1 /*
2  **********************************************************************
3  *   Copyright (C) 1999-2014, 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 "unicode/utf16.h"
23 #include "cstring.h"
24 #include "funcrepl.h"
25 #include "hash.h"
26 #include "quant.h"
27 #include "rbt.h"
28 #include "rbt_data.h"
29 #include "rbt_pars.h"
30 #include "rbt_rule.h"
31 #include "strmatch.h"
32 #include "strrepl.h"
33 #include "unicode/symtable.h"
34 #include "tridpars.h"
35 #include "uvector.h"
36 #include "hash.h"
37 #include "patternprops.h"
38 #include "util.h"
39 #include "cmemory.h"
40 #include "uprops.h"
41 #include "putilimp.h"
42 
43 // Operators
44 #define VARIABLE_DEF_OP ((UChar)0x003D) /*=*/
45 #define FORWARD_RULE_OP ((UChar)0x003E) /*>*/
46 #define REVERSE_RULE_OP ((UChar)0x003C) /*<*/
47 #define FWDREV_RULE_OP  ((UChar)0x007E) /*~*/ // internal rep of <> op
48 
49 // Other special characters
50 #define QUOTE             ((UChar)0x0027) /*'*/
51 #define ESCAPE            ((UChar)0x005C) /*\*/
52 #define END_OF_RULE       ((UChar)0x003B) /*;*/
53 #define RULE_COMMENT_CHAR ((UChar)0x0023) /*#*/
54 
55 #define SEGMENT_OPEN       ((UChar)0x0028) /*(*/
56 #define SEGMENT_CLOSE      ((UChar)0x0029) /*)*/
57 #define CONTEXT_ANTE       ((UChar)0x007B) /*{*/
58 #define CONTEXT_POST       ((UChar)0x007D) /*}*/
59 #define CURSOR_POS         ((UChar)0x007C) /*|*/
60 #define CURSOR_OFFSET      ((UChar)0x0040) /*@*/
61 #define ANCHOR_START       ((UChar)0x005E) /*^*/
62 #define KLEENE_STAR        ((UChar)0x002A) /***/
63 #define ONE_OR_MORE        ((UChar)0x002B) /*+*/
64 #define ZERO_OR_ONE        ((UChar)0x003F) /*?*/
65 
66 #define DOT                ((UChar)46)     /*.*/
67 
68 static const UChar DOT_SET[] = { // "[^[:Zp:][:Zl:]\r\n$]";
69     91, 94, 91, 58, 90, 112, 58, 93, 91, 58, 90,
70     108, 58, 93, 92, 114, 92, 110, 36, 93, 0
71 };
72 
73 // A function is denoted &Source-Target/Variant(text)
74 #define FUNCTION           ((UChar)38)     /*&*/
75 
76 // Aliases for some of the syntax characters. These are provided so
77 // transliteration rules can be expressed in XML without clashing with
78 // XML syntax characters '<', '>', and '&'.
79 #define ALT_REVERSE_RULE_OP ((UChar)0x2190) // Left Arrow
80 #define ALT_FORWARD_RULE_OP ((UChar)0x2192) // Right Arrow
81 #define ALT_FWDREV_RULE_OP  ((UChar)0x2194) // Left Right Arrow
82 #define ALT_FUNCTION        ((UChar)0x2206) // Increment (~Greek Capital Delta)
83 
84 // Special characters disallowed at the top level
85 static const UChar ILLEGAL_TOP[] = {41,0}; // ")"
86 
87 // Special characters disallowed within a segment
88 static const UChar ILLEGAL_SEG[] = {123,125,124,64,0}; // "{}|@"
89 
90 // Special characters disallowed within a function argument
91 static const UChar ILLEGAL_FUNC[] = {94,40,46,42,43,63,123,125,124,64,0}; // "^(.*+?{}|@"
92 
93 // By definition, the ANCHOR_END special character is a
94 // trailing SymbolTable.SYMBOL_REF character.
95 // private static final char ANCHOR_END       = '$';
96 
97 static const UChar gOPERATORS[] = { // "=><"
98     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
99     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
100     0
101 };
102 
103 static const UChar HALF_ENDERS[] = { // "=><;"
104     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
105     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
106     END_OF_RULE,
107     0
108 };
109 
110 // These are also used in Transliterator::toRules()
111 static const int32_t ID_TOKEN_LEN = 2;
112 static const UChar   ID_TOKEN[]   = { 0x3A, 0x3A }; // ':', ':'
113 
114 /*
115 commented out until we do real ::BEGIN/::END functionality
116 static const int32_t BEGIN_TOKEN_LEN = 5;
117 static const UChar BEGIN_TOKEN[] = { 0x42, 0x45, 0x47, 0x49, 0x4e }; // 'BEGIN'
118 
119 static const int32_t END_TOKEN_LEN = 3;
120 static const UChar END_TOKEN[] = { 0x45, 0x4e, 0x44 }; // 'END'
121 */
122 
123 U_NAMESPACE_BEGIN
124 
125 //----------------------------------------------------------------------
126 // BEGIN ParseData
127 //----------------------------------------------------------------------
128 
129 /**
130  * This class implements the SymbolTable interface.  It is used
131  * during parsing to give UnicodeSet access to variables that
132  * have been defined so far.  Note that it uses variablesVector,
133  * _not_ data.setVariables.
134  */
135 class ParseData : public UMemory, public SymbolTable {
136 public:
137     const TransliterationRuleData* data; // alias
138 
139     const UVector* variablesVector; // alias
140 
141     const Hashtable* variableNames; // alias
142 
143     ParseData(const TransliterationRuleData* data = 0,
144               const UVector* variablesVector = 0,
145               const Hashtable* variableNames = 0);
146 
147     virtual ~ParseData();
148 
149     virtual const UnicodeString* lookup(const UnicodeString& s) const;
150 
151     virtual const UnicodeFunctor* lookupMatcher(UChar32 ch) const;
152 
153     virtual UnicodeString parseReference(const UnicodeString& text,
154                                          ParsePosition& pos, int32_t limit) const;
155     /**
156      * Return true if the given character is a matcher standin or a plain
157      * character (non standin).
158      */
159     UBool isMatcher(UChar32 ch);
160 
161     /**
162      * Return true if the given character is a replacer standin or a plain
163      * character (non standin).
164      */
165     UBool isReplacer(UChar32 ch);
166 
167 private:
168     ParseData(const ParseData &other); // forbid copying of this class
169     ParseData &operator=(const ParseData &other); // forbid copying of this class
170 };
171 
ParseData(const TransliterationRuleData * d,const UVector * sets,const Hashtable * vNames)172 ParseData::ParseData(const TransliterationRuleData* d,
173                      const UVector* sets,
174                      const Hashtable* vNames) :
175     data(d), variablesVector(sets), variableNames(vNames) {}
176 
~ParseData()177 ParseData::~ParseData() {}
178 
179 /**
180  * Implement SymbolTable API.
181  */
lookup(const UnicodeString & name) const182 const UnicodeString* ParseData::lookup(const UnicodeString& name) const {
183     return (const UnicodeString*) variableNames->get(name);
184 }
185 
186 /**
187  * Implement SymbolTable API.
188  */
lookupMatcher(UChar32 ch) const189 const UnicodeFunctor* ParseData::lookupMatcher(UChar32 ch) const {
190     // Note that we cannot use data.lookupSet() because the
191     // set array has not been constructed yet.
192     const UnicodeFunctor* set = NULL;
193     int32_t i = ch - data->variablesBase;
194     if (i >= 0 && i < variablesVector->size()) {
195         int32_t i = ch - data->variablesBase;
196         set = (i < variablesVector->size()) ?
197             (UnicodeFunctor*) variablesVector->elementAt(i) : 0;
198     }
199     return set;
200 }
201 
202 /**
203  * Implement SymbolTable API.  Parse out a symbol reference
204  * name.
