------------------------------
248 // BEGIN RuleHalf
249 //----------------------------------------------------------------------
250 
251 /**
252  * A class representing one side of a rule.  This class knows how to
253  * parse half of a rule.  It is tightly coupled to the method
254  * RuleBasedTransliterator.Parser.parseRule().
255  */
256 class RuleHalf : public UMemory {
257 
258 public:
259 
260     UnicodeString text;
261 
262     int32_t cursor; // position of cursor in text
263     int32_t ante;   // position of ante context marker '{' in text
264     int32_t post;   // position of post context marker '}' in text
265 
266     // Record the offset to the cursor either to the left or to the
267     // right of the key.  This is indicated by characters on the output
268     // side that allow the cursor to be positioned arbitrarily within
269     // the matching text.  For example, abc{def} > | @@@ xyz; changes
270     // def to xyz and moves the cursor to before abc.  Offset characters
271     // must be at the start or end, and they cannot move the cursor past
272     // the ante- or postcontext text.  Placeholders are only valid in
273     // output text.  The length of the ante and post context is
274     // determined at runtime, because of supplementals and quantifiers.
275     int32_t cursorOffset; // only nonzero on output side
276 
277     // Position of first CURSOR_OFFSET on _right_.  This will be -1
278     // for |@, -2 for |@@, etc., and 1 for @|, 2 for @@|, etc.
279     int32_t cursorOffsetPos;
280 
281     UBool anchorStart;
282     UBool anchorEnd;
283 
284     /**
285      * The segment number from 1..n of the next '(' we see
286      * during parsing; 1-based.
287      */
288     int32_t nextSegmentNumber;
289 
290     TransliteratorParser& parser;
291 
292     //--------------------------------------------------
293     // Methods
294 
295     RuleHalf(TransliteratorParser& parser);
296     ~RuleHalf();
297 
298     int32_t parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status);
299 
300     int32_t parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
301                          UnicodeString& buf,
302                          const UnicodeString& illegal,
303                          UBool isSegment,
304                          UErrorCode& status);
305 
306     /**
307      * Remove context.
308      */
309     void removeContext();
310 
311     /**
312      * Return true if this half looks like valid output, that is, does not
313      * contain quantifiers or other special input-only elements.
314      */
315     UBool isValidOutput(TransliteratorParser& parser);
316 
317     /**
318      * Return true if this half looks like valid input, that is, does not
319      * contain functions or other special output-only elements.
320      */
321     UBool isValidInput(TransliteratorParser& parser);
322 
syntaxError(UErrorCode code,const UnicodeString & rule,int32_t start,UErrorCode & status)323     int syntaxError(UErrorCode code,
324                     const UnicodeString& rule,
325                     int32_t start,
326                     UErrorCode& status) {
327         return parser.syntaxError(code, rule, start, status);
328     }
329 
330 private:
331     // Disallowed methods; no impl.
332     RuleHalf(const RuleHalf&);
333     RuleHalf& operator=(const RuleHalf&);
334 };
335 
RuleHalf(TransliteratorParser & p)336 RuleHalf::RuleHalf(TransliteratorParser& p) :
337     parser(p)
338 {
339     cursor = -1;
340     ante = -1;
341     post = -1;
342     cursorOffset = 0;
343     cursorOffsetPos = 0;
344     anchorStart = anchorEnd = FALSE;
345     nextSegmentNumber = 1;
346 }
347 
~RuleHalf()348 RuleHalf::~RuleHalf() {
349 }
350 
351 /**
352  * Parse one side of a rule, stopping at either the limit,
353  * the END_OF_RULE character, or an operator.
354  * @return the index after the terminating character, or
355  * if limit was reached, limit
356  */
parse(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)357 int32_t RuleHalf::parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
358     int32_t start = pos;
359     text.truncate(0);
360     pos = parseSection(rule, pos, limit, text, ILLEGAL_TOP, FALSE, status);
361 
362     if (cursorOffset > 0 && cursor != cursorOffsetPos) {
363         return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
364     }
365 
366     return pos;
367 }
368 
369 /**
370  * Parse a section of one side of a rule, stopping at either
371  * the limit, the END_OF_RULE character, an operator, or a
372  * segment close character.  This method parses both a
373  * top-level rule half and a segment within such a rule half.
374  * It calls itself recursively to parse segments and nested
375  * segments.
376  * @param buf buffer into which to accumulate the rule pattern
377  * characters, either literal characters from the rule or
378  * standins for UnicodeMatcher objects including segments.
379  * @param illegal the set of special characters that is illegal during
380  * this parse.
381  * @param isSegment if true, then we've already seen a '(' and
382  * pos on entry points right after it.  Accumulate everything
383  * up to the closing ')', put it in a segment matcher object,
384  * generate a standin for it, and add the standin to buf.  As
385  * a side effect, update the segments vector with a reference
386  * to the segment matcher.  This works recursively for nested
387  * segments.  If isSegment is false, just accumulate
388  * characters into buf.
389  * @return the index after the terminating character, or
390  * if limit was reached, limit
391  */
parseSection(const UnicodeString & rule,int32_t pos,int32_t limit,UnicodeString & buf,const UnicodeString & illegal,UBool isSegment,UErrorCode & status)392 int32_t RuleHalf::parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
393                                UnicodeString& buf,
394                                const UnicodeString& illegal,
395                                UBool isSegment, UErrorCode& status) {
396     int32_t start = pos;
397     ParsePosition pp;
398     UnicodeString scratch;
399     UBool done = FALSE;
400     int32_t quoteStart = -1; // Most recent 'single quoted string'
401     int32_t quoteLimit = -1;
402     int32_t varStart = -1; // Most recent $variableReference
403     int32_t varLimit = -1;
404     int32_t bufStart = buf.length();
405 
406     while (pos < limit && !done) {
407         // Since all syntax characters are in the BMP, fetching
408         // 16-bit code units suffices here.
409         UChar c = rule.charAt(pos++);
410         if (PatternProps::isWhiteSpace(c)) {
411             // Ignore whitespace.  Note that this is not Unicode
412             // spaces, but Java spaces -- a subset, representing
413             // whitespace likely to be seen in code.
