]> git.saurik.com Git - apple/icu.git/blobdiff - icuSources/common/rbbiscan.cpp
ICU-59180.0.1.tar.gz
[apple/icu.git] / icuSources / common / rbbiscan.cpp
index 75c69f525e9b67596bec6cbec68e4692d0b73f2e..e68e0529d02e33974ac79e4d72850c7b7de004d2 100644 (file)
@@ -1,8 +1,9 @@
-
+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
 //
 //  file:  rbbiscan.cpp
 //
-//  Copyright (C) 2002-2006, International Business Machines Corporation and others.
+//  Copyright (C) 2002-2016, International Business Machines Corporation and others.
 //  All Rights Reserved.
 //
 //  This file contains the Rule Based Break Iterator Rule Builder functions for
@@ -23,7 +24,6 @@
 #include "unicode/uchriter.h"
 #include "unicode/parsepos.h"
 #include "unicode/parseerr.h"
-#include "uprops.h"
 #include "cmemory.h"
 #include "cstring.h"
 
@@ -36,8 +36,7 @@
 
 #include "uassert.h"
 
-
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 // Unicode Set init strings for each of the character classes needed for parsing a rule file.
 //               (Initialized with hex values for portability to EBCDIC based machines.
@@ -46,7 +45,7 @@
 //              The sets are referred to by name in the rbbirpt.txt, which is the
 //              source form of the state transition table for the RBBI rule parser.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 static const UChar gRuleSet_rule_char_pattern[]       = {
  //   [    ^      [    \     p     {      Z     }     \     u    0      0    2      0
     0x5b, 0x5e, 0x5b, 0x5c, 0x70, 0x7b, 0x5a, 0x7d, 0x5c, 0x75, 0x30, 0x30, 0x32, 0x30,
@@ -72,7 +71,7 @@ static const UChar kAny[] = {0x61, 0x6e, 0x79, 0x00};  // "any"
 
 U_CDECL_BEGIN
 static void U_CALLCONV RBBISetTable_deleter(void *p) {
-    RBBISetTableEl *px = (RBBISetTableEl *)p;
+    icu::RBBISetTableEl *px = (icu::RBBISetTableEl *)p;
     delete px->key;
     // Note:  px->val is owned by the linked list "fSetsListHead" in scanner.
     //        Don't delete the value nodes here.
@@ -82,37 +81,35 @@ U_CDECL_END
 
 U_NAMESPACE_BEGIN
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  Constructor.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 RBBIRuleScanner::RBBIRuleScanner(RBBIRuleBuilder *rb)
 {
     fRB                 = rb;
+    fScanIndex          = 0;
+    fNextIndex          = 0;
+    fQuoteMode          = FALSE;
+    fLineNum            = 1;
+    fCharNum            = 0;
+    fLastChar           = 0;
+    
+    fStateTable         = NULL;
+    fStack[0]           = 0;
     fStackPtr           = 0;
-    fStack[fStackPtr]   = 0;
-    fNodeStackPtr       = 0;
-    fRuleNum            = 0;
     fNodeStack[0]       = NULL;
-
-    fRuleSets[kRuleSet_rule_char-128]       = NULL;
-    fRuleSets[kRuleSet_white_space-128]     = NULL;
-    fRuleSets[kRuleSet_name_char-128]       = NULL;
-    fRuleSets[kRuleSet_name_start_char-128] = NULL;
-    fRuleSets[kRuleSet_digit_char-128]      = NULL;
-    fSymbolTable                            = NULL;
-    fSetTable                               = NULL;
-
-    fScanIndex = 0;
-    fNextIndex = 0;
+    fNodeStackPtr       = 0;
 
     fReverseRule        = FALSE;
     fLookAheadRule      = FALSE;
+    fNoChainInRule      = FALSE;
 
-    fLineNum    = 1;
-    fCharNum    = 0;
-    fQuoteMode  = FALSE;
+    fSymbolTable        = NULL;
+    fSetTable           = NULL;
+    fRuleNum            = 0;
+    fOptionStart        = 0;
 
