1//
2//  file:  regexcmp.cpp
3//
4//  Copyright (C) 2002-2010 International Business Machines Corporation and others.
5//  All Rights Reserved.
6//
7//  This file contains the ICU regular expression compiler, which is responsible
8//  for processing a regular expression pattern into the compiled form that
9//  is used by the match finding engine.
10//
11
12#include "unicode/utypes.h"
13
14#if !UCONFIG_NO_REGULAR_EXPRESSIONS
15
16#include "unicode/ustring.h"
17#include "unicode/unistr.h"
18#include "unicode/uniset.h"
19#include "unicode/uchar.h"
20#include "unicode/uchriter.h"
21#include "unicode/parsepos.h"
22#include "unicode/parseerr.h"
23#include "unicode/regex.h"
24#include "util.h"
25#include "putilimp.h"
26#include "cmemory.h"
27#include "cstring.h"
28#include "uvectr32.h"
29#include "uvectr64.h"
30#include "uassert.h"
31#include "ucln_in.h"
32#include "uinvchar.h"
33
34#include "regeximp.h"
35#include "regexcst.h"   // Contains state table for the regex pattern parser.
36                        //   generated by a Perl script.
37#include "regexcmp.h"
38#include "regexst.h"
39#include "regextxt.h"
40
41
42
43U_NAMESPACE_BEGIN
44
45
46//------------------------------------------------------------------------------
47//
48//  Constructor.
49//
50//------------------------------------------------------------------------------
51RegexCompile::RegexCompile(RegexPattern *rxp, UErrorCode &status) :
52   fParenStack(status), fSetStack(status), fSetOpStack(status)
53{
54    // Lazy init of all shared global sets (needed for init()'s empty text)
55    RegexStaticSets::initGlobals(&status);
56
57    fStatus           = &status;
58
59    fRXPat            = rxp;
60    fScanIndex        = 0;
61    fLastChar         = -1;
62    fPeekChar         = -1;
63    fLineNum          = 1;
64    fCharNum          = 0;
65    fQuoteMode        = FALSE;
66    fInBackslashQuote = FALSE;
67    fModeFlags        = fRXPat->fFlags | 0x80000000;
68    fEOLComments      = TRUE;
69
70    fMatchOpenParen   = -1;
71    fMatchCloseParen  = -1;
72    fStringOpStart    = -1;
73
74    if (U_SUCCESS(status) && U_FAILURE(rxp->fDeferredStatus)) {
75        status = rxp->fDeferredStatus;
76    }
77}
78
79static const UChar      chAmp       = 0x26;      // '&'
80static const UChar      chDash      = 0x2d;      // '-'
81
82
83//------------------------------------------------------------------------------
84//
85//  Destructor
86//
87//------------------------------------------------------------------------------
88RegexCompile::~RegexCompile() {
89}
90
91static inline void addCategory(UnicodeSet *set, int32_t value, UErrorCode& ec) {
92    set->addAll(UnicodeSet().applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, value, ec));
93}
94
95//------------------------------------------------------------------------------
96//
97//  Compile regex pattern.   The state machine for rexexp pattern parsing is here.
98//                           The state tables are hand-written in the file regexcst.txt,
99//                           and converted to the form used here by a perl
100//                           script regexcst.pl
101//
102//------------------------------------------------------------------------------
103void    RegexCompile::compile(
104                         const UnicodeString &pat,   // Source pat to be compiled.
105                         UParseError &pp,            // Error position info
106                         UErrorCode &e)              // Error Code
107{
108	fRXPat->fPatternString = new UnicodeString(pat);
109    UText patternText = UTEXT_INITIALIZER;
110    utext_openConstUnicodeString(&patternText, fRXPat->fPatternString, &e);
111
112    if (U_SUCCESS(e)) {
113        compile(&patternText, pp, e);
114        utext_close(&patternText);
115    }
116}
117
118//
119//   compile, UText mode
120//     All the work is actually done here.
121//
122void    RegexCompile::compile(
123                         UText *pat,                 // Source pat to be compiled.
124                         UParseError &pp,            // Error position info
125                         UErrorCode &e)              // Error Code
126{
127    fStatus             = &e;
128    fParseErr           = &pp;
129    fStackPtr           = 0;
130    fStack[fStackPtr]   = 0;
131
132    if (U_FAILURE(*fStatus)) {
133        return;
134    }
135
136    // There should be no pattern stuff in the RegexPattern object.  They can not be reused.
137    U_ASSERT(fRXPat->fPattern == NULL || utext_nativeLength(fRXPat->fPattern) == 0);
138
139    // Prepare the RegexPattern object to receive the compiled pattern.
140    fRXPat->fPattern        = utext_clone(fRXPat->fPattern, pat, FALSE, TRUE, fStatus);
141    fRXPat->fStaticSets     = RegexStaticSets::gStaticSets->fPropSets;
142    fRXPat->fStaticSets8    = RegexStaticSets::gStaticSets->fPropSets8;
143
144
145    // Initialize the pattern scanning state machine
146    fPatternLength = utext_nativeLength(pat);
147    uint16_t                state = 1;
148    const RegexTableEl      *tableEl;
149    nextChar(fC);                        // Fetch the first char from the pattern string.
150
151    //
152    // Main loop for the regex pattern parsing state machine.
153    //   Runs once per state transition.
154    //   Each time through optionally performs, depending on the state table,
155    //      - an advance to the the next pattern char
156    //      - an action to be performed.
157    //      - pushing or popping a state to/from the local state return stack.
158    //   file regexcst.txt is the source for the state table.  The logic behind
159    //     recongizing the pattern syntax is there, not here.
160    //
161    for (;;) {
162        //  Bail out if anything has gone wrong.
163        //  Regex pattern parsing stops on the first error encountered.
164        if (U_FAILURE(*fStatus)) {
165            break;
166        }
167
168        U_ASSERT(state != 0);
169
170        // Find the state table element that matches the input char from the pattern, or the
171        //    class of the input character.  Start with the first table row for this
172        //    state, then linearly scan forward until we find a row that matches the
173        //    character.  The last row for each state always matches all characters, so
174        //    the search will stop there, if not before.
175        //
176        tableEl = &gRuleParseStateTable[state];
177        REGEX_SCAN_DEBUG_PRINTF(("char, line, col = (\'%c\', %d, %d)    state=%s ",
178            fC.fChar, fLineNum, fCharNum, RegexStateNames[state]));
179
180        for (;;) {    // loop through table rows belonging to this state, looking for one
181                      //   that matches the current input char.
182            REGEX_SCAN_DEBUG_PRINTF(("."));
183            if (tableEl->fCharClass < 127 && fC.fQuoted == FALSE &&   tableEl->fCharClass == fC.fChar) {
184                // Table row specified an individual character, not a set, and
185                //   the input character is not quoted, and
186                //   the input character matched it.
187                break;
188            }
189            if (tableEl->fCharClass == 255) {
190                // Table row specified default, match anything character class.
191                break;
192            }
193            if (tableEl->fCharClass == 254 && fC.fQuoted)  {
194                // Table row specified "quoted" and the char was quoted.
195                break;
196            }
197            if (tableEl->fCharClass == 253 && fC.fChar == (UChar32)-1)  {
198                // Table row specified eof and we hit eof on the input.
199                break;
200            }
201
202            if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 &&   // Table specs a char class &&
203                fC.fQuoted == FALSE &&                                       //   char is not escaped &&
204                fC.fChar != (UChar32)-1) {                                   //   char is not EOF
205                if (RegexStaticSets::gStaticSets->fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) {
206                    // Table row specified a character class, or set of characters,
207                    //   and the current char matches it.
208                    break;
209                }
210            }
211
212            // No match on this row, advance to the next  row for this state,
213            tableEl++;
214        }
215        REGEX_SCAN_DEBUG_PRINTF(("\n"));
216
217        //
218        // We've found the row of the state table that matches the current input
219        //   character from the rules string.
220        // Perform any action specified  by this row in the state table.
221        if (doParseActions(tableEl->fAction) == FALSE) {
222            // Break out of the state machine loop if the
223            //   the action signalled some kind of error, or
224            //   the action was to exit, occurs on normal end-of-rules-input.
225            break;
226        }
227
228        if (tableEl->fPushState != 0) {
229            fStackPtr++;
230            if (fStackPtr >= kStackSize) {
231                error(U_REGEX_INTERNAL_ERROR);
232                REGEX_SCAN_DEBUG_PRINTF(("RegexCompile::parse() - state stack overflow.\n"));
233                fStackPtr--;
234            }
235            fStack[fStackPtr] = tableEl->fPushState;
236        }
237
238        //
239        //  NextChar.  This is where characters are actually fetched from the pattern.
240        //             Happens under control of the 'n' tag in the state table.
241        //
242        if (tableEl->fNextChar) {
243            nextChar(fC);
244        }
245
246        // Get the next state from the table entry, or from the
247        //   state stack if the next state was specified as "pop".
248        if (tableEl->fNextState != 255) {
249            state = tableEl->fNextState;
250        } else {
251            state = fStack[fStackPtr];
252            fStackPtr--;
253            if (fStackPtr < 0) {
254                // state stack underflow
255                // This will occur if the user pattern has mis-matched parentheses,
256                //   with extra close parens.
257                //
258                fStackPtr++;
259                error(U_REGEX_MISMATCHED_PAREN);
260            }
261        }
262
263    }
264
265    if (U_FAILURE(*fStatus)) {
266        // Bail out if the pattern had errors.
267        //   Set stack cleanup:  a successful compile would have left it empty,
268        //   but errors can leave temporary sets hanging around.
269        while (!fSetStack.empty()) {
270            delete (UnicodeSet *)fSetStack.pop();
271        }
272        return;
273    }
274
275    //
276    // The pattern has now been read and processed, and the compiled code generated.
277    //
278
279    //
280    // Compute the number of digits requried for the largest capture group number.
281    //
282    fRXPat->fMaxCaptureDigits = 1;
283    int32_t  n = 10;
284    int32_t  groupCount = fRXPat->fGroupMap->size();
285    while (n <= groupCount) {
286        fRXPat->fMaxCaptureDigits++;
287        n *= 10;
288    }
289
290    //
291    // The pattern's fFrameSize so far has accumulated the requirements for
292    //   storage for capture parentheses, counters, etc. that are encountered
293    //   in the pattern.  Add space for the two variables that are always
294    //   present in the saved state:  the input string position (int64_t) and
295    //   the position in the compiled pattern.
296    //
297    fRXPat->fFrameSize+=RESTACKFRAME_HDRCOUNT;
298
299    //
300    // Optimization pass 1: NOPs, back-references, and case-folding
301    //
302    stripNOPs();
303
304    //
305    // Get bounds for the minimum and maximum length of a string that this
306    //   pattern can match.  Used to avoid looking for matches in strings that
307    //   are too short.
308    //
309    fRXPat->fMinMatchLen = minMatchLength(3, fRXPat->fCompiledPat->size()-1);
310
311    //
312    // Optimization pass 2: match start type
313    //
314    matchStartType();
315
316    //
317    // Set up fast latin-1 range sets
318    //
319    int32_t numSets = fRXPat->fSets->size();
320    fRXPat->fSets8 = new Regex8BitSet[numSets];
321    // Null pointer check.
322    if (fRXPat->fSets8 == NULL) {
323        e = *fStatus = U_MEMORY_ALLOCATION_ERROR;
324        return;
325    }
326    int32_t i;
327    for (i=0; i<numSets; i++) {
328        UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(i);
329        fRXPat->fSets8[i].init(s);
330    }
331
332}
333
334
335
336
337
338//------------------------------------------------------------------------------
339//
340//  doParseAction        Do some action during regex pattern parsing.
341//                       Called by the parse state machine.
342//
343//                       Generation of the match engine PCode happens here, or
344//                       in functions called from the parse actions defined here.
345//
346//
347//------------------------------------------------------------------------------
348UBool RegexCompile::doParseActions(int32_t action)
349{
350    UBool   returnVal = TRUE;
351
352    switch ((Regex_PatternParseAction)action) {
353
354    case doPatStart:
355        // Start of pattern compiles to:
356        //0   SAVE   2        Fall back to position of FAIL
357        //1   jmp    3
358        //2   FAIL            Stop if we ever reach here.
359        //3   NOP             Dummy, so start of pattern looks the same as
360        //                    the start of an ( grouping.
361        //4   NOP             Resreved, will be replaced by a save if there are
362        //                    OR | operators at the top level
363        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_STATE_SAVE, 2), *fStatus);
364        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_JMP,  3), *fStatus);
365        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_FAIL, 0), *fStatus);
366
367        // Standard open nonCapture paren action emits the two NOPs and
368        //   sets up the paren stack frame.
369        doParseActions(doOpenNonCaptureParen);
370        break;
371
372    case doPatFinish:
373        // We've scanned to the end of the pattern
374        //  The end of pattern compiles to:
375        //        URX_END
376        //    which will stop the runtime match engine.
377        //  Encountering end of pattern also behaves like a close paren,
378        //   and forces fixups of the State Save at the beginning of the compiled pattern
379        //   and of any OR operations at the top level.
380        //
381        handleCloseParen();
382        if (fParenStack.size() > 0) {
383            // Missing close paren in pattern.
384            error(U_REGEX_MISMATCHED_PAREN);
385        }
386
387        // add the END operation to the compiled pattern.
388        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_END, 0), *fStatus);
389
390        // Terminate the pattern compilation state machine.
391        returnVal = FALSE;
392        break;
393
394
395
396    case doOrOperator:
397        // Scanning a '|', as in (A|B)
398        {
399            // Insert a SAVE operation at the start of the pattern section preceding
400            //   this OR at this level.  This SAVE will branch the match forward
401            //   to the right hand side of the OR in the event that the left hand
402            //   side fails to match and backtracks.  Locate the position for the
403            //   save from the location on the top of the parentheses stack.
404            int32_t savePosition = fParenStack.popi();
405            int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(savePosition);
406            U_ASSERT(URX_TYPE(op) == URX_NOP);  // original contents of reserved location
407            op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+1);
408            fRXPat->fCompiledPat->setElementAt(op, savePosition);
409
410            // Append an JMP operation into the compiled pattern.  The operand for
411            //  the JMP will eventually be the location following the ')' for the
412            //  group.  This will be patched in later, when the ')' is encountered.
413            op = URX_BUILD(URX_JMP, 0);
414            fRXPat->fCompiledPat->addElement(op, *fStatus);
415
416            // Push the position of the newly added JMP op onto the parentheses stack.
417            // This registers if for fixup when this block's close paren is encountered.
418            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);
419
420            // Append a NOP to the compiled pattern.  This is the slot reserved
421            //   for a SAVE in the event that there is yet another '|' following
422            //   this one.
423            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
424            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);
425        }
426        break;
427
428
429    case doOpenCaptureParen:
430        // Open Paren.
431        //   Compile to a
432        //      - NOP, which later may be replaced by a save-state if the
433        //         parenthesized group gets a * quantifier, followed by
434        //      - START_CAPTURE  n    where n is stack frame offset to the capture group variables.
435        //      - NOP, which may later be replaced by a save-state if there
436        //             is an '|' alternation within the parens.
437        //
438        //    Each capture group gets three slots in the save stack frame:
439        //         0: Capture Group start position (in input string being matched.)
440        //         1: Capture Group end position.
441        //         2: Start of Match-in-progress.
442        //    The first two locations are for a completed capture group, and are
443        //     referred to by back references and the like.
444        //    The third location stores the capture start position when an START_CAPTURE is
445        //      encountered.  This will be promoted to a completed capture when (and if) the corresponding
446        //      END_CAPTURE is encountered.
447        {
448            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
449            int32_t  varsLoc    = fRXPat->fFrameSize;    // Reserve three slots in match stack frame.
450            fRXPat->fFrameSize += 3;
451            int32_t  cop        = URX_BUILD(URX_START_CAPTURE, varsLoc);
452            fRXPat->fCompiledPat->addElement(cop, *fStatus);
453            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
454
455            // On the Parentheses stack, start a new frame and add the postions
456            //   of the two NOPs.  Depending on what follows in the pattern, the
457            //   NOPs may be changed to SAVE_STATE or JMP ops, with a target
458            //   address of the end of the parenthesized group.
459            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
460            fParenStack.push(capturing, *fStatus);                        // Frame type.
461            fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus);   // The first  NOP location
462            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP loc
463
464            // Save the mapping from group number to stack frame variable position.
465            fRXPat->fGroupMap->addElement(varsLoc, *fStatus);
466        }
467         break;
468
469    case doOpenNonCaptureParen:
470        // Open non-caputuring (grouping only) Paren.
471        //   Compile to a
472        //      - NOP, which later may be replaced by a save-state if the
473        //         parenthesized group gets a * quantifier, followed by
474        //      - NOP, which may later be replaced by a save-state if there
475        //             is an '|' alternation within the parens.
476        {
477            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
478            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
479
480            // On the Parentheses stack, start a new frame and add the postions
481            //   of the two NOPs.
482            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
483            fParenStack.push(plain,      *fStatus);                       // Begin a new frame.
484            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first  NOP location
485            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP loc
486        }
487         break;
488
489
490    case doOpenAtomicParen:
491        // Open Atomic Paren.  (?>
492        //   Compile to a
493        //      - NOP, which later may be replaced if the parenthesized group
494        //         has a quantifier, followed by
495        //      - STO_SP  save state stack position, so it can be restored at the ")"
496        //      - NOP, which may later be replaced by a save-state if there
497        //             is an '|' alternation within the parens.
498        {
499            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
500            int32_t  varLoc    = fRXPat->fDataSize;    // Reserve a data location for saving the
501            fRXPat->fDataSize += 1;                    //  state stack ptr.
502            int32_t  stoOp     = URX_BUILD(URX_STO_SP, varLoc);
503            fRXPat->fCompiledPat->addElement(stoOp, *fStatus);
504            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
505
506            // On the Parentheses stack, start a new frame and add the postions
507            //   of the two NOPs.  Depending on what follows in the pattern, the
508            //   NOPs may be changed to SAVE_STATE or JMP ops, with a target
509            //   address of the end of the parenthesized group.
510            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
511            fParenStack.push(atomic, *fStatus);                           // Frame type.
512            fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus);   // The first NOP
513            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP
514        }
515        break;
516
517
518    case doOpenLookAhead:
519        // Positive Look-ahead   (?=  stuff  )
520        //
521        //   Note:   Addition of transparent input regions, with the need to
522        //           restore the original regions when failing out of a lookahead
523        //           block, complicated this sequence.  Some conbined opcodes
524        //           might make sense - or might not, lookahead aren't that common.
525        //
526        //      Caution:  min match length optimization knows about this
527        //               sequence; don't change without making updates there too.
528        //
529        // Compiles to
530        //    1    START_LA     dataLoc     Saves SP, Input Pos
531        //    2.   STATE_SAVE   4            on failure of lookahead, goto 4
532        //    3    JMP          6           continue ...
533        //
534        //    4.   LA_END                   Look Ahead failed.  Restore regions.
535        //    5.   BACKTRACK                and back track again.
536        //
537        //    6.   NOP              reserved for use by quantifiers on the block.
538        //                          Look-ahead can't have quantifiers, but paren stack
539        //                             compile time conventions require the slot anyhow.
540        //    7.   NOP              may be replaced if there is are '|' ops in the block.
541        //    8.     code for parenthesized stuff.
542        //    9.   LA_END
543        //
544        //  Two data slots are reserved, for saving the stack ptr and the input position.
545        {
546            int32_t dataLoc = fRXPat->fDataSize;
547            fRXPat->fDataSize += 2;
548            int32_t op = URX_BUILD(URX_LA_START, dataLoc);
549            fRXPat->fCompiledPat->addElement(op, *fStatus);
550
551            op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+ 2);
552            fRXPat->fCompiledPat->addElement(op, *fStatus);
553
554            op = URX_BUILD(URX_JMP, fRXPat->fCompiledPat->size()+ 3);
555            fRXPat->fCompiledPat->addElement(op, *fStatus);
556
557            op = URX_BUILD(URX_LA_END, dataLoc);
558            fRXPat->fCompiledPat->addElement(op, *fStatus);
559
560            op = URX_BUILD(URX_BACKTRACK, 0);
561            fRXPat->fCompiledPat->addElement(op, *fStatus);
562
563            op = URX_BUILD(URX_NOP, 0);
564            fRXPat->fCompiledPat->addElement(op, *fStatus);
565            fRXPat->fCompiledPat->addElement(op, *fStatus);
566
567            // On the Parentheses stack, start a new frame and add the postions
568            //   of the NOPs.
569            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
570            fParenStack.push(lookAhead, *fStatus);                        // Frame type.
571            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first  NOP location
572            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP location
573        }
574        break;
575
576    case doOpenLookAheadNeg:
577        // Negated Lookahead.   (?! stuff )
578        // Compiles to
579        //    1.    START_LA    dataloc
580        //    2.    SAVE_STATE  7         // Fail within look-ahead block restores to this state,
581        //                                //   which continues with the match.