414             continue;
415         }
416         if (u_strchr(HALF_ENDERS, c) != NULL) {
417             if (isSegment) {
418                 // Unclosed segment
419                 return syntaxError(U_UNCLOSED_SEGMENT, rule, start, status);
420             }
421             break;
422         }
423         if (anchorEnd) {
424             // Text after a presumed end anchor is a syntax err
425             return syntaxError(U_MALFORMED_VARIABLE_REFERENCE, rule, start, status);
426         }
427         if (UnicodeSet::resemblesPattern(rule, pos-1)) {
428             pp.setIndex(pos-1); // Backup to opening '['
429             buf.append(parser.parseSet(rule, pp, status));
430             if (U_FAILURE(status)) {
431                 return syntaxError(U_MALFORMED_SET, rule, start, status);
432             }
433             pos = pp.getIndex();
434             continue;
435         }
436         // Handle escapes
437         if (c == ESCAPE) {
438             if (pos == limit) {
439                 return syntaxError(U_TRAILING_BACKSLASH, rule, start, status);
440             }
441             UChar32 escaped = rule.unescapeAt(pos); // pos is already past '\\'
442             if (escaped == (UChar32) -1) {
443                 return syntaxError(U_MALFORMED_UNICODE_ESCAPE, rule, start, status);
444             }
445             if (!parser.checkVariableRange(escaped)) {
446                 return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
447             }
448             buf.append(escaped);
449             continue;
450         }
451         // Handle quoted matter
452         if (c == QUOTE) {
453             int32_t iq = rule.indexOf(QUOTE, pos);
454             if (iq == pos) {
455                 buf.append(c); // Parse [''] outside quotes as [']
456                 ++pos;
457             } else {
458                 /* This loop picks up a run of quoted text of the
459                  * form 'aaaa' each time through.  If this run
460                  * hasn't really ended ('aaaa''bbbb') then it keeps
461                  * looping, each time adding on a new run.  When it
462                  * reaches the final quote it breaks.
463                  */
464                 quoteStart = buf.length();
465                 for (;;) {
466                     if (iq < 0) {
467                         return syntaxError(U_UNTERMINATED_QUOTE, rule, start, status);
468                     }
469                     scratch.truncate(0);
470                     rule.extractBetween(pos, iq, scratch);
471                     buf.append(scratch);
472                     pos = iq+1;
473                     if (pos < limit && rule.charAt(pos) == QUOTE) {
474                         // Parse [''] inside quotes as [']
475                         iq = rule.indexOf(QUOTE, pos+1);
476                         // Continue looping
477                     } else {
478                         break;
479                     }
480                 }
481                 quoteLimit = buf.length();
482 
483                 for (iq=quoteStart; iq<quoteLimit; ++iq) {
484                     if (!parser.checkVariableRange(buf.charAt(iq))) {
485                         return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
486                     }
487                 }
488             }
489             continue;
490         }
491 
492         if (!parser.checkVariableRange(c)) {
493             return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
494         }
495 
496         if (illegal.indexOf(c) >= 0) {
497             syntaxError(U_ILLEGAL_CHARACTER, rule, start, status);
498         }
499 
500         switch (c) {
501 
502         //------------------------------------------------------
503         // Elements allowed within and out of segments
504         //------------------------------------------------------
505         case ANCHOR_START:
506             if (buf.length() == 0 && !anchorStart) {
507                 anchorStart = TRUE;
508             } else {
509               return syntaxError(U_MISPLACED_ANCHOR_START,
510                                  rule, start, status);
511             }
512           break;
513         case SEGMENT_OPEN:
514             {
515                 // bufSegStart is the offset in buf to the first
516                 // character of the segment we are parsing.
517                 int32_t bufSegStart = buf.length();
518 
519                 // Record segment number now, since nextSegmentNumber
520                 // will be incremented during the call to parseSection
521                 // if there are nested segments.
522                 int32_t segmentNumber = nextSegmentNumber++; // 1-based
523 
524                 // Parse the segment
525                 pos = parseSection(rule, pos, limit, buf, ILLEGAL_SEG, TRUE, status);
526 
527                 // After parsing a segment, the relevant characters are
528                 // in buf, starting at offset bufSegStart.  Extract them
529                 // into a string matcher, and replace them with a
530                 // standin for that matcher.
531                 StringMatcher* m =
532                     new StringMatcher(buf, bufSegStart, buf.length(),
533                                       segmentNumber, *parser.curData);
534                 if (m == NULL) {
535                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
536                 }
537 
538                 // Record and associate object and segment number
539                 parser.setSegmentObject(segmentNumber, m, status);
540                 buf.truncate(bufSegStart);
541                 buf.append(parser.getSegmentStandin(segmentNumber, status));
542             }
543             break;
544         case FUNCTION:
545         case ALT_FUNCTION:
546             {
547                 int32_t iref = pos;
548                 TransliteratorIDParser::SingleID* single =
549                     TransliteratorIDParser::parseFilterID(rule, iref);
550                 // The next character MUST be a segment open
551                 if (single == NULL ||
552                     !ICU_Utility::parseChar(rule, iref, SEGMENT_OPEN)) {
553                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
554                 }
555 
556                 Transliterator *t = single->createInstance();
557                 delete single;
558                 if (t == NULL) {
559                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
560                 }
561 
562                 // bufSegStart is the offset in buf to the first
563                 // character of the segment we are parsing.
564                 int32_t bufSegStart = buf.length();
565 
566                 // Parse the segment
567                 pos = parseSection(rule, iref, limit, buf, ILLEGAL_FUNC, TRUE, status);
568 
569                 // After parsing a segment, the relevant characters are
570                 // in buf, starting at offset bufSegStart.
571                 UnicodeString output;
572                 buf.extractBetween(bufSegStart, buf.length(), output);
573                 FunctionReplacer *r =
574                     new FunctionReplacer(t, new StringReplacer(output, parser.curData));
575                 if (r == NULL) {
576                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
577                 }
578 
579                 // Replace the buffer contents with a stand-in
580                 buf.truncate(bufSegStart);
581                 buf.append(parser.generateStandInFor(r, status));
582             }
583             break;
584         case SymbolTable::SYMBOL_REF:
585             // Handle variable references and segment references "$1" .. "$9"
586             {
587                 // A variable reference must be followed immediately
588                 // by a Unicode identifier start and zero or more
589                 // Unicode identifier part characters, or by a digit
590                 // 1..9 if it is a segment reference.
591                 if (pos == limit) {
592                     // A variable ref character at the end acts as
593                     // an anchor to the context limit, as in perl.
594                     anchorEnd = TRUE;
595                     break;
596                 }
597                 // Parse "$1" "$2" .. "$9" .. (no upper limit)
598                 c = rule.charAt(pos);
599                 int32_t r = u_digit(c, 10);
600                 if (r >= 1 && r <= 9) {
601                     r = ICU_Utility::parseNumber(rule, pos, 10);
602                     if (r < 0) {
603                         return syntaxError(U_UNDEFINED_SEGMENT_REFERENCE,
604                                            rule, start, status);
605                     }
606                     buf.append(parser.getSegmentStandin(r, status));
607                 } else {
608                     pp.setIndex(pos);
609                     UnicodeString name = parser.parseData->
610                                     parseReference(rule, pp, limit);
611                     if (name.length() == 0) {
612                         // This means the '$' was not followed by a
613                         // valid name.  Try to interpret it as an
614                         // end anchor then.  If this also doesn't work
615                         // (if we see a following character) then signal
616                         // an error.
617                         anchorEnd = TRUE;
618                         break;
619                     }
620                     pos = pp.getIndex();
621                     // If this is a variable definition statement,
622                     // then the LHS variable will be undefined.  In
623                     // that case appendVariableDef() will append the
624                     // special placeholder char variableLimit-1.
