• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // Copyright (C) 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
3 /*
4  **********************************************************************
5  *   Copyright (C) 1999-2016, International Business Machines
6  *   Corporation and others.  All Rights Reserved.
7  **********************************************************************
8  *   Date        Name        Description
9  *   11/17/99    aliu        Creation.
10  **********************************************************************
11  */
12 
13 #include "unicode/utypes.h"
14 
15 #if !UCONFIG_NO_TRANSLITERATION
16 
17 #include "unicode/uobject.h"
18 #include "unicode/parseerr.h"
19 #include "unicode/parsepos.h"
20 #include "unicode/putil.h"
21 #include "unicode/uchar.h"
22 #include "unicode/ustring.h"
23 #include "unicode/uniset.h"
24 #include "unicode/utf16.h"
25 #include "cstring.h"
26 #include "funcrepl.h"
27 #include "hash.h"
28 #include "quant.h"
29 #include "rbt.h"
30 #include "rbt_data.h"
31 #include "rbt_pars.h"
32 #include "rbt_rule.h"
33 #include "strmatch.h"
34 #include "strrepl.h"
35 #include "unicode/symtable.h"
36 #include "tridpars.h"
37 #include "uvector.h"
38 #include "hash.h"
39 #include "patternprops.h"
40 #include "util.h"
41 #include "cmemory.h"
42 #include "uprops.h"
43 #include "putilimp.h"
44 
45 // Operators
46 #define VARIABLE_DEF_OP ((UChar)0x003D) /*=*/
47 #define FORWARD_RULE_OP ((UChar)0x003E) /*>*/
48 #define REVERSE_RULE_OP ((UChar)0x003C) /*<*/
49 #define FWDREV_RULE_OP  ((UChar)0x007E) /*~*/ // internal rep of <> op
50 
51 // Other special characters
52 #define QUOTE             ((UChar)0x0027) /*'*/
53 #define ESCAPE            ((UChar)0x005C) /*\*/
54 #define END_OF_RULE       ((UChar)0x003B) /*;*/
55 #define RULE_COMMENT_CHAR ((UChar)0x0023) /*#*/
56 
57 #define SEGMENT_OPEN       ((UChar)0x0028) /*(*/
58 #define SEGMENT_CLOSE      ((UChar)0x0029) /*)*/
59 #define CONTEXT_ANTE       ((UChar)0x007B) /*{*/
60 #define CONTEXT_POST       ((UChar)0x007D) /*}*/
61 #define CURSOR_POS         ((UChar)0x007C) /*|*/
62 #define CURSOR_OFFSET      ((UChar)0x0040) /*@*/
63 #define ANCHOR_START       ((UChar)0x005E) /*^*/
64 #define KLEENE_STAR        ((UChar)0x002A) /***/
65 #define ONE_OR_MORE        ((UChar)0x002B) /*+*/
66 #define ZERO_OR_ONE        ((UChar)0x003F) /*?*/
67 
68 #define DOT                ((UChar)46)     /*.*/
69 
70 static const UChar DOT_SET[] = { // "[^[:Zp:][:Zl:]\r\n$]";
71     91, 94, 91, 58, 90, 112, 58, 93, 91, 58, 90,
72     108, 58, 93, 92, 114, 92, 110, 36, 93, 0
73 };
74 
75 // A function is denoted &Source-Target/Variant(text)
76 #define FUNCTION           ((UChar)38)     /*&*/
77 
78 // Aliases for some of the syntax characters. These are provided so
79 // transliteration rules can be expressed in XML without clashing with
80 // XML syntax characters '<', '>', and '&'.
81 #define ALT_REVERSE_RULE_OP ((UChar)0x2190) // Left Arrow
82 #define ALT_FORWARD_RULE_OP ((UChar)0x2192) // Right Arrow
83 #define ALT_FWDREV_RULE_OP  ((UChar)0x2194) // Left Right Arrow
84 #define ALT_FUNCTION        ((UChar)0x2206) // Increment (~Greek Capital Delta)
85 
86 // Special characters disallowed at the top level
87 static const UChar ILLEGAL_TOP[] = {41,0}; // ")"
88 
89 // Special characters disallowed within a segment
90 static const UChar ILLEGAL_SEG[] = {123,125,124,64,0}; // "{}|@"
91 
92 // Special characters disallowed within a function argument
93 static const UChar ILLEGAL_FUNC[] = {94,40,46,42,43,63,123,125,124,64,0}; // "^(.*+?{}|@"
94 
95 // By definition, the ANCHOR_END special character is a
96 // trailing SymbolTable.SYMBOL_REF character.
97 // private static final char ANCHOR_END       = '$';
98 
99 static const UChar gOPERATORS[] = { // "=><"
100     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
101     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
102     0
103 };
104 
105 static const UChar HALF_ENDERS[] = { // "=><;"
106     VARIABLE_DEF_OP, FORWARD_RULE_OP, REVERSE_RULE_OP,
107     ALT_FORWARD_RULE_OP, ALT_REVERSE_RULE_OP, ALT_FWDREV_RULE_OP,
108     END_OF_RULE,
109     0
110 };
111 
112 // These are also used in Transliterator::toRules()
113 static const int32_t ID_TOKEN_LEN = 2;
114 static const UChar   ID_TOKEN[]   = { 0x3A, 0x3A }; // ':', ':'
115 
116 /*
117 commented out until we do real ::BEGIN/::END functionality
118 static const int32_t BEGIN_TOKEN_LEN = 5;
119 static const UChar BEGIN_TOKEN[] = { 0x42, 0x45, 0x47, 0x49, 0x4e }; // 'BEGIN'
120 
121 static const int32_t END_TOKEN_LEN = 3;
122 static const UChar END_TOKEN[] = { 0x45, 0x4e, 0x44 }; // 'END'
123 */
124 
125 U_NAMESPACE_BEGIN
126 
127 //----------------------------------------------------------------------
128 // BEGIN ParseData
129 //----------------------------------------------------------------------
130 
131 /**
132  * This class implements the SymbolTable interface.  It is used
133  * during parsing to give UnicodeSet access to variables that
134  * have been defined so far.  Note that it uses variablesVector,
135  * _not_ data.setVariables.
136  */
137 class ParseData : public UMemory, public SymbolTable {
138 public:
139     const TransliterationRuleData* data; // alias
140 
141     const UVector* variablesVector; // alias
142 
143     const Hashtable* variableNames; // alias
144 
145     ParseData(const TransliterationRuleData* data = 0,
146               const UVector* variablesVector = 0,
147               const Hashtable* variableNames = 0);
148 
149     virtual ~ParseData();
150 
151     virtual const UnicodeString* lookup(const UnicodeString& s) const;
152 
153     virtual const UnicodeFunctor* lookupMatcher(UChar32 ch) const;
154 
155     virtual UnicodeString parseReference(const UnicodeString& text,
156                                          ParsePosition& pos, int32_t limit) const;
157     /**
158      * Return true if the given character is a matcher standin or a plain
159      * character (non standin).
160      */
161     UBool isMatcher(UChar32 ch);
162 
163     /**
164      * Return true if the given character is a replacer standin or a plain
165      * character (non standin).
166      */
167     UBool isReplacer(UChar32 ch);
168 
169 private:
170     ParseData(const ParseData &other); // forbid copying of this class
171     ParseData &operator=(const ParseData &other); // forbid copying of this class
172 };
173 
ParseData(const TransliterationRuleData * d,const UVector * sets,const Hashtable * vNames)174 ParseData::ParseData(const TransliterationRuleData* d,
175                      const UVector* sets,
176                      const Hashtable* vNames) :
177     data(d), variablesVector(sets), variableNames(vNames) {}
178 
~ParseData()179 ParseData::~ParseData() {}
180 
181 /**
182  * Implement SymbolTable API.
183  */
lookup(const UnicodeString & name) const184 const UnicodeString* ParseData::lookup(const UnicodeString& name) const {
185     return (const UnicodeString*) variableNames->get(name);
186 }
187 
188 /**
189  * Implement SymbolTable API.
190  */
lookupMatcher(UChar32 ch) const191 const UnicodeFunctor* ParseData::lookupMatcher(UChar32 ch) const {
192     // Note that we cannot use data.lookupSet() because the
193     // set array has not been constructed yet.
194     const UnicodeFunctor* set = NULL;
195     int32_t i = ch - data->variablesBase;
196     if (i >= 0 && i < variablesVector->size()) {
197         int32_t i = ch - data->variablesBase;
198         set = (i < variablesVector->size()) ?