205  */
parseReference(const UnicodeString & text,ParsePosition & pos,int32_t limit) const206 UnicodeString ParseData::parseReference(const UnicodeString& text,
207                                         ParsePosition& pos, int32_t limit) const {
208     int32_t start = pos.getIndex();
209     int32_t i = start;
210     UnicodeString result;
211     while (i < limit) {
212         UChar c = text.charAt(i);
213         if ((i==start && !u_isIDStart(c)) || !u_isIDPart(c)) {
214             break;
215         }
216         ++i;
217     }
218     if (i == start) { // No valid name chars
219         return result; // Indicate failure with empty string
220     }
221     pos.setIndex(i);
222     text.extractBetween(start, i, result);
223     return result;
224 }
225 
isMatcher(UChar32 ch)226 UBool ParseData::isMatcher(UChar32 ch) {
227     // Note that we cannot use data.lookup() because the
228     // set array has not been constructed yet.
229     int32_t i = ch - data->variablesBase;
230     if (i >= 0 && i < variablesVector->size()) {
231         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
232         return f != NULL && f->toMatcher() != NULL;
233     }
234     return TRUE;
235 }
236 
237 /**
238  * Return true if the given character is a replacer standin or a plain
239  * character (non standin).
240  */
isReplacer(UChar32 ch)241 UBool ParseData::isReplacer(UChar32 ch) {
242     // Note that we cannot use data.lookup() because the
243     // set array has not been constructed yet.
244     int i = ch - data->variablesBase;
245     if (i >= 0 && i < variablesVector->size()) {
246         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
247         return f != NULL && f->toReplacer() != NULL;
248     }
249     return TRUE;
250 }
251 
252 //----------------------------------------------------------------------
253 // BEGIN RuleHalf
254 //----------------------------------------------------------------------
255 
256 /**
257  * A class representing one side of a rule.  This class knows how to
258  * parse half of a rule.  It is tightly coupled to the method
259  * RuleBasedTransliterator.Parser.parseRule().
260  */
261 class RuleHalf : public UMemory {
262 
263 public:
264 
265     UnicodeString text;
266 
267     int32_t cursor; // position of cursor in text
268     int32_t ante;   // position of ante context marker '{' in text
269     int32_t post;   // position of post context marker '}' in text
270 
271     // Record the offset to the cursor either to the left or to the
272     // right of the key.  This is indicated by characters on the output
273     // side that allow the cursor to be positioned arbitrarily within
274     // the matching text.  For example, abc{def} > | @@@ xyz; changes
275     // def to xyz and moves the cursor to before abc.  Offset characters
276     // must be at the start or end, and they cannot move the cursor past
277     // the ante- or postcontext text.  Placeholders are only valid in
278     // output text.  The length of the ante and post context is
279     // determined at runtime, because of supplementals and quantifiers.
280     int32_t cursorOffset; // only nonzero on output side
281 
282     // Position of first CURSOR_OFFSET on _right_.  This will be -1
283     // for |@, -2 for |@@, etc., and 1 for @|, 2 for @@|, etc.
284     int32_t cursorOffsetPos;
285 
286     UBool anchorStart;
287     UBool anchorEnd;
288 
289     /**
290      * The segment number from 1..n of the next '(' we see
291      * during parsing; 1-based.
292      */
293     int32_t nextSegmentNumber;
294 
295     TransliteratorParser& parser;
296 
297     //--------------------------------------------------
298     // Methods
299 
300     RuleHalf(TransliteratorParser& parser);
301     ~RuleHalf();
302 
303     int32_t parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status);
304 
305     int32_t parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
306                          UnicodeString& buf,
307                          const UnicodeString& illegal,
308                          UBool isSegment,
309                          UErrorCode& status);
310 
311     /**
312      * Remove context.
313      */
314     void removeContext();
315 
316     /**
317      * Return true if this half looks like valid output, that is, does not
318      * contain quantifiers or other special input-only elements.
319      */
320     UBool isValidOutput(TransliteratorParser& parser);
321 
322     /**
323      * Return true if this half looks like valid input, that is, does not
324      * contain functions or other special output-only elements.
325      */
326     UBool isValidInput(TransliteratorParser& parser);
327 
syntaxError(UErrorCode code,const UnicodeString & rule,int32_t start,UErrorCode & status)328     int syntaxError(UErrorCode code,
329                     const UnicodeString& rule,
330                     int32_t start,
331                     UErrorCode& status) {
332         return parser.syntaxError(code, rule, start, status);
333     }
334 
335 private:
336     // Disallowed methods; no impl.
337     RuleHalf(const RuleHalf&);
338     RuleHalf& operator=(const RuleHalf&);
339 };
340 
RuleHalf(TransliteratorParser & p)341 RuleHalf::RuleHalf(TransliteratorParser& p) :
342     parser(p)
343 {
344     cursor = -1;
345     ante = -1;
346     post = -1;
347     cursorOffset = 0;
348     cursorOffsetPos = 0;
349     anchorStart = anchorEnd = FALSE;
350     nextSegmentNumber = 1;
351 }
352 
~RuleHalf()353 RuleHalf::~RuleHalf() {
354 }
355 
356 /**
357  * Parse one side of a rule, stopping at either the limit,
358  * the END_OF_RULE character, or an operator.
359  * @return the index after the terminating character, or
360  * if limit was reached, limit
361  */
parse(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)362 int32_t RuleHalf::parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
363     int32_t start = pos;
364     text.truncate(0);
365     pos = parseSection(rule, pos, limit, text, UnicodeString(TRUE, ILLEGAL_TOP, -1), FALSE, status);
366 
367     if (cursorOffset > 0 && cursor != cursorOffsetPos) {
368         return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
369     }
370 
371     return pos;
372 }
373 
374 /**
375  * Parse a section of one side of a rule, stopping at either
376  * the limit, the END_OF_RULE character, an operator, or a
377  * segment close character.  This method parses both a
378  * top-level rule half and a segment within such a rule half.
379  * It calls itself recursively to parse segments and nested
380  * segments.
381  * @param buf buffer into which to accumulate the rule pattern
382  * characters, either literal characters from the rule or
383  * standins for UnicodeMatcher objects including segments.
384  * @param illegal the set of special characters that is illegal during
385  * this parse.
386  * @param isSegment if true, then we've already seen a '(' and
387  * pos on entry points right after it.  Accumulate everything
388  * up to the closing ')', put it in a segment matcher object,
389  * generate a standin for it, and add the standin to buf.  As
390  * a side effect, update the segments vector with a reference
391  * to the segment matcher.  This works recursively for nested
392  * segments.  If isSegment is false, just accumulate
393  * characters into buf.
394  * @return the index after the terminating character, or
395  * if limit was reached, limit
396  */
parseSection(const UnicodeString & rule,int32_t pos,int32_t limit,UnicodeString & buf,const UnicodeString & illegal,UBool isSegment,UErrorCode & status)397 int32_t RuleHalf::parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
398                                UnicodeString& buf,
399                                const UnicodeString& illegal,
400                                UBool isSegment, UErrorCode& status) {
401     int32_t start = pos;
402     ParsePosition pp;
403     UnicodeString scratch;
404     UBool done = FALSE;
405     int32_t quoteStart = -1; // Most recent 'single quoted string'
406     int32_t quoteLimit = -1;
407     int32_t varStart = -1; // Most recent $variableReference
408     int32_t varLimit = -1;
409     int32_t bufStart = buf.length();
410 
411     while (pos < limit && !done) {
412         // Since all syntax characters are in the BMP, fetching
413         // 16-bit code units suffices here.
414         UChar c = rule.charAt(pos++);
415         if (PatternProps::isWhiteSpace(c)) {
416             // Ignore whitespace.  Note that this is not Unicode
417             // spaces, but Java spaces -- a subset, representing
418             // whitespace likely to be seen in code.