625                     varStart = buf.length();
626                     parser.appendVariableDef(name, buf, status);
627                     varLimit = buf.length();
628                 }
629             }
630             break;
631         case DOT:
632             buf.append(parser.getDotStandIn(status));
633             break;
634         case KLEENE_STAR:
635         case ONE_OR_MORE:
636         case ZERO_OR_ONE:
637             // Quantifiers.  We handle single characters, quoted strings,
638             // variable references, and segments.
639             //  a+      matches  aaa
640             //  'foo'+  matches  foofoofoo
641             //  $v+     matches  xyxyxy if $v == xy
642             //  (seg)+  matches  segsegseg
643             {
644                 if (isSegment && buf.length() == bufStart) {
645                     // The */+ immediately follows '('
646                     return syntaxError(U_MISPLACED_QUANTIFIER, rule, start, status);
647                 }
648 
649                 int32_t qstart, qlimit;
650                 // The */+ follows an isolated character or quote
651                 // or variable reference
652                 if (buf.length() == quoteLimit) {
653                     // The */+ follows a 'quoted string'
654                     qstart = quoteStart;
655                     qlimit = quoteLimit;
656                 } else if (buf.length() == varLimit) {
657                     // The */+ follows a $variableReference
658                     qstart = varStart;
659                     qlimit = varLimit;
660                 } else {
661                     // The */+ follows a single character, possibly
662                     // a segment standin
663                     qstart = buf.length() - 1;
664                     qlimit = qstart + 1;
665                 }
666 
667                 UnicodeFunctor *m =
668                     new StringMatcher(buf, qstart, qlimit, 0, *parser.curData);
669                 if (m == NULL) {
670                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
671                 }
672                 int32_t min = 0;
673                 int32_t max = Quantifier::MAX;
674                 switch (c) {
675                 case ONE_OR_MORE:
676                     min = 1;
677                     break;
678                 case ZERO_OR_ONE:
679                     min = 0;
680                     max = 1;
681                     break;
682                 // case KLEENE_STAR:
683                 //    do nothing -- min, max already set
684                 }
685                 m = new Quantifier(m, min, max);
686                 if (m == NULL) {
687                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
688                 }
689                 buf.truncate(qstart);
690                 buf.append(parser.generateStandInFor(m, status));
691             }
692             break;
693 
694         //------------------------------------------------------
695         // Elements allowed ONLY WITHIN segments
696         //------------------------------------------------------
697         case SEGMENT_CLOSE:
698             // assert(isSegment);
699             // We're done parsing a segment.
700             done = TRUE;
701             break;
702 
703         //------------------------------------------------------
704         // Elements allowed ONLY OUTSIDE segments
705         //------------------------------------------------------
706         case CONTEXT_ANTE:
707             if (ante >= 0) {
708                 return syntaxError(U_MULTIPLE_ANTE_CONTEXTS, rule, start, status);
709             }
710             ante = buf.length();
711             break;
712         case CONTEXT_POST:
713             if (post >= 0) {
714                 return syntaxError(U_MULTIPLE_POST_CONTEXTS, rule, start, status);
715             }
716             post = buf.length();
717             break;
718         case CURSOR_POS:
719             if (cursor >= 0) {
720                 return syntaxError(U_MULTIPLE_CURSORS, rule, start, status);
721             }
722             cursor = buf.length();
723             break;
724         case CURSOR_OFFSET:
725             if (cursorOffset < 0) {
726                 if (buf.length() > 0) {
727                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
728                 }
729                 --cursorOffset;
730             } else if (cursorOffset > 0) {
731                 if (buf.length() != cursorOffsetPos || cursor >= 0) {
732                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
733                 }
734                 ++cursorOffset;
735             } else {
736                 if (cursor == 0 && buf.length() == 0) {
737                     cursorOffset = -1;
738                 } else if (cursor < 0) {
739                     cursorOffsetPos = buf.length();
740                     cursorOffset = 1;
741                 } else {
742                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
743                 }
744             }
745             break;
746 
747 
748         //------------------------------------------------------
749         // Non-special characters
750         //------------------------------------------------------
751         default:
752             // Disallow unquoted characters other than [0-9A-Za-z]
753             // in the printable ASCII range.  These characters are
754             // reserved for possible future use.
755             if (c >= 0x0021 && c <= 0x007E &&
756                 !((c >= 0x0030/*'0'*/ && c <= 0x0039/*'9'*/) ||
757                   (c >= 0x0041/*'A'*/ && c <= 0x005A/*'Z'*/) ||
758                   (c >= 0x0061/*'a'*/ && c <= 0x007A/*'z'*/))) {
759                 return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
760             }
761             buf.append(c);
762             break;
763         }
764     }
765 
766     return pos;
767 }
768 
769 /**
770  * Remove context.
771  */
removeContext()772 void RuleHalf::removeContext() {
773     //text = text.substring(ante < 0 ? 0 : ante,
774     //                      post < 0 ? text.length() : post);
775     if (post >= 0) {
776         text.remove(post);
777     }
778     if (ante >= 0) {
779         text.removeBetween(0, ante);
780     }
781     ante = post = -1;
782     anchorStart = anchorEnd = FALSE;
783 }
784 
785 /**
786  * Return true if this half looks like valid output, that is, does not
787  * contain quantifiers or other special input-only elements.
788  */
isValidOutput(TransliteratorParser & transParser)789 UBool RuleHalf::isValidOutput(TransliteratorParser& transParser) {
790     for (int32_t i=0; i<text.length(); ) {
791         UChar32 c = text.char32At(i);
792         i += UTF_CHAR_LENGTH(c);
793         if (!transParser.parseData->isReplacer(c)) {
794             return FALSE;
795         }
796     }
797     return TRUE;
798 }
799 
800 /**
801  * Return true if this half looks like valid input, that is, does not
802  * contain functions or other special output-only elements.
803  */
isValidInput(TransliteratorParser & transParser)804 UBool RuleHalf::isValidInput(TransliteratorParser& transParser) {
805     for (int32_t i=0; i<text.length(); ) {
806         UChar32 c = text.char32At(i);
807         i += UTF_CHAR_LENGTH(c);
808         if (!transParser.parseData->isMatcher(c)) {
809             return FALSE;
810         }
811     }
812     return TRUE;
813 }
814 
815 //----------------------------------------------------------------------
816 // PUBLIC API
817 //----------------------------------------------------------------------
818 
819 /**
820  * Constructor.
821  */
TransliteratorParser(UErrorCode & statusReturn)822 TransliteratorParser::TransliteratorParser(UErrorCode &statusReturn) :
823 dataVector(statusReturn),
824 idBlockVector(statusReturn),
825 variablesVector(statusReturn),
826 segmentObjects(statusReturn)
827 {
828     idBlockVector.setDeleter(uhash_deleteUnicodeString);
829     curData = NULL;
830     compoundFilter = NULL;
831     parseData = NULL;
832     variableNames.setValueDeleter(uhash_deleteUnicodeString);
833 }
834 
835 /**
836  * Destructor.