199             (UnicodeFunctor*) variablesVector->elementAt(i) : 0;
200     }
201     return set;
202 }
203 
204 /**
205  * Implement SymbolTable API.  Parse out a symbol reference
206  * name.
207  */
parseReference(const UnicodeString & text,ParsePosition & pos,int32_t limit) const208 UnicodeString ParseData::parseReference(const UnicodeString& text,
209                                         ParsePosition& pos, int32_t limit) const {
210     int32_t start = pos.getIndex();
211     int32_t i = start;
212     UnicodeString result;
213     while (i < limit) {
214         UChar c = text.charAt(i);
215         if ((i==start && !u_isIDStart(c)) || !u_isIDPart(c)) {
216             break;
217         }
218         ++i;
219     }
220     if (i == start) { // No valid name chars
221         return result; // Indicate failure with empty string
222     }
223     pos.setIndex(i);
224     text.extractBetween(start, i, result);
225     return result;
226 }
227 
isMatcher(UChar32 ch)228 UBool ParseData::isMatcher(UChar32 ch) {
229     // Note that we cannot use data.lookup() because the
230     // set array has not been constructed yet.
231     int32_t i = ch - data->variablesBase;
232     if (i >= 0 && i < variablesVector->size()) {
233         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
234         return f != NULL && f->toMatcher() != NULL;
235     }
236     return TRUE;
237 }
238 
239 /**
240  * Return true if the given character is a replacer standin or a plain
241  * character (non standin).
242  */
isReplacer(UChar32 ch)243 UBool ParseData::isReplacer(UChar32 ch) {
244     // Note that we cannot use data.lookup() because the
245     // set array has not been constructed yet.
246     int i = ch - data->variablesBase;
247     if (i >= 0 && i < variablesVector->size()) {
248         UnicodeFunctor *f = (UnicodeFunctor*) variablesVector->elementAt(i);
249         return f != NULL && f->toReplacer() != NULL;
250     }
251     return TRUE;
252 }
253 
254 //----------------------------------------------------------------------
255 // BEGIN RuleHalf
256 //----------------------------------------------------------------------
257 
258 /**
259  * A class representing one side of a rule.  This class knows how to
260  * parse half of a rule.  It is tightly coupled to the method
261  * RuleBasedTransliterator.Parser.parseRule().
262  */
263 class RuleHalf : public UMemory {
264 
265 public:
266 
267     UnicodeString text;
268 
269     int32_t cursor; // position of cursor in text
270     int32_t ante;   // position of ante context marker '{' in text
271     int32_t post;   // position of post context marker '}' in text
272 
273     // Record the offset to the cursor either to the left or to the
274     // right of the key.  This is indicated by characters on the output
275     // side that allow the cursor to be positioned arbitrarily within
276     // the matching text.  For example, abc{def} > | @@@ xyz; changes
277     // def to xyz and moves the cursor to before abc.  Offset characters
278     // must be at the start or end, and they cannot move the cursor past
279     // the ante- or postcontext text.  Placeholders are only valid in
280     // output text.  The length of the ante and post context is
281     // determined at runtime, because of supplementals and quantifiers.
282     int32_t cursorOffset; // only nonzero on output side
283 
284     // Position of first CURSOR_OFFSET on _right_.  This will be -1
285     // for |@, -2 for |@@, etc., and 1 for @|, 2 for @@|, etc.
286     int32_t cursorOffsetPos;
287 
288     UBool anchorStart;
289     UBool anchorEnd;
290 
291     /**
292      * The segment number from 1..n of the next '(' we see
293      * during parsing; 1-based.
294      */
295     int32_t nextSegmentNumber;
296 
297     TransliteratorParser& parser;
298 
299     //--------------------------------------------------
300     // Methods
301 
302     RuleHalf(TransliteratorParser& parser);
303     ~RuleHalf();
304 
305     int32_t parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status);
306 
307     int32_t parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
308                          UnicodeString& buf,
309                          const UnicodeString& illegal,
310                          UBool isSegment,
311                          UErrorCode& status);
312 
313     /**
314      * Remove context.
315      */
316     void removeContext();
317 
318     /**
319      * Return true if this half looks like valid output, that is, does not
320      * contain quantifiers or other special input-only elements.
321      */
322     UBool isValidOutput(TransliteratorParser& parser);
323 
324     /**
325      * Return true if this half looks like valid input, that is, does not
326      * contain functions or other special output-only elements.
327      */
328     UBool isValidInput(TransliteratorParser& parser);
329 
syntaxError(UErrorCode code,const UnicodeString & rule,int32_t start,UErrorCode & status)330     int syntaxError(UErrorCode code,
331                     const UnicodeString& rule,
332                     int32_t start,
333                     UErrorCode& status) {
334         return parser.syntaxError(code, rule, start, status);
335     }
336 
337 private:
338     // Disallowed methods; no impl.
339     RuleHalf(const RuleHalf&);
340     RuleHalf& operator=(const RuleHalf&);
341 };
342 
RuleHalf(TransliteratorParser & p)343 RuleHalf::RuleHalf(TransliteratorParser& p) :
344     parser(p)
345 {
346     cursor = -1;
347     ante = -1;
348     post = -1;
349     cursorOffset = 0;
350     cursorOffsetPos = 0;
351     anchorStart = anchorEnd = FALSE;
352     nextSegmentNumber = 1;
353 }
354 
~RuleHalf()355 RuleHalf::~RuleHalf() {
356 }
357 
358 /**
359  * Parse one side of a rule, stopping at either the limit,
360  * the END_OF_RULE character, or an operator.
361  * @return the index after the terminating character, or
362  * if limit was reached, limit
363  */
parse(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)364 int32_t RuleHalf::parse(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
365     int32_t start = pos;
366     text.truncate(0);
367     pos = parseSection(rule, pos, limit, text, UnicodeString(TRUE, ILLEGAL_TOP, -1), FALSE, status);
368 
369     if (cursorOffset > 0 && cursor != cursorOffsetPos) {
370         return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
371     }
372 
373     return pos;
374 }
375 
376 /**
377  * Parse a section of one side of a rule, stopping at either
378  * the limit, the END_OF_RULE character, an operator, or a
379  * segment close character.  This method parses both a
380  * top-level rule half and a segment within such a rule half.
381  * It calls itself recursively to parse segments and nested
382  * segments.
383  * @param buf buffer into which to accumulate the rule pattern
384  * characters, either literal characters from the rule or
385  * standins for UnicodeMatcher objects including segments.
386  * @param illegal the set of special characters that is illegal during
387  * this parse.
388  * @param isSegment if true, then we've already seen a '(' and
389  * pos on entry points right after it.  Accumulate everything
390  * up to the closing ')', put it in a segment matcher object,
391  * generate a standin for it, and add the standin to buf.  As
392  * a side effect, update the segments vector with a reference
393  * to the segment matcher.  This works recursively for nested
394  * segments.  If isSegment is false, just accumulate
395  * characters into buf.
396  * @return the index after the terminating character, or
397  * if limit was reached, limit
398  */
parseSection(const UnicodeString & rule,int32_t pos,int32_t limit,UnicodeString & buf,const UnicodeString & illegal,UBool isSegment,UErrorCode & status)399 int32_t RuleHalf::parseSection(const UnicodeString& rule, int32_t pos, int32_t limit,
400                                UnicodeString& buf,
401                                const UnicodeString& illegal,
402                                UBool isSegment, UErrorCode& status) {
403     int32_t start = pos;
404     ParsePosition pp;
405     UnicodeString scratch;
406     UBool done = FALSE;
407     int32_t quoteStart = -1; // Most recent 'single quoted string'
408     int32_t quoteLimit = -1;
409     int32_t varStart = -1; // Most recent $variableReference
410     int32_t varLimit = -1;
411     int32_t bufStart = buf.length();
412 
413     while (pos < limit && !done) {
414         // Since all syntax characters are in the BMP, fetching
415         // 16-bit code units suffices here.
416         UChar c = rule.charAt(pos++);
417         if (PatternProps::isWhiteSpace(c)) {
418             // Ignore whitespace.  Note that this is not Unicode
419             // spaces, but Java spaces -- a subset, representing
420             // whitespace likely to be seen in code.