582        //    3.    NOP                   // Std. Open Paren sequence, for possible '|'
583        //    4.       code for parenthesized stuff.
584        //    5.    END_LA                // Cut back stack, remove saved state from step 2.
585        //    6.    BACKTRACK             // code in block succeeded, so neg. lookahead fails.
586        //    7.    END_LA                // Restore match region, in case look-ahead was using
587        //                                        an alternate (transparent) region.
588        {
589            int32_t dataLoc = fRXPat->fDataSize;
590            fRXPat->fDataSize += 2;
591            int32_t op = URX_BUILD(URX_LA_START, dataLoc);
592            fRXPat->fCompiledPat->addElement(op, *fStatus);
593
594            op = URX_BUILD(URX_STATE_SAVE, 0);    // dest address will be patched later.
595            fRXPat->fCompiledPat->addElement(op, *fStatus);
596
597            op = URX_BUILD(URX_NOP, 0);
598            fRXPat->fCompiledPat->addElement(op, *fStatus);
599
600            // On the Parentheses stack, start a new frame and add the postions
601            //   of the StateSave and NOP.
602            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
603            fParenStack.push(negLookAhead, *fStatus);                    // Frame type
604            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The STATE_SAVE location
605            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP location
606
607            // Instructions #5 - #7 will be added when the ')' is encountered.
608        }
609        break;
610
611    case doOpenLookBehind:
612        {
613            //   Compile a (?<= look-behind open paren.
614            //
615            //          Compiles to
616            //              0       URX_LB_START     dataLoc
617            //              1       URX_LB_CONT      dataLoc
618            //              2                        MinMatchLen
619            //              3                        MaxMatchLen
620            //              4       URX_NOP          Standard '(' boilerplate.
621            //              5       URX_NOP          Reserved slot for use with '|' ops within (block).
622            //              6         <code for LookBehind expression>
623            //              7       URX_LB_END       dataLoc    # Check match len, restore input  len
624            //              8       URX_LA_END       dataLoc    # Restore stack, input pos
625            //
626            //          Allocate a block of matcher data, to contain (when running a match)
627            //              0:    Stack ptr on entry
628            //              1:    Input Index on entry
629            //              2:    Start index of match current match attempt.
630            //              3:    Original Input String len.
631
632            // Allocate data space
633            int32_t dataLoc = fRXPat->fDataSize;
634            fRXPat->fDataSize += 4;
635
636            // Emit URX_LB_START
637            int32_t op = URX_BUILD(URX_LB_START, dataLoc);
638            fRXPat->fCompiledPat->addElement(op, *fStatus);
639
640            // Emit URX_LB_CONT
641            op = URX_BUILD(URX_LB_CONT, dataLoc);
642            fRXPat->fCompiledPat->addElement(op, *fStatus);
643            fRXPat->fCompiledPat->addElement(0,  *fStatus);    // MinMatchLength.  To be filled later.
644            fRXPat->fCompiledPat->addElement(0,  *fStatus);    // MaxMatchLength.  To be filled later.
645
646            // Emit the NOP
647            op = URX_BUILD(URX_NOP, 0);
648            fRXPat->fCompiledPat->addElement(op, *fStatus);
649            fRXPat->fCompiledPat->addElement(op, *fStatus);
650
651            // On the Parentheses stack, start a new frame and add the postions
652            //   of the URX_LB_CONT and the NOP.
653            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
654            fParenStack.push(lookBehind, *fStatus);                       // Frame type
655            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first NOP location
656            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The 2nd   NOP location
657
658            // The final two instructions will be added when the ')' is encountered.
659        }
660
661        break;
662
663    case doOpenLookBehindNeg:
664        {
665            //   Compile a (?<! negated look-behind open paren.
666            //
667            //          Compiles to
668            //              0       URX_LB_START     dataLoc    # Save entry stack, input len
669            //              1       URX_LBN_CONT     dataLoc    # Iterate possible match positions
670            //              2                        MinMatchLen
671            //              3                        MaxMatchLen
672            //              4                        continueLoc (9)
673            //              5       URX_NOP          Standard '(' boilerplate.
674            //              6       URX_NOP          Reserved slot for use with '|' ops within (block).
675            //              7         <code for LookBehind expression>
676            //              8       URX_LBN_END      dataLoc    # Check match len, cause a FAIL
677            //              9       ...
678            //
679            //          Allocate a block of matcher data, to contain (when running a match)
680            //              0:    Stack ptr on entry
681            //              1:    Input Index on entry
682            //              2:    Start index of match current match attempt.
683            //              3:    Original Input String len.
684
685            // Allocate data space
686            int32_t dataLoc = fRXPat->fDataSize;
687            fRXPat->fDataSize += 4;
688
689            // Emit URX_LB_START
690            int32_t op = URX_BUILD(URX_LB_START, dataLoc);
691            fRXPat->fCompiledPat->addElement(op, *fStatus);
692
693            // Emit URX_LBN_CONT
694            op = URX_BUILD(URX_LBN_CONT, dataLoc);
695            fRXPat->fCompiledPat->addElement(op, *fStatus);
696            fRXPat->fCompiledPat->addElement(0,  *fStatus);    // MinMatchLength.  To be filled later.
697            fRXPat->fCompiledPat->addElement(0,  *fStatus);    // MaxMatchLength.  To be filled later.
698            fRXPat->fCompiledPat->addElement(0,  *fStatus);    // Continue Loc.    To be filled later.
699
700            // Emit the NOP
701            op = URX_BUILD(URX_NOP, 0);
702            fRXPat->fCompiledPat->addElement(op, *fStatus);
703            fRXPat->fCompiledPat->addElement(op, *fStatus);
704
705            // On the Parentheses stack, start a new frame and add the postions
706            //   of the URX_LB_CONT and the NOP.
707            fParenStack.push(fModeFlags, *fStatus);                       // Match mode state
708            fParenStack.push(lookBehindN, *fStatus);                      // Frame type
709            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first NOP location
710            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The 2nd   NOP location
711
712            // The final two instructions will be added when the ')' is encountered.
713        }
714        break;
715
716    case doConditionalExpr:
717        // Conditionals such as (?(1)a:b)
718    case doPerlInline:
719        // Perl inline-condtionals.  (?{perl code}a|b) We're not perl, no way to do them.
720        error(U_REGEX_UNIMPLEMENTED);
721        break;
722
723
724    case doCloseParen:
725        handleCloseParen();
726        if (fParenStack.size() <= 0) {
727            //  Extra close paren, or missing open paren.
728            error(U_REGEX_MISMATCHED_PAREN);
729        }
730        break;
731
732    case doNOP:
733        break;
734
735
736    case doBadOpenParenType:
737    case doRuleError:
738        error(U_REGEX_RULE_SYNTAX);
739        break;
740
741
742    case doMismatchedParenErr:
743        error(U_REGEX_MISMATCHED_PAREN);
744        break;
745
746    case doPlus:
747        //  Normal '+'  compiles to
748        //     1.   stuff to be repeated  (already built)
749        //     2.   jmp-sav 1
750        //     3.   ...
751        //
752        //  Or, if the item to be repeated can match a zero length string,
753        //     1.   STO_INP_LOC  data-loc
754        //     2.      body of stuff to be repeated
755        //     3.   JMP_SAV_X    2
756        //     4.   ...
757
758        //
759        //  Or, if the item to be repeated is simple
760        //     1.   Item to be repeated.
761        //     2.   LOOP_SR_I    set number  (assuming repeated item is a set ref)
762        //     3.   LOOP_C       stack location
763        {
764            int32_t  topLoc = blockTopLoc(FALSE);        // location of item #1
765            int32_t  frameLoc;
766
767            // Check for simple constructs, which may get special optimized code.
768            if (topLoc == fRXPat->fCompiledPat->size() - 1) {
769                int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc);
770
771                if (URX_TYPE(repeatedOp) == URX_SETREF) {
772                    // Emit optimized code for [char set]+
773                    int32_t loopOpI = URX_BUILD(URX_LOOP_SR_I, URX_VAL(repeatedOp));
774                    fRXPat->fCompiledPat->addElement(loopOpI, *fStatus);
775                    frameLoc = fRXPat->fFrameSize;
776                    fRXPat->fFrameSize++;
777                    int32_t loopOpC = URX_BUILD(URX_LOOP_C, frameLoc);
778                    fRXPat->fCompiledPat->addElement(loopOpC, *fStatus);
779                    break;
780                }
781
782                if (URX_TYPE(repeatedOp) == URX_DOTANY ||
783                    URX_TYPE(repeatedOp) == URX_DOTANY_ALL ||
784                    URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) {
785                    // Emit Optimized code for .+ operations.
786                    int32_t loopOpI = URX_BUILD(URX_LOOP_DOT_I, 0);
787                    if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) {
788                        // URX_LOOP_DOT_I operand is a flag indicating ". matches any" mode.
789                        loopOpI |= 1;
790                    }
791                    if (fModeFlags & UREGEX_UNIX_LINES) {
792                        loopOpI |= 2;
793                    }
794                    fRXPat->fCompiledPat->addElement(loopOpI, *fStatus);
795                    frameLoc = fRXPat->fFrameSize;
796                    fRXPat->fFrameSize++;
797                    int32_t loopOpC = URX_BUILD(URX_LOOP_C, frameLoc);
798                    fRXPat->fCompiledPat->addElement(loopOpC, *fStatus);
799                    break;
800                }
801
802            }
803
804            // General case.
805
806            // Check for minimum match length of zero, which requires
807            //    extra loop-breaking code.
808            if (minMatchLength(topLoc, fRXPat->fCompiledPat->size()-1) == 0) {
809                // Zero length match is possible.
810                // Emit the code sequence that can handle it.
811                insertOp(topLoc);
812                frameLoc =  fRXPat->fFrameSize;
813                fRXPat->fFrameSize++;
814
815                int32_t op = URX_BUILD(URX_STO_INP_LOC, frameLoc);
816                fRXPat->fCompiledPat->setElementAt(op, topLoc);
817
818                op = URX_BUILD(URX_JMP_SAV_X, topLoc+1);
819                fRXPat->fCompiledPat->addElement(op, *fStatus);
820            } else {
821                // Simpler code when the repeated body must match something non-empty
822                int32_t  jmpOp  = URX_BUILD(URX_JMP_SAV, topLoc);
823                fRXPat->fCompiledPat->addElement(jmpOp, *fStatus);
824            }
825        }
826        break;
827
828    case doNGPlus:
829        //  Non-greedy '+?'  compiles to
830        //     1.   stuff to be repeated  (already built)
831        //     2.   state-save  1
832        //     3.   ...
833        {
834            int32_t topLoc      = blockTopLoc(FALSE);
835            int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, topLoc);
836            fRXPat->fCompiledPat->addElement(saveStateOp, *fStatus);
837        }
838        break;
839
840
841    case doOpt:
842        // Normal (greedy) ? quantifier.
843        //  Compiles to
844        //     1. state save 3
845        //     2.    body of optional block
846        //     3. ...
847        // Insert the state save into the compiled pattern, and we're done.
848        {
849            int32_t   saveStateLoc = blockTopLoc(TRUE);
850            int32_t   saveStateOp  = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size());
851            fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc);
852        }
853        break;
854
855    case doNGOpt:
856        // Non-greedy ?? quantifier
857        //   compiles to
858        //    1.  jmp   4
859        //    2.     body of optional block
860        //    3   jmp   5
861        //    4.  state save 2
862        //    5    ...
863        //  This code is less than ideal, with two jmps instead of one, because we can only
864        //  insert one instruction at the top of the block being iterated.
865        {
866            int32_t  jmp1_loc = blockTopLoc(TRUE);
867            int32_t  jmp2_loc = fRXPat->fCompiledPat->size();
868
869            int32_t  jmp1_op  = URX_BUILD(URX_JMP, jmp2_loc+1);
870            fRXPat->fCompiledPat->setElementAt(jmp1_op, jmp1_loc);
871
872            int32_t  jmp2_op  = URX_BUILD(URX_JMP, jmp2_loc+2);
873            fRXPat->fCompiledPat->addElement(jmp2_op, *fStatus);
874
875            int32_t  save_op  = URX_BUILD(URX_STATE_SAVE, jmp1_loc+1);
876            fRXPat->fCompiledPat->addElement(save_op, *fStatus);
877        }
878        break;
879
880
881    case doStar:
882        // Normal (greedy) * quantifier.
883        // Compiles to
884        //       1.   STATE_SAVE   4
885        //       2.      body of stuff being iterated over
886        //       3.   JMP_SAV      2
887        //       4.   ...
888        //
889        // Or, if the body is a simple [Set],
890        //       1.   LOOP_SR_I    set number
891        //       2.   LOOP_C       stack location
892        //       ...
893        //
894        // Or if this is a .*
895        //       1.   LOOP_DOT_I    (. matches all mode flag)
896        //       2.   LOOP_C        stack location
897        //
898        // Or, if the body can match a zero-length string, to inhibit infinite loops,
899        //       1.   STATE_SAVE   5
900        //       2.   STO_INP_LOC  data-loc
901        //       3.      body of stuff
902        //       4.   JMP_SAV_X    2
903        //       5.   ...
904        {
905            // location of item #1, the STATE_SAVE
906            int32_t   topLoc = blockTopLoc(FALSE);
907            int32_t   dataLoc = -1;
908
909            // Check for simple *, where the construct being repeated
910            //   compiled to single opcode, and might be optimizable.
911            if (topLoc == fRXPat->fCompiledPat->size() - 1) {
912                int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc);
913
914                if (URX_TYPE(repeatedOp) == URX_SETREF) {
915                    // Emit optimized code for a [char set]*
916                    int32_t loopOpI = URX_BUILD(URX_LOOP_SR_I, URX_VAL(repeatedOp));
917                    fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc);
918                    dataLoc = fRXPat->fFrameSize;
919                    fRXPat->fFrameSize++;
920                    int32_t loopOpC = URX_BUILD(URX_LOOP_C, dataLoc);
921                    fRXPat->fCompiledPat->addElement(loopOpC, *fStatus);
922                    break;
923                }
924
925                if (URX_TYPE(repeatedOp) == URX_DOTANY ||
926                    URX_TYPE(repeatedOp) == URX_DOTANY_ALL ||
927                    URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) {
928                    // Emit Optimized code for .* operations.
929                    int32_t loopOpI = URX_BUILD(URX_LOOP_DOT_I, 0);
930                    if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) {
931                        // URX_LOOP_DOT_I operand is a flag indicating . matches any mode.
932                        loopOpI |= 1;
933                    }
934                    if ((fModeFlags & UREGEX_UNIX_LINES) != 0) {
935                        loopOpI |= 2;
936                    }
937                    fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc);
938                    dataLoc = fRXPat->fFrameSize;
939                    fRXPat->fFrameSize++;
940                    int32_t loopOpC = URX_BUILD(URX_LOOP_C, dataLoc);
941                    fRXPat->fCompiledPat->addElement(loopOpC, *fStatus);
942                    break;
943                }
944            }
945
946            // Emit general case code for this *
947            // The optimizations did not apply.
948
949            int32_t   saveStateLoc = blockTopLoc(TRUE);
950            int32_t   jmpOp        = URX_BUILD(URX_JMP_SAV, saveStateLoc+1);
951
952            // Check for minimum match length of zero, which requires
953            //    extra loop-breaking code.
954            if (minMatchLength(saveStateLoc, fRXPat->fCompiledPat->size()-1) == 0) {
955                insertOp(saveStateLoc);
956                dataLoc =  fRXPat->fFrameSize;
957                fRXPat->fFrameSize++;
958
959                int32_t op = URX_BUILD(URX_STO_INP_LOC, dataLoc);
960                fRXPat->fCompiledPat->setElementAt(op, saveStateLoc+1);
961                jmpOp      = URX_BUILD(URX_JMP_SAV_X, saveStateLoc+2);
962            }
963
964            // Locate the position in the compiled pattern where the match will continue
965            //   after completing the *.   (4 or 5 in the comment above)
966            int32_t continueLoc = fRXPat->fCompiledPat->size()+1;
967
968            // Put together the save state op store it into the compiled code.
969            int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, continueLoc);
970            fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc);
971
972            // Append the URX_JMP_SAV or URX_JMPX operation to the compiled pattern.
973            fRXPat->fCompiledPat->addElement(jmpOp, *fStatus);
974        }
975        break;
976
977    case doNGStar:
978        // Non-greedy *? quantifier
979        // compiles to
980        //     1.   JMP    3
981        //     2.      body of stuff being iterated over
982        //     3.   STATE_SAVE  2
983        //     4    ...
984        {
985            int32_t     jmpLoc  = blockTopLoc(TRUE);                   // loc  1.
986            int32_t     saveLoc = fRXPat->fCompiledPat->size();        // loc  3.
987            int32_t     jmpOp   = URX_BUILD(URX_JMP, saveLoc);
988            int32_t     stateSaveOp = URX_BUILD(URX_STATE_SAVE, jmpLoc+1);
989            fRXPat->fCompiledPat->setElementAt(jmpOp, jmpLoc);
990            fRXPat->fCompiledPat->addElement(stateSaveOp, *fStatus);
991        }
992        break;
993
994
995    case doIntervalInit:
996        // The '{' opening an interval quantifier was just scanned.
997        // Init the counter varaiables that will accumulate the values as the digits
998        //    are scanned.
999        fIntervalLow = 0;
1000        fIntervalUpper = -1;
1001        break;
1002
1003    case doIntevalLowerDigit:
1004        // Scanned a digit from the lower value of an {lower,upper} interval
1005        {
1006            int32_t digitValue = u_charDigitValue(fC.fChar);
1007            U_ASSERT(digitValue >= 0);
1008            fIntervalLow = fIntervalLow*10 + digitValue;
1009            if (fIntervalLow < 0) {
1010                error(U_REGEX_NUMBER_TOO_BIG);
1011            }
1012        }
1013        break;
1014
1015    case doIntervalUpperDigit:
1016        // Scanned a digit from the upper value of an {lower,upper} interval
1017        {
1018            if (fIntervalUpper < 0) {
1019                fIntervalUpper = 0;
1020            }
1021            int32_t digitValue = u_charDigitValue(fC.fChar);
1022            U_ASSERT(digitValue >= 0);
1023            fIntervalUpper = fIntervalUpper*10 + digitValue;
1024            if (fIntervalUpper < 0) {
1025                error(U_REGEX_NUMBER_TOO_BIG);
1026            }
1027        }
1028        break;
1029
1030    case doIntervalSame:
1031        // Scanned a single value interval like {27}.  Upper = Lower.
1032        fIntervalUpper = fIntervalLow;
1033        break;
1034
1035    case doInterval:
1036        // Finished scanning a normal {lower,upper} interval.  Generate the code for it.
1037        if (compileInlineInterval() == FALSE) {
1038            compileInterval(URX_CTR_INIT, URX_CTR_LOOP);
1039        }
1040        break;
1041
1042    case doPossessiveInterval:
1043        // Finished scanning a Possessive {lower,upper}+ interval.  Generate the code for it.
1044        {
1045            // Remember the loc for the top of the block being looped over.
1046            //   (Can not reserve a slot in the compiled pattern at this time, because
1047            //    compileInterval needs to reserve also, and blockTopLoc can only reserve
1048            //    once per block.)
1049            int32_t topLoc = blockTopLoc(FALSE);
1050
1051            // Produce normal looping code.
1052            compileInterval(URX_CTR_INIT, URX_CTR_LOOP);
1053
1054            // Surround the just-emitted normal looping code with a STO_SP ... LD_SP
1055            //  just as if the loop was inclosed in atomic parentheses.
1056
1057            // First the STO_SP before the start of the loop
1058            insertOp(topLoc);
1059            int32_t  varLoc    = fRXPat->fDataSize;    // Reserve a data location for saving the
1060            fRXPat->fDataSize += 1;                    //  state stack ptr.
1061            int32_t  op        = URX_BUILD(URX_STO_SP, varLoc);
1062            fRXPat->fCompiledPat->setElementAt(op, topLoc);
1063
1064            int32_t loopOp = (int32_t)fRXPat->fCompiledPat->popi();
1065            U_ASSERT(URX_TYPE(loopOp) == URX_CTR_LOOP && URX_VAL(loopOp) == topLoc);
1066            loopOp++;     // point LoopOp after the just-inserted STO_SP
1067            fRXPat->fCompiledPat->push(loopOp, *fStatus);
1068
1069            // Then the LD_SP after the end of the loop
1070            op = URX_BUILD(URX_LD_SP, varLoc);
1071            fRXPat->fCompiledPat->addElement(op, *fStatus);
1072        }
1073
1074        break;
1075
1076    case doNGInterval:
1077        // Finished scanning a non-greedy {lower,upper}? interval.  Generate the code for it.