837  */
~TransliteratorParser()838 TransliteratorParser::~TransliteratorParser() {
839     while (!dataVector.isEmpty())
840         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
841     delete compoundFilter;
842     delete parseData;
843     while (!variablesVector.isEmpty())
844         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
845 }
846 
847 void
parse(const UnicodeString & rules,UTransDirection transDirection,UParseError & pe,UErrorCode & ec)848 TransliteratorParser::parse(const UnicodeString& rules,
849                             UTransDirection transDirection,
850                             UParseError& pe,
851                             UErrorCode& ec) {
852     if (U_SUCCESS(ec)) {
853         parseRules(rules, transDirection, ec);
854         pe = parseError;
855     }
856 }
857 
858 /**
859  * Return the compound filter parsed by parse().  Caller owns result.
860  */
orphanCompoundFilter()861 UnicodeSet* TransliteratorParser::orphanCompoundFilter() {
862     UnicodeSet* f = compoundFilter;
863     compoundFilter = NULL;
864     return f;
865 }
866 
867 //----------------------------------------------------------------------
868 // Private implementation
869 //----------------------------------------------------------------------
870 
871 /**
872  * Parse the given string as a sequence of rules, separated by newline
873  * characters ('\n'), and cause this object to implement those rules.  Any
874  * previous rules are discarded.  Typically this method is called exactly
875  * once, during construction.
876  * @exception IllegalArgumentException if there is a syntax error in the
877  * rules
878  */
parseRules(const UnicodeString & rule,UTransDirection theDirection,UErrorCode & status)879 void TransliteratorParser::parseRules(const UnicodeString& rule,
880                                       UTransDirection theDirection,
881                                       UErrorCode& status)
882 {
883     // Clear error struct
884     uprv_memset(&parseError, 0, sizeof(parseError));
885     parseError.line = parseError.offset = -1;
886 
887     UBool parsingIDs = TRUE;
888     int32_t ruleCount = 0;
889 
890     while (!dataVector.isEmpty()) {
891         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
892     }
893     if (U_FAILURE(status)) {
894         return;
895     }
896 
897     idBlockVector.removeAllElements();
898     curData = NULL;
899     direction = theDirection;
900     ruleCount = 0;
901 
902     delete compoundFilter;
903     compoundFilter = NULL;
904 
905     while (!variablesVector.isEmpty()) {
906         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
907     }
908     variableNames.removeAll();
909     parseData = new ParseData(0, &variablesVector, &variableNames);
910     if (parseData == NULL) {
911         status = U_MEMORY_ALLOCATION_ERROR;
912         return;
913     }
914 
915     dotStandIn = (UChar) -1;
916 
917     UnicodeString *tempstr = NULL; // used for memory allocation error checking
918     UnicodeString str; // scratch
919     UnicodeString idBlockResult;
920     int32_t pos = 0;
921     int32_t limit = rule.length();
922 
923     // The compound filter offset is an index into idBlockResult.
924     // If it is 0, then the compound filter occurred at the start,
925     // and it is the offset to the _start_ of the compound filter
926     // pattern.  Otherwise it is the offset to the _limit_ of the
927     // compound filter pattern within idBlockResult.
928     compoundFilter = NULL;
929     int32_t compoundFilterOffset = -1;
930 
931     while (pos < limit && U_SUCCESS(status)) {
932         UChar c = rule.charAt(pos++);
933         if (PatternProps::isWhiteSpace(c)) {
934             // Ignore leading whitespace.
935             continue;
936         }
937         // Skip lines starting with the comment character
938         if (c == RULE_COMMENT_CHAR) {
939             pos = rule.indexOf((UChar)0x000A /*\n*/, pos) + 1;
940             if (pos == 0) {
941                 break; // No "\n" found; rest of rule is a commnet
942             }
943             continue; // Either fall out or restart with next line
944         }
945 
946         // skip empty rules
947         if (c == END_OF_RULE)
948             continue;
949 
950         // keep track of how many rules we've seen
951         ++ruleCount;
952 
953         // We've found the start of a rule or ID.  c is its first
954         // character, and pos points past c.
955         --pos;
956         // Look for an ID token.  Must have at least ID_TOKEN_LEN + 1
957         // chars left.
958         if ((pos + ID_TOKEN_LEN + 1) <= limit &&
959                 rule.compare(pos, ID_TOKEN_LEN, ID_TOKEN) == 0) {
960             pos += ID_TOKEN_LEN;
961             c = rule.charAt(pos);
962             while (PatternProps::isWhiteSpace(c) && pos < limit) {
963                 ++pos;
964                 c = rule.charAt(pos);
965             }
966 
967             int32_t p = pos;
968 
969             if (!parsingIDs) {
970                 if (curData != NULL) {
971                     if (direction == UTRANS_FORWARD)
972                         dataVector.addElement(curData, status);
973                     else
974                         dataVector.insertElementAt(curData, 0, status);
975                     curData = NULL;
976                 }
977                 parsingIDs = TRUE;
978             }
979 
980             TransliteratorIDParser::SingleID* id =
981                 TransliteratorIDParser::parseSingleID(rule, p, direction, status);
982             if (p != pos && ICU_Utility::parseChar(rule, p, END_OF_RULE)) {
983                 // Successful ::ID parse.
984 
985                 if (direction == UTRANS_FORWARD) {
986                     idBlockResult.append(id->canonID).append(END_OF_RULE);
987                 } else {
988                     idBlockResult.insert(0, END_OF_RULE);
989                     idBlockResult.insert(0, id->canonID);
990                 }
991 
992             } else {
993                 // Couldn't parse an ID.  Try to parse a global filter
994                 int32_t withParens = -1;
995                 UnicodeSet* f = TransliteratorIDParser::parseGlobalFilter(rule, p, direction, withParens, NULL);
996                 if (f != NULL) {
997                     if (ICU_Utility::parseChar(rule, p, END_OF_RULE)
998                         && (direction == UTRANS_FORWARD) == (withParens == 0))
999                     {
1000                         if (compoundFilter != NULL) {
1001                             // Multiple compound filters
1002                             syntaxError(U_MULTIPLE_COMPOUND_FILTERS, rule, pos, status);
1003                             delete f;
1004                         } else {
1005                             compoundFilter = f;
1006                             compoundFilterOffset = ruleCount;
1007                         }
1008                     } else {
1009                         delete f;
1010                     }
1011                 } else {
1012                     // Invalid ::id
1013                     // Can be parsed as neither an ID nor a global filter
1014                     syntaxError(U_INVALID_ID, rule, pos, status);
1015                 }
1016             }
1017             delete id;
1018             pos = p;
1019         } else {
1020             if (parsingIDs) {
1021                 tempstr = new UnicodeString(idBlockResult);
1022                 // NULL pointer check
1023                 if (tempstr == NULL) {
1024                     status = U_MEMORY_ALLOCATION_ERROR;
1025                     return;
1026                 }
1027                 if (direction == UTRANS_FORWARD)
1028                     idBlockVector.addElement(tempstr, status);
1029                 else
1030                     idBlockVector.insertElementAt(tempstr, 0, status);
1031                 idBlockResult.remove();
1032                 parsingIDs = FALSE;
1033                 curData = new TransliterationRuleData(status);
1034                 // NULL pointer check
1035                 if (curData == NULL) {
1036                     status = U_MEMORY_ALLOCATION_ERROR;
1037                     return;
1038                 }
1039                 parseData->data = curData;
1040 
1041                 // By default, rules use part of the private use area
1042                 // E000..F8FF for variables and other stand-ins.  Currently
1043                 // the range F000..F8FF is typically sufficient.  The 'use
1044                 // variable range' pragma allows rule sets to modify this.