421             continue;
422         }
423         if (u_strchr(HALF_ENDERS, c) != NULL) {
424             if (isSegment) {
425                 // Unclosed segment
426                 return syntaxError(U_UNCLOSED_SEGMENT, rule, start, status);
427             }
428             break;
429         }
430         if (anchorEnd) {
431             // Text after a presumed end anchor is a syntax err
432             return syntaxError(U_MALFORMED_VARIABLE_REFERENCE, rule, start, status);
433         }
434         if (UnicodeSet::resemblesPattern(rule, pos-1)) {
435             pp.setIndex(pos-1); // Backup to opening '['
436             buf.append(parser.parseSet(rule, pp, status));
437             if (U_FAILURE(status)) {
438                 return syntaxError(U_MALFORMED_SET, rule, start, status);
439             }
440             pos = pp.getIndex();
441             continue;
442         }
443         // Handle escapes
444         if (c == ESCAPE) {
445             if (pos == limit) {
446                 return syntaxError(U_TRAILING_BACKSLASH, rule, start, status);
447             }
448             UChar32 escaped = rule.unescapeAt(pos); // pos is already past '\\'
449             if (escaped == (UChar32) -1) {
450                 return syntaxError(U_MALFORMED_UNICODE_ESCAPE, rule, start, status);
451             }
452             if (!parser.checkVariableRange(escaped)) {
453                 return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
454             }
455             buf.append(escaped);
456             continue;
457         }
458         // Handle quoted matter
459         if (c == QUOTE) {
460             int32_t iq = rule.indexOf(QUOTE, pos);
461             if (iq == pos) {
462                 buf.append(c); // Parse [''] outside quotes as [']
463                 ++pos;
464             } else {
465                 /* This loop picks up a run of quoted text of the
466                  * form 'aaaa' each time through.  If this run
467                  * hasn't really ended ('aaaa''bbbb') then it keeps
468                  * looping, each time adding on a new run.  When it
469                  * reaches the final quote it breaks.
470                  */
471                 quoteStart = buf.length();
472                 for (;;) {
473                     if (iq < 0) {
474                         return syntaxError(U_UNTERMINATED_QUOTE, rule, start, status);
475                     }
476                     scratch.truncate(0);
477                     rule.extractBetween(pos, iq, scratch);
478                     buf.append(scratch);
479                     pos = iq+1;
480                     if (pos < limit && rule.charAt(pos) == QUOTE) {
481                         // Parse [''] inside quotes as [']
482                         iq = rule.indexOf(QUOTE, pos+1);
483                         // Continue looping
484                     } else {
485                         break;
486                     }
487                 }
488                 quoteLimit = buf.length();
489 
490                 for (iq=quoteStart; iq<quoteLimit; ++iq) {
491                     if (!parser.checkVariableRange(buf.charAt(iq))) {
492                         return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
493                     }
494                 }
495             }
496             continue;
497         }
498 
499         if (!parser.checkVariableRange(c)) {
500             return syntaxError(U_VARIABLE_RANGE_OVERLAP, rule, start, status);
501         }
502 
503         if (illegal.indexOf(c) >= 0) {
504             syntaxError(U_ILLEGAL_CHARACTER, rule, start, status);
505         }
506 
507         switch (c) {
508 
509         //------------------------------------------------------
510         // Elements allowed within and out of segments
511         //------------------------------------------------------
512         case ANCHOR_START:
513             if (buf.length() == 0 && !anchorStart) {
514                 anchorStart = TRUE;
515             } else {
516               return syntaxError(U_MISPLACED_ANCHOR_START,
517                                  rule, start, status);
518             }
519           break;
520         case SEGMENT_OPEN:
521             {
522                 // bufSegStart is the offset in buf to the first
523                 // character of the segment we are parsing.
524                 int32_t bufSegStart = buf.length();
525 
526                 // Record segment number now, since nextSegmentNumber
527                 // will be incremented during the call to parseSection
528                 // if there are nested segments.
529                 int32_t segmentNumber = nextSegmentNumber++; // 1-based
530 
531                 // Parse the segment
532                 pos = parseSection(rule, pos, limit, buf, UnicodeString(TRUE, ILLEGAL_SEG, -1), TRUE, status);
533 
534                 // After parsing a segment, the relevant characters are
535                 // in buf, starting at offset bufSegStart.  Extract them
536                 // into a string matcher, and replace them with a
537                 // standin for that matcher.
538                 StringMatcher* m =
539                     new StringMatcher(buf, bufSegStart, buf.length(),
540                                       segmentNumber, *parser.curData);
541                 if (m == NULL) {
542                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
543                 }
544 
545                 // Record and associate object and segment number
546                 parser.setSegmentObject(segmentNumber, m, status);
547                 buf.truncate(bufSegStart);
548                 buf.append(parser.getSegmentStandin(segmentNumber, status));
549             }
550             break;
551         case FUNCTION:
552         case ALT_FUNCTION:
553             {
554                 int32_t iref = pos;
555                 TransliteratorIDParser::SingleID* single =
556                     TransliteratorIDParser::parseFilterID(rule, iref);
557                 // The next character MUST be a segment open
558                 if (single == NULL ||
559                     !ICU_Utility::parseChar(rule, iref, SEGMENT_OPEN)) {
560                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
561                 }
562 
563                 Transliterator *t = single->createInstance();
564                 delete single;
565                 if (t == NULL) {
566                     return syntaxError(U_INVALID_FUNCTION, rule, start, status);
567                 }
568 
569                 // bufSegStart is the offset in buf to the first
570                 // character of the segment we are parsing.
571                 int32_t bufSegStart = buf.length();
572 
573                 // Parse the segment
574                 pos = parseSection(rule, iref, limit, buf, UnicodeString(TRUE, ILLEGAL_FUNC, -1), TRUE, status);
575 
576                 // After parsing a segment, the relevant characters are
577                 // in buf, starting at offset bufSegStart.
578                 UnicodeString output;
579                 buf.extractBetween(bufSegStart, buf.length(), output);
580                 FunctionReplacer *r =
581                     new FunctionReplacer(t, new StringReplacer(output, parser.curData));
582                 if (r == NULL) {
583                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
584                 }
585 
586                 // Replace the buffer contents with a stand-in
587                 buf.truncate(bufSegStart);
588                 buf.append(parser.generateStandInFor(r, status));
589             }
590             break;
591         case SymbolTable::SYMBOL_REF:
592             // Handle variable references and segment references "$1" .. "$9"
593             {
594                 // A variable reference must be followed immediately
595                 // by a Unicode identifier start and zero or more
596                 // Unicode identifier part characters, or by a digit
597                 // 1..9 if it is a segment reference.
598                 if (pos == limit) {
599                     // A variable ref character at the end acts as
600                     // an anchor to the context limit, as in perl.
601                     anchorEnd = TRUE;
602                     break;
603                 }
604                 // Parse "$1" "$2" .. "$9" .. (no upper limit)
605                 c = rule.charAt(pos);
606                 int32_t r = u_digit(c, 10);
607                 if (r >= 1 && r <= 9) {
608                     r = ICU_Utility::parseNumber(rule, pos, 10);
609                     if (r < 0) {
610                         return syntaxError(U_UNDEFINED_SEGMENT_REFERENCE,
611                                            rule, start, status);
612                     }
613                     buf.append(parser.getSegmentStandin(r, status));
614                 } else {
615                     pp.setIndex(pos);
616                     UnicodeString name = parser.parseData->
617                                     parseReference(rule, pp, limit);
618                     if (name.length() == 0) {
619                         // This means the '$' was not followed by a
620                         // valid name.  Try to interpret it as an
621                         // end anchor then.  If this also doesn't work
622                         // (if we see a following character) then signal
623                         // an error.
624                         anchorEnd = TRUE;
625                         break;
626                     }
627                     pos = pp.getIndex();
628                     // If this is a variable definition statement,
629                     // then the LHS variable will be undefined.  In
630                     // that case appendVariableDef() will append the
631                     // special placeholder char variableLimit-1.
632                     varStart = buf.length();
633                     parser.appendVariableDef(name, buf, status);
634                     varLimit = buf.length();
635                 }
636             }
637             break;
638         case DOT:
639             buf.append(parser.getDotStandIn(status));
640             break;
641         case KLEENE_STAR:
642         case ONE_OR_MORE:
643         case ZERO_OR_ONE:
644             // Quantifiers.  We handle single characters, quoted strings,
645             // variable references, and segments.