1078        compileInterval(URX_CTR_INIT_NG, URX_CTR_LOOP_NG);
1079        break;
1080
1081    case doIntervalError:
1082        error(U_REGEX_BAD_INTERVAL);
1083        break;
1084
1085    case doLiteralChar:
1086        // We've just scanned a "normal" character from the pattern,
1087        literalChar(fC.fChar);
1088        break;
1089
1090
1091    case doEscapedLiteralChar:
1092        // We've just scanned an backslashed escaped character with  no
1093        //   special meaning.  It represents itself.
1094        if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 &&
1095            ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) ||     // in [A-Z]
1096            (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) {   // in [a-z]
1097               error(U_REGEX_BAD_ESCAPE_SEQUENCE);
1098             }
1099        literalChar(fC.fChar);
1100        break;
1101
1102
1103    case doDotAny:
1104        // scanned a ".",  match any single character.
1105        {
1106            int32_t   op;
1107            if (fModeFlags & UREGEX_DOTALL) {
1108                op = URX_BUILD(URX_DOTANY_ALL, 0);
1109            } else if (fModeFlags & UREGEX_UNIX_LINES) {
1110                op = URX_BUILD(URX_DOTANY_UNIX, 0);
1111            } else {
1112                op = URX_BUILD(URX_DOTANY, 0);
1113            }
1114            fRXPat->fCompiledPat->addElement(op, *fStatus);
1115        }
1116        break;
1117
1118    case doCaret:
1119        {
1120            int32_t op = 0;
1121            if (       (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
1122                op = URX_CARET;
1123            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
1124                op = URX_CARET_M;
1125            } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
1126                op = URX_CARET;   // Only testing true start of input.
1127            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
1128                op = URX_CARET_M_UNIX;
1129            }
1130            fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus);
1131        }
1132        break;
1133
1134    case doDollar:
1135        {
1136            int32_t op = 0;
1137            if (       (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
1138                op = URX_DOLLAR;
1139            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) {
1140                op = URX_DOLLAR_M;
1141            } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
1142                op = URX_DOLLAR_D;
1143            } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) {
1144                op = URX_DOLLAR_MD;
1145            }
1146            fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus);
1147        }
1148        break;
1149
1150    case doBackslashA:
1151        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_CARET, 0), *fStatus);
1152        break;
1153
1154    case doBackslashB:
1155        {
1156            #if  UCONFIG_NO_BREAK_ITERATION==1
1157            if (fModeFlags & UREGEX_UWORD) {
1158                error(U_UNSUPPORTED_ERROR);
1159            }
1160            #endif
1161            int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B;
1162            fRXPat->fCompiledPat->addElement(URX_BUILD(op, 1), *fStatus);
1163        }
1164        break;
1165
1166    case doBackslashb:
1167        {
1168            #if  UCONFIG_NO_BREAK_ITERATION==1
1169            if (fModeFlags & UREGEX_UWORD) {
1170                error(U_UNSUPPORTED_ERROR);
1171            }
1172            #endif
1173            int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B;
1174            fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus);
1175        }
1176        break;
1177
1178    case doBackslashD:
1179        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_D, 1), *fStatus);
1180        break;
1181
1182    case doBackslashd:
1183        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_D, 0), *fStatus);
1184        break;
1185
1186    case doBackslashG:
1187        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_G, 0), *fStatus);
1188        break;
1189
1190    case doBackslashS:
1191        fRXPat->fCompiledPat->addElement(
1192            URX_BUILD(URX_STAT_SETREF_N, URX_ISSPACE_SET), *fStatus);
1193        break;
1194
1195    case doBackslashs:
1196        fRXPat->fCompiledPat->addElement(
1197            URX_BUILD(URX_STATIC_SETREF, URX_ISSPACE_SET), *fStatus);
1198        break;
1199
1200    case doBackslashW:
1201        fRXPat->fCompiledPat->addElement(
1202            URX_BUILD(URX_STAT_SETREF_N, URX_ISWORD_SET), *fStatus);
1203        break;
1204
1205    case doBackslashw:
1206        fRXPat->fCompiledPat->addElement(
1207            URX_BUILD(URX_STATIC_SETREF, URX_ISWORD_SET), *fStatus);
1208        break;
1209
1210    case doBackslashX:
1211        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_X, 0), *fStatus);
1212        break;
1213
1214
1215    case doBackslashZ:
1216        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_DOLLAR, 0), *fStatus);
1217        break;
1218
1219    case doBackslashz:
1220        fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_Z, 0), *fStatus);
1221        break;
1222
1223    case doEscapeError:
1224        error(U_REGEX_BAD_ESCAPE_SEQUENCE);
1225        break;
1226
1227    case doExit:
1228        returnVal = FALSE;
1229        break;
1230
1231    case doProperty:
1232        {
1233            UnicodeSet *theSet = scanProp();
1234            compileSet(theSet);
1235        }
1236        break;
1237
1238    case doNamedChar:
1239        {
1240            UChar32 c = scanNamedChar();
1241            literalChar(c);
1242        }
1243        break;
1244
1245
1246    case doBackRef:
1247        // BackReference.  Somewhat unusual in that the front-end can not completely parse
1248        //                 the regular expression, because the number of digits to be consumed
1249        //                 depends on the number of capture groups that have been defined.  So
1250        //                 we have to do it here instead.
1251        {
1252            int32_t  numCaptureGroups = fRXPat->fGroupMap->size();
1253            int32_t  groupNum = 0;
1254            UChar32  c        = fC.fChar;
1255
1256            for (;;) {
1257                // Loop once per digit, for max allowed number of digits in a back reference.
1258                int32_t digit = u_charDigitValue(c);
1259                groupNum = groupNum * 10 + digit;
1260                if (groupNum >= numCaptureGroups) {
1261                    break;
1262                }
1263                c = peekCharLL();
1264                if (RegexStaticSets::gStaticSets->fRuleDigitsAlias->contains(c) == FALSE) {
1265                    break;
1266                }
1267                nextCharLL();
1268            }
1269
1270            // Scan of the back reference in the source regexp is complete.  Now generate
1271            //  the compiled code for it.
1272            // Because capture groups can be forward-referenced by back-references,
1273            //  we fill the operand with the capture group number.  At the end
1274            //  of compilation, it will be changed to the variable's location.
1275            U_ASSERT(groupNum > 0);
1276            int32_t  op;
1277            if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
1278                op = URX_BUILD(URX_BACKREF_I, groupNum);
1279            } else {
1280                op = URX_BUILD(URX_BACKREF, groupNum);
1281            }
1282            fRXPat->fCompiledPat->addElement(op, *fStatus);
1283        }
1284        break;
1285
1286
1287    case doPossessivePlus:
1288        // Possessive ++ quantifier.
1289        // Compiles to
1290        //       1.   STO_SP
1291        //       2.      body of stuff being iterated over
1292        //       3.   STATE_SAVE 5
1293        //       4.   JMP        2
1294        //       5.   LD_SP
1295        //       6.   ...
1296        //
1297        //  Note:  TODO:  This is pretty inefficient.  A mass of saved state is built up
1298        //                then unconditionally discarded.  Perhaps introduce a new opcode.  Ticket 6056
1299        //
1300        {
1301            // Emit the STO_SP
1302            int32_t   topLoc = blockTopLoc(TRUE);
1303            int32_t   stoLoc = fRXPat->fDataSize;
1304            fRXPat->fDataSize++;       // Reserve the data location for storing save stack ptr.
1305            int32_t   op     = URX_BUILD(URX_STO_SP, stoLoc);
1306            fRXPat->fCompiledPat->setElementAt(op, topLoc);
1307
1308            // Emit the STATE_SAVE
1309            op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+2);
1310            fRXPat->fCompiledPat->addElement(op, *fStatus);
1311
1312            // Emit the JMP
1313            op = URX_BUILD(URX_JMP, topLoc+1);
1314            fRXPat->fCompiledPat->addElement(op, *fStatus);
1315
1316            // Emit the LD_SP
1317            op = URX_BUILD(URX_LD_SP, stoLoc);
1318            fRXPat->fCompiledPat->addElement(op, *fStatus);
1319        }
1320        break;
1321
1322    case doPossessiveStar:
1323        // Possessive *+ quantifier.
1324        // Compiles to
1325        //       1.   STO_SP       loc
1326        //       2.   STATE_SAVE   5
1327        //       3.      body of stuff being iterated over
1328        //       4.   JMP          2
1329        //       5.   LD_SP        loc
1330        //       6    ...
1331        // TODO:  do something to cut back the state stack each time through the loop.
1332        {
1333            // Reserve two slots at the top of the block.
1334            int32_t   topLoc = blockTopLoc(TRUE);
1335            insertOp(topLoc);
1336
1337            // emit   STO_SP     loc
1338            int32_t   stoLoc = fRXPat->fDataSize;
1339            fRXPat->fDataSize++;       // Reserve the data location for storing save stack ptr.
1340            int32_t   op     = URX_BUILD(URX_STO_SP, stoLoc);
1341            fRXPat->fCompiledPat->setElementAt(op, topLoc);
1342
1343            // Emit the SAVE_STATE   5
1344            int32_t L7 = fRXPat->fCompiledPat->size()+1;
1345            op = URX_BUILD(URX_STATE_SAVE, L7);
1346            fRXPat->fCompiledPat->setElementAt(op, topLoc+1);
1347
1348            // Append the JMP operation.
1349            op = URX_BUILD(URX_JMP, topLoc+1);
1350            fRXPat->fCompiledPat->addElement(op, *fStatus);
1351
1352            // Emit the LD_SP       loc
1353            op = URX_BUILD(URX_LD_SP, stoLoc);
1354            fRXPat->fCompiledPat->addElement(op, *fStatus);
1355        }
1356        break;
1357
1358    case doPossessiveOpt:
1359        // Possessive  ?+ quantifier.
1360        //  Compiles to
1361        //     1. STO_SP      loc
1362        //     2. SAVE_STATE  5
1363        //     3.    body of optional block
1364        //     4. LD_SP       loc
1365        //     5. ...
1366        //
1367        {
1368            // Reserve two slots at the top of the block.
1369            int32_t   topLoc = blockTopLoc(TRUE);
1370            insertOp(topLoc);
1371
1372            // Emit the STO_SP
1373            int32_t   stoLoc = fRXPat->fDataSize;
1374            fRXPat->fDataSize++;       // Reserve the data location for storing save stack ptr.
1375            int32_t   op     = URX_BUILD(URX_STO_SP, stoLoc);
1376            fRXPat->fCompiledPat->setElementAt(op, topLoc);
1377
1378            // Emit the SAVE_STATE
1379            int32_t   continueLoc = fRXPat->fCompiledPat->size()+1;
1380            op = URX_BUILD(URX_STATE_SAVE, continueLoc);
1381            fRXPat->fCompiledPat->setElementAt(op, topLoc+1);
1382
1383            // Emit the LD_SP
1384            op = URX_BUILD(URX_LD_SP, stoLoc);
1385            fRXPat->fCompiledPat->addElement(op, *fStatus);
1386        }
1387        break;
1388
1389
1390    case doBeginMatchMode:
1391        fNewModeFlags = fModeFlags;
1392        fSetModeFlag  = TRUE;
1393        break;
1394
1395    case doMatchMode:   //  (?i)    and similar
1396        {
1397            int32_t  bit = 0;
1398            switch (fC.fChar) {
1399            case 0x69: /* 'i' */   bit = UREGEX_CASE_INSENSITIVE; break;
1400            case 0x64: /* 'd' */   bit = UREGEX_UNIX_LINES;       break;
1401            case 0x6d: /* 'm' */   bit = UREGEX_MULTILINE;        break;
1402            case 0x73: /* 's' */   bit = UREGEX_DOTALL;           break;
1403            case 0x75: /* 'u' */   bit = 0; /* Unicode casing */  break;
1404            case 0x77: /* 'w' */   bit = UREGEX_UWORD;            break;
1405            case 0x78: /* 'x' */   bit = UREGEX_COMMENTS;         break;
1406            case 0x2d: /* '-' */   fSetModeFlag = FALSE;          break;
1407            default:
1408                U_ASSERT(FALSE);   // Should never happen.  Other chars are filtered out
1409                                   // by the scanner.
1410            }
1411            if (fSetModeFlag) {
1412                fNewModeFlags |= bit;
1413            } else {
1414                fNewModeFlags &= ~bit;
1415            }
1416        }
1417        break;
1418
1419    case doSetMatchMode:
1420        // We've got a (?i) or similar.  The match mode is being changed, but
1421        //   the change is not scoped to a parenthesized block.
1422        U_ASSERT(fNewModeFlags < 0);
1423        fModeFlags = fNewModeFlags;
1424
1425        // Prevent any string from spanning across the change of match mode.
1426        //   Otherwise the pattern "abc(?i)def" would make a single string of "abcdef"
1427        fixLiterals();
1428        break;
1429
1430
1431    case doMatchModeParen:
1432        // We've got a (?i: or similar.  Begin a parenthesized block, save old
1433        //   mode flags so they can be restored at the close of the block.
1434        //
1435        //   Compile to a
1436        //      - NOP, which later may be replaced by a save-state if the
1437        //         parenthesized group gets a * quantifier, followed by
1438        //      - NOP, which may later be replaced by a save-state if there
1439        //             is an '|' alternation within the parens.
1440        {
1441            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
1442            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus);
1443
1444            // On the Parentheses stack, start a new frame and add the postions
1445            //   of the two NOPs (a normal non-capturing () frame, except for the
1446            //   saving of the orignal mode flags.)
1447            fParenStack.push(fModeFlags, *fStatus);
1448            fParenStack.push(flags, *fStatus);                            // Frame Marker
1449            fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus);   // The first NOP
1450            fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus);   // The second NOP
1451
1452            // Set the current mode flags to the new values.
1453            U_ASSERT(fNewModeFlags < 0);
1454            fModeFlags = fNewModeFlags;
1455        }
1456        break;
1457
1458    case doBadModeFlag:
1459        error(U_REGEX_INVALID_FLAG);
1460        break;
1461
1462    case doSuppressComments:
1463        // We have just scanned a '(?'.  We now need to prevent the character scanner from
1464        // treating a '#' as a to-the-end-of-line comment.
1465        //   (This Perl compatibility just gets uglier and uglier to do...)
1466        fEOLComments = FALSE;
1467        break;
1468
1469
1470    case doSetAddAmp:
1471        {
1472          UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1473          set->add(chAmp);
1474        }
1475        break;
1476
1477    case doSetAddDash:
1478        {
1479          UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1480          set->add(chDash);
1481        }
1482        break;
1483
1484     case doSetBackslash_s:
1485        {
1486         UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1487         set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]);
1488         break;
1489        }
1490
1491     case doSetBackslash_S:
1492        {
1493            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1494            UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]);
1495            SSet.complement();
1496            set->addAll(SSet);
1497            break;
1498        }
1499
1500    case doSetBackslash_d:
1501        {
1502            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1503            // TODO - make a static set, ticket 6058.
1504            addCategory(set, U_GC_ND_MASK, *fStatus);
1505            break;
1506        }
1507
1508    case doSetBackslash_D:
1509        {
1510            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1511            UnicodeSet digits;
1512            // TODO - make a static set, ticket 6058.
1513            digits.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus);
1514            digits.complement();
1515            set->addAll(digits);
1516            break;
1517        }
1518
1519    case doSetBackslash_w:
1520        {
1521            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1522            set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]);
1523            break;
1524        }
1525
1526    case doSetBackslash_W:
1527        {
1528            UnicodeSet *set = (UnicodeSet *)fSetStack.peek();
1529            UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]);
1530            SSet.complement();
1531            set->addAll(SSet);
1532            break;
1533        }
1534
1535    case doSetBegin:
1536        fSetStack.push(new UnicodeSet(), *fStatus);
1537        fSetOpStack.push(setStart, *fStatus);
1538        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
1539            fSetOpStack.push(setCaseClose, *fStatus);
1540        }
1541        break;
1542
1543    case doSetBeginDifference1:
1544        //  We have scanned something like [[abc]-[
1545        //  Set up a new UnicodeSet for the set beginning with the just-scanned '['
1546        //  Push a Difference operator, which will cause the new set to be subtracted from what
1547        //    went before once it is created.
1548        setPushOp(setDifference1);
1549        fSetOpStack.push(setStart, *fStatus);
1550        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
1551            fSetOpStack.push(setCaseClose, *fStatus);
1552        }
1553        break;
1554
1555    case doSetBeginIntersection1:
1556        //  We have scanned something like  [[abc]&[
1557        //   Need both the '&' operator and the open '[' operator.
1558        setPushOp(setIntersection1);
1559        fSetOpStack.push(setStart, *fStatus);
1560        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
1561            fSetOpStack.push(setCaseClose, *fStatus);
1562        }
1563        break;
1564
1565    case doSetBeginUnion:
1566        //  We have scanned something like  [[abc][
1567        //     Need to handle the union operation explicitly [[abc] | [
1568        setPushOp(setUnion);
1569        fSetOpStack.push(setStart, *fStatus);
1570        if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) {
1571            fSetOpStack.push(setCaseClose, *fStatus);
1572        }
1573        break;
1574
1575    case doSetDifference2:
1576        // We have scanned something like [abc--
1577        //   Consider this to unambiguously be a set difference operator.
1578        setPushOp(setDifference2);
1579        break;
1580
1581    case doSetEnd:
1582        // Have encountered the ']' that closes a set.
1583        //    Force the evaluation of any pending operations within this set,
1584        //    leave the completed set on the top of the set stack.
1585        setEval(setEnd);
1586        U_ASSERT(fSetOpStack.peeki()==setStart);
1587        fSetOpStack.popi();
1588        break;
1589
1590    case doSetFinish:
1591        {
1592        // Finished a complete set expression, including all nested sets.
1593        //   The close bracket has already triggered clearing out pending set operators,
1594        //    the operator stack should be empty and the operand stack should have just
1595        //    one entry, the result set.
1596        U_ASSERT(fSetOpStack.empty());
1597        UnicodeSet *theSet = (UnicodeSet *)fSetStack.pop();
1598        U_ASSERT(fSetStack.empty());
1599        compileSet(theSet);
1600        break;
1601        }
1602
1603    case doSetIntersection2:
1604        // Have scanned something like [abc&&
1605        setPushOp(setIntersection2);
1606        break;
1607
1608    case doSetLiteral:
1609        // Union the just-scanned literal character into the set being built.
1610        //    This operation is the highest precedence set operation, so we can always do
1611        //    it immediately, without waiting to see what follows.  It is necessary to perform
1612        //    any pending '-' or '&' operation first, because these have the same precedence
1613        //    as union-ing in a literal'
1614        {
1615            setEval(setUnion);
1616            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
1617            s->add(fC.fChar);
1618            fLastSetLiteral = fC.fChar;
1619            break;
1620        }
1621
1622    case doSetLiteralEscaped:
1623        // A back-slash escaped literal character was encountered.
1624        // Processing is the same as with setLiteral, above, with the addition of
1625        //  the optional check for errors on escaped ASCII letters.
1626        {
1627            if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 &&
1628                ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) ||     // in [A-Z]
1629                 (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) {   // in [a-z]
1630                error(U_REGEX_BAD_ESCAPE_SEQUENCE);
1631            }
1632            setEval(setUnion);
1633            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
1634            s->add(fC.fChar);
1635            fLastSetLiteral = fC.fChar;
1636            break;
1637        }
1638
1639        case doSetNamedChar:
1640        // Scanning a \N{UNICODE CHARACTER NAME}
1641        //  Aside from the source of the character, the processing is identical to doSetLiteral,
1642        //    above.
1643        {
1644            UChar32  c = scanNamedChar();
1645            setEval(setUnion);
1646            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
1647            s->add(c);
1648            fLastSetLiteral = c;
1649            break;
1650        }
1651
1652    case doSetNamedRange:
1653        // We have scanned literal-\N{CHAR NAME}.  Add the range to the set.
1654        // The left character is already in the set, and is saved in fLastSetLiteral.
1655        // The right side needs to be picked up, the scan is at the 'N'.
1656        // Lower Limit > Upper limit being an error matches both Java
1657        //        and ICU UnicodeSet behavior.
1658        {
1659            UChar32  c = scanNamedChar();
1660            if (U_SUCCESS(*fStatus) && fLastSetLiteral > c) {
1661                error(U_REGEX_INVALID_RANGE);
1662            }
1663            UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
1664            s->add(fLastSetLiteral, c);
1665            fLastSetLiteral = c;
1666            break;
1667        }
1668
1669
1670    case  doSetNegate:
1671        // Scanned a '^' at the start of a set.
1672        // Push the negation operator onto the set op stack.
1673        // A twist for case-insensitive matching:
1674        //   the case closure operation must happen _before_ negation.
1675        //   But the case closure operation will already be on the stack if it's required.
1676        //   This requires checking for case closure, and swapping the stack order
1677        //    if it is present.