1045                 setVariableRange(0xF000, 0xF8FF, status);
1046             }
1047 
1048             if (resemblesPragma(rule, pos, limit)) {
1049                 int32_t ppp = parsePragma(rule, pos, limit, status);
1050                 if (ppp < 0) {
1051                     syntaxError(U_MALFORMED_PRAGMA, rule, pos, status);
1052                 }
1053                 pos = ppp;
1054             // Parse a rule
1055             } else {
1056                 pos = parseRule(rule, pos, limit, status);
1057             }
1058         }
1059     }
1060 
1061     if (parsingIDs && idBlockResult.length() > 0) {
1062         tempstr = new UnicodeString(idBlockResult);
1063         // NULL pointer check
1064         if (tempstr == NULL) {
1065             status = U_MEMORY_ALLOCATION_ERROR;
1066             return;
1067         }
1068         if (direction == UTRANS_FORWARD)
1069             idBlockVector.addElement(tempstr, status);
1070         else
1071             idBlockVector.insertElementAt(tempstr, 0, status);
1072     }
1073     else if (!parsingIDs && curData != NULL) {
1074         if (direction == UTRANS_FORWARD)
1075             dataVector.addElement(curData, status);
1076         else
1077             dataVector.insertElementAt(curData, 0, status);
1078     }
1079 
1080     if (U_SUCCESS(status)) {
1081         // Convert the set vector to an array
1082         int32_t i, dataVectorSize = dataVector.size();
1083         for (i = 0; i < dataVectorSize; i++) {
1084             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1085             data->variablesLength = variablesVector.size();
1086             if (data->variablesLength == 0) {
1087                 data->variables = 0;
1088             } else {
1089                 data->variables = (UnicodeFunctor**)uprv_malloc(data->variablesLength * sizeof(UnicodeFunctor*));
1090                 // NULL pointer check
1091                 if (data->variables == NULL) {
1092                     status = U_MEMORY_ALLOCATION_ERROR;
1093                     return;
1094                 }
1095                 data->variablesAreOwned = (i == 0);
1096             }
1097 
1098             for (int32_t j = 0; j < data->variablesLength; j++) {
1099                 data->variables[j] =
1100                     ((UnicodeSet*)variablesVector.elementAt(j));
1101             }
1102 
1103             data->variableNames.removeAll();
1104             int32_t pos = -1;
1105             const UHashElement* he = variableNames.nextElement(pos);
1106             while (he != NULL) {
1107                 UnicodeString* tempus = (UnicodeString*)(((UnicodeString*)(he->value.pointer))->clone());
1108                 if (tempus == NULL) {
1109                     status = U_MEMORY_ALLOCATION_ERROR;
1110                     return;
1111                 }
1112                 data->variableNames.put(*((UnicodeString*)(he->key.pointer)),
1113                     tempus, status);
1114                 he = variableNames.nextElement(pos);
1115             }
1116         }
1117         variablesVector.removeAllElements();   // keeps them from getting deleted when we succeed
1118 
1119         // Index the rules
1120         if (compoundFilter != NULL) {
1121             if ((direction == UTRANS_FORWARD && compoundFilterOffset != 1) ||
1122                 (direction == UTRANS_REVERSE && compoundFilterOffset != ruleCount)) {
1123                 status = U_MISPLACED_COMPOUND_FILTER;
1124             }
1125         }
1126 
1127         for (i = 0; i < dataVectorSize; i++) {
1128             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1129             data->ruleSet.freeze(parseError, status);
1130         }
1131         if (idBlockVector.size() == 1 && ((UnicodeString*)idBlockVector.elementAt(0))->isEmpty()) {
1132             idBlockVector.removeElementAt(0);
1133         }
1134     }
1135 }
1136 
1137 /**
1138  * Set the variable range to [start, end] (inclusive).
1139  */
setVariableRange(int32_t start,int32_t end,UErrorCode & status)1140 void TransliteratorParser::setVariableRange(int32_t start, int32_t end, UErrorCode& status) {
1141     if (start > end || start < 0 || end > 0xFFFF) {
1142         status = U_MALFORMED_PRAGMA;
1143         return;
1144     }
1145 
1146     curData->variablesBase = (UChar) start;
1147     if (dataVector.size() == 0) {
1148         variableNext = (UChar) start;
1149         variableLimit = (UChar) (end + 1);
1150     }
1151 }
1152 
1153 /**
1154  * Assert that the given character is NOT within the variable range.
1155  * If it is, return FALSE.  This is neccesary to ensure that the
1156  * variable range does not overlap characters used in a rule.
1157  */
checkVariableRange(UChar32 ch) const1158 UBool TransliteratorParser::checkVariableRange(UChar32 ch) const {
1159     return !(ch >= curData->variablesBase && ch < variableLimit);
1160 }
1161 
1162 /**
1163  * Set the maximum backup to 'backup', in response to a pragma
1164  * statement.
1165  */
pragmaMaximumBackup(int32_t)1166 void TransliteratorParser::pragmaMaximumBackup(int32_t /*backup*/) {
1167     //TODO Finish
1168 }
1169 
1170 /**
1171  * Begin normalizing all rules using the given mode, in response
1172  * to a pragma statement.
1173  */
pragmaNormalizeRules(UNormalizationMode)1174 void TransliteratorParser::pragmaNormalizeRules(UNormalizationMode /*mode*/) {
1175     //TODO Finish
1176 }
1177 
1178 static const UChar PRAGMA_USE[] = {0x75,0x73,0x65,0x20,0}; // "use "
1179 
1180 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 # #~;"
1181 
1182 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 #~;"
1183 
1184 static const UChar PRAGMA_NFD_RULES[] = {0x7E,0x6E,0x66,0x64,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfd rules~;"
1185 
1186 static const UChar PRAGMA_NFC_RULES[] = {0x7E,0x6E,0x66,0x63,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfc rules~;"
1187 
1188 /**
1189  * Return true if the given rule looks like a pragma.
1190  * @param pos offset to the first non-whitespace character
1191  * of the rule.
1192  * @param limit pointer past the last character of the rule.
1193  */
resemblesPragma(const UnicodeString & rule,int32_t pos,int32_t limit)1194 UBool TransliteratorParser::resemblesPragma(const UnicodeString& rule, int32_t pos, int32_t limit) {
1195     // Must start with /use\s/i
1196     return ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_USE, NULL) >= 0;
1197 }
1198 
1199 /**
1200  * Parse a pragma.  This method assumes resemblesPragma() has
1201  * already returned true.