     // Do not check status until after all critical fields are sufficiently initialized
     //   that the destructor can run cleanly.
@@ -126,14 +123,20 @@ RBBIRuleScanner::RBBIRuleScanner(RBBIRuleBuilder *rb)
     //            all instances of RBBIRuleScanners.  BUT this is quite a bit simpler,
     //            and the time to build these few sets should be small compared to a
     //            full break iterator build.
-    fRuleSets[kRuleSet_rule_char-128]       = new UnicodeSet(gRuleSet_rule_char_pattern,       *rb->fStatus);
-    fRuleSets[kRuleSet_white_space-128]     = (UnicodeSet*) uprv_openRuleWhiteSpaceSet(rb->fStatus);
-    fRuleSets[kRuleSet_name_char-128]       = new UnicodeSet(gRuleSet_name_char_pattern,       *rb->fStatus);
-    fRuleSets[kRuleSet_name_start_char-128] = new UnicodeSet(gRuleSet_name_start_char_pattern, *rb->fStatus);
-    fRuleSets[kRuleSet_digit_char-128]      = new UnicodeSet(gRuleSet_digit_char_pattern,      *rb->fStatus);
+    fRuleSets[kRuleSet_rule_char-128]
+        = UnicodeSet(UnicodeString(gRuleSet_rule_char_pattern),       *rb->fStatus);
+    // fRuleSets[kRuleSet_white_space-128] = [:Pattern_White_Space:]
+    fRuleSets[kRuleSet_white_space-128].
+        add(9, 0xd).add(0x20).add(0x85).add(0x200e, 0x200f).add(0x2028, 0x2029);
+    fRuleSets[kRuleSet_name_char-128]
+        = UnicodeSet(UnicodeString(gRuleSet_name_char_pattern),       *rb->fStatus);
+    fRuleSets[kRuleSet_name_start_char-128]
+        = UnicodeSet(UnicodeString(gRuleSet_name_start_char_pattern), *rb->fStatus);
+    fRuleSets[kRuleSet_digit_char-128]
+        = UnicodeSet(UnicodeString(gRuleSet_digit_char_pattern),      *rb->fStatus);
     if (*rb->fStatus == U_ILLEGAL_ARGUMENT_ERROR) {
         // This case happens if ICU's data is missing.  UnicodeSet tries to look up property
-        //   names from the init string, can't find them, and claims an illegal arguement.
+        //   names from the init string, can't find them, and claims an illegal argument.
         //   Change the error so that the actual problem will be clearer to users.
         *rb->fStatus = U_BRK_INIT_ERROR;
     }
@@ -142,24 +145,25 @@ RBBIRuleScanner::RBBIRuleScanner(RBBIRuleBuilder *rb)
     }
 
     fSymbolTable = new RBBISymbolTable(this, rb->fRules, *rb->fStatus);
+    if (fSymbolTable == NULL) {
+        *rb->fStatus = U_MEMORY_ALLOCATION_ERROR;
+        return;
+    }
     fSetTable    = uhash_open(uhash_hashUnicodeString, uhash_compareUnicodeString, NULL, rb->fStatus);
+    if (U_FAILURE(*rb->fStatus)) {
+        return;
+    }
     uhash_setValueDeleter(fSetTable, RBBISetTable_deleter);
 }
 
 
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  Destructor
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 RBBIRuleScanner::~RBBIRuleScanner() {
-    delete fRuleSets[kRuleSet_rule_char-128];
-    delete fRuleSets[kRuleSet_white_space-128];
-    delete fRuleSets[kRuleSet_name_char-128];
-    delete fRuleSets[kRuleSet_name_start_char-128];
-    delete fRuleSets[kRuleSet_digit_char-128];
-
     delete fSymbolTable;
     if (fSetTable != NULL) {
          uhash_close(fSetTable);
@@ -178,7 +182,7 @@ RBBIRuleScanner::~RBBIRuleScanner() {
 