1678        {
1679            int32_t  tosOp = fSetOpStack.peeki();
1680            if (tosOp == setCaseClose) {
1681                fSetOpStack.popi();
1682                fSetOpStack.push(setNegation, *fStatus);
1683                fSetOpStack.push(setCaseClose, *fStatus);
1684            } else {
1685                fSetOpStack.push(setNegation, *fStatus);
1686            }
1687        }
1688        break;
1689
1690    case doSetNoCloseError:
1691        error(U_REGEX_MISSING_CLOSE_BRACKET);
1692        break;
1693
1694    case doSetOpError:
1695        error(U_REGEX_RULE_SYNTAX);   //  -- or && at the end of a set.  Illegal.
1696        break;
1697
1698    case doSetPosixProp:
1699        {
1700            UnicodeSet *s = scanPosixProp();
1701            if (s != NULL) {
1702                UnicodeSet *tos = (UnicodeSet *)fSetStack.peek();
1703                tos->addAll(*s);
1704                delete s;
1705            }  // else error.  scanProp() reported the error status already.
1706        }
1707        break;
1708
1709    case doSetProp:
1710        //  Scanned a \p \P within [brackets].
1711        {
1712            UnicodeSet *s = scanProp();
1713            if (s != NULL) {
1714                UnicodeSet *tos = (UnicodeSet *)fSetStack.peek();
1715                tos->addAll(*s);
1716                delete s;
1717            }  // else error.  scanProp() reported the error status already.
1718        }
1719        break;
1720
1721
1722    case doSetRange:
1723        // We have scanned literal-literal.  Add the range to the set.
1724        // The left character is already in the set, and is saved in fLastSetLiteral.
1725        // The right side is the current character.
1726        // Lower Limit > Upper limit being an error matches both Java
1727        //        and ICU UnicodeSet behavior.
1728        {
1729        if (fLastSetLiteral > fC.fChar) {
1730            error(U_REGEX_INVALID_RANGE);
1731        }
1732        UnicodeSet *s = (UnicodeSet *)fSetStack.peek();
1733        s->add(fLastSetLiteral, fC.fChar);
1734        break;
1735        }
1736
1737
1738    default:
1739        U_ASSERT(FALSE);
1740        error(U_REGEX_INTERNAL_ERROR);
1741        break;
1742    }
1743
1744    if (U_FAILURE(*fStatus)) {
1745        returnVal = FALSE;
1746    }
1747
1748    return returnVal;
1749}
1750
1751
1752
1753//------------------------------------------------------------------------------
1754//
1755//   literalChar           We've encountered a literal character from the pattern,
1756//                             or an escape sequence that reduces to a character.
1757//                         Add it to the string containing all literal chars/strings from
1758//                             the pattern.
1759//                         If we are in a pattern string already, add the new char to it.
1760//                         If we aren't in a pattern string, begin one now.
1761//
1762//------------------------------------------------------------------------------
1763void RegexCompile::literalChar(UChar32 c)  {
1764    int32_t           op;            // An operation in the compiled pattern.
1765    int32_t           opType;
1766    int32_t           patternLoc;   // A position in the compiled pattern.
1767    int32_t           stringLen;
1768
1769
1770    // If the last thing compiled into the pattern was not a literal char,
1771    //   force this new literal char to begin a new string, and not append to the previous.
1772    op     = (int32_t)fRXPat->fCompiledPat->lastElementi();
1773    opType = URX_TYPE(op);
1774    if (!(opType == URX_STRING_LEN || opType == URX_ONECHAR || opType == URX_ONECHAR_I)) {
1775        fixLiterals();
1776    }
1777
1778    if (fStringOpStart == -1) {
1779        // First char of a string in the pattern.
1780        // Emit a OneChar op into the compiled pattern.
1781        emitONE_CHAR(c);
1782
1783        // Mark that we might actually be starting a string here
1784        fStringOpStart = fRXPat->fLiteralText.length();
1785        return;
1786    }
1787
1788    op     = (int32_t)fRXPat->fCompiledPat->lastElementi();
1789    opType = URX_TYPE(op);
1790    U_ASSERT(opType == URX_ONECHAR || opType == URX_ONECHAR_I || opType == URX_STRING_LEN);
1791
1792    // If the most recently emitted op is a URX_ONECHAR,
1793    if (opType == URX_ONECHAR || opType == URX_ONECHAR_I) {
1794        if (U16_IS_TRAIL(c) && U16_IS_LEAD(URX_VAL(op))) {
1795            // The most recently emitted op is a ONECHAR that was the first half
1796            //   of a surrogate pair.  Update the ONECHAR's operand to be the
1797            //   supplementary code point resulting from both halves of the pair.
1798            c = U16_GET_SUPPLEMENTARY(URX_VAL(op), c);
1799            op = URX_BUILD(opType, c);
1800            patternLoc = fRXPat->fCompiledPat->size() - 1;
1801            fRXPat->fCompiledPat->setElementAt(op, patternLoc);
1802            return;
1803        }
1804
1805        // The most recently emitted op is a ONECHAR.
1806        //  We've now received another adjacent char.  Change the ONECHAR op
1807        //   to a string op.
1808        fRXPat->fLiteralText.append(URX_VAL(op));
1809
1810        if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
1811            op     = URX_BUILD(URX_STRING_I, fStringOpStart);
1812        } else {
1813            op     = URX_BUILD(URX_STRING, fStringOpStart);
1814        }
1815        patternLoc = fRXPat->fCompiledPat->size() - 1;
1816        fRXPat->fCompiledPat->setElementAt(op, patternLoc);
1817        op         = URX_BUILD(URX_STRING_LEN, 0);
1818        fRXPat->fCompiledPat->addElement(op, *fStatus);
1819    }
1820
1821    // We are adding onto an existing string
1822    fRXPat->fLiteralText.append(c);
1823
1824    // The pattern contains a URX_SRING / URX_STRING_LEN.  Update the
1825    //  string length to reflect the new char we just added to the string.
1826    stringLen  = fRXPat->fLiteralText.length() - fStringOpStart;
1827    op         = URX_BUILD(URX_STRING_LEN, stringLen);
1828    patternLoc = fRXPat->fCompiledPat->size() - 1;
1829    fRXPat->fCompiledPat->setElementAt(op, patternLoc);
1830}
1831
1832
1833
1834//------------------------------------------------------------------------------
1835//
1836//    emitONE_CHAR         emit a ONE_CHAR op into the generated code.
1837//                         Choose cased or uncased version, depending on the
1838//                         match mode and whether the character itself is cased.
1839//
1840//------------------------------------------------------------------------------
1841void RegexCompile::emitONE_CHAR(UChar32  c) {
1842    int32_t op;
1843    if ((fModeFlags & UREGEX_CASE_INSENSITIVE) &&
1844        u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) {
1845        // We have a cased character, and are in case insensitive matching mode.
1846        //c  = u_foldCase(c, U_FOLD_CASE_DEFAULT);  // !!!: handled in stripNOPs() now
1847        op = URX_BUILD(URX_ONECHAR_I, c);
1848    } else {
1849        // Uncased char, or case sensitive match mode.
1850        //  Either way, just generate a literal compare of the char.
1851        op = URX_BUILD(URX_ONECHAR, c);
1852    }
1853    fRXPat->fCompiledPat->addElement(op, *fStatus);
1854}
1855
1856
1857//------------------------------------------------------------------------------
1858//
1859//    fixLiterals           When compiling something that can follow a literal
1860//                          string in a pattern, we need to "fix" any preceding
1861//                          string, which will cause any subsequent literals to
1862//                          begin a new string, rather than appending to the
1863//                          old one.
1864//
1865//                          Optionally, split the last char of the string off into
1866//                          a single "ONE_CHAR" operation, so that quantifiers can
1867//                          apply to that char alone.  Example:   abc*
1868//                          The * must apply to the 'c' only.
1869//
1870//------------------------------------------------------------------------------
1871void    RegexCompile::fixLiterals(UBool split) {
1872    int32_t  stringStart = fStringOpStart;    // start index of the current literal string
1873    int32_t  op;                              // An op from/for the compiled pattern.
1874    int32_t  opType;                          // An opcode type from the compiled pattern.
1875    int32_t  stringLastCharIdx;
1876    UChar32  lastChar;
1877    int32_t  stringNextToLastCharIdx;
1878    UChar32  nextToLastChar;
1879    int32_t  stringLen;
1880
1881    fStringOpStart = -1;
1882    if (!split) {
1883        return;
1884    }
1885
1886    // Split:  We need to  ensure that the last item in the compiled pattern does
1887    //   not refer to a literal string of more than one char.  If it does,
1888    //   separate the last char from the rest of the string.
1889
1890    // If the last operation from the compiled pattern is not a string,
1891    //   nothing needs to be done
1892    op     = (int32_t)fRXPat->fCompiledPat->lastElementi();
1893    opType = URX_TYPE(op);
1894    if (opType != URX_STRING_LEN) {
1895        return;
1896    }
1897    stringLen = URX_VAL(op);
1898
1899    //
1900    // Find the position of the last code point in the string  (might be a surrogate pair)
1901    //
1902    stringLastCharIdx = fRXPat->fLiteralText.length();
1903    stringLastCharIdx = fRXPat->fLiteralText.moveIndex32(stringLastCharIdx, -1);
1904    lastChar          = fRXPat->fLiteralText.char32At(stringLastCharIdx);
1905
1906    // The string should always be at least two code points long, meaning that there
1907    //   should be something before the last char position that we just found.
1908    U_ASSERT(stringLastCharIdx > stringStart);
1909    stringNextToLastCharIdx = fRXPat->fLiteralText.moveIndex32(stringLastCharIdx, -1);
1910    U_ASSERT(stringNextToLastCharIdx >= stringStart);
1911    nextToLastChar          = fRXPat->fLiteralText.char32At(stringNextToLastCharIdx);
1912
1913    if (stringNextToLastCharIdx > stringStart) {
1914        // The length of string remaining after removing one char is two or more.
1915        // Leave the string in the compiled pattern, shorten it by one char,
1916        //   and append a URX_ONECHAR op for the last char.
1917        stringLen -= (fRXPat->fLiteralText.length() - stringLastCharIdx);
1918        op = URX_BUILD(URX_STRING_LEN, stringLen);
1919        fRXPat->fCompiledPat->setElementAt(op, fRXPat->fCompiledPat->size() -1);
1920        emitONE_CHAR(lastChar);
1921    } else {
1922        // The original string consisted of exactly two characters.  Replace
1923        // the existing compiled URX_STRING/URX_STRING_LEN ops with a pair
1924        // of URX_ONECHARs.
1925        fRXPat->fCompiledPat->setSize(fRXPat->fCompiledPat->size() -2);
1926        emitONE_CHAR(nextToLastChar);
1927        emitONE_CHAR(lastChar);
1928    }
1929}
1930
1931
1932
1933
1934
1935
1936//------------------------------------------------------------------------------
1937//
1938//   insertOp()             Insert a slot for a new opcode into the already
1939//                          compiled pattern code.
1940//
1941//                          Fill the slot with a NOP.  Our caller will replace it
1942//                          with what they really wanted.
1943//
1944//------------------------------------------------------------------------------
1945void   RegexCompile::insertOp(int32_t where) {
1946    UVector64 *code = fRXPat->fCompiledPat;
1947    U_ASSERT(where>0 && where < code->size());
1948
1949    int32_t  nop = URX_BUILD(URX_NOP, 0);
1950    code->insertElementAt(nop, where, *fStatus);
1951
1952    // Walk through the pattern, looking for any ops with targets that
1953    //  were moved down by the insert.  Fix them.
1954    int32_t loc;
1955    for (loc=0; loc<code->size(); loc++) {
1956        int32_t op = (int32_t)code->elementAti(loc);
1957        int32_t opType = URX_TYPE(op);
1958        int32_t opValue = URX_VAL(op);
1959        if ((opType == URX_JMP         ||
1960            opType == URX_JMPX         ||
1961            opType == URX_STATE_SAVE   ||
1962            opType == URX_CTR_LOOP     ||
1963            opType == URX_CTR_LOOP_NG  ||
1964            opType == URX_JMP_SAV      ||
1965            opType == URX_JMP_SAV_X    ||
1966            opType == URX_RELOC_OPRND)    && opValue > where) {
1967            // Target location for this opcode is after the insertion point and
1968            //   needs to be incremented to adjust for the insertion.
1969            opValue++;
1970            op = URX_BUILD(opType, opValue);
1971            code->setElementAt(op, loc);
1972        }
1973    }
1974
1975    // Now fix up the parentheses stack.  All positive values in it are locations in
1976    //  the compiled pattern.   (Negative values are frame boundaries, and don't need fixing.)
1977    for (loc=0; loc<fParenStack.size(); loc++) {
1978        int32_t x = fParenStack.elementAti(loc);
1979        U_ASSERT(x < code->size());
1980        if (x>where) {
1981            x++;
1982            fParenStack.setElementAt(x, loc);
1983        }
1984    }
1985
1986    if (fMatchCloseParen > where) {
1987        fMatchCloseParen++;
1988    }
1989    if (fMatchOpenParen > where) {
1990        fMatchOpenParen++;
1991    }
1992}
1993
1994
1995
1996//------------------------------------------------------------------------------
1997//
1998//   blockTopLoc()          Find or create a location in the compiled pattern
1999//                          at the start of the operation or block that has
2000//                          just been compiled.  Needed when a quantifier (* or
2001//                          whatever) appears, and we need to add an operation
2002//                          at the start of the thing being quantified.
2003//
2004//                          (Parenthesized Blocks) have a slot with a NOP that
2005//                          is reserved for this purpose.  .* or similar don't
2006//                          and a slot needs to be added.
2007//
2008//       parameter reserveLoc   :  TRUE -  ensure that there is space to add an opcode
2009//                                         at the returned location.
2010//                                 FALSE - just return the address,
2011//                                         do not reserve a location there.
2012//
2013//------------------------------------------------------------------------------
2014int32_t   RegexCompile::blockTopLoc(UBool reserveLoc) {
2015    int32_t   theLoc;
2016    if (fRXPat->fCompiledPat->size() == fMatchCloseParen)
2017    {
2018        // The item just processed is a parenthesized block.
2019        theLoc = fMatchOpenParen;   // A slot is already reserved for us.
2020        U_ASSERT(theLoc > 0);
2021        U_ASSERT(URX_TYPE(((uint32_t)fRXPat->fCompiledPat->elementAti(theLoc))) == URX_NOP);
2022    }
2023    else {
2024        // Item just compiled is a single thing, a ".", or a single char, or a set reference.
2025        // No slot for STATE_SAVE was pre-reserved in the compiled code.
2026        // We need to make space now.
2027        fixLiterals(TRUE);  // If last item was a string, separate the last char.
2028        theLoc = fRXPat->fCompiledPat->size()-1;
2029        if (reserveLoc) {
2030            /*int32_t opAtTheLoc = fRXPat->fCompiledPat->elementAti(theLoc);*/
2031            int32_t  nop = URX_BUILD(URX_NOP, 0);
2032            fRXPat->fCompiledPat->insertElementAt(nop, theLoc, *fStatus);
2033        }
2034    }
2035    return theLoc;
2036}
2037
2038
2039
2040//------------------------------------------------------------------------------
2041//
2042//    handleCloseParen      When compiling a close paren, we need to go back
2043//                          and fix up any JMP or SAVE operations within the
2044//                          parenthesized block that need to target the end
2045//                          of the block.  The locations of these are kept on
2046//                          the paretheses stack.
2047//
2048//                          This function is called both when encountering a
2049//                          real ) and at the end of the pattern.
2050//
2051//------------------------------------------------------------------------------
2052void  RegexCompile::handleCloseParen() {
2053    int32_t   patIdx;
2054    int32_t   patOp;
2055    if (fParenStack.size() <= 0) {
2056        error(U_REGEX_MISMATCHED_PAREN);
2057        return;
2058    }
2059
2060    // Force any literal chars that may follow the close paren to start a new string,
2061    //   and not attach to any preceding it.
2062    fixLiterals(FALSE);
2063
2064    // Fixup any operations within the just-closed parenthesized group
2065    //    that need to reference the end of the (block).
2066    //    (The first one popped from the stack is an unused slot for
2067    //     alternation (OR) state save, but applying the fixup to it does no harm.)
2068    for (;;) {
2069        patIdx = fParenStack.popi();
2070        if (patIdx < 0) {
2071            // value < 0 flags the start of the frame on the paren stack.
2072            break;
2073        }
2074        U_ASSERT(patIdx>0 && patIdx <= fRXPat->fCompiledPat->size());
2075        patOp = (int32_t)fRXPat->fCompiledPat->elementAti(patIdx);
2076        U_ASSERT(URX_VAL(patOp) == 0);          // Branch target for JMP should not be set.
2077        patOp |= fRXPat->fCompiledPat->size();  // Set it now.
2078        fRXPat->fCompiledPat->setElementAt(patOp, patIdx);
2079        fMatchOpenParen     = patIdx;
2080    }
2081
2082    //  At the close of any parenthesized block, restore the match mode flags  to
2083    //  the value they had at the open paren.  Saved value is
2084    //  at the top of the paren stack.
2085    fModeFlags = fParenStack.popi();
2086    U_ASSERT(fModeFlags < 0);
2087
2088    // DO any additional fixups, depending on the specific kind of
2089    // parentesized grouping this is
2090
2091    switch (patIdx) {
2092    case plain:
2093    case flags:
2094        // No additional fixups required.
2095        //   (Grouping-only parentheses)
2096        break;
2097    case capturing:
2098        // Capturing Parentheses.
2099        //   Insert a End Capture op into the pattern.
2100        //   The frame offset of the variables for this cg is obtained from the
2101        //       start capture op and put it into the end-capture op.
2102        {
2103            int32_t   captureOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1);
2104            U_ASSERT(URX_TYPE(captureOp) == URX_START_CAPTURE);
2105
2106            int32_t   frameVarLocation = URX_VAL(captureOp);
2107            int32_t   endCaptureOp = URX_BUILD(URX_END_CAPTURE, frameVarLocation);
2108            fRXPat->fCompiledPat->addElement(endCaptureOp, *fStatus);
2109        }
2110        break;
2111    case atomic:
2112        // Atomic Parenthesis.
2113        //   Insert a LD_SP operation to restore the state stack to the position
2114        //   it was when the atomic parens were entered.
2115        {
2116            int32_t   stoOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1);
2117            U_ASSERT(URX_TYPE(stoOp) == URX_STO_SP);
2118            int32_t   stoLoc = URX_VAL(stoOp);
2119            int32_t   ldOp   = URX_BUILD(URX_LD_SP, stoLoc);
2120            fRXPat->fCompiledPat->addElement(ldOp, *fStatus);
2121        }
2122        break;
2123
2124    case lookAhead:
2125        {
2126            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5);
2127            U_ASSERT(URX_TYPE(startOp) == URX_LA_START);
2128            int32_t dataLoc  = URX_VAL(startOp);
2129            int32_t op       = URX_BUILD(URX_LA_END, dataLoc);
2130            fRXPat->fCompiledPat->addElement(op, *fStatus);
2131        }
2132        break;
2133
2134    case negLookAhead:
2135        {
2136            // See comment at doOpenLookAheadNeg
2137            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-1);
2138            U_ASSERT(URX_TYPE(startOp) == URX_LA_START);
2139            int32_t dataLoc  = URX_VAL(startOp);
2140            int32_t op       = URX_BUILD(URX_LA_END, dataLoc);
2141            fRXPat->fCompiledPat->addElement(op, *fStatus);
2142            op               = URX_BUILD(URX_BACKTRACK, 0);
2143            fRXPat->fCompiledPat->addElement(op, *fStatus);
2144            op               = URX_BUILD(URX_LA_END, dataLoc);
2145            fRXPat->fCompiledPat->addElement(op, *fStatus);
2146
2147            // Patch the URX_SAVE near the top of the block.
2148            // The destination of the SAVE is the final LA_END that was just added.
2149            int32_t saveOp   = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen);
2150            U_ASSERT(URX_TYPE(saveOp) == URX_STATE_SAVE);
2151            int32_t dest     = fRXPat->fCompiledPat->size()-1;
2152            saveOp           = URX_BUILD(URX_STATE_SAVE, dest);
2153            fRXPat->fCompiledPat->setElementAt(saveOp, fMatchOpenParen);
2154        }
2155        break;
2156
2157    case lookBehind:
2158        {
2159            // See comment at doOpenLookBehind.
2160
2161            // Append the URX_LB_END and URX_LA_END to the compiled pattern.
2162            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-4);
2163            U_ASSERT(URX_TYPE(startOp) == URX_LB_START);
2164            int32_t dataLoc  = URX_VAL(startOp);
2165            int32_t op       = URX_BUILD(URX_LB_END, dataLoc);
2166            fRXPat->fCompiledPat->addElement(op, *fStatus);
2167                    op       = URX_BUILD(URX_LA_END, dataLoc);
2168            fRXPat->fCompiledPat->addElement(op, *fStatus);
2169
2170            // Determine the min and max bounds for the length of the
2171            //  string that the pattern can match.