419             continue;
420         }
421         if (u_strchr(HALF_ENDERS, c) != NULL) {
422             if (isSegment) {
423                 // Unclosed segment
424                 return syntaxError(U_UNCLOSED_SEGMENT, rule, start, status);
425             }
426             break;
427         }
428         if (anchorEnd) {
429             // Text after a presumed end anchor is a syntax err
430             return syntaxError(U_MALFORMED_VARIABLE_REFERENCE, rule, start, status);
431         }
432         if (UnicodeSet::resemblesPattern(rule, pos-1)) {
433             pp.setIndex(pos-1); // Backup to opening '['
434             buf.append(parser.parseSet(rule, pp, status));
435             if (U_FAILURE(status)) {
436                 return syntaxError(U_MALFORMED_SET, rule, start, status);
437             }
438             pos = pp.getIndex();
439             continue;
440         }
441         // Handle escapes
442         if (c == ESCAPE) {
443             if (pos == limit) {
444                 return syntaxError(U_TRAILING_BACKSLASH, rule, start, status);
445             }
446             UChar32 escaped = rule.unescapeAt(pos); // pos is already past '\\'
447             if (escaped == (UChar32) -1) {
448                 return syntaxError(U_MALFORMED_UNICODE_ESCAPE, rule, start, status);
449             }
450             if (!parser.checkVariableRange(escaped)) {
451                 return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
452             }
453             buf.append(escaped);
454             continue;
455         }
456         // Handle quoted matter
457         if (c == QUOTE) {
458             int32_t iq = rule.indexOf(QUOTE, pos);
459             if (iq == pos) {
460                 buf.append(c); // Parse [''] outside quotes as [']
461                 ++pos;
462             } else {
463                 /* This loop picks up a run of quoted text of the
464                  * form 'aaaa' each time through.  If this run
465                  * hasn't really ended ('aaaa''bbbb') then it keeps
466                  * looping, each time adding on a new run.  When it
467                  * reaches the final quote it breaks.
468                  */
469                 quoteStart = buf.length();
470                 for (;;) {
471                     if (iq < 0) {
472                         return syntaxError(U_UNTERMINATED_QUOTE, rule, start, status);
473                     }
474                     scratch.truncate(0);
475                     rule.extractBetween(pos, iq, scratch);
476                     buf.append(scratch);
477                     pos = iq+1;
478                     if (pos < limit && rule.charAt(pos) == QUOTE) {
479                         // Parse [''] inside quotes as [']
480                         iq = rule.indexOf(QUOTE, pos+1);
481                         // Continue looping
482                     } else {
483                         break;
484                     }
485                 }
486                 quoteLimit = buf.length();
487 
488                 for (iq=quoteStart; iq<quoteLimit; ++iq) {
489                     if (!parser.checkVariableRange(buf.charAt(iq))) {
490                         return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
491                     }
492                 }
493             }
494             continue;
495         }
496 
497         if (!parser.checkVariableRange(c)) {
498             return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
499         }
500 
501         if (illegal.indexOf(c) >= 0) {
502             syntaxError(U_ILLEGAL_CHARACTER, rule, start, status);
503         }
504 
505         switch (c) {
506 
507         //------------------------------------------------------
508         // Elements allowed within and out of segments
509         //------------------------------------------------------
510         case ANCHOR_START:
511             if (buf.length() == 0 && !anchorStart) {
512                 anchorStart = TRUE;
513             } else {
514               return syntaxError(U_MISPLACED_ANCHOR_START,
515                                  rule, start, status);
516             }
517           break;
518         case SEGMENT_OPEN:
519             {
520                 // bufSegStart is the offset in buf to the first
521                 // character of the segment we are parsing.
522                 int32_t bufSegStart = buf.length();
523 
524                 // Record segment number now, since nextSegmentNumber
525                 // will be incremented during the call to parseSection
526                 // if there are nested segments.
527                 int32_t segmentNumber = nextSegmentNumber++; // 1-based
528 
529                 // Parse the segment
530                 pos = parseSection(rule, pos, limit, buf, UnicodeString(TRUE, ILLEGAL_SEG, -1), TRUE, status);
531 
532                 // After parsing a segment, the relevant characters are
533                 // in buf, starting at offset bufSegStart.  Extract them
534                 // into a string matcher, and replace them with a
535                 // standin for that matcher.
536                 StringMatcher* m =
537                     new StringMatcher(buf, bufSegStart, buf.length(),
538                                       segmentNumber, *parser.curData);
539                 if (m == NULL) {
540                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
541                 }
542 
543                 // Record and associate object and segment number
544                 parser.setSegmentObject(segmentNumber, m, status);
545                 buf.truncate(bufSegStart);
546                 buf.append(parser.getSegmentStandin(segmentNumber, status));
547             }
548             break;
549         case FUNCTION:
550         case ALT_FUNCTION:
551             {
552                 int32_t iref = pos;
553                 TransliteratorIDParser::SingleID* single =
554                     TransliteratorIDParser::parseFilterID(rule, iref);
555                 // The next character MUST be a segment open
556                 if (single == NULL ||
557                     !ICU_Utility::parseChar(rule, iref, SEGMENT_OPEN)) {
558                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
559                 }
560 
561                 Transliterator *t = single->createInstance();
562                 delete single;
563                 if (t == NULL) {
564                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
565                 }
566 
567                 // bufSegStart is the offset in buf to the first
568                 // character of the segment we are parsing.
569                 int32_t bufSegStart = buf.length();
570 
571                 // Parse the segment
572                 pos = parseSection(rule, iref, limit, buf, UnicodeString(TRUE, ILLEGAL_FUNC, -1), TRUE, status);
573 
574                 // After parsing a segment, the relevant characters are
575                 // in buf, starting at offset bufSegStart.
576                 UnicodeString output;
577                 buf.extractBetween(bufSegStart, buf.length(), output);
578                 FunctionReplacer *r =
579                     new FunctionReplacer(t, new StringReplacer(output, parser.curData));
580                 if (r == NULL) {
581                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
582                 }
583 
584                 // Replace the buffer contents with a stand-in
585                 buf.truncate(bufSegStart);
586                 buf.append(parser.generateStandInFor(r, status));
587             }
588             break;
589         case SymbolTable::SYMBOL_REF:
590             // Handle variable references and segment references "$1" .. "$9"
591             {
592                 // A variable reference must be followed immediately
593                 // by a Unicode identifier start and zero or more
594                 // Unicode identifier part characters, or by a digit
595                 // 1..9 if it is a segment reference.
596                 if (pos == limit) {
597                     // A variable ref character at the end acts as
598                     // an anchor to the context limit, as in perl.
599                     anchorEnd = TRUE;
600                     break;
601                 }
602                 // Parse "$1" "$2" .. "$9" .. (no upper limit)
603                 c = rule.charAt(pos);
604                 int32_t r = u_digit(c, 10);
605                 if (r >= 1 && r <= 9) {
606                     r = ICU_Utility::parseNumber(rule, pos, 10);
607                     if (r < 0) {
608                         return syntaxError(U_UNDEFINED_SEGMENT_REFERENCE,
609                                            rule, start, status);
610                     }
611                     buf.append(parser.getSegmentStandin(r, status));
612                 } else {
613                     pp.setIndex(pos);
614                     UnicodeString name = parser.parseData->
615                                     parseReference(rule, pp, limit);
616                     if (name.length() == 0) {
617                         // This means the '$' was not followed by a
618                         // valid name.  Try to interpret it as an
619                         // end anchor then.  If this also doesn't work
620                         // (if we see a following character) then signal
621                         // an error.
622                         anchorEnd = TRUE;
623                         break;
624                     }
625                     pos = pp.getIndex();
626                     // If this is a variable definition statement,
627                     // then the LHS variable will be undefined.  In
628                     // that case appendVariableDef() will append the
629                     // special placeholder char variableLimit-1.
630                     varStart = buf.length();
631                     parser.appendVariableDef(name, buf, status);
632                     varLimit = buf.length();
633                 }
634             }
635             break;
636         case DOT:
637             buf.append(parser.getDotStandIn(status));
638             break;
639         case KLEENE_STAR:
640         case ONE_OR_MORE:
641         case ZERO_OR_ONE:
642             // Quantifiers.  We handle single characters, quoted strings,
643             // variable references, and segments.