1202  * @param pos offset to the first non-whitespace character
1203  * of the rule.
1204  * @param limit pointer past the last character of the rule.
1205  * @return the position index after the final ';' of the pragma,
1206  * or -1 on failure.
1207  */
parsePragma(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1208 int32_t TransliteratorParser::parsePragma(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1209     int32_t array[2];
1210 
1211     // resemblesPragma() has already returned true, so we
1212     // know that pos points to /use\s/i; we can skip 4 characters
1213     // immediately
1214     pos += 4;
1215 
1216     // Here are the pragmas we recognize:
1217     // use variable range 0xE000 0xEFFF;
1218     // use maximum backup 16;
1219     // use nfd rules;
1220     // use nfc rules;
1221     int p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_VARIABLE_RANGE, array);
1222     if (p >= 0) {
1223         setVariableRange(array[0], array[1], status);
1224         return p;
1225     }
1226 
1227     p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_MAXIMUM_BACKUP, array);
1228     if (p >= 0) {
1229         pragmaMaximumBackup(array[0]);
1230         return p;
1231     }
1232 
1233     p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_NFD_RULES, NULL);
1234     if (p >= 0) {
1235         pragmaNormalizeRules(UNORM_NFD);
1236         return p;
1237     }
1238 
1239     p = ICU_Utility::parsePattern(rule, pos, limit, PRAGMA_NFC_RULES, NULL);
1240     if (p >= 0) {
1241         pragmaNormalizeRules(UNORM_NFC);
1242         return p;
1243     }
1244 
1245     // Syntax error: unable to parse pragma
1246     return -1;
1247 }
1248 
1249 /**
1250  * MAIN PARSER.  Parse the next rule in the given rule string, starting
1251  * at pos.  Return the index after the last character parsed.  Do not
1252  * parse characters at or after limit.
1253  *
1254  * Important:  The character at pos must be a non-whitespace character
1255  * that is not the comment character.
1256  *
1257  * This method handles quoting, escaping, and whitespace removal.  It
1258  * parses the end-of-rule character.  It recognizes context and cursor
1259  * indicators.  Once it does a lexical breakdown of the rule at pos, it
1260  * creates a rule object and adds it to our rule list.
1261  */
parseRule(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1262 int32_t TransliteratorParser::parseRule(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1263     // Locate the left side, operator, and right side
1264     int32_t start = pos;
1265     UChar op = 0;
1266     int32_t i;
1267 
1268     // Set up segments data
1269     segmentStandins.truncate(0);
1270     segmentObjects.removeAllElements();
1271 
1272     // Use pointers to automatics to make swapping possible.
1273     RuleHalf _left(*this), _right(*this);
1274     RuleHalf* left = &_left;
1275     RuleHalf* right = &_right;
1276 
1277     undefinedVariableName.remove();
1278     pos = left->parse(rule, pos, limit, status);
1279     if (U_FAILURE(status)) {
1280         return start;
1281     }
1282 
1283     if (pos == limit || u_strchr(gOPERATORS, (op = rule.charAt(--pos))) == NULL) {
1284         return syntaxError(U_MISSING_OPERATOR, rule, start, status);
1285     }
1286     ++pos;
1287 
1288     // Found an operator char.  Check for forward-reverse operator.
1289     if (op == REVERSE_RULE_OP &&
1290         (pos < limit && rule.charAt(pos) == FORWARD_RULE_OP)) {
1291         ++pos;
1292         op = FWDREV_RULE_OP;
1293     }
1294 
1295     // Translate alternate op characters.
1296     switch (op) {
1297     case ALT_FORWARD_RULE_OP:
1298         op = FORWARD_RULE_OP;
1299         break;
1300     case ALT_REVERSE_RULE_OP:
1301         op = REVERSE_RULE_OP;
1302         break;
1303     case ALT_FWDREV_RULE_OP:
1304         op = FWDREV_RULE_OP;
1305         break;
1306     }
1307 
1308     pos = right->parse(rule, pos, limit, status);
1309     if (U_FAILURE(status)) {
1310         return start;
1311     }
1312 
1313     if (pos < limit) {
1314         if (rule.charAt(--pos) == END_OF_RULE) {
1315             ++pos;
1316         } else {
1317             // RuleHalf parser must have terminated at an operator
1318             return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
1319         }
1320     }
1321 
1322     if (op == VARIABLE_DEF_OP) {
1323         // LHS is the name.  RHS is a single character, either a literal
1324         // or a set (already parsed).  If RHS is longer than one
1325         // character, it is either a multi-character string, or multiple
1326         // sets, or a mixture of chars and sets -- syntax error.
1327 
1328         // We expect to see a single undefined variable (the one being
1329         // defined).
1330         if (undefinedVariableName.length() == 0) {
1331             // "Missing '$' or duplicate definition"
1332             return syntaxError(U_BAD_VARIABLE_DEFINITION, rule, start, status);
1333         }
1334         if (left->text.length() != 1 || left->text.charAt(0) != variableLimit) {
1335             // "Malformed LHS"
1336             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1337         }
1338         if (left->anchorStart || left->anchorEnd ||
1339             right->anchorStart || right->anchorEnd) {
1340             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1341         }
1342         // We allow anything on the right, including an empty string.
1343         UnicodeString* value = new UnicodeString(right->text);
1344         // NULL pointer check
1345         if (value == NULL) {
1346             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1347         }
1348         variableNames.put(undefinedVariableName, value, status);
1349         ++variableLimit;
1350         return pos;
1351     }
1352 
1353     // If this is not a variable definition rule, we shouldn't have
1354     // any undefined variable names.
1355     if (undefinedVariableName.length() != 0) {
1356         return syntaxError(// "Undefined variable $" + undefinedVariableName,
1357                     U_UNDEFINED_VARIABLE,
1358                     rule, start, status);
1359     }
1360 
1361     // Verify segments
1362     if (segmentStandins.length() > segmentObjects.size()) {
1363         syntaxError(U_UNDEFINED_SEGMENT_REFERENCE, rule, start, status);
1364     }
1365     for (i=0; i<segmentStandins.length(); ++i) {
1366         if (segmentStandins.charAt(i) == 0) {
1367             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1368         }
1369     }
1370     for (i=0; i<segmentObjects.size(); ++i) {
1371         if (segmentObjects.elementAt(i) == NULL) {
1372             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1373         }
1374     }
1375 
1376     // If the direction we want doesn't match the rule
1377     // direction, do nothing.
1378     if (op != FWDREV_RULE_OP &&
1379         ((direction == UTRANS_FORWARD) != (op == FORWARD_RULE_OP))) {
1380         return pos;
1381     }
1382 
1383     // Transform the rule into a forward rule by swapping the
1384     // sides if necessary.
1385     if (direction == UTRANS_REVERSE) {
1386         left = &_right;
1387         right = &_left;
1388     }
1389 
1390     // Remove non-applicable elements in forward-reverse
1391     // rules.  Bidirectional rules ignore elements that do not
1392     // apply.