2172            //  An unbounded upper limit is an error.
2173            int32_t patEnd   = fRXPat->fCompiledPat->size() - 1;
2174            int32_t minML    = minMatchLength(fMatchOpenParen, patEnd);
2175            int32_t maxML    = maxMatchLength(fMatchOpenParen, patEnd);
2176            if (maxML == INT32_MAX) {
2177                error(U_REGEX_LOOK_BEHIND_LIMIT);
2178                break;
2179            }
2180            U_ASSERT(minML <= maxML);
2181
2182            // Insert the min and max match len bounds into the URX_LB_CONT op that
2183            //  appears at the top of the look-behind block, at location fMatchOpenParen+1
2184            fRXPat->fCompiledPat->setElementAt(minML,  fMatchOpenParen-2);
2185            fRXPat->fCompiledPat->setElementAt(maxML,  fMatchOpenParen-1);
2186
2187        }
2188        break;
2189
2190
2191
2192    case lookBehindN:
2193        {
2194            // See comment at doOpenLookBehindNeg.
2195
2196            // Append the URX_LBN_END to the compiled pattern.
2197            int32_t  startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5);
2198            U_ASSERT(URX_TYPE(startOp) == URX_LB_START);
2199            int32_t dataLoc  = URX_VAL(startOp);
2200            int32_t op       = URX_BUILD(URX_LBN_END, dataLoc);
2201            fRXPat->fCompiledPat->addElement(op, *fStatus);
2202
2203            // Determine the min and max bounds for the length of the
2204            //  string that the pattern can match.
2205            //  An unbounded upper limit is an error.
2206            int32_t patEnd   = fRXPat->fCompiledPat->size() - 1;
2207            int32_t minML    = minMatchLength(fMatchOpenParen, patEnd);
2208            int32_t maxML    = maxMatchLength(fMatchOpenParen, patEnd);
2209            if (maxML == INT32_MAX) {
2210                error(U_REGEX_LOOK_BEHIND_LIMIT);
2211                break;
2212            }
2213            U_ASSERT(minML <= maxML);
2214
2215            // Insert the min and max match len bounds into the URX_LB_CONT op that
2216            //  appears at the top of the look-behind block, at location fMatchOpenParen+1
2217            fRXPat->fCompiledPat->setElementAt(minML,  fMatchOpenParen-3);
2218            fRXPat->fCompiledPat->setElementAt(maxML,  fMatchOpenParen-2);
2219
2220            // Insert the pattern location to continue at after a successful match
2221            //  as the last operand of the URX_LBN_CONT
2222            op = URX_BUILD(URX_RELOC_OPRND, fRXPat->fCompiledPat->size());
2223            fRXPat->fCompiledPat->setElementAt(op,  fMatchOpenParen-1);
2224        }
2225        break;
2226
2227
2228
2229    default:
2230        U_ASSERT(FALSE);
2231    }
2232
2233    // remember the next location in the compiled pattern.
2234    // The compilation of Quantifiers will look at this to see whether its looping
2235    //   over a parenthesized block or a single item
2236    fMatchCloseParen = fRXPat->fCompiledPat->size();
2237}
2238
2239
2240
2241//------------------------------------------------------------------------------
2242//
2243//   compileSet       Compile the pattern operations for a reference to a
2244//                    UnicodeSet.
2245//
2246//------------------------------------------------------------------------------
2247void        RegexCompile::compileSet(UnicodeSet *theSet)
2248{
2249    if (theSet == NULL) {
2250        return;
2251    }
2252    //  Remove any strings from the set.
2253    //  There shoudn't be any, but just in case.
2254    //     (Case Closure can add them; if we had a simple case closure avaialble that
2255    //      ignored strings, that would be better.)
2256    theSet->removeAllStrings();
2257    int32_t  setSize = theSet->size();
2258
2259    switch (setSize) {
2260    case 0:
2261        {
2262            // Set of no elements.   Always fails to match.
2263            fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKTRACK, 0), *fStatus);
2264            delete theSet;
2265        }
2266        break;
2267
2268    case 1:
2269        {
2270            // The set contains only a single code point.  Put it into
2271            //   the compiled pattern as a single char operation rather
2272            //   than a set, and discard the set itself.
2273            literalChar(theSet->charAt(0));
2274            delete theSet;
2275        }
2276        break;
2277
2278    default:
2279        {
2280            //  The set contains two or more chars.  (the normal case)
2281            //  Put it into the compiled pattern as a set.
2282            int32_t setNumber = fRXPat->fSets->size();
2283            fRXPat->fSets->addElement(theSet, *fStatus);
2284            int32_t setOp = URX_BUILD(URX_SETREF, setNumber);
2285            fRXPat->fCompiledPat->addElement(setOp, *fStatus);
2286        }
2287    }
2288}
2289
2290
2291//------------------------------------------------------------------------------
2292//
2293//   compileInterval    Generate the code for a {min, max} style interval quantifier.
2294//                      Except for the specific opcodes used, the code is the same
2295//                      for all three types (greedy, non-greedy, possessive) of
2296//                      intervals.  The opcodes are supplied as parameters.
2297//
2298//                      The code for interval loops has this form:
2299//                         0  CTR_INIT   counter loc (in stack frame)
2300//                         1             5  patt address of CTR_LOOP at bottom of block
2301//                         2             min count
2302//                         3             max count   (-1 for unbounded)
2303//                         4  ...        block to be iterated over
2304//                         5  CTR_LOOP
2305//
2306//                       In
2307//------------------------------------------------------------------------------
2308void        RegexCompile::compileInterval(int32_t InitOp,  int32_t LoopOp)
2309{
2310    // The CTR_INIT op at the top of the block with the {n,m} quantifier takes
2311    //   four slots in the compiled code.  Reserve them.
2312    int32_t   topOfBlock = blockTopLoc(TRUE);
2313    insertOp(topOfBlock);
2314    insertOp(topOfBlock);
2315    insertOp(topOfBlock);
2316
2317    // The operands for the CTR_INIT opcode include the index in the matcher data
2318    //   of the counter.  Allocate it now.
2319    int32_t   counterLoc = fRXPat->fFrameSize;
2320    fRXPat->fFrameSize++;
2321
2322    int32_t   op = URX_BUILD(InitOp, counterLoc);
2323    fRXPat->fCompiledPat->setElementAt(op, topOfBlock);
2324
2325    // The second operand of CTR_INIT is the location following the end of the loop.
2326    //   Must put in as a URX_RELOC_OPRND so that the value will be adjusted if the
2327    //   compilation of something later on causes the code to grow and the target
2328    //   position to move.
2329    int32_t loopEnd = fRXPat->fCompiledPat->size();
2330    op = URX_BUILD(URX_RELOC_OPRND, loopEnd);
2331    fRXPat->fCompiledPat->setElementAt(op, topOfBlock+1);
2332
2333    // Followed by the min and max counts.
2334    fRXPat->fCompiledPat->setElementAt(fIntervalLow, topOfBlock+2);
2335    fRXPat->fCompiledPat->setElementAt(fIntervalUpper, topOfBlock+3);
2336
2337    // Apend the CTR_LOOP op.  The operand is the location of the CTR_INIT op.
2338    //   Goes at end of the block being looped over, so just append to the code so far.
2339    op = URX_BUILD(LoopOp, topOfBlock);
2340    fRXPat->fCompiledPat->addElement(op, *fStatus);
2341
2342    if ((fIntervalLow & 0xff000000) != 0 ||
2343        (fIntervalUpper > 0 && (fIntervalUpper & 0xff000000) != 0)) {
2344            error(U_REGEX_NUMBER_TOO_BIG);
2345        }
2346
2347    if (fIntervalLow > fIntervalUpper && fIntervalUpper != -1) {
2348        error(U_REGEX_MAX_LT_MIN);
2349    }
2350}
2351
2352
2353
2354UBool RegexCompile::compileInlineInterval() {
2355    if (fIntervalUpper > 10 || fIntervalUpper < fIntervalLow) {
2356        // Too big to inline.  Fail, which will cause looping code to be generated.
2357        //   (Upper < Lower picks up unbounded upper and errors, both.)
2358        return FALSE;
2359    }
2360
2361    int32_t   topOfBlock = blockTopLoc(FALSE);
2362    if (fIntervalUpper == 0) {
2363        // Pathological case.  Attempt no matches, as if the block doesn't exist.
2364        fRXPat->fCompiledPat->setSize(topOfBlock);
2365        return TRUE;
2366    }
2367
2368    if (topOfBlock != fRXPat->fCompiledPat->size()-1 && fIntervalUpper != 1) {
2369        // The thing being repeated is not a single op, but some
2370        //   more complex block.  Do it as a loop, not inlines.
2371        //   Note that things "repeated" a max of once are handled as inline, because
2372        //     the one copy of the code already generated is just fine.
2373        return FALSE;
2374    }
2375
2376    // Pick up the opcode that is to be repeated
2377    //
2378    int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(topOfBlock);
2379
2380    // Compute the pattern location where the inline sequence
2381    //   will end, and set up the state save op that will be needed.
2382    //
2383    int32_t endOfSequenceLoc = fRXPat->fCompiledPat->size()-1
2384                                + fIntervalUpper + (fIntervalUpper-fIntervalLow);
2385    int32_t saveOp = URX_BUILD(URX_STATE_SAVE, endOfSequenceLoc);
2386    if (fIntervalLow == 0) {
2387        insertOp(topOfBlock);
2388        fRXPat->fCompiledPat->setElementAt(saveOp, topOfBlock);
2389    }
2390
2391
2392
2393    //  Loop, emitting the op for the thing being repeated each time.
2394    //    Loop starts at 1 because one instance of the op already exists in the pattern,
2395    //    it was put there when it was originally encountered.
2396    int32_t i;
2397    for (i=1; i<fIntervalUpper; i++ ) {
2398        if (i == fIntervalLow) {
2399            fRXPat->fCompiledPat->addElement(saveOp, *fStatus);
2400        }
2401        if (i > fIntervalLow) {
2402            fRXPat->fCompiledPat->addElement(saveOp, *fStatus);
2403        }
2404        fRXPat->fCompiledPat->addElement(op, *fStatus);
2405    }
2406    return TRUE;
2407}
2408
2409
2410
2411//------------------------------------------------------------------------------
2412//
2413//   matchStartType    Determine how a match can start.
2414//                     Used to optimize find() operations.
2415//
2416//                     Operation is very similar to minMatchLength().  Walk the compiled
2417//                     pattern, keeping an on-going minimum-match-length.  For any
2418//                     op where the min match coming in is zero, add that ops possible
2419//                     starting matches to the possible starts for the overall pattern.
2420//
2421//------------------------------------------------------------------------------
2422void   RegexCompile::matchStartType() {
2423    if (U_FAILURE(*fStatus)) {
2424        return;
2425    }
2426
2427
2428    int32_t    loc;                    // Location in the pattern of the current op being processed.
2429    int32_t    op;                     // The op being processed
2430    int32_t    opType;                 // The opcode type of the op
2431    int32_t    currentLen = 0;         // Minimum length of a match to this point (loc) in the pattern
2432    int32_t    numInitialStrings = 0;  // Number of strings encountered that could match at start.
2433
2434    UBool      atStart = TRUE;         // True if no part of the pattern yet encountered
2435                                       //   could have advanced the position in a match.
2436                                       //   (Maximum match length so far == 0)
2437
2438    // forwardedLength is a vector holding minimum-match-length values that
2439    //   are propagated forward in the pattern by JMP or STATE_SAVE operations.
2440    //   It must be one longer than the pattern being checked because some  ops
2441    //   will jmp to a end-of-block+1 location from within a block, and we must
2442    //   count those when checking the block.
2443    int32_t end = fRXPat->fCompiledPat->size();
2444    UVector32  forwardedLength(end+1, *fStatus);
2445    forwardedLength.setSize(end+1);
2446    for (loc=3; loc<end; loc++) {
2447        forwardedLength.setElementAt(INT32_MAX, loc);
2448    }
2449
2450    for (loc = 3; loc<end; loc++) {
2451        op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
2452        opType = URX_TYPE(op);
2453
2454        // The loop is advancing linearly through the pattern.
2455        // If the op we are now at was the destination of a branch in the pattern,
2456        // and that path has a shorter minimum length than the current accumulated value,
2457        // replace the current accumulated value.
2458        if (forwardedLength.elementAti(loc) < currentLen) {
2459            currentLen = forwardedLength.elementAti(loc);
2460            U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);
2461        }
2462
2463        switch (opType) {
2464            // Ops that don't change the total length matched
2465        case URX_RESERVED_OP:
2466        case URX_END:
2467        case URX_FAIL:
2468        case URX_STRING_LEN:
2469        case URX_NOP:
2470        case URX_START_CAPTURE:
2471        case URX_END_CAPTURE:
2472        case URX_BACKSLASH_B:
2473        case URX_BACKSLASH_BU:
2474        case URX_BACKSLASH_G:
2475        case URX_BACKSLASH_Z:
2476        case URX_DOLLAR:
2477        case URX_DOLLAR_M:
2478        case URX_DOLLAR_D:
2479        case URX_DOLLAR_MD:
2480        case URX_RELOC_OPRND:
2481        case URX_STO_INP_LOC:
2482        case URX_BACKREF:         // BackRef.  Must assume that it might be a zero length match
2483        case URX_BACKREF_I:
2484
2485        case URX_STO_SP:          // Setup for atomic or possessive blocks.  Doesn't change what can match.
2486        case URX_LD_SP:
2487            break;
2488
2489        case URX_CARET:
2490            if (atStart) {
2491                fRXPat->fStartType = START_START;
2492            }
2493            break;
2494
2495        case URX_CARET_M:
2496        case URX_CARET_M_UNIX:
2497            if (atStart) {
2498                fRXPat->fStartType = START_LINE;
2499            }
2500            break;
2501
2502        case URX_ONECHAR:
2503            if (currentLen == 0) {
2504                // This character could appear at the start of a match.
2505                //   Add it to the set of possible starting characters.
2506                fRXPat->fInitialChars->add(URX_VAL(op));
2507                numInitialStrings += 2;
2508            }
2509            currentLen++;
2510            atStart = FALSE;
2511            break;
2512
2513
2514        case URX_SETREF:
2515            if (currentLen == 0) {
2516                int32_t  sn = URX_VAL(op);
2517                U_ASSERT(sn > 0 && sn < fRXPat->fSets->size());
2518                const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn);
2519                fRXPat->fInitialChars->addAll(*s);
2520                numInitialStrings += 2;
2521            }
2522            currentLen++;
2523            atStart = FALSE;
2524            break;
2525
2526        case URX_LOOP_SR_I:
2527            // [Set]*, like a SETREF, above, in what it can match,
2528            //  but may not match at all, so currentLen is not incremented.
2529            if (currentLen == 0) {
2530                int32_t  sn = URX_VAL(op);
2531                U_ASSERT(sn > 0 && sn < fRXPat->fSets->size());
2532                const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn);
2533                fRXPat->fInitialChars->addAll(*s);
2534                numInitialStrings += 2;
2535            }
2536            atStart = FALSE;
2537            break;
2538
2539        case URX_LOOP_DOT_I:
2540            if (currentLen == 0) {
2541                // .* at the start of a pattern.
2542                //    Any character can begin the match.
2543                fRXPat->fInitialChars->clear();
2544                fRXPat->fInitialChars->complement();
2545                numInitialStrings += 2;
2546            }
2547            atStart = FALSE;
2548            break;
2549
2550
2551        case URX_STATIC_SETREF:
2552            if (currentLen == 0) {
2553                int32_t  sn = URX_VAL(op);
2554                U_ASSERT(sn>0 && sn<URX_LAST_SET);
2555                const UnicodeSet *s = fRXPat->fStaticSets[sn];
2556                fRXPat->fInitialChars->addAll(*s);
2557                numInitialStrings += 2;
2558            }
2559            currentLen++;
2560            atStart = FALSE;
2561            break;
2562
2563
2564
2565        case URX_STAT_SETREF_N:
2566            if (currentLen == 0) {
2567                int32_t  sn = URX_VAL(op);
2568                const UnicodeSet *s = fRXPat->fStaticSets[sn];
2569                UnicodeSet sc(*s);
2570                sc.complement();
2571                fRXPat->fInitialChars->addAll(sc);
2572                numInitialStrings += 2;
2573            }
2574            currentLen++;
2575            atStart = FALSE;
2576            break;
2577
2578
2579
2580        case URX_BACKSLASH_D:
2581            // Digit Char
2582             if (currentLen == 0) {
2583                 UnicodeSet s;
2584                 s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus);
2585                 if (URX_VAL(op) != 0) {
2586                     s.complement();
2587                 }
2588                 fRXPat->fInitialChars->addAll(s);
2589                 numInitialStrings += 2;
2590            }
2591            currentLen++;
2592            atStart = FALSE;
2593            break;
2594
2595
2596        case URX_ONECHAR_I:
2597            // Case Insensitive Single Character.
2598            if (currentLen == 0) {
2599                UChar32  c = URX_VAL(op);
2600                if (u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) {
2601                    // character may have distinct cased forms.  Add all of them
2602                    //   to the set of possible starting match chars.
2603                    UnicodeSet s(c, c);
2604                    s.closeOver(USET_CASE_INSENSITIVE);
2605                    fRXPat->fInitialChars->addAll(s);
2606                } else {
2607                    // Char has no case variants.  Just add it as-is to the
2608                    //   set of possible starting chars.
2609                    fRXPat->fInitialChars->add(c);
2610                }
2611                numInitialStrings += 2;
2612            }
2613            currentLen++;
2614            atStart = FALSE;
2615            break;
2616
2617
2618        case URX_BACKSLASH_X:   // Grahpeme Cluster.  Minimum is 1, max unbounded.
2619        case URX_DOTANY_ALL:    // . matches one or two.
2620        case URX_DOTANY:
2621        case URX_DOTANY_UNIX:
2622            if (currentLen == 0) {
2623                // These constructs are all bad news when they appear at the start
2624                //   of a match.  Any character can begin the match.
2625                fRXPat->fInitialChars->clear();
2626                fRXPat->fInitialChars->complement();
2627                numInitialStrings += 2;
2628            }
2629            currentLen++;
2630            atStart = FALSE;
2631            break;
2632
2633
2634        case URX_JMPX:
2635            loc++;             // Except for extra operand on URX_JMPX, same as URX_JMP.
2636        case URX_JMP:
2637            {
2638                int32_t  jmpDest = URX_VAL(op);
2639                if (jmpDest < loc) {
2640                    // Loop of some kind.  Can safely ignore, the worst that will happen
2641                    //  is that we understate the true minimum length
2642                    currentLen = forwardedLength.elementAti(loc+1);
2643
2644                } else {
2645                    // Forward jump.  Propagate the current min length to the target loc of the jump.
2646                    U_ASSERT(jmpDest <= end+1);
2647                    if (forwardedLength.elementAti(jmpDest) > currentLen) {
2648                        forwardedLength.setElementAt(currentLen, jmpDest);
2649                    }
2650                }
2651            }
2652            atStart = FALSE;
2653            break;
2654
2655        case URX_JMP_SAV:
2656        case URX_JMP_SAV_X:
2657            // Combo of state save to the next loc, + jmp backwards.
2658            //   Net effect on min. length computation is nothing.
2659            atStart = FALSE;
2660            break;
2661
2662        case URX_BACKTRACK:
2663            // Fails are kind of like a branch, except that the min length was
2664            //   propagated already, by the state save.
2665            currentLen = forwardedLength.elementAti(loc+1);
2666            atStart = FALSE;
2667            break;
2668
2669
2670        case URX_STATE_SAVE:
2671            {
2672                // State Save, for forward jumps, propagate the current minimum.
2673                //             of the state save.
2674                int32_t  jmpDest = URX_VAL(op);
2675                if (jmpDest > loc) {
2676                    if (currentLen < forwardedLength.elementAti(jmpDest)) {
2677                        forwardedLength.setElementAt(currentLen, jmpDest);
2678                    }
2679                }
2680            }
2681            atStart = FALSE;
2682            break;
2683
2684
2685
2686
2687        case URX_STRING:
2688            {
2689                loc++;
2690                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
2691                int32_t stringLen   = URX_VAL(stringLenOp);
2692                U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN);
2693                U_ASSERT(stringLenOp >= 2);
2694                if (currentLen == 0) {
2695                    // Add the starting character of this string to the set of possible starting
2696                    //   characters for this pattern.
2697                    int32_t stringStartIdx = URX_VAL(op);
2698                    UChar32  c = fRXPat->fLiteralText.char32At(stringStartIdx);
2699                    fRXPat->fInitialChars->add(c);
2700
2701                    // Remember this string.  After the entire pattern has been checked,
2702                    //  if nothing else is identified that can start a match, we'll use it.