646             //  a+      matches  aaa
647             //  'foo'+  matches  foofoofoo
648             //  $v+     matches  xyxyxy if $v == xy
649             //  (seg)+  matches  segsegseg
650             {
651                 if (isSegment && buf.length() == bufStart) {
652                     // The */+ immediately follows '('
653                     return syntaxError(U_MISPLACED_QUANTIFIER, rule, start, status);
654                 }
655 
656                 int32_t qstart, qlimit;
657                 // The */+ follows an isolated character or quote
658                 // or variable reference
659                 if (buf.length() == quoteLimit) {
660                     // The */+ follows a 'quoted string'
661                     qstart = quoteStart;
662                     qlimit = quoteLimit;
663                 } else if (buf.length() == varLimit) {
664                     // The */+ follows a $variableReference
665                     qstart = varStart;
666                     qlimit = varLimit;
667                 } else {
668                     // The */+ follows a single character, possibly
669                     // a segment standin
670                     qstart = buf.length() - 1;
671                     qlimit = qstart + 1;
672                 }
673 
674                 UnicodeFunctor *m =
675                     new StringMatcher(buf, qstart, qlimit, 0, *parser.curData);
676                 if (m == NULL) {
677                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
678                 }
679                 int32_t min = 0;
680                 int32_t max = Quantifier::MAX;
681                 switch (c) {
682                 case ONE_OR_MORE:
683                     min = 1;
684                     break;
685                 case ZERO_OR_ONE:
686                     min = 0;
687                     max = 1;
688                     break;
689                 // case KLEENE_STAR:
690                 //    do nothing -- min, max already set
691                 }
692                 m = new Quantifier(m, min, max);
693                 if (m == NULL) {
694                     return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
695                 }
696                 buf.truncate(qstart);
697                 buf.append(parser.generateStandInFor(m, status));
698             }
699             break;
700 
701         //------------------------------------------------------
702         // Elements allowed ONLY WITHIN segments
703         //------------------------------------------------------
704         case SEGMENT_CLOSE:
705             // assert(isSegment);
706             // We're done parsing a segment.
707             done = TRUE;
708             break;
709 
710         //------------------------------------------------------
711         // Elements allowed ONLY OUTSIDE segments
712         //------------------------------------------------------
713         case CONTEXT_ANTE:
714             if (ante >= 0) {
715                 return syntaxError(U_MULTIPLE_ANTE_CONTEXTS, rule, start, status);
716             }
717             ante = buf.length();
718             break;
719         case CONTEXT_POST:
720             if (post >= 0) {
721                 return syntaxError(U_MULTIPLE_POST_CONTEXTS, rule, start, status);
722             }
723             post = buf.length();
724             break;
725         case CURSOR_POS:
726             if (cursor >= 0) {
727                 return syntaxError(U_MULTIPLE_CURSORS, rule, start, status);
728             }
729             cursor = buf.length();
730             break;
731         case CURSOR_OFFSET:
732             if (cursorOffset < 0) {
733                 if (buf.length() > 0) {
734                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
735                 }
736                 --cursorOffset;
737             } else if (cursorOffset > 0) {
738                 if (buf.length() != cursorOffsetPos || cursor >= 0) {
739                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
740                 }
741                 ++cursorOffset;
742             } else {
743                 if (cursor == 0 && buf.length() == 0) {
744                     cursorOffset = -1;
745                 } else if (cursor < 0) {
746                     cursorOffsetPos = buf.length();
747                     cursorOffset = 1;
748                 } else {
749                     return syntaxError(U_MISPLACED_CURSOR_OFFSET, rule, start, status);
750                 }
751             }
752             break;
753 
754 
755         //------------------------------------------------------
756         // Non-special characters
757         //------------------------------------------------------
758         default:
759             // Disallow unquoted characters other than [0-9A-Za-z]
760             // in the printable ASCII range.  These characters are
761             // reserved for possible future use.
762             if (c >= 0x0021 && c <= 0x007E &&
763                 !((c >= 0x0030/*'0'*/ && c <= 0x0039/*'9'*/) ||
764                   (c >= 0x0041/*'A'*/ && c <= 0x005A/*'Z'*/) ||
765                   (c >= 0x0061/*'a'*/ && c <= 0x007A/*'z'*/))) {
766                 return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
767             }
768             buf.append(c);
769             break;
770         }
771     }
772 
773     return pos;
774 }
775 
776 /**
777  * Remove context.
778  */
removeContext()779 void RuleHalf::removeContext() {
780     //text = text.substring(ante < 0 ? 0 : ante,
781     //                      post < 0 ? text.length() : post);
782     if (post >= 0) {
783         text.remove(post);
784     }
785     if (ante >= 0) {
786         text.removeBetween(0, ante);
787     }
788     ante = post = -1;
789     anchorStart = anchorEnd = FALSE;
790 }
791 
792 /**
793  * Return true if this half looks like valid output, that is, does not
794  * contain quantifiers or other special input-only elements.
795  */
isValidOutput(TransliteratorParser & transParser)796 UBool RuleHalf::isValidOutput(TransliteratorParser& transParser) {
797     for (int32_t i=0; i<text.length(); ) {
798         UChar32 c = text.char32At(i);
799         i += U16_LENGTH(c);
800         if (!transParser.parseData->isReplacer(c)) {
801             return FALSE;
802         }
803     }
804     return TRUE;
805 }
806 
807 /**
808  * Return true if this half looks like valid input, that is, does not
809  * contain functions or other special output-only elements.
810  */
isValidInput(TransliteratorParser & transParser)811 UBool RuleHalf::isValidInput(TransliteratorParser& transParser) {
812     for (int32_t i=0; i<text.length(); ) {
813         UChar32 c = text.char32At(i);
814         i += U16_LENGTH(c);
815         if (!transParser.parseData->isMatcher(c)) {
816             return FALSE;
817         }
818     }
819     return TRUE;
820 }
821 
822 //----------------------------------------------------------------------
823 // PUBLIC API
824 //----------------------------------------------------------------------
825 
826 /**
827  * Constructor.
828  */
TransliteratorParser(UErrorCode & statusReturn)829 TransliteratorParser::TransliteratorParser(UErrorCode &statusReturn) :
830 dataVector(statusReturn),
831 idBlockVector(statusReturn),
832 variablesVector(statusReturn),
833 segmentObjects(statusReturn)
834 {
835     idBlockVector.setDeleter(uprv_deleteUObject);
836     curData = NULL;
837     compoundFilter = NULL;
838     parseData = NULL;
839     variableNames.setValueDeleter(uprv_deleteUObject);
840 }
841 
842 /**
843  * Destructor.
844  */
~TransliteratorParser()845 TransliteratorParser::~TransliteratorParser() {
846     while (!dataVector.isEmpty())
847         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
848     delete compoundFilter;
849     delete parseData;
850     while (!variablesVector.isEmpty())
851         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
852 }
853 
854 void
parse(const UnicodeString & rules,UTransDirection transDirection,UParseError & pe,UErrorCode & ec)855 TransliteratorParser::parse(const UnicodeString& rules,
856                             UTransDirection transDirection,
857                             UParseError& pe,
858                             UErrorCode& ec) {
859     if (U_SUCCESS(ec)) {
860         parseRules(rules, transDirection, ec);
861         pe = parseError;
862     }
863 }
864 
865 /**
866  * Return the compound filter parsed by parse().  Caller owns result.
867  */
orphanCompoundFilter()868 UnicodeSet* TransliteratorParser::orphanCompoundFilter() {
869     UnicodeSet* f = compoundFilter;
870     compoundFilter = NULL;
871     return f;
872 }
873 
874 //----------------------------------------------------------------------
875 // Private implementation
876 //----------------------------------------------------------------------
877 
878 /**
879  * Parse the given string as a sequence of rules, separated by newline
880  * characters ('\n'), and cause this object to implement those rules.  Any
881  * previous rules are discarded.  Typically this method is called exactly
882  * once, during construction.
883  * @exception IllegalArgumentException if there is a syntax error in the
884  * rules
885  */
parseRules(const UnicodeString & rule,UTransDirection theDirection,UErrorCode & status)886 void TransliteratorParser::parseRules(const UnicodeString& rule,
887                                       UTransDirection theDirection,
888                                       UErrorCode& status)
889 {
890     // Clear error struct
891     uprv_memset(&parseError, 0, sizeof(parseError));
892     parseError.line = parseError.offset = -1;
893 
894     UBool parsingIDs = TRUE;
895     int32_t ruleCount = 0;
896 
897     while (!dataVector.isEmpty()) {
898         delete (TransliterationRuleData*)(dataVector.orphanElementAt(0));
899     }
900     if (U_FAILURE(status)) {
901         return;
902     }
903 
904     idBlockVector.removeAllElements();
905     curData = NULL;
906     direction = theDirection;
907     ruleCount = 0;
908 
909     delete compoundFilter;
910     compoundFilter = NULL;
911 
912     while (!variablesVector.isEmpty()) {
913         delete (UnicodeFunctor*)variablesVector.orphanElementAt(0);
914     }
915     variableNames.removeAll();
916     parseData = new ParseData(0, &variablesVector, &variableNames);
917     if (parseData == NULL) {
918         status = U_MEMORY_ALLOCATION_ERROR;
919         return;
920     }
921 
922     dotStandIn = (UChar) -1;
923 
924     UnicodeString *tempstr = NULL; // used for memory allocation error checking
925     UnicodeString str; // scratch
926     UnicodeString idBlockResult;
927     int32_t pos = 0;
928     int32_t limit = rule.length();
929 
930     // The compound filter offset is an index into idBlockResult.