644             //  a+      matches  aaa
645             //  'foo'+  matches  foofoofoo
646             //  $v+     matches  xyxyxy if $v == xy
647             //  (seg)+  matches  segsegseg
648             {
649                 if (isSegment && buf.length() == bufStart) {
650                     // The */+ immediately follows '('
651                     return syntaxError(U_MISPLACED_QUANTIFIER, rule, start, status);
652                 }
653 
654                 int32_t qstart, qlimit;
655                 // The */+ follows an isolated character or quote
656                 // or variable reference
657                 if (buf.length() == quoteLimit) {
658                     // The */+ follows a 'quoted string'
659                     qstart = quoteStart;
660                     qlimit = quoteLimit;
661                 } else if (buf.length() == varLimit) {
662                     // The */+ follows a $variableReference
663                     qstart = varStart;
664                     qlimit = varLimit;
665                 } else {
666                     // The */+ follows a single character, possibly
667                     // a segment standin
668                     qstart = buf.length() - 1;
669                     qlimit = qstart + 1;
670                 }
671 
672                 UnicodeFunctor *m =
673                     new StringMatcher(buf, qstart, qlimit, 0, *parser.curData);
674                 if (m == NULL) {
675                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
676                 }
677                 int32_t min = 0;
678                 int32_t max = Quantifier::MAX;
679                 switch (c) {
680                 case ONE_OR_MORE:
681                     min = 1;
682                     break;
683                 case ZERO_OR_ONE:
684                     min = 0;
685                     max = 1;
686                     break;
687                 // case KLEENE_STAR:
688                 //    do nothing -- min, max already set
689                 }
690                 m = new Quantifier(m, min, max);
691                 if (m == NULL) {
692                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
693                 }
694                 buf.truncate(qstart);
695                 buf.append(parser.generateStandInFor(m, status));
696             }
697             break;
698 
699         //------------------------------------------------------
700         // Elements allowed ONLY WITHIN segments
701         //------------------------------------------------------
702         case SEGMENT_CLOSE:
703             // assert(isSegment);
704             // We're done parsing a segment.
705             done = TRUE;
706             break;
707 
708         //------------------------------------------------------
709         // Elements allowed ONLY OUTSIDE segments
710         //------------------------------------------------------
711         case CONTEXT_ANTE:
712             if (ante >= 0) {
713                 return syntaxError(U_MULTIPLE_ANTE_CONTEXTS, rule, start, status);
714             }
715             ante = buf.length();
716             break;
717         case CONTEXT_POST:
718             if (post >= 0) {
719                 return syntaxError(U_MULTIPLE_POST_CONTEXTS, rule, start, status);
720             }
721             post = buf.length();
722             break;
723         case CURSOR_POS:
724             if (cursor >= 0) {
725                 return syntaxError(U_MULTIPLE_CURSORS, rule, start, status);
726             }
727             cursor = buf.length();
728             break;
729         case CURSOR_OFFSET:
730             if (cursorOffset < 0) {
731                 if (buf.length() > 0) {
732                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
733                 }
734                 --cursorOffset;
735             } else if (cursorOffset > 0) {
736                 if (buf.length() != cursorOffsetPos || cursor >= 0) {
737                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
738                 }
739                 ++cursorOffset;
740             } else {
741                 if (cursor == 0 && buf.length() == 0) {
742                     cursorOffset = -1;
743                 } else if (cursor < 0) {
744                     cursorOffsetPos = buf.length();
745                     cursorOffset = 1;
746                 } else {
747                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
748                 }
749             }
750             break;
751 
752 
753         //------------------------------------------------------
754         // Non-special characters
755         //------------------------------------------------------
756         default:
757             // Disallow unquoted characters other than [0-9A-Za-z]
758             // in the printable ASCII range.  These characters are
759             // reserved for possible future use.
760             if (c >= 0x0021 && c <= 0x007E &&
761                 !((c >= 0x0030/*'0'*/ && c <= 0x0039/*'9'*/) ||
762                   (c >= 0x0041/*'A'*/ && c <= 0x005A/*'Z'*/) ||
763                   (c >= 0x0061/*'a'*/ && c <= 0x007A/*'z'*/))) {
764                 return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
765             }
766             buf.append(c);
767             break;
768         }
769     }
770 
771     return pos;
772 }
773 
774 /**
775  * Remove context.
776  */
removeContext()777 void RuleHalf::removeContext() {
778     //text = text.substring(ante < 0 ? 0 : ante,
779     //                      post < 0 ? text.length() : post);
780     if (post >= 0) {
781         text.remove(post);
782     }
783     if (ante >= 0) {
784         text.removeBetween(0, ante);
785     }
786     ante = post = -1;
787     anchorStart = anchorEnd = FALSE;
788 }
789 
790 /**
791  * Return true if this half looks like valid output, that is, does not
792  * contain quantifiers or other special input-only elements.
793  */
isValidOutput(TransliteratorParser & transParser)794 UBool RuleHalf::isValidOutput(TransliteratorParser& transParser) {
795     for (int32_t i=0; i<text.length(); ) {
796         UChar32 c = text.char32At(i);
797         i += U16_LENGTH(c);
798         if (!transParser.parseData->isReplacer(c)) {
799             return FALSE;
800         }
801     }
802     return TRUE;
803 }
804 
805 /**
806  * Return true if this half looks like valid input, that is, does not
807  * contain functions or other special output-only elements.
808  */
isValidInput(TransliteratorParser & transParser)809 UBool RuleHalf::isValidInput(TransliteratorParser& transParser) {
810     for (int32_t i=0; i<text.length(); ) {
811         UChar32 c = text.char32At(i);
812         i += U16_LENGTH(c);
813         if (!transParser.parseData->isMatcher(c)) {
814             return FALSE;
815         }
816     }
817     return TRUE;
818 }
819 
820 //----------------------------------------------------------------------
821 // PUBLIC API
822 //----------------------------------------------------------------------
823 
824 /**
825  * Constructor.
826  */
TransliteratorParser(UErrorCode & statusReturn)827 TransliteratorParser::TransliteratorParser(UErrorCode &statusReturn) :
828 dataVector(statusReturn),
829 idBlockVector(statusReturn),
830 variablesVector(statusReturn),
831 segmentObjects(statusReturn)
832 {
833     idBlockVector.setDeleter(uprv_deleteUObject);
834     curData = NULL;
835     compoundFilter = NULL;
836     parseData = NULL;
837     variableNames.setValueDeleter(uprv_deleteUObject);
838 }
839 
840 /**
841  * Destructor.
842  */
~TransliteratorParser()843 TransliteratorParser::~TransliteratorParser() {
844     while (!dataVector.isEmpty())
845         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
846     delete compoundFilter;
847     delete parseData;
848     while (!variablesVector.isEmpty())
849         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
850 }
851 
852 void
parse(const UnicodeString & rules,UTransDirection transDirection,UParseError & pe,UErrorCode & ec)853 TransliteratorParser::parse(const UnicodeString& rules,
854                             UTransDirection transDirection,
855                             UParseError& pe,
856                             UErrorCode& ec) {
857     if (U_SUCCESS(ec)) {
858         parseRules(rules, transDirection, ec);
859         pe = parseError;
860     }
861 }
862 
863 /**
864  * Return the compound filter parsed by parse().  Caller owns result.
865  */
orphanCompoundFilter()866 UnicodeSet* TransliteratorParser::orphanCompoundFilter() {
867     UnicodeSet* f = compoundFilter;
868     compoundFilter = NULL;
869     return f;
870 }
871 
872 //----------------------------------------------------------------------
873 // Private implementation
874 //----------------------------------------------------------------------
875 
876 /**
877  * Parse the given string as a sequence of rules, separated by newline
878  * characters ('\n'), and cause this object to implement those rules.  Any
879  * previous rules are discarded.  Typically this method is called exactly
880  * once, during construction.
881  * @exception IllegalArgumentException if there is a syntax error in the
882  * rules
883  */
parseRules(const UnicodeString & rule,UTransDirection theDirection,UErrorCode & status)884 void TransliteratorParser::parseRules(const UnicodeString& rule,
885                                       UTransDirection theDirection,
886                                       UErrorCode& status)
887 {
888     // Clear error struct
889     uprv_memset(&parseError, 0, sizeof(parseError));
890     parseError.line = parseError.offset = -1;
891 
892     UBool parsingIDs = TRUE;
893     int32_t ruleCount = 0;
894 
895     while (!dataVector.isEmpty()) {
896         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
897     }
898     if (U_FAILURE(status)) {
899         return;
900     }
901 
902     idBlockVector.removeAllElements();
903     curData = NULL;
904     direction = theDirection;
905     ruleCount = 0;
906 
907     delete compoundFilter;
908     compoundFilter = NULL;
909 
910     while (!variablesVector.isEmpty()) {
911         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
912     }
913     variableNames.removeAll();
914     parseData = new ParseData(0, &variablesVector, &variableNames);
915     if (parseData == NULL) {
916         status = U_MEMORY_ALLOCATION_ERROR;
917         return;
918     }
919 
920     dotStandIn = (UChar) -1;
921 
922     UnicodeString *tempstr = NULL; // used for memory allocation error checking
923     UnicodeString str; // scratch
924     UnicodeString idBlockResult;
925     int32_t pos = 0;
926     int32_t limit = rule.length();
927 
928     // The compound filter offset is an index into idBlockResult.