2703                    numInitialStrings++;
2704                    fRXPat->fInitialStringIdx = stringStartIdx;
2705                    fRXPat->fInitialStringLen = stringLen;
2706                }
2707
2708                currentLen += stringLen;
2709                atStart = FALSE;
2710            }
2711            break;
2712
2713        case URX_STRING_I:
2714            {
2715                // Case-insensitive string.  Unlike exact-match strings, we won't
2716                //   attempt a string search for possible match positions.  But we
2717                //   do update the set of possible starting characters.
2718                loc++;
2719                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
2720                int32_t stringLen   = URX_VAL(stringLenOp);
2721                U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN);
2722                U_ASSERT(stringLenOp >= 2);
2723                if (currentLen == 0) {
2724                    // Add the starting character of this string to the set of possible starting
2725                    //   characters for this pattern.
2726                    int32_t stringStartIdx = URX_VAL(op);
2727                    UChar32  c = fRXPat->fLiteralText.char32At(stringStartIdx);
2728                    UnicodeSet s(c, c);
2729                    s.closeOver(USET_CASE_INSENSITIVE);
2730                    fRXPat->fInitialChars->addAll(s);
2731                    numInitialStrings += 2;  // Matching on an initial string not possible.
2732                }
2733                currentLen += stringLen;
2734                atStart = FALSE;
2735            }
2736            break;
2737
2738        case URX_CTR_INIT:
2739        case URX_CTR_INIT_NG:
2740            {
2741                // Loop Init Ops.  These don't change the min length, but they are 4 word ops
2742                //   so location must be updated accordingly.
2743                // Loop Init Ops.
2744                //   If the min loop count == 0
2745                //      move loc forwards to the end of the loop, skipping over the body.
2746                //   If the min count is > 0,
2747                //      continue normal processing of the body of the loop.
2748                int32_t loopEndLoc   = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1);
2749                        loopEndLoc   = URX_VAL(loopEndLoc);
2750                int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2);
2751                if (minLoopCount == 0) {
2752                    // Min Loop Count of 0, treat like a forward branch and
2753                    //   move the current minimum length up to the target
2754                    //   (end of loop) location.
2755                    U_ASSERT(loopEndLoc <= end+1);
2756                    if (forwardedLength.elementAti(loopEndLoc) > currentLen) {
2757                        forwardedLength.setElementAt(currentLen, loopEndLoc);
2758                    }
2759                }
2760                loc+=3;  // Skips over operands of CTR_INIT
2761            }
2762            atStart = FALSE;
2763            break;
2764
2765
2766        case URX_CTR_LOOP:
2767        case URX_CTR_LOOP_NG:
2768            // Loop ops.
2769            //  The jump is conditional, backwards only.
2770            atStart = FALSE;
2771            break;
2772
2773        case URX_LOOP_C:
2774            // More loop ops.  These state-save to themselves.
2775            //   don't change the minimum match
2776            atStart = FALSE;
2777            break;
2778
2779
2780        case URX_LA_START:
2781        case URX_LB_START:
2782            {
2783                // Look-around.  Scan forward until the matching look-ahead end,
2784                //   without processing the look-around block.  This is overly pessimistic.
2785
2786                // Keep track of the nesting depth of look-around blocks.  Boilerplate code for
2787                //   lookahead contains two LA_END instructions, so count goes up by two
2788                //   for each LA_START.
2789                int32_t  depth = (opType == URX_LA_START? 2: 1);
2790                for (;;) {
2791                    loc++;
2792                    op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
2793                    if (URX_TYPE(op) == URX_LA_START) {
2794                        depth+=2;
2795                    }
2796                    if (URX_TYPE(op) == URX_LB_START) {
2797                        depth++;
2798                    }
2799                    if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) {
2800                        depth--;
2801                        if (depth == 0) {
2802                            break;
2803                        }
2804                    }
2805                    if (URX_TYPE(op) == URX_STATE_SAVE) {
2806                        // Need this because neg lookahead blocks will FAIL to outside
2807                        //   of the block.
2808                        int32_t  jmpDest = URX_VAL(op);
2809                        if (jmpDest > loc) {
2810                            if (currentLen < forwardedLength.elementAti(jmpDest)) {
2811                                forwardedLength.setElementAt(currentLen, jmpDest);
2812                            }
2813                        }
2814                    }
2815                    U_ASSERT(loc <= end);
2816                }
2817            }
2818            break;
2819
2820        case URX_LA_END:
2821        case URX_LB_CONT:
2822        case URX_LB_END:
2823        case URX_LBN_CONT:
2824        case URX_LBN_END:
2825            U_ASSERT(FALSE);     // Shouldn't get here.  These ops should be
2826                                 //  consumed by the scan in URX_LA_START and LB_START
2827
2828            break;
2829
2830        default:
2831            U_ASSERT(FALSE);
2832            }
2833
2834        }
2835
2836
2837    // We have finished walking through the ops.  Check whether some forward jump
2838    //   propagated a shorter length to location end+1.
2839    if (forwardedLength.elementAti(end+1) < currentLen) {
2840        currentLen = forwardedLength.elementAti(end+1);
2841    }
2842
2843
2844    fRXPat->fInitialChars8->init(fRXPat->fInitialChars);
2845
2846
2847    // Sort out what we should check for when looking for candidate match start positions.
2848    // In order of preference,
2849    //     1.   Start of input text buffer.
2850    //     2.   A literal string.
2851    //     3.   Start of line in multi-line mode.
2852    //     4.   A single literal character.
2853    //     5.   A character from a set of characters.
2854    //
2855    if (fRXPat->fStartType == START_START) {
2856        // Match only at the start of an input text string.
2857        //    start type is already set.  We're done.
2858    } else if (numInitialStrings == 1 && fRXPat->fMinMatchLen > 0) {
2859        // Match beginning only with a literal string.
2860        UChar32  c = fRXPat->fLiteralText.char32At(fRXPat->fInitialStringIdx);
2861        U_ASSERT(fRXPat->fInitialChars->contains(c));
2862        fRXPat->fStartType   = START_STRING;
2863        fRXPat->fInitialChar = c;
2864    } else if (fRXPat->fStartType == START_LINE) {
2865        // Match at start of line in Multi-Line mode.
2866        // Nothing to do here; everything is already set.
2867    } else if (fRXPat->fMinMatchLen == 0) {
2868        // Zero length match possible.  We could start anywhere.
2869        fRXPat->fStartType = START_NO_INFO;
2870    } else if (fRXPat->fInitialChars->size() == 1) {
2871        // All matches begin with the same char.
2872        fRXPat->fStartType   = START_CHAR;
2873        fRXPat->fInitialChar = fRXPat->fInitialChars->charAt(0);
2874        U_ASSERT(fRXPat->fInitialChar != (UChar32)-1);
2875    } else if (fRXPat->fInitialChars->contains((UChar32)0, (UChar32)0x10ffff) == FALSE &&
2876        fRXPat->fMinMatchLen > 0) {
2877        // Matches start with a set of character smaller than the set of all chars.
2878        fRXPat->fStartType = START_SET;
2879    } else {
2880        // Matches can start with anything
2881        fRXPat->fStartType = START_NO_INFO;
2882    }
2883
2884    return;
2885}
2886
2887
2888
2889//------------------------------------------------------------------------------
2890//
2891//   minMatchLength    Calculate the length of the shortest string that could
2892//                     match the specified pattern.
2893//                     Length is in 16 bit code units, not code points.
2894//
2895//                     The calculated length may not be exact.  The returned
2896//                     value may be shorter than the actual minimum; it must
2897//                     never be longer.
2898//
2899//                     start and end are the range of p-code operations to be
2900//                     examined.  The endpoints are included in the range.
2901//
2902//------------------------------------------------------------------------------
2903int32_t   RegexCompile::minMatchLength(int32_t start, int32_t end) {
2904    if (U_FAILURE(*fStatus)) {
2905        return 0;
2906    }
2907
2908    U_ASSERT(start <= end);
2909    U_ASSERT(end < fRXPat->fCompiledPat->size());
2910
2911
2912    int32_t    loc;
2913    int32_t    op;
2914    int32_t    opType;
2915    int32_t    currentLen = 0;
2916
2917
2918    // forwardedLength is a vector holding minimum-match-length values that
2919    //   are propagated forward in the pattern by JMP or STATE_SAVE operations.
2920    //   It must be one longer than the pattern being checked because some  ops
2921    //   will jmp to a end-of-block+1 location from within a block, and we must
2922    //   count those when checking the block.
2923    UVector32  forwardedLength(end+2, *fStatus);
2924    forwardedLength.setSize(end+2);
2925    for (loc=start; loc<=end+1; loc++) {
2926        forwardedLength.setElementAt(INT32_MAX, loc);
2927    }
2928
2929    for (loc = start; loc<=end; loc++) {
2930        op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
2931        opType = URX_TYPE(op);
2932
2933        // The loop is advancing linearly through the pattern.
2934        // If the op we are now at was the destination of a branch in the pattern,
2935        // and that path has a shorter minimum length than the current accumulated value,
2936        // replace the current accumulated value.
2937        // U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);  // MinLength == INT32_MAX for some
2938                                                               //   no-match-possible cases.
2939        if (forwardedLength.elementAti(loc) < currentLen) {
2940            currentLen = forwardedLength.elementAti(loc);
2941            U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);
2942        }
2943
2944        switch (opType) {
2945            // Ops that don't change the total length matched
2946        case URX_RESERVED_OP:
2947        case URX_END:
2948        case URX_STRING_LEN:
2949        case URX_NOP:
2950        case URX_START_CAPTURE:
2951        case URX_END_CAPTURE:
2952        case URX_BACKSLASH_B:
2953        case URX_BACKSLASH_BU:
2954        case URX_BACKSLASH_G:
2955        case URX_BACKSLASH_Z:
2956        case URX_CARET:
2957        case URX_DOLLAR:
2958        case URX_DOLLAR_M:
2959        case URX_DOLLAR_D:
2960        case URX_DOLLAR_MD:
2961        case URX_RELOC_OPRND:
2962        case URX_STO_INP_LOC:
2963        case URX_CARET_M:
2964        case URX_CARET_M_UNIX:
2965        case URX_BACKREF:         // BackRef.  Must assume that it might be a zero length match
2966        case URX_BACKREF_I:
2967
2968        case URX_STO_SP:          // Setup for atomic or possessive blocks.  Doesn't change what can match.
2969        case URX_LD_SP:
2970
2971        case URX_JMP_SAV:
2972        case URX_JMP_SAV_X:
2973            break;
2974
2975
2976            // Ops that match a minimum of one character (one or two 16 bit code units.)
2977            //
2978        case URX_ONECHAR:
2979        case URX_STATIC_SETREF:
2980        case URX_STAT_SETREF_N:
2981        case URX_SETREF:
2982        case URX_BACKSLASH_D:
2983        case URX_ONECHAR_I:
2984        case URX_BACKSLASH_X:   // Grahpeme Cluster.  Minimum is 1, max unbounded.
2985        case URX_DOTANY_ALL:    // . matches one or two.
2986        case URX_DOTANY:
2987        case URX_DOTANY_UNIX:
2988            currentLen++;
2989            break;
2990
2991
2992        case URX_JMPX:
2993            loc++;              // URX_JMPX has an extra operand, ignored here,
2994                                //   otherwise processed identically to URX_JMP.
2995        case URX_JMP:
2996            {
2997                int32_t  jmpDest = URX_VAL(op);
2998                if (jmpDest < loc) {
2999                    // Loop of some kind.  Can safely ignore, the worst that will happen
3000                    //  is that we understate the true minimum length
3001                    currentLen = forwardedLength.elementAti(loc+1);
3002                } else {
3003                    // Forward jump.  Propagate the current min length to the target loc of the jump.
3004                    U_ASSERT(jmpDest <= end+1);
3005                    if (forwardedLength.elementAti(jmpDest) > currentLen) {
3006                        forwardedLength.setElementAt(currentLen, jmpDest);
3007                    }
3008                }
3009            }
3010            break;
3011
3012        case URX_BACKTRACK:
3013            {
3014                // Back-tracks are kind of like a branch, except that the min length was
3015                //   propagated already, by the state save.
3016                currentLen = forwardedLength.elementAti(loc+1);
3017            }
3018            break;
3019
3020
3021        case URX_STATE_SAVE:
3022            {
3023                // State Save, for forward jumps, propagate the current minimum.
3024                //             of the state save.
3025                int32_t  jmpDest = URX_VAL(op);
3026                if (jmpDest > loc) {
3027                    if (currentLen < forwardedLength.elementAti(jmpDest)) {
3028                        forwardedLength.setElementAt(currentLen, jmpDest);
3029                    }
3030                }
3031            }
3032            break;
3033
3034
3035        case URX_STRING:
3036        case URX_STRING_I:
3037            {
3038                loc++;
3039                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
3040                currentLen += URX_VAL(stringLenOp);
3041            }
3042            break;
3043
3044
3045        case URX_CTR_INIT:
3046        case URX_CTR_INIT_NG:
3047            {
3048                // Loop Init Ops.
3049                //   If the min loop count == 0
3050                //      move loc forwards to the end of the loop, skipping over the body.
3051                //   If the min count is > 0,
3052                //      continue normal processing of the body of the loop.
3053                int32_t loopEndLoc   = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1);
3054                        loopEndLoc   = URX_VAL(loopEndLoc);
3055                int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2);
3056                if (minLoopCount == 0) {
3057                    loc = loopEndLoc;
3058                } else {
3059                    loc+=3;  // Skips over operands of CTR_INIT
3060                }
3061            }
3062            break;
3063
3064
3065        case URX_CTR_LOOP:
3066        case URX_CTR_LOOP_NG:
3067            // Loop ops.
3068            //  The jump is conditional, backwards only.
3069            break;
3070
3071        case URX_LOOP_SR_I:
3072        case URX_LOOP_DOT_I:
3073        case URX_LOOP_C:
3074            // More loop ops.  These state-save to themselves.
3075            //   don't change the minimum match - could match nothing at all.
3076            break;
3077
3078
3079        case URX_LA_START:
3080        case URX_LB_START:
3081            {
3082                // Look-around.  Scan forward until the matching look-ahead end,
3083                //   without processing the look-around block.  This is overly pessimistic for look-ahead,
3084                //   it assumes that the look-ahead match might be zero-length.
3085                //   TODO:  Positive lookahead could recursively do the block, then continue
3086                //          with the longer of the block or the value coming in.  Ticket 6060
3087                int32_t  depth = (opType == URX_LA_START? 2: 1);;
3088                for (;;) {
3089                    loc++;
3090                    op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
3091                    if (URX_TYPE(op) == URX_LA_START) {
3092                        // The boilerplate for look-ahead includes two LA_END insturctions,
3093                        //    Depth will be decremented by each one when it is seen.
3094                        depth += 2;
3095                    }
3096                    if (URX_TYPE(op) == URX_LB_START) {
3097                        depth++;
3098                    }
3099                    if (URX_TYPE(op) == URX_LA_END) {
3100                        depth--;
3101                        if (depth == 0) {
3102                            break;
3103                        }
3104                    }
3105                    if (URX_TYPE(op)==URX_LBN_END) {
3106                        depth--;
3107                        if (depth == 0) {
3108                            break;
3109                        }
3110                    }
3111                    if (URX_TYPE(op) == URX_STATE_SAVE) {
3112                        // Need this because neg lookahead blocks will FAIL to outside
3113                        //   of the block.
3114                        int32_t  jmpDest = URX_VAL(op);
3115                        if (jmpDest > loc) {
3116                            if (currentLen < forwardedLength.elementAti(jmpDest)) {
3117                                forwardedLength.setElementAt(currentLen, jmpDest);
3118                            }
3119                        }
3120                    }
3121                    U_ASSERT(loc <= end);
3122                }
3123            }
3124            break;
3125
3126        case URX_LA_END:
3127        case URX_LB_CONT:
3128        case URX_LB_END:
3129        case URX_LBN_CONT:
3130        case URX_LBN_END:
3131            // Only come here if the matching URX_LA_START or URX_LB_START was not in the
3132            //   range being sized, which happens when measuring size of look-behind blocks.
3133            break;
3134
3135        default:
3136            U_ASSERT(FALSE);
3137            }
3138
3139        }
3140
3141    // We have finished walking through the ops.  Check whether some forward jump
3142    //   propagated a shorter length to location end+1.
3143    if (forwardedLength.elementAti(end+1) < currentLen) {
3144        currentLen = forwardedLength.elementAti(end+1);
3145        U_ASSERT(currentLen>=0 && currentLen < INT32_MAX);
3146    }
3147
3148    return currentLen;
3149}
3150
3151
3152
3153//------------------------------------------------------------------------------
3154//
3155//   maxMatchLength    Calculate the length of the longest string that could
3156//                     match the specified pattern.
3157//                     Length is in 16 bit code units, not code points.
3158//
3159//                     The calculated length may not be exact.  The returned
3160//                     value may be longer than the actual maximum; it must
3161//                     never be shorter.
3162//
3163//------------------------------------------------------------------------------
3164int32_t   RegexCompile::maxMatchLength(int32_t start, int32_t end) {
3165    if (U_FAILURE(*fStatus)) {
3166        return 0;
3167    }
3168    U_ASSERT(start <= end);
3169    U_ASSERT(end < fRXPat->fCompiledPat->size());
3170
3171
3172    int32_t    loc;
3173    int32_t    op;
3174    int32_t    opType;
3175    int32_t    currentLen = 0;
3176    UVector32  forwardedLength(end+1, *fStatus);
3177    forwardedLength.setSize(end+1);
3178
3179    for (loc=start; loc<=end; loc++) {
3180        forwardedLength.setElementAt(0, loc);
3181    }
3182
3183    for (loc = start; loc<=end; loc++) {
3184        op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
3185        opType = URX_TYPE(op);
3186
3187        // The loop is advancing linearly through the pattern.
3188        // If the op we are now at was the destination of a branch in the pattern,
3189        // and that path has a longer maximum length than the current accumulated value,
3190        // replace the current accumulated value.
3191        if (forwardedLength.elementAti(loc) > currentLen) {
3192            currentLen = forwardedLength.elementAti(loc);
3193        }
3194
3195        switch (opType) {
3196            // Ops that don't change the total length matched
3197        case URX_RESERVED_OP:
3198        case URX_END:
3199        case URX_STRING_LEN:
3200        case URX_NOP:
3201        case URX_START_CAPTURE:
3202        case URX_END_CAPTURE:
3203        case URX_BACKSLASH_B:
3204        case URX_BACKSLASH_BU:
3205        case URX_BACKSLASH_G:
3206        case URX_BACKSLASH_Z:
3207        case URX_CARET:
3208        case URX_DOLLAR:
3209        case URX_DOLLAR_M:
3210        case URX_DOLLAR_D:
3211        case URX_DOLLAR_MD:
3212        case URX_RELOC_OPRND:
3213        case URX_STO_INP_LOC:
3214        case URX_CARET_M:
3215        case URX_CARET_M_UNIX:
3216
3217        case URX_STO_SP:          // Setup for atomic or possessive blocks.  Doesn't change what can match.
3218        case URX_LD_SP:
3219
3220        case URX_LB_END:
3221        case URX_LB_CONT:
3222        case URX_LBN_CONT:
3223        case URX_LBN_END:
3224            break;
3225
3226
3227            // Ops that increase that cause an unbounded increase in the length
3228            //   of a matched string, or that increase it a hard to characterize way.
3229            //   Call the max length unbounded, and stop further checking.
3230        case URX_BACKREF:         // BackRef.  Must assume that it might be a zero length match
3231        case URX_BACKREF_I:
3232        case URX_BACKSLASH_X:   // Grahpeme Cluster.  Minimum is 1, max unbounded.
3233            currentLen = INT32_MAX;
3234            break;
3235
3236
3237            // Ops that match a max of one character (possibly two 16 bit code units.)
3238            //
3239        case URX_STATIC_SETREF:
3240        case URX_STAT_SETREF_N:
3241        case URX_SETREF:
3242        case URX_BACKSLASH_D:
3243        case URX_ONECHAR_I:
3244        case URX_DOTANY_ALL:
3245        case URX_DOTANY:
3246        case URX_DOTANY_UNIX:
3247            currentLen+=2;
3248            break;
3249
3250            // Single literal character.  Increase current max length by one or two,
3251            //       depending on whether the char is in the supplementary range.
3252        case URX_ONECHAR:
3253            currentLen++;
3254            if (URX_VAL(op) > 0x10000) {
3255                currentLen++;
3256            }
3257            break;
3258
3259            // Jumps.
3260            //
3261        case URX_JMP:
3262        case URX_JMPX:
3263        case URX_JMP_SAV:
3264        case URX_JMP_SAV_X:
3265            {
3266                int32_t  jmpDest = URX_VAL(op);
3267                if (jmpDest < loc) {
3268                    // Loop of some kind.  Max match length is unbounded.