931     // If it is 0, then the compound filter occurred at the start,
932     // and it is the offset to the _start_ of the compound filter
933     // pattern.  Otherwise it is the offset to the _limit_ of the
934     // compound filter pattern within idBlockResult.
935     compoundFilter = NULL;
936     int32_t compoundFilterOffset = -1;
937 
938     while (pos < limit && U_SUCCESS(status)) {
939         UChar c = rule.charAt(pos++);
940         if (PatternProps::isWhiteSpace(c)) {
941             // Ignore leading whitespace.
942             continue;
943         }
944         // Skip lines starting with the comment character
945         if (c == RULE_COMMENT_CHAR) {
946             pos = rule.indexOf((UChar)0x000A /*\n*/, pos) + 1;
947             if (pos == 0) {
948                 break; // No "\n" found; rest of rule is a commnet
949             }
950             continue; // Either fall out or restart with next line
951         }
952 
953         // skip empty rules
954         if (c == END_OF_RULE)
955             continue;
956 
957         // keep track of how many rules we've seen
958         ++ruleCount;
959 
960         // We've found the start of a rule or ID.  c is its first
961         // character, and pos points past c.
962         --pos;
963         // Look for an ID token.  Must have at least ID_TOKEN_LEN + 1
964         // chars left.
965         if ((pos + ID_TOKEN_LEN + 1) <= limit &&
966                 rule.compare(pos, ID_TOKEN_LEN, ID_TOKEN) == 0) {
967             pos += ID_TOKEN_LEN;
968             c = rule.charAt(pos);
969             while (PatternProps::isWhiteSpace(c) && pos < limit) {
970                 ++pos;
971                 c = rule.charAt(pos);
972             }
973 
974             int32_t p = pos;
975 
976             if (!parsingIDs) {
977                 if (curData != NULL) {
978                     if (direction == UTRANS_FORWARD)
979                         dataVector.addElement(curData, status);
980                     else
981                         dataVector.insertElementAt(curData, 0, status);
982                     curData = NULL;
983                 }
984                 parsingIDs = TRUE;
985             }
986 
987             TransliteratorIDParser::SingleID* id =
988                 TransliteratorIDParser::parseSingleID(rule, p, direction, status);
989             if (p != pos && ICU_Utility::parseChar(rule, p, END_OF_RULE)) {
990                 // Successful ::ID parse.
991 
992                 if (direction == UTRANS_FORWARD) {
993                     idBlockResult.append(id->canonID).append(END_OF_RULE);
994                 } else {
995                     idBlockResult.insert(0, END_OF_RULE);
996                     idBlockResult.insert(0, id->canonID);
997                 }
998 
999             } else {
1000                 // Couldn't parse an ID.  Try to parse a global filter
1001                 int32_t withParens = -1;
1002                 UnicodeSet* f = TransliteratorIDParser::parseGlobalFilter(rule, p, direction, withParens, NULL);
1003                 if (f != NULL) {
1004                     if (ICU_Utility::parseChar(rule, p, END_OF_RULE)
1005                         && (direction == UTRANS_FORWARD) == (withParens == 0))
1006                     {
1007                         if (compoundFilter != NULL) {
1008                             // Multiple compound filters
1009                             syntaxError(U_MULTIPLE_COMPOUND_FILTERS, rule, pos, status);
1010                             delete f;
1011                         } else {
1012                             compoundFilter = f;
1013                             compoundFilterOffset = ruleCount;
1014                         }
1015                     } else {
1016                         delete f;
1017                     }
1018                 } else {
1019                     // Invalid ::id
1020                     // Can be parsed as neither an ID nor a global filter
1021                     syntaxError(U_INVALID_ID, rule, pos, status);
1022                 }
1023             }
1024             delete id;
1025             pos = p;
1026         } else {
1027             if (parsingIDs) {
1028                 tempstr = new UnicodeString(idBlockResult);
1029                 // NULL pointer check
1030                 if (tempstr == NULL) {
1031                     status = U_MEMORY_ALLOCATION_ERROR;
1032                     return;
1033                 }
1034                 if (direction == UTRANS_FORWARD)
1035                     idBlockVector.addElement(tempstr, status);
1036                 else
1037                     idBlockVector.insertElementAt(tempstr, 0, status);
1038                 idBlockResult.remove();
1039                 parsingIDs = FALSE;
1040                 curData = new TransliterationRuleData(status);
1041                 // NULL pointer check
1042                 if (curData == NULL) {
1043                     status = U_MEMORY_ALLOCATION_ERROR;
1044                     return;
1045                 }
1046                 parseData->data = curData;
1047 
1048                 // By default, rules use part of the private use area
1049                 // E000..F8FF for variables and other stand-ins.  Currently
1050                 // the range F000..F8FF is typically sufficient.  The 'use
1051                 // variable range' pragma allows rule sets to modify this.
1052                 setVariableRange(0xF000, 0xF8FF, status);
1053             }
1054 
1055             if (resemblesPragma(rule, pos, limit)) {
1056                 int32_t ppp = parsePragma(rule, pos, limit, status);
1057                 if (ppp < 0) {
1058                     syntaxError(U_MALFORMED_PRAGMA, rule, pos, status);
1059                 }
1060                 pos = ppp;
1061             // Parse a rule
1062             } else {
1063                 pos = parseRule(rule, pos, limit, status);
1064             }
1065         }
1066     }
1067 
1068     if (parsingIDs && idBlockResult.length() > 0) {
1069         tempstr = new UnicodeString(idBlockResult);
1070         // NULL pointer check
1071         if (tempstr == NULL) {
1072             status = U_MEMORY_ALLOCATION_ERROR;
1073             return;
1074         }
1075         if (direction == UTRANS_FORWARD)
1076             idBlockVector.addElement(tempstr, status);
1077         else
1078             idBlockVector.insertElementAt(tempstr, 0, status);
1079     }
1080     else if (!parsingIDs && curData != NULL) {
1081         if (direction == UTRANS_FORWARD)
1082             dataVector.addElement(curData, status);
1083         else
1084             dataVector.insertElementAt(curData, 0, status);
1085     }
1086 
1087     if (U_SUCCESS(status)) {
1088         // Convert the set vector to an array
1089         int32_t i, dataVectorSize = dataVector.size();
1090         for (i = 0; i < dataVectorSize; i++) {
1091             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1092             data->variablesLength = variablesVector.size();
1093             if (data->variablesLength == 0) {
1094                 data->variables = 0;
1095             } else {
1096                 data->variables = (UnicodeFunctor**)uprv_malloc(data->variablesLength * sizeof(UnicodeFunctor*));
1097                 // NULL pointer check
1098                 if (data->variables == NULL) {
1099                     status = U_MEMORY_ALLOCATION_ERROR;
1100                     return;
1101                 }
1102                 data->variablesAreOwned = (i == 0);
1103             }
1104 
1105             for (int32_t j = 0; j < data->variablesLength; j++) {
1106                 data->variables[j] =
1107                     static_cast<UnicodeFunctor *>(variablesVector.elementAt(j));
1108             }
1109 
1110             data->variableNames.removeAll();
1111             int32_t pos = UHASH_FIRST;
1112             const UHashElement* he = variableNames.nextElement(pos);
1113             while (he != NULL) {
1114                 UnicodeString* tempus = (UnicodeString*)(((UnicodeString*)(he->value.pointer))->clone());
1115                 if (tempus == NULL) {
1116                     status = U_MEMORY_ALLOCATION_ERROR;
1117                     return;
1118                 }
1119                 data->variableNames.put(*((UnicodeString*)(he->key.pointer)),
1120                     tempus, status);
1121                 he = variableNames.nextElement(pos);
1122             }
1123         }
1124         variablesVector.removeAllElements();   // keeps them from getting deleted when we succeed
1125 
1126         // Index the rules
1127         if (compoundFilter != NULL) {
1128             if ((direction == UTRANS_FORWARD && compoundFilterOffset != 1) ||
1129                 (direction == UTRANS_REVERSE && compoundFilterOffset != ruleCount)) {
1130                 status = U_MISPLACED_COMPOUND_FILTER;
1131             }
1132         }
1133 
1134         for (i = 0; i < dataVectorSize; i++) {
1135             TransliterationRuleData* data = (TransliterationRuleData*)dataVector.elementAt(i);
1136             data->ruleSet.freeze(parseError, status);
1137         }
1138         if (idBlockVector.size() == 1 && ((UnicodeString*)idBlockVector.elementAt(0))->isEmpty()) {
1139             idBlockVector.removeElementAt(0);
1140         }
1141     }
1142 }
1143 
1144 /**
1145  * Set the variable range to [start, end] (inclusive).