929     // If it is 0, then the compound filter occurred at the start,
930     // and it is the offset to the _start_ of the compound filter
931     // pattern.  Otherwise it is the offset to the _limit_ of the
932     // compound filter pattern within idBlockResult.
933     compoundFilter = NULL;
934     int32_t compoundFilterOffset = -1;
935 
936     while (pos < limit && U_SUCCESS(status)) {
937         UChar c = rule.charAt(pos++);
938         if (PatternProps::isWhiteSpace(c)) {
939             // Ignore leading whitespace.
940             continue;
941         }
942         // Skip lines starting with the comment character
943         if (c == RULE_COMMENT_CHAR) {
944             pos = rule.indexOf((UChar)0x000A /*\n*/, pos) + 1;
945             if (pos == 0) {
946                 break; // No "\n" found; rest of rule is a commnet
947             }
948             continue; // Either fall out or restart with next line
949         }
950 
951         // skip empty rules
952         if (c == END_OF_RULE)
953             continue;
954 
955         // keep track of how many rules we've seen
956         ++ruleCount;
957 
958         // We've found the start of a rule or ID.  c is its first
959         // character, and pos points past c.
960         --pos;
961         // Look for an ID token.  Must have at least ID_TOKEN_LEN + 1
962         // chars left.
963         if ((pos + ID_TOKEN_LEN + 1) <= limit &&
964                 rule.compare(pos, ID_TOKEN_LEN, ID_TOKEN) == 0) {
965             pos += ID_TOKEN_LEN;
966             c = rule.charAt(pos);
967             while (PatternProps::isWhiteSpace(c) && pos < limit) {
968                 ++pos;
969                 c = rule.charAt(pos);
970             }
971 
972             int32_t p = pos;
973 
974             if (!parsingIDs) {
975                 if (curData != NULL) {
976                     if (direction == UTRANS_FORWARD)
977                         dataVector.addElement(curData, status);
978                     else
979                         dataVector.insertElementAt(curData, 0, status);
980                     curData = NULL;
981                 }
982                 parsingIDs = TRUE;
983             }
984 
985             TransliteratorIDParser::SingleID* id =
986                 TransliteratorIDParser::parseSingleID(rule, p, direction, status);
987             if (p != pos && ICU_Utility::parseChar(rule, p, END_OF_RULE)) {
988                 // Successful ::ID parse.
989 
990                 if (direction == UTRANS_FORWARD) {
991                     idBlockResult.append(id->canonID).append(END_OF_RULE);
992                 } else {
993                     idBlockResult.insert(0, END_OF_RULE);
994                     idBlockResult.insert(0, id->canonID);
995                 }
996 
997             } else {
998                 // Couldn't parse an ID.  Try to parse a global filter
999                 int32_t withParens = -1;
1000                 UnicodeSet* f = TransliteratorIDParser::parseGlobalFilter(rule, p, direction, withParens, NULL);
1001                 if (f != NULL) {
1002                     if (ICU_Utility::parseChar(rule, p, END_OF_RULE)
1003                         && (direction == UTRANS_FORWARD) == (withParens == 0))
1004                     {
1005                         if (compoundFilter != NULL) {
1006                             // Multiple compound filters
1007                             syntaxError(U_MULTIPLE_COMPOUND_FILTERS, rule, pos, status);
1008                             delete f;
1009                         } else {
1010                             compoundFilter = f;
1011                             compoundFilterOffset = ruleCount;
1012                         }
1013                     } else {
1014                         delete f;
1015                     }
1016                 } else {
1017                     // Invalid ::id
1018                     // Can be parsed as neither an ID nor a global filter
1019                     syntaxError(U_INVALID_ID, rule, pos, status);
1020                 }
1021             }
1022             delete id;
1023             pos = p;
1024         } else {
1025             if (parsingIDs) {
1026                 tempstr = new UnicodeString(idBlockResult);
1027                 // NULL pointer check
1028                 if (tempstr == NULL) {
1029                     status = U_MEMORY_ALLOCATION_ERROR;
1030                     return;
1031                 }
1032                 if (direction == UTRANS_FORWARD)
1033                     idBlockVector.addElement(tempstr, status);
1034                 else
1035                     idBlockVector.insertElementAt(tempstr, 0, status);
1036                 idBlockResult.remove();
1037                 parsingIDs = FALSE;
1038                 curData = new TransliterationRuleData(status);
1039                 // NULL pointer check
1040                 if (curData == NULL) {
1041                     status = U_MEMORY_ALLOCATION_ERROR;
1042                     return;
1043                 }
1044                 parseData->data = curData;
1045 
1046                 // By default, rules use part of the private use area
1047                 // E000..F8FF for variables and other stand-ins.  Currently
1048                 // the range F000..F8FF is typically sufficient.  The 'use
1049                 // variable range' pragma allows rule sets to modify this.
1050                 setVariableRange(0xF000, 0xF8FF, status);
1051             }
1052 
1053             if (resemblesPragma(rule, pos, limit)) {
1054                 int32_t ppp = parsePragma(rule, pos, limit, status);
1055                 if (ppp < 0) {
1056                     syntaxError(U_MALFORMED_PRAGMA, rule, pos, status);
1057                 }
1058                 pos = ppp;
1059             // Parse a rule
1060             } else {
1061                 pos = parseRule(rule, pos, limit, status);
1062             }
1063         }
1064     }
1065 
1066     if (parsingIDs && idBlockResult.length() > 0) {
1067         tempstr = new UnicodeString(idBlockResult);
1068         // NULL pointer check
1069         if (tempstr == NULL) {
1070             status = U_MEMORY_ALLOCATION_ERROR;
1071             return;
1072         }
1073         if (direction == UTRANS_FORWARD)
1074             idBlockVector.addElement(tempstr, status);
1075         else
1076             idBlockVector.insertElementAt(tempstr, 0, status);
1077     }
1078     else if (!parsingIDs && curData != NULL) {
1079         if (direction == UTRANS_FORWARD)
1080             dataVector.addElement(curData, status);
1081         else
1082             dataVector.insertElementAt(curData, 0, status);
1083     }
1084 
1085     if (U_SUCCESS(status)) {
1086         // Convert the set vector to an array
1087         int32_t i, dataVectorSize = dataVector.size();
1088         for (i = 0; i < dataVectorSize; i++) {
1089             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1090             data->variablesLength = variablesVector.size();
1091             if (data->variablesLength == 0) {
1092                 data->variables = 0;
1093             } else {
1094                 data->variables = (UnicodeFunctor**)uprv_malloc(data->variablesLength * sizeof(UnicodeFunctor*));
1095                 // NULL pointer check
1096                 if (data->variables == NULL) {
1097                     status = U_MEMORY_ALLOCATION_ERROR;
1098                     return;
1099                 }
1100                 data->variablesAreOwned = (i == 0);
1101             }
1102 
1103             for (int32_t j = 0; j < data->variablesLength; j++) {
1104                 data->variables[j] =
1105                     ((UnicodeSet*)variablesVector.elementAt(j));
1106             }
1107 
1108             data->variableNames.removeAll();
1109             int32_t pos = UHASH_FIRST;
1110             const UHashElement* he = variableNames.nextElement(pos);
1111             while (he != NULL) {
1112                 UnicodeString* tempus = (UnicodeString*)(((UnicodeString*)(he->value.pointer))->clone());
1113                 if (tempus == NULL) {
1114                     status = U_MEMORY_ALLOCATION_ERROR;
1115                     return;
1116                 }
1117                 data->variableNames.put(*((UnicodeString*)(he->key.pointer)),
1118                     tempus, status);
1119                 he = variableNames.nextElement(pos);
1120             }
1121         }
1122         variablesVector.removeAllElements();   // keeps them from getting deleted when we succeed
1123 
1124         // Index the rules
1125         if (compoundFilter != NULL) {
1126             if ((direction == UTRANS_FORWARD && compoundFilterOffset != 1) ||
1127                 (direction == UTRANS_REVERSE && compoundFilterOffset != ruleCount)) {
1128                 status = U_MISPLACED_COMPOUND_FILTER;
1129             }
1130         }
1131 
1132         for (i = 0; i < dataVectorSize; i++) {
1133             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1134             data->ruleSet.freeze(parseError, status);
1135         }
1136         if (idBlockVector.size() == 1 && ((UnicodeString*)idBlockVector.elementAt(0))->isEmpty()) {
1137             idBlockVector.removeElementAt(0);
1138         }
1139     }
1140 }
1141 
1142 /**
1143  * Set the variable range to [start, end] (inclusive).