 }
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  doParseAction        Do some action during rule parsing.
 //                       Called by the parse state machine.
@@ -191,14 +195,14 @@ RBBIRuleScanner::~RBBIRuleScanner() {
 //                              in some compilers, while at the same time avoiding multiple
 //                              definitions problems.  I'm sure that there's a better way.
 //
-//----------------------------------------------------------------------------------------
-UBool RBBIRuleScanner::doParseActions(EParseAction action)
+//------------------------------------------------------------------------------
+UBool RBBIRuleScanner::doParseActions(int32_t action)
 {
     RBBINode *n       = NULL;
 
     UBool   returnVal = TRUE;
 
-    switch ((RBBI_RuleParseAction)action) {
+    switch (action) {
 
     case doExprStart:
         pushNewNode(RBBINode::opStart);
@@ -206,11 +210,20 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         break;
 
 
+    case doNoChain:
+        // Scanned a '^' while on the rule start state.
+        fNoChainInRule = TRUE;
+        break;
+
+
     case doExprOrOperator:
         {
             fixOpStack(RBBINode::precOpCat);
             RBBINode  *operandNode = fNodeStack[fNodeStackPtr--];
             RBBINode  *orNode      = pushNewNode(RBBINode::opOr);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             orNode->fLeftChild     = operandNode;
             operandNode->fParent   = orNode;
         }
@@ -225,6 +238,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
             fixOpStack(RBBINode::precOpCat);
             RBBINode  *operandNode = fNodeStack[fNodeStackPtr--];
             RBBINode  *catNode     = pushNewNode(RBBINode::opCat);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             catNode->fLeftChild    = operandNode;
             operandNode->fParent   = catNode;
         }
@@ -313,21 +329,40 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "rtree")) {printNodeStack("end of rule");}
 #endif
         U_ASSERT(fNodeStackPtr == 1);
+        RBBINode *thisRule = fNodeStack[fNodeStackPtr];
 
         // If this rule includes a look-ahead '/', add a endMark node to the
         //   expression tree.
         if (fLookAheadRule) {
-            RBBINode  *thisRule       = fNodeStack[fNodeStackPtr];
             RBBINode  *endNode        = pushNewNode(RBBINode::endMark);
             RBBINode  *catNode        = pushNewNode(RBBINode::opCat);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             fNodeStackPtr -= 2;
             catNode->fLeftChild       = thisRule;
             catNode->fRightChild      = endNode;
             fNodeStack[fNodeStackPtr] = catNode;
             endNode->fVal             = fRuleNum;
             endNode->fLookAheadEnd    = TRUE;
+            thisRule                  = catNode;
+
+            // TODO: Disable chaining out of look-ahead (hard break) rules.
+            //   The break on rule match is forced, so there is no point in building up
+            //   the state table to chain into another rule for a longer match.
         }
 