1146  */
setVariableRange(int32_t start,int32_t end,UErrorCode & status)1147 void TransliteratorParser::setVariableRange(int32_t start, int32_t end, UErrorCode& status) {
1148     if (start > end || start < 0 || end > 0xFFFF) {
1149         status = U_MALFORMED_PRAGMA;
1150         return;
1151     }
1152 
1153     curData->variablesBase = (UChar) start;
1154     if (dataVector.size() == 0) {
1155         variableNext = (UChar) start;
1156         variableLimit = (UChar) (end + 1);
1157     }
1158 }
1159 
1160 /**
1161  * Assert that the given character is NOT within the variable range.
1162  * If it is, return FALSE.  This is neccesary to ensure that the
1163  * variable range does not overlap characters used in a rule.
1164  */
checkVariableRange(UChar32 ch) const1165 UBool TransliteratorParser::checkVariableRange(UChar32 ch) const {
1166     return !(ch >= curData->variablesBase && ch < variableLimit);
1167 }
1168 
1169 /**
1170  * Set the maximum backup to 'backup', in response to a pragma
1171  * statement.
1172  */
pragmaMaximumBackup(int32_t)1173 void TransliteratorParser::pragmaMaximumBackup(int32_t /*backup*/) {
1174     //TODO Finish
1175 }
1176 
1177 /**
1178  * Begin normalizing all rules using the given mode, in response
1179  * to a pragma statement.
1180  */
pragmaNormalizeRules(UNormalizationMode)1181 void TransliteratorParser::pragmaNormalizeRules(UNormalizationMode /*mode*/) {
1182     //TODO Finish
1183 }
1184 
1185 static const UChar PRAGMA_USE[] = {0x75,0x73,0x65,0x20,0}; // "use "
1186 
1187 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 # #~;"
1188 
1189 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 #~;"
1190 
1191 static const UChar PRAGMA_NFD_RULES[] = {0x7E,0x6E,0x66,0x64,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfd rules~;"
1192 
1193 static const UChar PRAGMA_NFC_RULES[] = {0x7E,0x6E,0x66,0x63,0x20,0x72,0x75,0x6C,0x65,0x73,0x7E,0x3B,0}; // "~nfc rules~;"
1194 
1195 /**
1196  * Return true if the given rule looks like a pragma.
1197  * @param pos offset to the first non-whitespace character
1198  * of the rule.
1199  * @param limit pointer past the last character of the rule.
1200  */
resemblesPragma(const UnicodeString & rule,int32_t pos,int32_t limit)1201 UBool TransliteratorParser::resemblesPragma(const UnicodeString& rule, int32_t pos, int32_t limit) {
1202     // Must start with /use\s/i
1203     return ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_USE, 4), NULL) >= 0;
1204 }
1205 
1206 /**
1207  * Parse a pragma.  This method assumes resemblesPragma() has
1208  * already returned true.
1209  * @param pos offset to the first non-whitespace character
1210  * of the rule.
1211  * @param limit pointer past the last character of the rule.
1212  * @return the position index after the final ';' of the pragma,
1213  * or -1 on failure.
1214  */
parsePragma(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1215 int32_t TransliteratorParser::parsePragma(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1216     int32_t array[2];
1217 
1218     // resemblesPragma() has already returned true, so we
1219     // know that pos points to /use\s/i; we can skip 4 characters
1220     // immediately
1221     pos += 4;
1222 
1223     // Here are the pragmas we recognize:
1224     // use variable range 0xE000 0xEFFF;
1225     // use maximum backup 16;
1226     // use nfd rules;
1227     // use nfc rules;
1228     int p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_VARIABLE_RANGE, -1), array);
1229     if (p >= 0) {
1230         setVariableRange(array[0], array[1], status);
1231         return p;
1232     }
1233 
1234     p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_MAXIMUM_BACKUP, -1), array);
1235     if (p >= 0) {
1236         pragmaMaximumBackup(array[0]);
1237         return p;
1238     }
1239 
1240     p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_NFD_RULES, -1), NULL);
1241     if (p >= 0) {
1242         pragmaNormalizeRules(UNORM_NFD);
1243         return p;
1244     }
1245 
1246     p = ICU_Utility::parsePattern(rule, pos, limit, UnicodeString(TRUE, PRAGMA_NFC_RULES, -1), NULL);
1247     if (p >= 0) {
1248         pragmaNormalizeRules(UNORM_NFC);
1249         return p;
1250     }
1251 
1252     // Syntax error: unable to parse pragma
1253     return -1;
1254 }
1255 
1256 /**
1257  * MAIN PARSER.  Parse the next rule in the given rule string, starting
1258  * at pos.  Return the index after the last character parsed.  Do not
1259  * parse characters at or after limit.
1260  *
1261  * Important:  The character at pos must be a non-whitespace character
1262  * that is not the comment character.
1263  *
1264  * This method handles quoting, escaping, and whitespace removal.  It
1265  * parses the end-of-rule character.  It recognizes context and cursor
1266  * indicators.  Once it does a lexical breakdown of the rule at pos, it
1267  * creates a rule object and adds it to our rule list.
1268  */
parseRule(const UnicodeString & rule,int32_t pos,int32_t limit,UErrorCode & status)1269 int32_t TransliteratorParser::parseRule(const UnicodeString& rule, int32_t pos, int32_t limit, UErrorCode& status) {
1270     // Locate the left side, operator, and right side
1271     int32_t start = pos;
1272     UChar op = 0;
1273     int32_t i;
1274 
1275     // Set up segments data
1276     segmentStandins.truncate(0);
1277     segmentObjects.removeAllElements();
1278 
1279     // Use pointers to automatics to make swapping possible.
1280     RuleHalf _left(*this), _right(*this);
1281     RuleHalf* left = &_left;
1282     RuleHalf* right = &_right;
1283 
1284     undefinedVariableName.remove();
1285     pos = left->parse(rule, pos, limit, status);
1286     if (U_FAILURE(status)) {
1287         return start;
1288     }
1289 
1290     if (pos == limit || u_strchr(gOPERATORS, (op = rule.charAt(--pos))) == NULL) {
1291         return syntaxError(U_MISSING_OPERATOR, rule, start, status);
1292     }
1293     ++pos;
1294 
1295     // Found an operator char.  Check for forward-reverse operator.
1296     if (op == REVERSE_RULE_OP &&
1297         (pos < limit && rule.charAt(pos) == FORWARD_RULE_OP)) {
1298         ++pos;
1299         op = FWDREV_RULE_OP;
1300     }
1301 
1302     // Translate alternate op characters.
1303     switch (op) {
1304     case ALT_FORWARD_RULE_OP:
1305         op = FORWARD_RULE_OP;
1306         break;
1307     case ALT_REVERSE_RULE_OP:
1308         op = REVERSE_RULE_OP;
1309         break;
1310     case ALT_FWDREV_RULE_OP:
1311         op = FWDREV_RULE_OP;
1312         break;
1313     }
1314 
1315     pos = right->parse(rule, pos, limit, status);
1316     if (U_FAILURE(status)) {
1317         return start;
1318     }
1319 
1320     if (pos < limit) {
1321         if (rule.charAt(--pos) == END_OF_RULE) {
1322             ++pos;
1323         } else {
1324             // RuleHalf parser must have terminated at an operator
1325             return syntaxError(U_UNQUOTED_SPECIAL, rule, start, status);
1326         }
1327     }
1328 
1329     if (op == VARIABLE_DEF_OP) {
1330         // LHS is the name.  RHS is a single character, either a literal
1331         // or a set (already parsed).  If RHS is longer than one
1332         // character, it is either a multi-character string, or multiple
1333         // sets, or a mixture of chars and sets -- syntax error.