1144  */
setVariableRange(int32_t start,int32_t end,UErrorCode & status)1145 void TransliteratorParser::setVariableRange(int32_t start, int32_t end, UErrorCode& status) {
1146     if (start > end || start < 0 || end > 0xFFFF) {
1147         status = U_MALFORMED_PRAGMA;
1148         return;
1149     }
1150 
1151     curData->variablesBase = (UChar) start;
1152     if (dataVector.size() == 0) {
1153         variableNext = (UChar) start;
1154         variableLimit = (UChar) (end + 1);
1155     }
1156 }
1157 
1158 /**
1159  * Assert that the given character is NOT within the variable range.
1160  * If it is, return FALSE.  This is neccesary to ensure that the
1161  * variable range does not overlap characters used in a rule.
1162  */
checkVariableRange(UChar32 ch) const1163 UBool TransliteratorParser::checkVariableRange(UChar32 ch) const {
1164     return !(ch >= curData->variablesBase && ch < variableLimit);
1165 }
1166 
1167 /**
1168  * Set the maximum backup to 'backup', in response to a pragma
1169  * statement.
1170  */
pragmaMaximumBackup(int32_t)1171 void TransliteratorParser::pragmaMaximumBackup(int32_t /*backup*/) {
1172     //TODO Finish
1173 }
1174 
1175 /**
1176  * Begin normalizing all rules using the given mode, in response
1177  * to a pragma statement.
1178  */
pragmaNormalizeRules(UNormalizationMode)1179 void TransliteratorParser::pragmaNormalizeRules(UNormalizationMode /*mode*/) {
1180     //TODO Finish
1181 }
1182 
1183 static const UChar PRAGMA_USE[] = {0x75,0x73,0x65,0x20,0}; // "use "
1184 
1185 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 # #~;"
1186 
1187 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 #~;"
1188 
1189 static const UChar PRAGMA_NFD_RULES[] = {0x7E,0x6E,0x66,0x64,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfd rules~;"
1190 
1191 static const UChar PRAGMA_NFC_RULES[] = {0x7E,0x6E,0x66,0x63,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfc rules~;"
1192 
1193 /**
1194  * Return true if the given rule looks like a pragma.
1195  * @param pos offset to the first non-whitespace character
1196  * of the rule.
1197  * @param limit pointer past the last character of the rule.
1198  */
resemblesPragma(const UnicodeString & rule,int32_t pos,int32_t limit)1199 UBool TransliteratorParser::resemblesPragma(const UnicodeString& rule, int32_t pos, int32_t limit) {
1200     // Must start with /use\s/i
1201     return ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_USE, 4), NULL) >= 0;
1202 }
1203 
1204 /**
1205  * Parse a pragma.  This method assumes resemblesPragma() has
1206  * already returned true.
1207  * @param pos offset to the first non-whitespace character
1208  * of the rule.
1209  * @param limit pointer past the last character of the rule.
1210  * @return the position index after the final ';' of the pragma,
1211  * or -1 on failure.
1212  */
parsePragma(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1213 int32_t TransliteratorParser::parsePragma(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1214     int32_t array[2];
1215 
1216     // resemblesPragma() has already returned true, so we
1217     // know that pos points to /use\s/i; we can skip 4 characters
1218     // immediately
1219     pos += 4;
1220 
1221     // Here are the pragmas we recognize:
1222     // use variable range 0xE000 0xEFFF;
1223     // use maximum backup 16;
1224     // use nfd rules;
1225     // use nfc rules;
1226     int p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_VARIABLE_RANGE, -1), array);
1227     if (p >= 0) {
1228         setVariableRange(array[0], array[1], status);
1229         return p;
1230     }
1231 
1232     p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_MAXIMUM_BACKUP, -1), array);
1233     if (p >= 0) {
1234         pragmaMaximumBackup(array[0]);
1235         return p;
1236     }
1237 
1238     p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_NFD_RULES, -1), NULL);
1239     if (p >= 0) {
1240         pragmaNormalizeRules(UNORM_NFD);
1241         return p;
1242     }
1243 
1244     p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_NFC_RULES, -1), NULL);
1245     if (p >= 0) {
1246         pragmaNormalizeRules(UNORM_NFC);
1247         return p;
1248     }
1249 
1250     // Syntax error: unable to parse pragma
1251     return -1;
1252 }
1253 
1254 /**
1255  * MAIN PARSER.  Parse the next rule in the given rule string, starting
1256  * at pos.  Return the index after the last character parsed.  Do not
1257  * parse characters at or after limit.
1258  *
1259  * Important:  The character at pos must be a non-whitespace character
1260  * that is not the comment character.
1261  *
1262  * This method handles quoting, escaping, and whitespace removal.  It
1263  * parses the end-of-rule character.  It recognizes context and cursor
1264  * indicators.  Once it does a lexical breakdown of the rule at pos, it
1265  * creates a rule object and adds it to our rule list.
1266  */
parseRule(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1267 int32_t TransliteratorParser::parseRule(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1268     // Locate the left side, operator, and right side
1269     int32_t start = pos;
1270     UChar op = 0;
1271     int32_t i;
1272 
1273     // Set up segments data
1274     segmentStandins.truncate(0);
1275     segmentObjects.removeAllElements();
1276 
1277     // Use pointers to automatics to make swapping possible.
1278     RuleHalf _left(*this), _right(*this);
1279     RuleHalf* left = &_left;
1280     RuleHalf* right = &_right;
1281 
1282     undefinedVariableName.remove();
1283     pos = left->parse(rule, pos, limit, status);
1284     if (U_FAILURE(status)) {
1285         return start;
1286     }
1287 
1288     if (pos == limit || u_strchr(gOPERATORS, (op = rule.charAt(--pos))) == NULL) {
1289         return syntaxError(U_MISSING_OPERATOR, rule, start, status);
1290     }
1291     ++pos;
1292 
1293     // Found an operator char.  Check for forward-reverse operator.
1294     if (op == REVERSE_RULE_OP &&
1295         (pos < limit && rule.charAt(pos) == FORWARD_RULE_OP)) {
1296         ++pos;
1297         op = FWDREV_RULE_OP;
1298     }
1299 
1300     // Translate alternate op characters.
1301     switch (op) {
1302     case ALT_FORWARD_RULE_OP:
1303         op = FORWARD_RULE_OP;
1304         break;
1305     case ALT_REVERSE_RULE_OP:
1306         op = REVERSE_RULE_OP;
1307         break;
1308     case ALT_FWDREV_RULE_OP:
1309         op = FWDREV_RULE_OP;
1310         break;
1311     }
1312 
1313     pos = right->parse(rule, pos, limit, status);
1314     if (U_FAILURE(status)) {
1315         return start;
1316     }
1317 
1318     if (pos < limit) {
1319         if (rule.charAt(--pos) == END_OF_RULE) {
1320             ++pos;
1321         } else {
1322             // RuleHalf parser must have terminated at an operator
1323             return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
1324         }
1325     }
1326 
1327     if (op == VARIABLE_DEF_OP) {
1328         // LHS is the name.  RHS is a single character, either a literal
1329         // or a set (already parsed).  If RHS is longer than one
1330         // character, it is either a multi-character string, or multiple
1331         // sets, or a mixture of chars and sets -- syntax error.