1393     if (op == FWDREV_RULE_OP) {
1394         right->removeContext();
1395         left->cursor = -1;
1396         left->cursorOffset = 0;
1397     }
1398 
1399     // Normalize context
1400     if (left->ante < 0) {
1401         left->ante = 0;
1402     }
1403     if (left->post < 0) {
1404         left->post = left->text.length();
1405     }
1406 
1407     // Context is only allowed on the input side.  Cursors are only
1408     // allowed on the output side.  Segment delimiters can only appear
1409     // on the left, and references on the right.  Cursor offset
1410     // cannot appear without an explicit cursor.  Cursor offset
1411     // cannot place the cursor outside the limits of the context.
1412     // Anchors are only allowed on the input side.
1413     if (right->ante >= 0 || right->post >= 0 || left->cursor >= 0 ||
1414         (right->cursorOffset != 0 && right->cursor < 0) ||
1415         // - The following two checks were used to ensure that the
1416         // - the cursor offset stayed within the ante- or postcontext.
1417         // - However, with the addition of quantifiers, we have to
1418         // - allow arbitrary cursor offsets and do runtime checking.
1419         //(right->cursorOffset > (left->text.length() - left->post)) ||
1420         //(-right->cursorOffset > left->ante) ||
1421         right->anchorStart || right->anchorEnd ||
1422         !left->isValidInput(*this) || !right->isValidOutput(*this) ||
1423         left->ante > left->post) {
1424 
1425         return syntaxError(U_MALFORMED_RULE, rule, start, status);
1426     }
1427 
1428     // Flatten segment objects vector to an array
1429     UnicodeFunctor** segmentsArray = NULL;
1430     if (segmentObjects.size() > 0) {
1431         segmentsArray = (UnicodeFunctor **)uprv_malloc(segmentObjects.size() * sizeof(UnicodeFunctor *));
1432         // Null pointer check
1433         if (segmentsArray == NULL) {
1434             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1435         }
1436         segmentObjects.toArray((void**) segmentsArray);
1437     }
1438     TransliterationRule* temptr = new TransliterationRule(
1439             left->text, left->ante, left->post,
1440             right->text, right->cursor, right->cursorOffset,
1441             segmentsArray,
1442             segmentObjects.size(),
1443             left->anchorStart, left->anchorEnd,
1444             curData,
1445             status);
1446     //Null pointer check
1447     if (temptr == NULL) {
1448         uprv_free(segmentsArray);
1449         return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1450     }
1451 
1452     curData->ruleSet.addRule(temptr, status);
1453 
1454     return pos;
1455 }
1456 
1457 /**
1458  * Called by main parser upon syntax error.  Search the rule string
1459  * for the probable end of the rule.  Of course, if the error is that
1460  * the end of rule marker is missing, then the rule end will not be found.
1461  * In any case the rule start will be correctly reported.
1462  * @param msg error description
1463  * @param rule pattern string
1464  * @param start position of first character of current rule
1465  */
syntaxError(UErrorCode parseErrorCode,const UnicodeString & rule,int32_t pos,UErrorCode & status)1466 int32_t TransliteratorParser::syntaxError(UErrorCode parseErrorCode,
1467                                           const UnicodeString& rule,
1468                                           int32_t pos,
1469                                           UErrorCode& status)
1470 {
1471     parseError.offset = pos;
1472     parseError.line = 0 ; /* we are not using line numbers */
1473 
1474     // for pre-context
1475     const int32_t LEN = U_PARSE_CONTEXT_LEN - 1;
1476     int32_t start = uprv_max(pos - LEN, 0);
1477     int32_t stop  = pos;
1478 
1479     rule.extract(start,stop-start,parseError.preContext);
1480     //null terminate the buffer
1481     parseError.preContext[stop-start] = 0;
1482 
1483     //for post-context
1484     start = pos;
1485     stop  = uprv_min(pos + LEN, rule.length());
1486 
1487     rule.extract(start,stop-start,parseError.postContext);
1488     //null terminate the buffer
1489     parseError.postContext[stop-start]= 0;
1490 
1491     status = (UErrorCode)parseErrorCode;
1492     return pos;
1493 
1494 }
1495 
1496 /**
1497  * Parse a UnicodeSet out, store it, and return the stand-in character
1498  * used to represent it.
1499  */
parseSet(const UnicodeString & rule,ParsePosition & pos,UErrorCode & status)1500 UChar TransliteratorParser::parseSet(const UnicodeString& rule,
1501                                           ParsePosition& pos,
1502                                           UErrorCode& status) {
1503     UnicodeSet* set = new UnicodeSet(rule, pos, USET_IGNORE_SPACE, parseData, status);
1504     // Null pointer check
1505     if (set == NULL) {
1506         status = U_MEMORY_ALLOCATION_ERROR;
1507         return (UChar)0x0000; // Return empty character with error.
1508     }
1509     set->compact();
1510     return generateStandInFor(set, status);
1511 }
1512 
1513 /**
1514  * Generate and return a stand-in for a new UnicodeFunctor.  Store
1515  * the matcher (adopt it).
1516  */
generateStandInFor(UnicodeFunctor * adopted,UErrorCode & status)1517 UChar TransliteratorParser::generateStandInFor(UnicodeFunctor* adopted, UErrorCode& status) {
1518     // assert(obj != null);
1519 
1520     // Look up previous stand-in, if any.  This is a short list
1521     // (typical n is 0, 1, or 2); linear search is optimal.
1522     for (int32_t i=0; i<variablesVector.size(); ++i) {
1523         if (variablesVector.elementAt(i) == adopted) { // [sic] pointer comparison
1524             return (UChar) (curData->variablesBase + i);
1525         }
1526     }
1527 
1528     if (variableNext >= variableLimit) {
1529         delete adopted;
1530         status = U_VARIABLE_RANGE_EXHAUSTED;
1531         return 0;
1532     }
1533     variablesVector.addElement(adopted, status);
1534     return variableNext++;
1535 }
1536 
1537 /**
1538  * Return the standin for segment seg (1-based).
1539  */
getSegmentStandin(int32_t seg,UErrorCode & status)1540 UChar TransliteratorParser::getSegmentStandin(int32_t seg, UErrorCode& status) {
1541     // Special character used to indicate an empty spot
1542     UChar empty = curData->variablesBase - 1;
1543     while (segmentStandins.length() < seg) {
1544         segmentStandins.append(empty);
1545     }
1546     UChar c = segmentStandins.charAt(seg-1);
1547     if (c == empty) {
1548         if (variableNext >= variableLimit) {
1549             status = U_VARIABLE_RANGE_EXHAUSTED;
1550             return 0;
1551         }
1552         c = variableNext++;
1553         // Set a placeholder in the master variables vector that will be
1554         // filled in later by setSegmentObject().  We know that we will get
1555         // called first because setSegmentObject() will call us.