1334 
1335         // We expect to see a single undefined variable (the one being
1336         // defined).
1337         if (undefinedVariableName.length() == 0) {
1338             // "Missing '$' or duplicate definition"
1339             return syntaxError(U_BAD_VARIABLE_DEFINITION, rule, start, status);
1340         }
1341         if (left->text.length() != 1 || left->text.charAt(0) != variableLimit) {
1342             // "Malformed LHS"
1343             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1344         }
1345         if (left->anchorStart || left->anchorEnd ||
1346             right->anchorStart || right->anchorEnd) {
1347             return syntaxError(U_MALFORMED_VARIABLE_DEFINITION, rule, start, status);
1348         }
1349         // We allow anything on the right, including an empty string.
1350         UnicodeString* value = new UnicodeString(right->text);
1351         // NULL pointer check
1352         if (value == NULL) {
1353             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1354         }
1355         variableNames.put(undefinedVariableName, value, status);
1356         ++variableLimit;
1357         return pos;
1358     }
1359 
1360     // If this is not a variable definition rule, we shouldn't have
1361     // any undefined variable names.
1362     if (undefinedVariableName.length() != 0) {
1363         return syntaxError(// "Undefined variable $" + undefinedVariableName,
1364                     U_UNDEFINED_VARIABLE,
1365                     rule, start, status);
1366     }
1367 
1368     // Verify segments
1369     if (segmentStandins.length() > segmentObjects.size()) {
1370         syntaxError(U_UNDEFINED_SEGMENT_REFERENCE, rule, start, status);
1371     }
1372     for (i=0; i<segmentStandins.length(); ++i) {
1373         if (segmentStandins.charAt(i) == 0) {
1374             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1375         }
1376     }
1377     for (i=0; i<segmentObjects.size(); ++i) {
1378         if (segmentObjects.elementAt(i) == NULL) {
1379             syntaxError(U_INTERNAL_TRANSLITERATOR_ERROR, rule, start, status); // will never happen
1380         }
1381     }
1382 
1383     // If the direction we want doesn't match the rule
1384     // direction, do nothing.
1385     if (op != FWDREV_RULE_OP &&
1386         ((direction == UTRANS_FORWARD) != (op == FORWARD_RULE_OP))) {
1387         return pos;
1388     }
1389 
1390     // Transform the rule into a forward rule by swapping the
1391     // sides if necessary.
1392     if (direction == UTRANS_REVERSE) {
1393         left = &_right;
1394         right = &_left;
1395     }
1396 
1397     // Remove non-applicable elements in forward-reverse
1398     // rules.  Bidirectional rules ignore elements that do not
1399     // apply.
1400     if (op == FWDREV_RULE_OP) {
1401         right->removeContext();
1402         left->cursor = -1;
1403         left->cursorOffset = 0;
1404     }
1405 
1406     // Normalize context
1407     if (left->ante < 0) {
1408         left->ante = 0;
1409     }
1410     if (left->post < 0) {
1411         left->post = left->text.length();
1412     }
1413 
1414     // Context is only allowed on the input side.  Cursors are only
1415     // allowed on the output side.  Segment delimiters can only appear
1416     // on the left, and references on the right.  Cursor offset
1417     // cannot appear without an explicit cursor.  Cursor offset
1418     // cannot place the cursor outside the limits of the context.
1419     // Anchors are only allowed on the input side.
1420     if (right->ante >= 0 || right->post >= 0 || left->cursor >= 0 ||
1421         (right->cursorOffset != 0 && right->cursor < 0) ||
1422         // - The following two checks were used to ensure that the
1423         // - the cursor offset stayed within the ante- or postcontext.
1424         // - However, with the addition of quantifiers, we have to
1425         // - allow arbitrary cursor offsets and do runtime checking.
1426         //(right->cursorOffset > (left->text.length() - left->post)) ||
1427         //(-right->cursorOffset > left->ante) ||
1428         right->anchorStart || right->anchorEnd ||
1429         !left->isValidInput(*this) || !right->isValidOutput(*this) ||
1430         left->ante > left->post) {
1431 
1432         return syntaxError(U_MALFORMED_RULE, rule, start, status);
1433     }
1434 
1435     // Flatten segment objects vector to an array
1436     UnicodeFunctor** segmentsArray = NULL;
1437     if (segmentObjects.size() > 0) {
1438         segmentsArray = (UnicodeFunctor **)uprv_malloc(segmentObjects.size() * sizeof(UnicodeFunctor *));
1439         // Null pointer check
1440         if (segmentsArray == NULL) {
1441             return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1442         }
1443         segmentObjects.toArray((void**) segmentsArray);
1444     }
1445     TransliterationRule* temptr = new TransliterationRule(
1446             left->text, left->ante, left->post,
1447             right->text, right->cursor, right->cursorOffset,
1448             segmentsArray,
1449             segmentObjects.size(),
1450             left->anchorStart, left->anchorEnd,
1451             curData,
1452             status);
1453     //Null pointer check
1454     if (temptr == NULL) {
1455         uprv_free(segmentsArray);
1456         return syntaxError(U_MEMORY_ALLOCATION_ERROR, rule, start, status);
1457     }
1458 
1459     curData->ruleSet.addRule(temptr, status);
1460 
1461     return pos;
1462 }
1463 
1464 /**
1465  * Called by main parser upon syntax error.  Search the rule string
1466  * for the probable end of the rule.  Of course, if the error is that
1467  * the end of rule marker is missing, then the rule end will not be found.
1468  * In any case the rule start will be correctly reported.
1469  * @param msg error description
1470  * @param rule pattern string
1471  * @param start position of first character of current rule
1472  */
syntaxError(UErrorCode parseErrorCode,const UnicodeString & rule,int32_t pos,UErrorCode & status)1473 int32_t TransliteratorParser::syntaxError(UErrorCode parseErrorCode,
1474                                           const UnicodeString& rule,
1475                                           int32_t pos,
1476                                           UErrorCode& status)
1477 {
1478     parseError.offset = pos;
1479     parseError.line = 0 ; /* we are not using line numbers */
1480 
1481     // for pre-context
1482     const int32_t LEN = U_PARSE_CONTEXT_LEN - 1;
1483     int32_t start = uprv_max(pos - LEN, 0);
1484     int32_t stop  = pos;
1485 
1486     rule.extract(start,stop-start,parseError.preContext);
1487     //null terminate the buffer
1488     parseError.preContext[stop-start] = 0;
1489 
1490     //for post-context
1491     start = pos;
1492     stop  = uprv_min(pos + LEN, rule.length());
1493 
1494     rule.extract(start,stop-start,parseError.postContext);
1495     //null terminate the buffer
1496     parseError.postContext[stop-start]= 0;
1497 
1498     status = (UErrorCode)parseErrorCode;
1499     return pos;
1500 
1501 }
1502 
1503 /**
1504  * Parse a UnicodeSet out, store it, and return the stand-in character
1505  * used to represent it.
1506  */
parseSet(const UnicodeString & rule,ParsePosition & pos,UErrorCode & status)1507 UChar TransliteratorParser::parseSet(const UnicodeString& rule,
1508                                           ParsePosition& pos,
1509                                           UErrorCode& status) {
1510     UnicodeSet* set = new UnicodeSet(rule, pos, USET_IGNORE_SPACE, parseData, status);
1511     // Null pointer check
1512     if (set == NULL) {
1513         status = U_MEMORY_ALLOCATION_ERROR;
1514         return (UChar)0x0000; // Return empty character with error.
1515     }
1516     set->compact();
1517     return generateStandInFor(set, status);
1518 }
1519 
1520 /**
1521  * Generate and return a stand-in for a new UnicodeFunctor.  Store
1522  * the matcher (adopt it).
1523  */
generateStandInFor(UnicodeFunctor * adopted,UErrorCode & status)1524 UChar TransliteratorParser::generateStandInFor(UnicodeFunctor* adopted, UErrorCode& status) {
1525     // assert(obj != null);
1526 
1527     // Look up previous stand-in, if any.  This is a short list
1528     // (typical n is 0, 1, or 2); linear search is optimal.