1332 
1333         // We expect to see a single undefined variable (the one being
1334         // defined).
1335         if (undefinedVariableName.length() == 0) {
1336             // "Missing '$' or duplicate definition"
1337             return syntaxError(U_BAD_VARIABLE_DEFINITION, rule, start, status);
1338         }
1339         if (left->text.length() != 1 || left->text.charAt(0) != variableLimit) {
1340             // "Malformed LHS"
1341             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1342         }
1343         if (left->anchorStart || left->anchorEnd ||
1344             right->anchorStart || right->anchorEnd) {
1345             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1346         }
1347         // We allow anything on the right, including an empty string.
1348         UnicodeString* value = new UnicodeString(right->text);
1349         // NULL pointer check
1350         if (value == NULL) {
1351             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1352         }
1353         variableNames.put(undefinedVariableName, value, status);
1354         ++variableLimit;
1355         return pos;
1356     }
1357 
1358     // If this is not a variable definition rule, we shouldn't have
1359     // any undefined variable names.
1360     if (undefinedVariableName.length() != 0) {
1361         return syntaxError(// "Undefined variable $" + undefinedVariableName,
1362                     U_UNDEFINED_VARIABLE,
1363                     rule, start, status);
1364     }
1365 
1366     // Verify segments
1367     if (segmentStandins.length() > segmentObjects.size()) {
1368         syntaxError(U_UNDEFINED_SEGMENT_REFERENCE, rule, start, status);
1369     }
1370     for (i=0; i<segmentStandins.length(); ++i) {
1371         if (segmentStandins.charAt(i) == 0) {
1372             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1373         }
1374     }
1375     for (i=0; i<segmentObjects.size(); ++i) {
1376         if (segmentObjects.elementAt(i) == NULL) {
1377             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1378         }
1379     }
1380 
1381     // If the direction we want doesn't match the rule
1382     // direction, do nothing.
1383     if (op != FWDREV_RULE_OP &&
1384         ((direction == UTRANS_FORWARD) != (op == FORWARD_RULE_OP))) {
1385         return pos;
1386     }
1387 
1388     // Transform the rule into a forward rule by swapping the
1389     // sides if necessary.
1390     if (direction == UTRANS_REVERSE) {
1391         left = &_right;
1392         right = &_left;
1393     }
1394 
1395     // Remove non-applicable elements in forward-reverse
1396     // rules.  Bidirectional rules ignore elements that do not
1397     // apply.
1398     if (op == FWDREV_RULE_OP) {
1399         right->removeContext();
1400         left->cursor = -1;
1401         left->cursorOffset = 0;
1402     }
1403 
1404     // Normalize context
1405     if (left->ante < 0) {
1406         left->ante = 0;
1407     }
1408     if (left->post < 0) {
1409         left->post = left->text.length();
1410     }
1411 
1412     // Context is only allowed on the input side.  Cursors are only
1413     // allowed on the output side.  Segment delimiters can only appear
1414     // on the left, and references on the right.  Cursor offset
1415     // cannot appear without an explicit cursor.  Cursor offset
1416     // cannot place the cursor outside the limits of the context.
1417     // Anchors are only allowed on the input side.
1418     if (right->ante >= 0 || right->post >= 0 || left->cursor >= 0 ||
1419         (right->cursorOffset != 0 && right->cursor < 0) ||
1420         // - The following two checks were used to ensure that the
1421         // - the cursor offset stayed within the ante- or postcontext.
1422         // - However, with the addition of quantifiers, we have to
1423         // - allow arbitrary cursor offsets and do runtime checking.
1424         //(right->cursorOffset > (left->text.length() - left->post)) ||
1425         //(-right->cursorOffset > left->ante) ||
1426         right->anchorStart || right->anchorEnd ||
1427         !left->isValidInput(*this) || !right->isValidOutput(*this) ||
1428         left->ante > left->post) {
1429 
1430         return syntaxError(U_MALFORMED_RULE, rule, start, status);
1431     }
1432 
1433     // Flatten segment objects vector to an array
1434     UnicodeFunctor** segmentsArray = NULL;
1435     if (segmentObjects.size() > 0) {
1436         segmentsArray = (UnicodeFunctor **)uprv_malloc(segmentObjects.size() * sizeof(UnicodeFunctor *));
1437         // Null pointer check
1438         if (segmentsArray == NULL) {
1439             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1440         }
1441         segmentObjects.toArray((void**) segmentsArray);
1442     }
1443     TransliterationRule* temptr = new TransliterationRule(
1444             left->text, left->ante, left->post,
1445             right->text, right->cursor, right->cursorOffset,
1446             segmentsArray,
1447             segmentObjects.size(),
1448             left->anchorStart, left->anchorEnd,
1449             curData,
1450             status);
1451     //Null pointer check
1452     if (temptr == NULL) {
1453         uprv_free(segmentsArray);
1454         return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1455     }
1456 
1457     curData->ruleSet.addRule(temptr, status);
1458 
1459     return pos;
1460 }
1461 
1462 /**
1463  * Called by main parser upon syntax error.  Search the rule string
1464  * for the probable end of the rule.  Of course, if the error is that
1465  * the end of rule marker is missing, then the rule end will not be found.
1466  * In any case the rule start will be correctly reported.
1467  * @param msg error description
1468  * @param rule pattern string
1469  * @param start position of first character of current rule
1470  */
syntaxError(UErrorCode parseErrorCode,const UnicodeString & rule,int32_t pos,UErrorCode & status)1471 int32_t TransliteratorParser::syntaxError(UErrorCode parseErrorCode,
1472                                           const UnicodeString& rule,
1473                                           int32_t pos,
1474                                           UErrorCode& status)
1475 {
1476     parseError.offset = pos;
1477     parseError.line = 0 ; /* we are not using line numbers */
1478 
1479     // for pre-context
1480     const int32_t LEN = U_PARSE_CONTEXT_LEN - 1;
1481     int32_t start = uprv_max(pos - LEN, 0);
1482     int32_t stop  = pos;
1483 
1484     rule.extract(start,stop-start,parseError.preContext);
1485     //null terminate the buffer
1486     parseError.preContext[stop-start] = 0;
1487 
1488     //for post-context
1489     start = pos;
1490     stop  = uprv_min(pos + LEN, rule.length());
1491 
1492     rule.extract(start,stop-start,parseError.postContext);
1493     //null terminate the buffer
1494     parseError.postContext[stop-start]= 0;
1495 
1496     status = (UErrorCode)parseErrorCode;
1497     return pos;
1498 
1499 }
1500 
1501 /**
1502  * Parse a UnicodeSet out, store it, and return the stand-in character
1503  * used to represent it.
1504  */
parseSet(const UnicodeString & rule,ParsePosition & pos,UErrorCode & status)1505 UChar TransliteratorParser::parseSet(const UnicodeString& rule,
1506                                           ParsePosition& pos,
1507                                           UErrorCode& status) {
1508     UnicodeSet* set = new UnicodeSet(rule, pos, USET_IGNORE_SPACE, parseData, status);
1509     // Null pointer check
1510     if (set == NULL) {
1511         status = U_MEMORY_ALLOCATION_ERROR;
1512         return (UChar)0x0000; // Return empty character with error.
1513     }
1514     set->compact();
1515     return generateStandInFor(set, status);
1516 }
1517 
1518 /**
1519  * Generate and return a stand-in for a new UnicodeFunctor.  Store
1520  * the matcher (adopt it).
1521  */
generateStandInFor(UnicodeFunctor * adopted,UErrorCode & status)1522 UChar TransliteratorParser::generateStandInFor(UnicodeFunctor* adopted, UErrorCode& status) {
1523     // assert(obj != null);
1524 
1525     // Look up previous stand-in, if any.  This is a short list
1526     // (typical n is 0, 1, or 2); linear search is optimal.