+        // Mark this node as being the root of a rule.
+        thisRule->fRuleRoot = TRUE;
+
+        // Flag if chaining into this rule is wanted.
+        //    
+        if (fRB->fChainRules &&         // If rule chaining is enabled globally via !!chain
+                !fNoChainInRule) {      //     and no '^' chain-in inhibit was on this rule
+            thisRule->fChainIn = TRUE;
+        }
+
+
         // All rule expressions are ORed together.
         // The ';' that terminates an expression really just functions as a '|' with
         //   a low operator prededence.
@@ -347,6 +382,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
             RBBINode  *thisRule    = fNodeStack[fNodeStackPtr];
             RBBINode  *prevRules   = *destRules;
             RBBINode  *orNode      = pushNewNode(RBBINode::opOr);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             orNode->fLeftChild     = prevRules;
             prevRules->fParent     = orNode;
             orNode->fRightChild    = thisRule;
@@ -361,6 +399,7 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         }
         fReverseRule   = FALSE;   // in preparation for the next rule.
         fLookAheadRule = FALSE;
+        fNoChainInRule = FALSE;
         fNodeStackPtr  = 0;
         }
         break;
@@ -387,6 +426,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         {
             RBBINode  *operandNode = fNodeStack[fNodeStackPtr--];
             RBBINode  *plusNode    = pushNewNode(RBBINode::opPlus);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             plusNode->fLeftChild   = operandNode;
             operandNode->fParent   = plusNode;
         }
@@ -396,6 +438,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         {
             RBBINode  *operandNode = fNodeStack[fNodeStackPtr--];
             RBBINode  *qNode       = pushNewNode(RBBINode::opQuestion);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             qNode->fLeftChild      = operandNode;
             operandNode->fParent   = qNode;
         }
@@ -405,6 +450,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         {
             RBBINode  *operandNode = fNodeStack[fNodeStackPtr--];
             RBBINode  *starNode    = pushNewNode(RBBINode::opStar);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
             starNode->fLeftChild   = operandNode;
             operandNode->fParent   = starNode;
         }
@@ -418,7 +466,10 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         // sets that just happen to contain only one character.
         {
             n = pushNewNode(RBBINode::setRef);
-            findSetFor(fC.fChar, n);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
+            findSetFor(UnicodeString(fC.fChar), n);
             n->fFirstPos = fScanIndex;
             n->fLastPos  = fNextIndex;
             fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText);
@@ -429,7 +480,10 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         // scanned a ".", meaning match any single character.
         {
             n = pushNewNode(RBBINode::setRef);
-            findSetFor(kAny, n);
+            if (U_FAILURE(*fRB->fStatus)) {
+                break;
+            }
+            findSetFor(UnicodeString(TRUE, kAny, 3), n);
             n->fFirstPos = fScanIndex;
             n->fLastPos  = fNextIndex;
             fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText);
@@ -439,6 +493,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
     case doSlash:
         // Scanned a '/', which identifies a look-ahead break position in a rule.
         n = pushNewNode(RBBINode::lookAhead);
+        if (U_FAILURE(*fRB->fStatus)) {
+            break;
+        }
         n->fVal      = fRuleNum;
         n->fFirstPos = fScanIndex;
         n->fLastPos  = fNextIndex;
@@ -450,6 +507,9 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
     case doStartTagValue:
         // Scanned a '{', the opening delimiter for a tag value within a rule.
         n = pushNewNode(RBBINode::tag);
+        if (U_FAILURE(*fRB->fStatus)) {
+            break;
+        }
         n->fVal      = 0;
         n->fFirstPos = fScanIndex;
         n->fLastPos  = fNextIndex;
@@ -560,32 +620,34 @@ UBool RBBIRuleScanner::doParseActions(EParseAction action)
         returnVal = FALSE;
         break;
     }
-    return returnVal;
+    return returnVal && U_SUCCESS(*fRB->fStatus);
 }
 
 
 
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  Error         Report a rule parse error.
 //                Only report it if no previous error has been recorded.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 void RBBIRuleScanner::error(UErrorCode e) {
     if (U_SUCCESS(*fRB->fStatus)) {
         *fRB->fStatus = e;
-        fRB->fParseError->line  = fLineNum;
-        fRB->fParseError->offset = fCharNum;
-        fRB->fParseError->preContext[0] = 0;
-        fRB->fParseError->preContext[0] = 0;
+        if (fRB->fParseError) {
+            fRB->fParseError->line  = fLineNum;
+            fRB->fParseError->offset = fCharNum;
+            fRB->fParseError->preContext[0] = 0;
+            fRB->fParseError->postContext[0] = 0;
+        }
     }
 }
 
 
 
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  fixOpStack   The parse stack holds partially assembled chunks of the parse tree.
 //               An entry on the stack may be as small as a single setRef node,
@@ -599,7 +661,7 @@ void RBBIRuleScanner::error(UErrorCode e) {
 //               the precedence of the current operator, binds the operand left,
 //               to the previously encountered operator.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 void RBBIRuleScanner::fixOpStack(RBBINode::OpPrecedence p) {
     RBBINode *n;
     // printNodeStack("entering fixOpStack()");
@@ -646,7 +708,7 @@ void RBBIRuleScanner::fixOpStack(RBBINode::OpPrecedence p) {
 