1529     for (int32_t i=0; i<variablesVector.size(); ++i) {
1530         if (variablesVector.elementAt(i) == adopted) { // [sic] pointer comparison
1531             return (UChar) (curData->variablesBase + i);
1532         }
1533     }
1534 
1535     if (variableNext >= variableLimit) {
1536         delete adopted;
1537         status = U_VARIABLE_RANGE_EXHAUSTED;
1538         return 0;
1539     }
1540     variablesVector.addElement(adopted, status);
1541     return variableNext++;
1542 }
1543 
1544 /**
1545  * Return the standin for segment seg (1-based).
1546  */
getSegmentStandin(int32_t seg,UErrorCode & status)1547 UChar TransliteratorParser::getSegmentStandin(int32_t seg, UErrorCode& status) {
1548     // Special character used to indicate an empty spot
1549     UChar empty = curData->variablesBase - 1;
1550     while (segmentStandins.length() < seg) {
1551         segmentStandins.append(empty);
1552     }
1553     UChar c = segmentStandins.charAt(seg-1);
1554     if (c == empty) {
1555         if (variableNext >= variableLimit) {
1556             status = U_VARIABLE_RANGE_EXHAUSTED;
1557             return 0;
1558         }
1559         c = variableNext++;
1560         // Set a placeholder in the master variables vector that will be
1561         // filled in later by setSegmentObject().  We know that we will get
1562         // called first because setSegmentObject() will call us.
1563         variablesVector.addElement((void*) NULL, status);
1564         segmentStandins.setCharAt(seg-1, c);
1565     }
1566     return c;
1567 }
1568 
1569 /**
1570  * Set the object for segment seg (1-based).
1571  */
setSegmentObject(int32_t seg,StringMatcher * adopted,UErrorCode & status)1572 void TransliteratorParser::setSegmentObject(int32_t seg, StringMatcher* adopted, UErrorCode& status) {
1573     // Since we call parseSection() recursively, nested
1574     // segments will result in segment i+1 getting parsed
1575     // and stored before segment i; be careful with the
1576     // vector handling here.
1577     if (segmentObjects.size() < seg) {
1578         segmentObjects.setSize(seg, status);
1579     }
1580     int32_t index = getSegmentStandin(seg, status) - curData->variablesBase;
1581     if (segmentObjects.elementAt(seg-1) != NULL ||
1582         variablesVector.elementAt(index) != NULL) {
1583         // should never happen
1584         status = U_INTERNAL_TRANSLITERATOR_ERROR;
1585         return;
1586     }
1587     segmentObjects.setElementAt(adopted, seg-1);
1588     variablesVector.setElementAt(adopted, index);
1589 }
1590 
1591 /**
1592  * Return the stand-in for the dot set.  It is allocated the first
1593  * time and reused thereafter.
1594  */
getDotStandIn(UErrorCode & status)1595 UChar TransliteratorParser::getDotStandIn(UErrorCode& status) {
1596     if (dotStandIn == (UChar) -1) {
1597         UnicodeSet* tempus = new UnicodeSet(UnicodeString(TRUE, DOT_SET, -1), status);
1598         // Null pointer check.
1599         if (tempus == NULL) {
1600             status = U_MEMORY_ALLOCATION_ERROR;
1601             return (UChar)0x0000;
1602         }
1603         dotStandIn = generateStandInFor(tempus, status);
1604     }
1605     return dotStandIn;
1606 }
1607 
1608 /**
1609  * Append the value of the given variable name to the given
1610  * UnicodeString.
1611  */
appendVariableDef(const UnicodeString & name,UnicodeString & buf,UErrorCode & status)1612 void TransliteratorParser::appendVariableDef(const UnicodeString& name,
1613                                                   UnicodeString& buf,
1614                                                   UErrorCode& status) {
1615     const UnicodeString* s = (const UnicodeString*) variableNames.get(name);
1616     if (s == NULL) {
1617         // We allow one undefined variable so that variable definition
1618         // statements work.  For the first undefined variable we return
1619         // the special placeholder variableLimit-1, and save the variable
1620         // name.
1621         if (undefinedVariableName.length() == 0) {
1622             undefinedVariableName = name;
1623             if (variableNext >= variableLimit) {
1624                 // throw new RuntimeException("Private use variables exhausted");
1625                 status = U_ILLEGAL_ARGUMENT_ERROR;
1626                 return;
1627             }
1628             buf.append((UChar) --variableLimit);
1629         } else {
1630             //throw new IllegalArgumentException("Undefined variable $"
1631             //                                   + name);
1632             status = U_ILLEGAL_ARGUMENT_ERROR;
1633             return;
1634         }
1635     } else {
1636         buf.append(*s);
1637     }
1638 }
1639 
1640 /**
1641  * Glue method to get around access restrictions in C++.
1642  */
1643 /*Transliterator* TransliteratorParser::createBasicInstance(const UnicodeString& id, const UnicodeString* canonID) {
1644     return Transliterator::createBasicInstance(id, canonID);
1645 }*/
1646 
1647 U_NAMESPACE_END
1648 
1649 U_CAPI int32_t
utrans_stripRules(const UChar * source,int32_t sourceLen,UChar * target,UErrorCode * status)1650 utrans_stripRules(const UChar *source, int32_t sourceLen, UChar *target, UErrorCode *status) {
1651     U_NAMESPACE_USE
1652 
1653     //const UChar *sourceStart = source;
1654     const UChar *targetStart = target;
1655     const UChar *sourceLimit = source+sourceLen;
1656     UChar *targetLimit = target+sourceLen;
1657     UChar32 c = 0;
1658     UBool quoted = FALSE;
1659     int32_t index;
1660 
1661     uprv_memset(target, 0, sourceLen*U_SIZEOF_UCHAR);
1662 
1663     /* read the rules into the buffer */
1664     while (source < sourceLimit)
1665     {
1666         index=0;
1667         U16_NEXT_UNSAFE(source, index, c);
1668         source+=index;
1669         if(c == QUOTE) {
1670             quoted = (UBool)!quoted;
1671         }
1672         else if (!quoted) {
1673             if (c == RULE_COMMENT_CHAR) {
1674                 /* skip comments and all preceding spaces */
1675                 while (targetStart < target && *(target - 1) == 0x0020) {
1676                     target--;
1677                 }
1678                 do {
1679                     if (source == sourceLimit) {
1680                         c = U_SENTINEL;
1681                         break;
1682                     }
1683                     c = *(source++);
1684                 }
1685                 while (c != CR && c != LF);
1686                 if (c < 0) {
1687                     break;
1688                 }
1689             }
1690             else if (c == ESCAPE && source < sourceLimit) {
1691                 UChar32   c2 = *source;
1692                 if (c2 == CR || c2 == LF) {
1693                     /* A backslash at the end of a line. */
1694                     /* Since we're stripping lines, ignore the backslash. */
1695                     source++;
1696                     continue;
1697                 }
1698                 if (c2 == 0x0075 && source+5 < sourceLimit) { /* \u seen. \U isn't unescaped. */
1699                     int32_t escapeOffset = 0;
1700                     UnicodeString escapedStr(source, 5);
1701                     c2 = escapedStr.unescapeAt(escapeOffset);
1702 
1703                     if (c2 == (UChar32)0xFFFFFFFF || escapeOffset == 0)
1704                     {
1705                         *status = U_PARSE_ERROR;
1706                         return 0;
1707                     }
1708                     if (!PatternProps::isWhiteSpace(c2) && !u_iscntrl(c2) && !u_ispunct(c2)) {
1709                         /* It was escaped for a reason. Write what it was suppose to be. */
1710                         source+=5;
1711                         c = c2;
1712                     }
1713                 }
1714                 else if (c2 == QUOTE) {
1715                     /* \' seen. Make sure we don't do anything when we see it again. */
1716                     quoted = (UBool)!quoted;
1717                 }
1718             }
1719         }
1720         if (c == CR || c == LF)
1721         {
1722             /* ignore spaces carriage returns, and all leading spaces on the next line.
1723             * and line feed unless in the form \uXXXX
1724             */
1725             quoted = FALSE;
1726             while (source < sourceLimit) {
1727                 c = *(source);
1728                 if (c != CR && c != LF && c != 0x0020) {
1729                     break;
1730                 }
1731                 source++;
1732             }
1733             continue;
1734         }
1735 
1736         /* Append UChar * after dissembling if c > 0xffff*/
1737         index=0;
1738         U16_APPEND_UNSAFE(target, index, c);
1739         target+=index;
1740     }
1741     if (target < targetLimit) {
1742         *target = 0;
1743     }
1744     return (int32_t)(target-targetStart);
1745 }
1746 
1747 #endif /* #if !UCONFIG_NO_TRANSLITERATION */
1748