1527     for (int32_t i=0; i<variablesVector.size(); ++i) {
1528         if (variablesVector.elementAt(i) == adopted) { // [sic] pointer comparison
1529             return (UChar) (curData->variablesBase + i);
1530         }
1531     }
1532 
1533     if (variableNext >= variableLimit) {
1534         delete adopted;
1535         status = U_VARIABLE_RANGE_EXHAUSTED;
1536         return 0;
1537     }
1538     variablesVector.addElement(adopted, status);
1539     return variableNext++;
1540 }
1541 
1542 /**
1543  * Return the standin for segment seg (1-based).
1544  */
getSegmentStandin(int32_t seg,UErrorCode & status)1545 UChar TransliteratorParser::getSegmentStandin(int32_t seg, UErrorCode& status) {
1546     // Special character used to indicate an empty spot
1547     UChar empty = curData->variablesBase - 1;
1548     while (segmentStandins.length() < seg) {
1549         segmentStandins.append(empty);
1550     }
1551     UChar c = segmentStandins.charAt(seg-1);
1552     if (c == empty) {
1553         if (variableNext >= variableLimit) {
1554             status = U_VARIABLE_RANGE_EXHAUSTED;
1555             return 0;
1556         }
1557         c = variableNext++;
1558         // Set a placeholder in the master variables vector that will be
1559         // filled in later by setSegmentObject().  We know that we will get
1560         // called first because setSegmentObject() will call us.
1561         variablesVector.addElement((void*) NULL, status);
1562         segmentStandins.setCharAt(seg-1, c);
1563     }
1564     return c;
1565 }
1566 
1567 /**
1568  * Set the object for segment seg (1-based).
1569  */
setSegmentObject(int32_t seg,StringMatcher * adopted,UErrorCode & status)1570 void TransliteratorParser::setSegmentObject(int32_t seg, StringMatcher* adopted, UErrorCode& status) {
1571     // Since we call parseSection() recursively, nested
1572     // segments will result in segment i+1 getting parsed
1573     // and stored before segment i; be careful with the
1574     // vector handling here.
1575     if (segmentObjects.size() < seg) {
1576         segmentObjects.setSize(seg, status);
1577     }
1578     int32_t index = getSegmentStandin(seg, status) - curData->variablesBase;
1579     if (segmentObjects.elementAt(seg-1) != NULL ||
1580         variablesVector.elementAt(index) != NULL) {
1581         // should never happen
1582         status = U_INTERNAL_TRANSLITERATOR_ERROR;
1583         return;
1584     }
1585     segmentObjects.setElementAt(adopted, seg-1);
1586     variablesVector.setElementAt(adopted, index);
1587 }
1588 
1589 /**
1590  * Return the stand-in for the dot set.  It is allocated the first
1591  * time and reused thereafter.
1592  */
getDotStandIn(UErrorCode & status)1593 UChar TransliteratorParser::getDotStandIn(UErrorCode& status) {
1594     if (dotStandIn == (UChar) -1) {
1595         UnicodeSet* tempus = new UnicodeSet(UnicodeString(TRUE, DOT_SET, -1), status);
1596         // Null pointer check.
1597         if (tempus == NULL) {
1598             status = U_MEMORY_ALLOCATION_ERROR;
1599             return (UChar)0x0000;
1600         }
1601         dotStandIn = generateStandInFor(tempus, status);
1602     }
1603     return dotStandIn;
1604 }
1605 
1606 /**
1607  * Append the value of the given variable name to the given
1608  * UnicodeString.
1609  */
appendVariableDef(const UnicodeString & name,UnicodeString & buf,UErrorCode & status)1610 void TransliteratorParser::appendVariableDef(const UnicodeString& name,
1611                                                   UnicodeString& buf,
1612                                                   UErrorCode& status) {
1613     const UnicodeString* s = (const UnicodeString*) variableNames.get(name);
1614     if (s == NULL) {
1615         // We allow one undefined variable so that variable definition
1616         // statements work.  For the first undefined variable we return
1617         // the special placeholder variableLimit-1, and save the variable
1618         // name.
1619         if (undefinedVariableName.length() == 0) {
1620             undefinedVariableName = name;
1621             if (variableNext >= variableLimit) {
1622                 // throw new RuntimeException("Private use variables exhausted");
1623                 status = U_ILLEGAL_ARGUMENT_ERROR;
1624                 return;
1625             }
1626             buf.append((UChar) --variableLimit);
1627         } else {
1628             //throw new IllegalArgumentException("Undefined variable $"
1629             //                                   + name);
1630             status = U_ILLEGAL_ARGUMENT_ERROR;
1631             return;
1632         }
1633     } else {
1634         buf.append(*s);
1635     }
1636 }
1637 
1638 /**
1639  * Glue method to get around access restrictions in C++.
1640  */
1641 /*Transliterator* TransliteratorParser::createBasicInstance(const UnicodeString& id, const UnicodeString* canonID) {
1642     return Transliterator::createBasicInstance(id, canonID);
1643 }*/
1644 
1645 U_NAMESPACE_END
1646 
1647 U_CAPI int32_t
utrans_stripRules(const UChar * source,int32_t sourceLen,UChar * target,UErrorCode * status)1648 utrans_stripRules(const UChar *source, int32_t sourceLen, UChar *target, UErrorCode *status) {
1649     U_NAMESPACE_USE
1650 
1651     //const UChar *sourceStart = source;
1652     const UChar *targetStart = target;
1653     const UChar *sourceLimit = source+sourceLen;
1654     UChar *targetLimit = target+sourceLen;
1655     UChar32 c = 0;
1656     UBool quoted = FALSE;
1657     int32_t index;
1658 
1659     uprv_memset(target, 0, sourceLen*U_SIZEOF_UCHAR);
1660 
1661     /* read the rules into the buffer */
1662     while (source < sourceLimit)
1663     {
1664         index=0;
1665         U16_NEXT_UNSAFE(source, index, c);
1666         source+=index;
1667         if(c == QUOTE) {
1668             quoted = (UBool)!quoted;
1669         }
1670         else if (!quoted) {
1671             if (c == RULE_COMMENT_CHAR) {
1672                 /* skip comments and all preceding spaces */
1673                 while (targetStart < target && *(target - 1) == 0x0020) {
1674                     target--;
1675                 }
1676                 do {
1677                     c = *(source++);
1678                 }
1679                 while (c != CR && c != LF);
1680             }
1681             else if (c == ESCAPE) {
1682                 UChar32   c2 = *source;
1683                 if (c2 == CR || c2 == LF) {
1684                     /* A backslash at the end of a line. */
1685                     /* Since we're stripping lines, ignore the backslash. */
1686                     source++;
1687                     continue;
1688                 }
1689                 if (c2 == 0x0075 && source+5 < sourceLimit) { /* \u seen. \U isn't unescaped. */
1690                     int32_t escapeOffset = 0;
1691                     UnicodeString escapedStr(source, 5);
1692                     c2 = escapedStr.unescapeAt(escapeOffset);
1693 
1694                     if (c2 == (UChar32)0xFFFFFFFF || escapeOffset == 0)
1695                     {
1696                         *status = U_PARSE_ERROR;
1697                         return 0;
1698                     }
1699                     if (!PatternProps::isWhiteSpace(c2) && !u_iscntrl(c2) && !u_ispunct(c2)) {
1700                         /* It was escaped for a reason. Write what it was suppose to be. */
1701                         source+=5;
1702                         c = c2;
1703                     }
1704                 }
1705                 else if (c2 == QUOTE) {
1706                     /* \' seen. Make sure we don't do anything when we see it again. */
1707                     quoted = (UBool)!quoted;
1708                 }
1709             }
1710         }
1711         if (c == CR || c == LF)
1712         {
1713             /* ignore spaces carriage returns, and all leading spaces on the next line.
1714             * and line feed unless in the form \uXXXX
1715             */
1716             quoted = FALSE;
1717             while (source < sourceLimit) {
1718                 c = *(source);
1719                 if (c != CR && c != LF && c != 0x0020) {
1720                     break;
1721                 }
1722                 source++;
1723             }
1724             continue;
1725         }
1726 
1727         /* Append UChar * after dissembling if c > 0xffff*/
1728         index=0;
1729         U16_APPEND_UNSAFE(target, index, c);
1730         target+=index;
1731     }
1732     if (target < targetLimit) {
1733         *target = 0;
1734     }
1735     return (int32_t)(target-targetStart);
1736 }
1737 
1738 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
1739