1556         variablesVector.addElement((void*) NULL, status);
1557         segmentStandins.setCharAt(seg-1, c);
1558     }
1559     return c;
1560 }
1561 
1562 /**
1563  * Set the object for segment seg (1-based).
1564  */
setSegmentObject(int32_t seg,StringMatcher * adopted,UErrorCode & status)1565 void TransliteratorParser::setSegmentObject(int32_t seg, StringMatcher* adopted, UErrorCode& status) {
1566     // Since we call parseSection() recursively, nested
1567     // segments will result in segment i+1 getting parsed
1568     // and stored before segment i; be careful with the
1569     // vector handling here.
1570     if (segmentObjects.size() < seg) {
1571         segmentObjects.setSize(seg, status);
1572     }
1573     int32_t index = getSegmentStandin(seg, status) - curData->variablesBase;
1574     if (segmentObjects.elementAt(seg-1) != NULL ||
1575         variablesVector.elementAt(index) != NULL) {
1576         // should never happen
1577         status = U_INTERNAL_TRANSLITERATOR_ERROR;
1578         return;
1579     }
1580     segmentObjects.setElementAt(adopted, seg-1);
1581     variablesVector.setElementAt(adopted, index);
1582 }
1583 
1584 /**
1585  * Return the stand-in for the dot set.  It is allocated the first
1586  * time and reused thereafter.
1587  */
getDotStandIn(UErrorCode & status)1588 UChar TransliteratorParser::getDotStandIn(UErrorCode& status) {
1589     if (dotStandIn == (UChar) -1) {
1590         UnicodeSet* tempus = new UnicodeSet(DOT_SET, status);
1591         // Null pointer check.
1592         if (tempus == NULL) {
1593             status = U_MEMORY_ALLOCATION_ERROR;
1594             return (UChar)0x0000;
1595         }
1596         dotStandIn = generateStandInFor(tempus, status);
1597     }
1598     return dotStandIn;
1599 }
1600 
1601 /**
1602  * Append the value of the given variable name to the given
1603  * UnicodeString.
1604  */
appendVariableDef(const UnicodeString & name,UnicodeString & buf,UErrorCode & status)1605 void TransliteratorParser::appendVariableDef(const UnicodeString& name,
1606                                                   UnicodeString& buf,
1607                                                   UErrorCode& status) {
1608     const UnicodeString* s = (const UnicodeString*) variableNames.get(name);
1609     if (s == NULL) {
1610         // We allow one undefined variable so that variable definition
1611         // statements work.  For the first undefined variable we return
1612         // the special placeholder variableLimit-1, and save the variable
1613         // name.
1614         if (undefinedVariableName.length() == 0) {
1615             undefinedVariableName = name;
1616             if (variableNext >= variableLimit) {
1617                 // throw new RuntimeException("Private use variables exhausted");
1618                 status = U_ILLEGAL_ARGUMENT_ERROR;
1619                 return;
1620             }
1621             buf.append((UChar) --variableLimit);
1622         } else {
1623             //throw new IllegalArgumentException("Undefined variable $"
1624             //                                   + name);
1625             status = U_ILLEGAL_ARGUMENT_ERROR;
1626             return;
1627         }
1628     } else {
1629         buf.append(*s);
1630     }
1631 }
1632 
1633 /**
1634  * Glue method to get around access restrictions in C++.
1635  */
1636 /*Transliterator* TransliteratorParser::createBasicInstance(const UnicodeString& id, const UnicodeString* canonID) {
1637     return Transliterator::createBasicInstance(id, canonID);
1638 }*/
1639 
1640 U_NAMESPACE_END
1641 
1642 U_CAPI int32_t
utrans_stripRules(const UChar * source,int32_t sourceLen,UChar * target,UErrorCode * status)1643 utrans_stripRules(const UChar *source, int32_t sourceLen, UChar *target, UErrorCode *status) {
1644     U_NAMESPACE_USE
1645 
1646     //const UChar *sourceStart = source;
1647     const UChar *targetStart = target;
1648     const UChar *sourceLimit = source+sourceLen;
1649     UChar *targetLimit = target+sourceLen;
1650     UChar32 c = 0;
1651     UBool quoted = FALSE;
1652     int32_t index;
1653 
1654     uprv_memset(target, 0, sourceLen*U_SIZEOF_UCHAR);
1655 
1656     /* read the rules into the buffer */
1657     while (source < sourceLimit)
1658     {
1659         index=0;
1660         U16_NEXT_UNSAFE(source, index, c);
1661         source+=index;
1662         if(c == QUOTE) {
1663             quoted = (UBool)!quoted;
1664         }
1665         else if (!quoted) {
1666             if (c == RULE_COMMENT_CHAR) {
1667                 /* skip comments and all preceding spaces */
1668                 while (targetStart < target && *(target - 1) == 0x0020) {
1669                     target--;
1670                 }
1671                 do {
1672                     c = *(source++);
1673                 }
1674                 while (c != CR && c != LF);
1675             }
1676             else if (c == ESCAPE) {
1677                 UChar32   c2 = *source;
1678                 if (c2 == CR || c2 == LF) {
1679                     /* A backslash at the end of a line. */
1680                     /* Since we're stripping lines, ignore the backslash. */
1681                     source++;
1682                     continue;
1683                 }
1684                 if (c2 == 0x0075 && source+5 < sourceLimit) { /* \u seen. \U isn't unescaped. */
1685                     int32_t escapeOffset = 0;
1686                     UnicodeString escapedStr(source, 5);
1687                     c2 = escapedStr.unescapeAt(escapeOffset);
1688 
1689                     if (c2 == (UChar32)0xFFFFFFFF || escapeOffset == 0)
1690                     {
1691                         *status = U_PARSE_ERROR;
1692                         return 0;
1693                     }
1694                     if (!PatternProps::isWhiteSpace(c2) && !u_iscntrl(c2) && !u_ispunct(c2)) {
1695                         /* It was escaped for a reason. Write what it was suppose to be. */
1696                         source+=5;
1697                         c = c2;
1698                     }
1699                 }
1700                 else if (c2 == QUOTE) {
1701                     /* \' seen. Make sure we don't do anything when we see it again. */
1702                     quoted = (UBool)!quoted;
1703                 }
1704             }
1705         }
1706         if (c == CR || c == LF)
1707         {
1708             /* ignore spaces carriage returns, and all leading spaces on the next line.
1709             * and line feed unless in the form \uXXXX
1710             */
1711             quoted = FALSE;
1712             while (source < sourceLimit) {
1713                 c = *(source);
1714                 if (c != CR && c != LF && c != 0x0020) {
1715                     break;
1716                 }
1717                 source++;
1718             }
1719             continue;
1720         }
1721 
1722         /* Append UChar * after dissembling if c > 0xffff*/
1723         index=0;
1724         U16_APPEND_UNSAFE(target, index, c);
1725         target+=index;
1726     }
1727     if (target < targetLimit) {
1728         *target = 0;
1729     }
1730     return (int32_t)(target-targetStart);
1731 }
1732 
1733 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
1734