 
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //   findSetFor    given a UnicodeString,
 //                  - find the corresponding Unicode Set  (uset node)
@@ -661,7 +723,7 @@ void RBBIRuleScanner::fixOpStack(RBBINode::OpPrecedence p) {
 //                    just one element which is the char in question.
 //                 If the string is "any", return a set containing all chars.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 void RBBIRuleScanner::findSetFor(const UnicodeString &s, RBBINode *node, UnicodeSet *setToAdopt) {
 
     RBBISetTableEl   *el;
@@ -695,6 +757,10 @@ void RBBIRuleScanner::findSetFor(const UnicodeString &s, RBBINode *node, Unicode
     // This new uset node becomes the child of the caller's setReference node.
     //
     RBBINode *usetNode    = new RBBINode(RBBINode::uset);
+    if (usetNode == NULL) {
+        error(U_MEMORY_ALLOCATION_ERROR);
+        return;
+    }
     usetNode->fInputSet   = setToAdopt;
     usetNode->fParent     = node;
     node->fLeftChild      = usetNode;
@@ -713,6 +779,14 @@ void RBBIRuleScanner::findSetFor(const UnicodeString &s, RBBINode *node, Unicode
     el      = (RBBISetTableEl *)uprv_malloc(sizeof(RBBISetTableEl));
     UnicodeString *tkey = new UnicodeString(s);
     if (tkey == NULL || el == NULL || setToAdopt == NULL) {
+        // Delete to avoid memory leak
+        delete tkey;
+        tkey = NULL;
+        uprv_free(el);
+        el = NULL;
+        delete setToAdopt;
+        setToAdopt = NULL;
+
         error(U_MEMORY_ALLOCATION_ERROR);
         return;
     }
@@ -741,12 +815,12 @@ static const UChar      chLParen    = 0x28;
 static const UChar      chRParen    = 0x29;
 
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  stripRules    Return a rules string without unnecessary
 //                characters.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 UnicodeString RBBIRuleScanner::stripRules(const UnicodeString &rules) {
     UnicodeString strippedRules;
     int rulesLength = rules.length();
@@ -768,13 +842,13 @@ UnicodeString RBBIRuleScanner::stripRules(const UnicodeString &rules) {
 }
 
 
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  nextCharLL    Low Level Next Char from rule input source.
 //                Get a char from the input character iterator,
 //                keep track of input position for error reporting.
 //
-//----------------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 UChar32  RBBIRuleScanner::nextCharLL() {
     UChar32  ch;
 
@@ -787,7 +861,7 @@ UChar32  RBBIRuleScanner::nextCharLL() {
     if (ch == chCR ||
         ch == chNEL ||
         ch == chLS   ||
-        ch == chLF && fLastChar != chCR) {
+        (ch == chLF && fLastChar != chCR)) {
         // Character is starting a new line.  Bump up the line number, and
         //  reset the column to 0.
         fLineNum++;
@@ -809,13 +883,13 @@ UChar32  RBBIRuleScanner::nextCharLL() {
 }
 
 
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //   nextChar     for rules scanning.  At this level, we handle stripping
 //                out comments and processing backslash character escapes.
 //                The rest of the rules grammar is handled at the next level up.
 //
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 void RBBIRuleScanner::nextChar(RBBIRuleChar &c) {
 
     // Unicode Character constants needed for the processing done by nextChar(),
@@ -893,14 +967,14 @@ void RBBIRuleScanner::nextChar(RBBIRuleChar &c) {
     // putc(c.fChar, stdout);
 }
 
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  Parse RBBI rules.   The state machine for rules parsing is here.
 //                      The state tables are hand-written in the file rbbirpt.txt,
 //                      and converted to the form used here by a perl
 //                      script rbbicst.pl
 //
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 void RBBIRuleScanner::parse() {
     uint16_t                state;
     const RBBIRuleTableEl  *tableEl;
@@ -948,7 +1022,7 @@ void RBBIRuleScanner::parse() {
 