3269                    currentLen = INT32_MAX;
3270                } else {
3271                    // Forward jump.  Propagate the current min length to the target loc of the jump.
3272                    if (forwardedLength.elementAti(jmpDest) < currentLen) {
3273                        forwardedLength.setElementAt(currentLen, jmpDest);
3274                    }
3275                    currentLen = 0;
3276                }
3277            }
3278            break;
3279
3280        case URX_BACKTRACK:
3281            // back-tracks are kind of like a branch, except that the max length was
3282            //   propagated already, by the state save.
3283            currentLen = forwardedLength.elementAti(loc+1);
3284            break;
3285
3286
3287        case URX_STATE_SAVE:
3288            {
3289                // State Save, for forward jumps, propagate the current minimum.
3290                //               of the state save.
3291                //             For backwards jumps, they create a loop, maximum
3292                //               match length is unbounded.
3293                int32_t  jmpDest = URX_VAL(op);
3294                if (jmpDest > loc) {
3295                    if (currentLen > forwardedLength.elementAti(jmpDest)) {
3296                        forwardedLength.setElementAt(currentLen, jmpDest);
3297                    }
3298                } else {
3299                    currentLen = INT32_MAX;
3300                }
3301            }
3302            break;
3303
3304
3305
3306
3307        case URX_STRING:
3308        case URX_STRING_I:
3309            {
3310                loc++;
3311                int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
3312                currentLen += URX_VAL(stringLenOp);
3313            }
3314            break;
3315
3316
3317        case URX_CTR_INIT:
3318        case URX_CTR_INIT_NG:
3319        case URX_CTR_LOOP:
3320        case URX_CTR_LOOP_NG:
3321        case URX_LOOP_SR_I:
3322        case URX_LOOP_DOT_I:
3323        case URX_LOOP_C:
3324            // For anything to do with loops, make the match length unbounded.
3325            //   Note:  INIT instructions are multi-word.  Can ignore because
3326            //          INT32_MAX length will stop the per-instruction loop.
3327            currentLen = INT32_MAX;
3328            break;
3329
3330
3331
3332        case URX_LA_START:
3333        case URX_LA_END:
3334            // Look-ahead.  Just ignore, treat the look-ahead block as if
3335            // it were normal pattern.  Gives a too-long match length,
3336            //  but good enough for now.
3337            break;
3338
3339            // End of look-ahead ops should always be consumed by the processing at
3340            //  the URX_LA_START op.
3341            // U_ASSERT(FALSE);
3342            // break;
3343
3344        case URX_LB_START:
3345            {
3346                // Look-behind.  Scan forward until the matching look-around end,
3347                //   without processing the look-behind block.
3348                int32_t  depth = 0;
3349                for (;;) {
3350                    loc++;
3351                    op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
3352                    if (URX_TYPE(op) == URX_LA_START || URX_TYPE(op) == URX_LB_START) {
3353                        depth++;
3354                    }
3355                    if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) {
3356                        if (depth == 0) {
3357                            break;
3358                        }
3359                        depth--;
3360                    }
3361                    U_ASSERT(loc < end);
3362                }
3363            }
3364            break;
3365
3366        default:
3367            U_ASSERT(FALSE);
3368        }
3369
3370
3371        if (currentLen == INT32_MAX) {
3372            //  The maximum length is unbounded.
3373            //  Stop further processing of the pattern.
3374            break;
3375        }
3376
3377    }
3378    return currentLen;
3379
3380}
3381
3382
3383//------------------------------------------------------------------------------
3384//
3385//   stripNOPs    Remove any NOP operations from the compiled pattern code.
3386//                Extra NOPs are inserted for some constructs during the initial
3387//                code generation to provide locations that may be patched later.
3388//                Many end up unneeded, and are removed by this function.
3389//
3390//                In order to minimize the number of passes through the pattern,
3391//                back-reference fixup is also performed here (adjusting
3392//                back-reference operands to point to the correct frame offsets).
3393//
3394//                In addition, case-insensitive character and string literals are
3395//                now case-folded here, rather than when first parsed or at match
3396//                time.
3397//
3398//------------------------------------------------------------------------------
3399void RegexCompile::stripNOPs() {
3400
3401    if (U_FAILURE(*fStatus)) {
3402        return;
3403    }
3404
3405    int32_t    end = fRXPat->fCompiledPat->size();
3406    UVector32  deltas(end, *fStatus);
3407
3408    // Make a first pass over the code, computing the amount that things
3409    //   will be offset at each location in the original code.
3410    int32_t   loc;
3411    int32_t   d = 0;
3412    for (loc=0; loc<end; loc++) {
3413        deltas.addElement(d, *fStatus);
3414        int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
3415        if (URX_TYPE(op) == URX_NOP) {
3416            d++;
3417        }
3418    }
3419
3420    UnicodeString caseStringBuffer;
3421    int32_t stringDelta = 0;
3422
3423    // Make a second pass over the code, removing the NOPs by moving following
3424    //  code up, and patching operands that refer to code locations that
3425    //  are being moved.  The array of offsets from the first step is used
3426    //  to compute the new operand values.
3427    int32_t src;
3428    int32_t dst = 0;
3429    for (src=0; src<end; src++) {
3430        int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(src);
3431        int32_t opType = URX_TYPE(op);
3432        switch (opType) {
3433        case URX_NOP:
3434            break;
3435
3436        case URX_STATE_SAVE:
3437        case URX_JMP:
3438        case URX_CTR_LOOP:
3439        case URX_CTR_LOOP_NG:
3440        case URX_RELOC_OPRND:
3441        case URX_JMPX:
3442        case URX_JMP_SAV:
3443        case URX_JMP_SAV_X:
3444            // These are instructions with operands that refer to code locations.
3445            {
3446                int32_t  operandAddress = URX_VAL(op);
3447                U_ASSERT(operandAddress>=0 && operandAddress<deltas.size());
3448                int32_t fixedOperandAddress = operandAddress - deltas.elementAti(operandAddress);
3449                op = URX_BUILD(opType, fixedOperandAddress);
3450                fRXPat->fCompiledPat->setElementAt(op, dst);
3451                dst++;
3452                break;
3453            }
3454
3455        case URX_ONECHAR_I:
3456            {
3457                UChar32 c = URX_VAL(op);
3458                if (u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) {
3459                    // We have a cased character to fold
3460                    c  = u_foldCase(c, U_FOLD_CASE_DEFAULT);
3461                    op = URX_BUILD(URX_ONECHAR_I, c);
3462                }
3463
3464                fRXPat->fCompiledPat->setElementAt(op, dst);
3465                dst++;
3466                break;
3467            }
3468        case URX_STRING_I:
3469            {
3470                op = URX_BUILD(URX_STRING_I, URX_VAL(op)+stringDelta);
3471
3472                src++;
3473                int32_t lengthOp = (int32_t)fRXPat->fCompiledPat->elementAti(src);
3474
3475                caseStringBuffer.setTo(fRXPat->fLiteralText, URX_VAL(op), URX_VAL(lengthOp));
3476                caseStringBuffer.foldCase(U_FOLD_CASE_DEFAULT);
3477
3478                int32_t newLen = caseStringBuffer.length();
3479                if (newLen <= URX_VAL(lengthOp)) {
3480                    // don't shift if we don't have to, take the tiny memory hit of a smaller string
3481                    fRXPat->fLiteralText.replace(URX_VAL(op), newLen, caseStringBuffer);
3482                } else {
3483                    // shift other strings over...at least UnicodeString handles this for us!
3484                    fRXPat->fLiteralText.replace(URX_VAL(op), URX_VAL(lengthOp), caseStringBuffer);
3485                    stringDelta += newLen - URX_VAL(lengthOp);
3486                }
3487                lengthOp = URX_BUILD(URX_STRING_LEN, newLen);
3488
3489                fRXPat->fCompiledPat->setElementAt(op, dst);
3490                fRXPat->fCompiledPat->setElementAt(lengthOp, dst+1);
3491                dst += 2;
3492                break;
3493            }
3494        case URX_BACKREF:
3495        case URX_BACKREF_I:
3496            {
3497                int32_t where = URX_VAL(op);
3498                if (where > fRXPat->fGroupMap->size()) {
3499                    error(U_REGEX_INVALID_BACK_REF);
3500                    break;
3501                }
3502                where = fRXPat->fGroupMap->elementAti(where-1);
3503                op    = URX_BUILD(opType, where);
3504                fRXPat->fCompiledPat->setElementAt(op, dst);
3505                dst++;
3506
3507                fRXPat->fNeedsAltInput = TRUE;
3508                break;
3509            }
3510        case URX_STRING:
3511            op = URX_BUILD(URX_STRING, URX_VAL(op)+stringDelta);
3512            // continue
3513        case URX_RESERVED_OP:
3514        case URX_RESERVED_OP_N:
3515        case URX_BACKTRACK:
3516        case URX_END:
3517        case URX_ONECHAR:
3518        case URX_STRING_LEN:
3519        case URX_START_CAPTURE:
3520        case URX_END_CAPTURE:
3521        case URX_STATIC_SETREF:
3522        case URX_STAT_SETREF_N:
3523        case URX_SETREF:
3524        case URX_DOTANY:
3525        case URX_FAIL:
3526        case URX_BACKSLASH_B:
3527        case URX_BACKSLASH_BU:
3528        case URX_BACKSLASH_G:
3529        case URX_BACKSLASH_X:
3530        case URX_BACKSLASH_Z:
3531        case URX_DOTANY_ALL:
3532        case URX_BACKSLASH_D:
3533        case URX_CARET:
3534        case URX_DOLLAR:
3535        case URX_CTR_INIT:
3536        case URX_CTR_INIT_NG:
3537        case URX_DOTANY_UNIX:
3538        case URX_STO_SP:
3539        case URX_LD_SP:
3540        case URX_STO_INP_LOC:
3541        case URX_LA_START:
3542        case URX_LA_END:
3543        case URX_DOLLAR_M:
3544        case URX_CARET_M:
3545        case URX_CARET_M_UNIX:
3546        case URX_LB_START:
3547        case URX_LB_CONT:
3548        case URX_LB_END:
3549        case URX_LBN_CONT:
3550        case URX_LBN_END:
3551        case URX_LOOP_SR_I:
3552        case URX_LOOP_DOT_I:
3553        case URX_LOOP_C:
3554        case URX_DOLLAR_D:
3555        case URX_DOLLAR_MD:
3556            // These instructions are unaltered by the relocation.
3557            fRXPat->fCompiledPat->setElementAt(op, dst);
3558            dst++;
3559            break;
3560
3561        default:
3562            // Some op is unaccounted for.
3563            U_ASSERT(FALSE);
3564            error(U_REGEX_INTERNAL_ERROR);
3565        }
3566    }
3567
3568    fRXPat->fCompiledPat->setSize(dst);
3569}
3570
3571
3572
3573
3574//------------------------------------------------------------------------------
3575//
3576//  Error         Report a rule parse error.
3577//                Only report it if no previous error has been recorded.
3578//
3579//------------------------------------------------------------------------------
3580void RegexCompile::error(UErrorCode e) {
3581    if (U_SUCCESS(*fStatus)) {
3582        *fStatus = e;
3583        // Hmm. fParseErr (UParseError) line & offset fields are int32_t in public
3584        // API (see common/unicode/parseerr.h), while fLineNum and fCharNum are
3585        // int64_t. If the values of the latter are out of range for the former,
3586        // set them to the appropriate "field not supported" values.
3587        if (fLineNum > 0x7FFFFFFF) {
3588            fParseErr->line   = 0;
3589            fParseErr->offset = -1;
3590        } else if (fCharNum > 0x7FFFFFFF) {
3591            fParseErr->line   = (int32_t)fLineNum;
3592            fParseErr->offset = -1;
3593        } else {
3594            fParseErr->line   = (int32_t)fLineNum;
3595            fParseErr->offset = (int32_t)fCharNum;
3596        }
3597
3598        UErrorCode status = U_ZERO_ERROR; // throwaway status for extracting context
3599
3600        // Fill in the context.
3601        //   Note: extractBetween() pins supplied indicies to the string bounds.
3602        uprv_memset(fParseErr->preContext,  0, sizeof(fParseErr->preContext));
3603        uprv_memset(fParseErr->postContext, 0, sizeof(fParseErr->postContext));
3604        utext_extract(fRXPat->fPattern, fScanIndex-U_PARSE_CONTEXT_LEN+1, fScanIndex, fParseErr->preContext, U_PARSE_CONTEXT_LEN, &status);
3605        utext_extract(fRXPat->fPattern, fScanIndex, fScanIndex+U_PARSE_CONTEXT_LEN-1, fParseErr->postContext, U_PARSE_CONTEXT_LEN, &status);
3606    }
3607}
3608
3609
3610//
3611//  Assorted Unicode character constants.
3612//     Numeric because there is no portable way to enter them as literals.
3613//     (Think EBCDIC).
3614//
3615static const UChar      chCR        = 0x0d;      // New lines, for terminating comments.
3616static const UChar      chLF        = 0x0a;      // Line Feed
3617static const UChar      chPound     = 0x23;      // '#', introduces a comment.
3618static const UChar      chDigit0    = 0x30;      // '0'
3619static const UChar      chDigit7    = 0x37;      // '9'
3620static const UChar      chColon     = 0x3A;      // ':'
3621static const UChar      chE         = 0x45;      // 'E'
3622static const UChar      chQ         = 0x51;      // 'Q'
3623static const UChar      chN         = 0x4E;      // 'N'
3624static const UChar      chP         = 0x50;      // 'P'
3625static const UChar      chBackSlash = 0x5c;      // '\'  introduces a char escape
3626static const UChar      chLBracket  = 0x5b;      // '['
3627static const UChar      chRBracket  = 0x5d;      // ']'
3628static const UChar      chUp        = 0x5e;      // '^'
3629static const UChar      chLowerP    = 0x70;
3630static const UChar      chLBrace    = 0x7b;      // '{'
3631static const UChar      chRBrace    = 0x7d;      // '}'
3632static const UChar      chNEL       = 0x85;      //    NEL newline variant
3633static const UChar      chLS        = 0x2028;    //    Unicode Line Separator
3634
3635
3636//------------------------------------------------------------------------------
3637//
3638//  nextCharLL    Low Level Next Char from the regex pattern.
3639//                Get a char from the string, keep track of input position
3640//                     for error reporting.
3641//
3642//------------------------------------------------------------------------------
3643UChar32  RegexCompile::nextCharLL() {
3644    UChar32       ch;
3645
3646    if (fPeekChar != -1) {
3647        ch = fPeekChar;
3648        fPeekChar = -1;
3649        return ch;
3650    }
3651
3652    // assume we're already in the right place
3653    ch = UTEXT_NEXT32(fRXPat->fPattern);
3654    if (ch == U_SENTINEL) {
3655        return ch;
3656    }
3657
3658    if (ch == chCR ||
3659        ch == chNEL ||
3660        ch == chLS   ||
3661        (ch == chLF && fLastChar != chCR)) {
3662        // Character is starting a new line.  Bump up the line number, and
3663        //  reset the column to 0.
3664        fLineNum++;
3665        fCharNum=0;
3666    }
3667    else {
3668        // Character is not starting a new line.  Except in the case of a
3669        //   LF following a CR, increment the column position.
3670        if (ch != chLF) {
3671            fCharNum++;
3672        }
3673    }
3674    fLastChar = ch;
3675    return ch;
3676}
3677
3678//------------------------------------------------------------------------------
3679//
3680//   peekCharLL    Low Level Character Scanning, sneak a peek at the next
3681//                 character without actually getting it.
3682//
3683//------------------------------------------------------------------------------
3684UChar32  RegexCompile::peekCharLL() {
3685    if (fPeekChar == -1) {
3686        fPeekChar = nextCharLL();
3687    }
3688    return fPeekChar;
3689}
3690
3691
3692//------------------------------------------------------------------------------
3693//
3694//   nextChar     for pattern scanning.  At this level, we handle stripping
3695//                out comments and processing some backslash character escapes.
3696//                The rest of the pattern grammar is handled at the next level up.
3697//
3698//------------------------------------------------------------------------------
3699void RegexCompile::nextChar(RegexPatternChar &c) {
3700
3701    fScanIndex = UTEXT_GETNATIVEINDEX(fRXPat->fPattern);
3702    c.fChar    = nextCharLL();
3703    c.fQuoted  = FALSE;
3704
3705    if (fQuoteMode) {
3706        c.fQuoted = TRUE;
3707        if ((c.fChar==chBackSlash && peekCharLL()==chE) || c.fChar == (UChar32)-1) {
3708            fQuoteMode = FALSE;  //  Exit quote mode,
3709            nextCharLL();       // discard the E
3710            nextChar(c);        // recurse to get the real next char
3711        }
3712    }
3713    else if (fInBackslashQuote) {
3714        // The current character immediately follows a '\'
3715        // Don't check for any further escapes, just return it as-is.
3716        // Don't set c.fQuoted, because that would prevent the state machine from
3717        //    dispatching on the character.
3718        fInBackslashQuote = FALSE;
3719    }
3720    else
3721    {
3722        // We are not in a \Q quoted region \E of the source.
3723        //
3724        if (fModeFlags & UREGEX_COMMENTS) {
3725            //
3726            // We are in free-spacing and comments mode.
3727            //  Scan through any white space and comments, until we
3728            //  reach a significant character or the end of inut.
3729            for (;;) {
3730                if (c.fChar == (UChar32)-1) {
3731                    break;     // End of Input
3732                }
3733                if  (c.fChar == chPound && fEOLComments == TRUE) {
3734                    // Start of a comment.  Consume the rest of it, until EOF or a new line
3735                    for (;;) {
3736                        c.fChar = nextCharLL();
3737                        if (c.fChar == (UChar32)-1 ||  // EOF
3738                            c.fChar == chCR        ||
3739                            c.fChar == chLF        ||
3740                            c.fChar == chNEL       ||
3741                            c.fChar == chLS)       {
3742                            break;
3743                        }
3744                    }
3745                }
3746                // TODO:  check what Java & Perl do with non-ASCII white spaces.  Ticket 6061.
3747                if (uprv_isRuleWhiteSpace(c.fChar) == FALSE) {
3748                    break;
3749                }
3750                c.fChar = nextCharLL();
3751            }
3752        }
3753
3754        //
3755        //  check for backslash escaped characters.
3756        //
3757        if (c.fChar == chBackSlash) {
3758            int64_t pos = UTEXT_GETNATIVEINDEX(fRXPat->fPattern);
3759            if (RegexStaticSets::gStaticSets->fUnescapeCharSet.contains(peekCharLL())) {
3760                //
3761                // A '\' sequence that is handled by ICU's standard unescapeAt function.
3762                //   Includes \uxxxx, \n, \r, many others.
3763                //   Return the single equivalent character.
3764                //
3765                nextCharLL();                 // get & discard the peeked char.
3766                c.fQuoted = TRUE;
3767
3768                if (UTEXT_FULL_TEXT_IN_CHUNK(fRXPat->fPattern, fPatternLength)) {
3769                    int32_t endIndex = (int32_t)pos;
3770                    c.fChar = u_unescapeAt(uregex_ucstr_unescape_charAt, &endIndex, (int32_t)fPatternLength, (void *)fRXPat->fPattern->chunkContents);
3771
3772                    if (endIndex == pos) {
3773                        error(U_REGEX_BAD_ESCAPE_SEQUENCE);
3774                    }
3775                    fCharNum += endIndex - pos;
3776                    UTEXT_SETNATIVEINDEX(fRXPat->fPattern, endIndex);
3777                } else {
3778                    int32_t offset = 0;
3779                    struct URegexUTextUnescapeCharContext context = U_REGEX_UTEXT_UNESCAPE_CONTEXT(fRXPat->fPattern);
3780
3781                    UTEXT_SETNATIVEINDEX(fRXPat->fPattern, pos);
3782                    c.fChar = u_unescapeAt(uregex_utext_unescape_charAt, &offset, INT32_MAX, &context);
3783
3784                    if (offset == 0) {
3785                        error(U_REGEX_BAD_ESCAPE_SEQUENCE);
3786                    } else if (context.lastOffset == offset) {
3787                        UTEXT_PREVIOUS32(fRXPat->fPattern);
3788                    } else if (context.lastOffset != offset-1) {
3789                        utext_moveIndex32(fRXPat->fPattern, offset - context.lastOffset - 1);
3790                    }
3791                    fCharNum += offset;
3792                }
3793            }
3794            else if (peekCharLL() == chDigit0) {
3795                //  Octal Escape, using Java Regexp Conventions
3796                //    which are \0 followed by 1-3 octal digits.
3797                //    Different from ICU Unescape handling of Octal, which does not
3798                //    require the leading 0.
3799                //  Java also has the convention of only consuming 2 octal digits if
3800                //    the three digit number would be > 0xff
3801                //
3802                c.fChar = 0;
3803                nextCharLL();    // Consume the initial 0.
3804                int index;
3805                for (index=0; index<3; index++) {
3806                    int32_t ch = peekCharLL();
3807                    if (ch<chDigit0 || ch>chDigit7) {
3808                        if (index==0) {
3809                           // \0 is not followed by any octal digits.
3810                           error(U_REGEX_BAD_ESCAPE_SEQUENCE);
3811                        }
3812                        break;
3813                    }
3814                    c.fChar <<= 3;
3815                    c.fChar += ch&7;
3816                    if (c.fChar <= 255) {
3817                        nextCharLL();
3818                    } else {
3819                        // The last digit made the number too big.  Forget we saw it.
3820                        c.fChar >>= 3;
3821                    }
3822                }
3823                c.fQuoted = TRUE;
3824            }
3825            else if (peekCharLL() == chQ) {
3826                //  "\Q"  enter quote mode, which will continue until "\E"
3827                fQuoteMode = TRUE;
3828                nextCharLL();       // discard the 'Q'.
3829                nextChar(c);        // recurse to get the real next char.
3830            }
3831            else
3832            {
3833                // We are in a '\' escape that will be handled by the state table scanner.
3834                // Just return the backslash, but remember that the following char is to
3835                //  be taken literally.
3836                fInBackslashQuote = TRUE;
3837            }
3838        }
3839    }
3840
3841    // re-enable # to end-of-line comments, in case they were disabled.
3842    // They are disabled by the parser upon seeing '(?', but this lasts for
3843    //  the fetching of the next character only.
3844    fEOLComments = TRUE;
3845
3846    // putc(c.fChar, stdout);
3847}
3848
3849
3850
3851//------------------------------------------------------------------------------
3852//
3853//  scanNamedChar
3854 //            Get a UChar32 from a \N{UNICODE CHARACTER NAME} in the pattern.
3855//
3856//             The scan position will be at the 'N'.  On return
3857//             the scan position should be just after the '}'
3858//
3859//             Return the UChar32
3860//
3861//------------------------------------------------------------------------------
3862UChar32  RegexCompile::scanNamedChar() {
3863    if (U_FAILURE(*fStatus)) {
3864        return 0;
3865    }
3866
3867    nextChar(fC);
3868    if (fC.fChar != chLBrace) {
3869        error(U_REGEX_PROPERTY_SYNTAX);
3870        return 0;
3871    }
3872
3873    UnicodeString  charName;
3874    for (;;) {
3875        nextChar(fC);
3876        if (fC.fChar == chRBrace) {
3877            break;
3878        }
3879        if (fC.fChar == -1) {
3880            error(U_REGEX_PROPERTY_SYNTAX);
3881            return 0;
3882        }
3883        charName.append(fC.fChar);
3884    }
3885
3886    char name[100];
3887    if (!uprv_isInvariantUString(charName.getBuffer(), charName.length()) ||
3888         (uint32_t)charName.length()>=sizeof(name)) {
3889        // All Unicode character names have only invariant characters.
3890        // The API to get a character, given a name, accepts only char *, forcing us to convert,
3891        //   which requires this error check
3892        error(U_REGEX_PROPERTY_SYNTAX);
3893        return 0;
3894    }
3895    charName.extract(0, charName.length(), name, sizeof(name), US_INV);
3896
3897    UChar32  theChar = u_charFromName(U_UNICODE_CHAR_NAME, name, fStatus);
3898    if (U_FAILURE(*fStatus)) {
3899        error(U_REGEX_PROPERTY_SYNTAX);
3900    }
3901
3902    nextChar(fC);      // Continue overall regex pattern processing with char after the '}'
3903    return theChar;
3904}
3905
3906//------------------------------------------------------------------------------
3907//
3908//  scanProp   Construct a UnicodeSet from the text at the current scan
3909//             position, which will be of the form \p{whaterver}
3910//
3911//             The scan position will be at the 'p' or 'P'.  On return
3912//             the scan position should be just after the '}'
3913//
3914//             Return a UnicodeSet, constructed from the \P pattern,
3915//             or NULL if the pattern is invalid.
3916//
3917//------------------------------------------------------------------------------
3918UnicodeSet *RegexCompile::scanProp() {
3919    UnicodeSet    *uset = NULL;
3920
3921    if (U_FAILURE(*fStatus)) {
3922        return NULL;
3923    }
3924    U_ASSERT(fC.fChar == chLowerP || fC.fChar == chP);
3925    UBool negated = (fC.fChar == chP);
3926
3927    UnicodeString propertyName;
3928    nextChar(fC);
3929    if (fC.fChar != chLBrace) {
3930        error(U_REGEX_PROPERTY_SYNTAX);
3931        return NULL;
3932    }
3933    for (;;) {
3934        nextChar(fC);
3935        if (fC.fChar == chRBrace) {
3936            break;
3937        }
3938        if (fC.fChar == -1) {
3939            // Hit the end of the input string without finding the closing '}'
3940            error(U_REGEX_PROPERTY_SYNTAX);
3941            return NULL;
3942        }
3943        propertyName.append(fC.fChar);
3944    }
3945    uset = createSetForProperty(propertyName, negated);
3946    nextChar(fC);    // Move input scan to position following the closing '}'
3947    return uset;
3948}
3949
3950//------------------------------------------------------------------------------
3951//
3952//  scanPosixProp   Construct a UnicodeSet from the text at the current scan
3953//             position, which is expected be of the form [:property expression:]
3954//
3955//             The scan position will be at the opening ':'.  On return
3956//             the scan position must be on the closing ']'
3957//
3958//             Return a UnicodeSet constructed from the pattern,
3959//             or NULL if this is not a valid POSIX-style set expression.
3960//             If not a property expression, restore the initial scan position
3961//                (to the opening ':')
3962//
3963//               Note:  the opening '[:' is not sufficient to guarantee that
3964//                      this is a [:property:] expression.
3965//                      [:'+=,] is a perfectly good ordinary set expression that
3966//                              happens to include ':' as one of its characters.
3967//
3968//------------------------------------------------------------------------------
3969UnicodeSet *RegexCompile::scanPosixProp() {
3970    UnicodeSet    *uset = NULL;
3971
3972    if (U_FAILURE(*fStatus)) {
3973        return NULL;
3974    }
3975
3976    U_ASSERT(fC.fChar == chColon);
3977
3978    // Save the scanner state.
3979    // TODO:  move this into the scanner, with the state encapsulated in some way.  Ticket 6062
3980    int64_t     savedScanIndex        = fScanIndex;
3981    int64_t     savedNextIndex        = UTEXT_GETNATIVEINDEX(fRXPat->fPattern);
3982    UBool       savedQuoteMode        = fQuoteMode;
3983    UBool       savedInBackslashQuote = fInBackslashQuote;
3984    UBool       savedEOLComments      = fEOLComments;
3985    int64_t     savedLineNum          = fLineNum;
3986    int64_t     savedCharNum          = fCharNum;
3987    UChar32     savedLastChar         = fLastChar;
3988    UChar32     savedPeekChar         = fPeekChar;
3989    RegexPatternChar savedfC          = fC;
3990
3991    // Scan for a closing ].   A little tricky because there are some perverse
3992    //   edge cases possible.  "[:abc\Qdef:] \E]"  is a valid non-property expression,
3993    //   ending on the second closing ].
3994
3995    UnicodeString propName;
3996    UBool         negated  = FALSE;
3997
3998    // Check for and consume the '^' in a negated POSIX property, e.g.  [:^Letter:]
3999    nextChar(fC);
4000    if (fC.fChar == chUp) {
4001       negated = TRUE;
4002       nextChar(fC);
4003    }
4004
4005    // Scan for the closing ":]", collecting the property name along the way.
4006    UBool  sawPropSetTerminator = FALSE;
4007    for (;;) {
4008        propName.append(fC.fChar);
4009        nextChar(fC);
4010        if (fC.fQuoted || fC.fChar == -1) {
4011            // Escaped characters or end of input - either says this isn't a [:Property:]
4012            break;
4013        }
4014        if (fC.fChar == chColon) {
4015            nextChar(fC);
4016            if (fC.fChar == chRBracket) {
4017                sawPropSetTerminator = TRUE;
4018            }
4019            break;
4020        }
4021    }
4022
4023    if (sawPropSetTerminator) {
4024        uset = createSetForProperty(propName, negated);
4025    }
4026    else
4027    {
4028        // No closing ":]".
4029        //  Restore the original scan position.
4030        //  The main scanner will retry the input as a normal set expression,
4031        //    not a [:Property:] expression.
4032        fScanIndex        = savedScanIndex;
4033        fQuoteMode        = savedQuoteMode;
4034        fInBackslashQuote = savedInBackslashQuote;
4035        fEOLComments      = savedEOLComments;
4036        fLineNum          = savedLineNum;
4037        fCharNum          = savedCharNum;
4038        fLastChar         = savedLastChar;
4039        fPeekChar         = savedPeekChar;
4040        fC                = savedfC;
4041        UTEXT_SETNATIVEINDEX(fRXPat->fPattern, savedNextIndex);
4042    }
4043    return uset;
4044}
4045
4046static inline void addIdentifierIgnorable(UnicodeSet *set, UErrorCode& ec) {
4047    set->add(0, 8).add(0x0e, 0x1b).add(0x7f, 0x9f);
4048    addCategory(set, U_GC_CF_MASK, ec);
4049}
4050
4051//
4052//  Create a Unicode Set from a Unicode Property expression.
4053//     This is common code underlying both \p{...} ane [:...:] expressions.
4054//     Includes trying the Java "properties" that aren't supported as
4055//     normal ICU UnicodeSet properties
4056//
4057static const UChar posSetPrefix[] = {0x5b, 0x5c, 0x70, 0x7b, 0}; // "[\p{"
4058static const UChar negSetPrefix[] = {0x5b, 0x5c, 0x50, 0x7b, 0}; // "[\P{"
4059UnicodeSet *RegexCompile::createSetForProperty(const UnicodeString &propName, UBool negated) {
4060    UnicodeString   setExpr;
4061    UnicodeSet      *set;
4062    uint32_t        usetFlags = 0;
4063
4064    if (U_FAILURE(*fStatus)) {
4065        return NULL;
4066    }
4067
4068    //
4069    //  First try the property as we received it
4070    //
4071    if (negated) {
4072        setExpr.append(negSetPrefix, -1);
4073    } else {
4074        setExpr.append(posSetPrefix, -1);
4075    }
4076    setExpr.append(propName);
4077    setExpr.append(chRBrace);
4078    setExpr.append(chRBracket);
4079    if (fModeFlags & UREGEX_CASE_INSENSITIVE) {
4080        usetFlags |= USET_CASE_INSENSITIVE;
4081    }
4082    set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus);
4083    if (U_SUCCESS(*fStatus)) {
4084       return set;
4085    }
4086    delete set;
4087    set = NULL;
4088
4089    //
4090    //  The property as it was didn't work.
4091
4092    //  Do [:word:]. It is not recognized as a property by UnicodeSet.  "word" not standard POSIX
4093    //     or standard Java, but many other regular expression packages do recognize it.
4094
4095    if (propName.caseCompare(UNICODE_STRING_SIMPLE("word"), 0) == 0) {
4096        *fStatus = U_ZERO_ERROR;
4097        set = new UnicodeSet(*(fRXPat->fStaticSets[URX_ISWORD_SET]));
4098        if (set == NULL) {
4099            *fStatus = U_MEMORY_ALLOCATION_ERROR;
4100            return set;
4101        }
4102        if (negated) {
4103            set->complement();
4104        }
4105        return set;
4106    }
4107
4108
4109    //    Do Java fixes -
4110    //       InGreek -> InGreek or Coptic, that being the official Unicode name for that block.
4111    //       InCombiningMarksforSymbols -> InCombiningDiacriticalMarksforSymbols.
4112    //
4113    //       Note on Spaces:  either "InCombiningMarksForSymbols" or "InCombining Marks for Symbols"
4114    //                        is accepted by Java.  The property part of the name is compared
4115    //                        case-insenstively.  The spaces must be exactly as shown, either
4116    //                        all there, or all omitted, with exactly one at each position
4117    //                        if they are present.  From checking against JDK 1.6
4118    //
4119    //       This code should be removed when ICU properties support the Java  compatibility names
4120    //          (ICU 4.0?)
4121    //
4122    UnicodeString mPropName = propName;
4123    if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InGreek"), 0) == 0) {
4124        mPropName = UNICODE_STRING_SIMPLE("InGreek and Coptic");
4125    }
4126    if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombining Marks for Symbols"), 0) == 0 ||
4127        mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombiningMarksforSymbols"), 0) == 0) {
4128        mPropName = UNICODE_STRING_SIMPLE("InCombining Diacritical Marks for Symbols");
4129    }
4130    else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) {
4131        mPropName = UNICODE_STRING_SIMPLE("javaValidCodePoint");
4132    }
4133
4134    //    See if the property looks like a Java "InBlockName", which
4135    //    we will recast as "Block=BlockName"
4136    //
4137    static const UChar IN[] = {0x49, 0x6E, 0};  // "In"
4138    static const UChar BLOCK[] = {0x42, 0x6C, 0x6f, 0x63, 0x6b, 0x3d, 00};  // "Block="
4139    if (mPropName.startsWith(IN, 2) && propName.length()>=3) {
4140        setExpr.truncate(4);   // Leaves "[\p{", or "[\P{"
4141        setExpr.append(BLOCK, -1);
4142        setExpr.append(UnicodeString(mPropName, 2));  // Property with the leading "In" removed.
4143        setExpr.append(chRBrace);
4144        setExpr.append(chRBracket);
4145        *fStatus = U_ZERO_ERROR;
4146        set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus);
4147        if (U_SUCCESS(*fStatus)) {
4148            return set;
4149        }
4150        delete set;
4151        set = NULL;
4152    }
4153
4154    if (propName.startsWith(UNICODE_STRING_SIMPLE("java")) ||
4155        propName.compare(UNICODE_STRING_SIMPLE("all")) == 0)
4156    {
4157        UErrorCode localStatus = U_ZERO_ERROR;
4158        //setExpr.remove();
4159        set = new UnicodeSet();
4160        //
4161        //  Try the various Java specific properties.
4162        //   These all begin with "java"
4163        //
4164        if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDefined")) == 0) {
4165            addCategory(set, U_GC_CN_MASK, localStatus);
4166            set->complement();
4167        }
4168        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDigit")) == 0) {
4169            addCategory(set, U_GC_ND_MASK, localStatus);
4170        }
4171        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaIdentifierIgnorable")) == 0) {
4172            addIdentifierIgnorable(set, localStatus);
4173        }
4174        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaISOControl")) == 0) {
4175            set->add(0, 0x1F).add(0x7F, 0x9F);
4176        }
4177        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierPart")) == 0) {
4178            addCategory(set, U_GC_L_MASK, localStatus);
4179            addCategory(set, U_GC_SC_MASK, localStatus);
4180            addCategory(set, U_GC_PC_MASK, localStatus);
4181            addCategory(set, U_GC_ND_MASK, localStatus);
4182            addCategory(set, U_GC_NL_MASK, localStatus);
4183            addCategory(set, U_GC_MC_MASK, localStatus);
4184            addCategory(set, U_GC_MN_MASK, localStatus);
4185            addIdentifierIgnorable(set, localStatus);
4186        }
4187        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierStart")) == 0) {
4188            addCategory(set, U_GC_L_MASK, localStatus);
4189            addCategory(set, U_GC_NL_MASK, localStatus);
4190            addCategory(set, U_GC_SC_MASK, localStatus);
4191            addCategory(set, U_GC_PC_MASK, localStatus);
4192        }
4193        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetter")) == 0) {
4194            addCategory(set, U_GC_L_MASK, localStatus);
4195        }
4196        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetterOrDigit")) == 0) {
4197            addCategory(set, U_GC_L_MASK, localStatus);
4198            addCategory(set, U_GC_ND_MASK, localStatus);
4199        }
4200        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLowerCase")) == 0) {
4201            addCategory(set, U_GC_LL_MASK, localStatus);
4202        }
4203        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaMirrored")) == 0) {
4204            set->applyIntPropertyValue(UCHAR_BIDI_MIRRORED, 1, localStatus);
4205        }
4206        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSpaceChar")) == 0) {
4207            addCategory(set, U_GC_Z_MASK, localStatus);
4208        }
4209        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSupplementaryCodePoint")) == 0) {
4210            set->add(0x10000, UnicodeSet::MAX_VALUE);
4211        }
4212        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaTitleCase")) == 0) {
4213            addCategory(set, U_GC_LT_MASK, localStatus);
4214        }
4215        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierStart")) == 0) {
4216            addCategory(set, U_GC_L_MASK, localStatus);
4217            addCategory(set, U_GC_NL_MASK, localStatus);
4218        }
4219        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierPart")) == 0) {
4220            addCategory(set, U_GC_L_MASK, localStatus);
4221            addCategory(set, U_GC_PC_MASK, localStatus);
4222            addCategory(set, U_GC_ND_MASK, localStatus);
4223            addCategory(set, U_GC_NL_MASK, localStatus);
4224            addCategory(set, U_GC_MC_MASK, localStatus);
4225            addCategory(set, U_GC_MN_MASK, localStatus);
4226            addIdentifierIgnorable(set, localStatus);
4227        }
4228        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUpperCase")) == 0) {
4229            addCategory(set, U_GC_LU_MASK, localStatus);
4230        }
4231        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaValidCodePoint")) == 0) {
4232            set->add(0, UnicodeSet::MAX_VALUE);
4233        }
4234        else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaWhitespace")) == 0) {
4235            addCategory(set, U_GC_Z_MASK, localStatus);
4236            set->removeAll(UnicodeSet().add(0xa0).add(0x2007).add(0x202f));
4237            set->add(9, 0x0d).add(0x1c, 0x1f);
4238        }
4239        else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) {
4240            set->add(0, UnicodeSet::MAX_VALUE);
4241        }
4242
4243        if (U_SUCCESS(localStatus) && !set->isEmpty()) {
4244            *fStatus = U_ZERO_ERROR;
4245            if (usetFlags & USET_CASE_INSENSITIVE) {
4246                set->closeOver(USET_CASE_INSENSITIVE);
4247            }
4248            if (negated) {
4249                set->complement();
4250            }
4251            return set;
4252        }
4253        delete set;
4254        set = NULL;
4255    }
4256    error(*fStatus);
4257    return NULL;
4258}
4259
4260
4261
4262//
4263//  SetEval   Part of the evaluation of [set expressions].
4264//            Perform any pending (stacked) operations with precedence
4265//            equal or greater to that of the next operator encountered
4266//            in the expression.
4267//
4268void RegexCompile::setEval(int32_t nextOp) {
4269    UnicodeSet *rightOperand = NULL;
4270    UnicodeSet *leftOperand  = NULL;
4271    for (;;) {
4272        U_ASSERT(fSetOpStack.empty()==FALSE);
4273        int32_t pendingSetOperation = fSetOpStack.peeki();
4274        if ((pendingSetOperation&0xffff0000) < (nextOp&0xffff0000)) {
4275            break;
4276        }
4277        fSetOpStack.popi();
4278        U_ASSERT(fSetStack.empty() == FALSE);
4279        rightOperand = (UnicodeSet *)fSetStack.peek();
4280        switch (pendingSetOperation) {
4281            case setNegation:
4282                rightOperand->complement();
4283                break;
4284            case setCaseClose:
4285                // TODO: need a simple close function.  Ticket 6065
4286                rightOperand->closeOver(USET_CASE_INSENSITIVE);
4287                rightOperand->removeAllStrings();
4288                break;
4289            case setDifference1:
4290            case setDifference2:
4291                fSetStack.pop();
4292                leftOperand = (UnicodeSet *)fSetStack.peek();
4293                leftOperand->removeAll(*rightOperand);
4294                delete rightOperand;
4295                break;
4296            case setIntersection1:
4297            case setIntersection2:
4298                fSetStack.pop();
4299                leftOperand = (UnicodeSet *)fSetStack.peek();
4300                leftOperand->retainAll(*rightOperand);
4301                delete rightOperand;
4302                break;
4303            case setUnion:
4304                fSetStack.pop();
4305                leftOperand = (UnicodeSet *)fSetStack.peek();
4306                leftOperand->addAll(*rightOperand);
4307                delete rightOperand;
4308                break;
4309            default:
4310                U_ASSERT(FALSE);
4311                break;
4312            }
4313        }
4314    }
4315
4316void RegexCompile::setPushOp(int32_t op) {
4317    setEval(op);
4318    fSetOpStack.push(op, *fStatus);
4319    fSetStack.push(new UnicodeSet(), *fStatus);
4320}
4321
4322U_NAMESPACE_END
4323#endif  // !UCONFIG_NO_REGULAR_EXPRESSIONS
4324
4325