         for (;;) {
             #ifdef RBBI_DEBUG
-                if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "scan")) { RBBIDebugPrintf(".");}
+                if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "scan")) { RBBIDebugPrintf("."); fflush(stdout);}
             #endif
             if (tableEl->fCharClass < 127 && fC.fEscaped == FALSE &&   tableEl->fCharClass == fC.fChar) {
                 // Table row specified an individual character, not a set, and
@@ -977,8 +1051,8 @@ void RBBIRuleScanner::parse() {
             if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 &&   // Table specs a char class &&
                 fC.fEscaped == FALSE &&                                      //   char is not escaped &&
                 fC.fChar != (UChar32)-1) {                                   //   char is not EOF
-                UnicodeSet *uniset = fRuleSets[tableEl->fCharClass-128];
-                if (uniset->contains(fC.fChar)) {
+                U_ASSERT((tableEl->fCharClass-128) < UPRV_LENGTHOF(fRuleSets));
+                if (fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) {
                     // Table row specified a character class, or set of characters,
                     //   and the current char matches it.
                     break;
@@ -994,7 +1068,7 @@ void RBBIRuleScanner::parse() {
         // We've found the row of the state table that matches the current input
         //   character from the rules string.
         // Perform any action specified  by this row in the state table.
-        if (doParseActions((EParseAction)tableEl->fAction) == FALSE) {
+        if (doParseActions((int32_t)tableEl->fAction) == FALSE) {
             // Break out of the state machine loop if the
             //   the action signalled some kind of error, or
             //   the action was to exit, occurs on normal end-of-rules-input.
@@ -1031,13 +1105,27 @@ void RBBIRuleScanner::parse() {
 
     }
 
+    if (U_FAILURE(*fRB->fStatus)) {
+        return;
+    }
+    
+    // If there are no forward rules set an error.
+    //
+    if (fRB->fForwardTree == NULL) {
+        error(U_BRK_RULE_SYNTAX);
+        return;
+    }
+
     //
     // If there were NO user specified reverse rules, set up the equivalent of ".*;"
     //
     if (fRB->fReverseTree == NULL) {
         fRB->fReverseTree  = pushNewNode(RBBINode::opStar);
         RBBINode  *operand = pushNewNode(RBBINode::setRef);
-        findSetFor(kAny, operand);
+        if (U_FAILURE(*fRB->fStatus)) {
+            return;
+        }
+        findSetFor(UnicodeString(TRUE, kAny, 3), operand);
         fRB->fReverseTree->fLeftChild = operand;
         operand->fParent              = fRB->fReverseTree;
         fNodeStackPtr -= 2;
@@ -1051,51 +1139,52 @@ void RBBIRuleScanner::parse() {
     //
 #ifdef RBBI_DEBUG
     if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "symbols")) {fSymbolTable->rbbiSymtablePrint();}
-    if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ptree"))
-    {
+    if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ptree")) {
         RBBIDebugPrintf("Completed Forward Rules Parse Tree...\n");
-        fRB->fForwardTree->printTree(TRUE);
+        RBBINode::printTree(fRB->fForwardTree, TRUE);
         RBBIDebugPrintf("\nCompleted Reverse Rules Parse Tree...\n");
-        fRB->fReverseTree->printTree(TRUE);
+        RBBINode::printTree(fRB->fReverseTree, TRUE);
         RBBIDebugPrintf("\nCompleted Safe Point Forward Rules Parse Tree...\n");
-        fRB->fSafeFwdTree->printTree(TRUE);
+        RBBINode::printTree(fRB->fSafeFwdTree, TRUE);
         RBBIDebugPrintf("\nCompleted Safe Point Reverse Rules Parse Tree...\n");
-        fRB->fSafeRevTree->printTree(TRUE);
+        RBBINode::printTree(fRB->fSafeRevTree, TRUE);
     }
 #endif
 }
 
 
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  printNodeStack     for debugging...
 //
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 #ifdef RBBI_DEBUG
 void RBBIRuleScanner::printNodeStack(const char *title) {
     int i;
     RBBIDebugPrintf("%s.  Dumping node stack...\n", title);
-    for (i=fNodeStackPtr; i>0; i--) {fNodeStack[i]->printTree(TRUE);}
+    for (i=fNodeStackPtr; i>0; i--) {RBBINode::printTree(fNodeStack[i], TRUE);}
 }
 #endif
 
 
 
 
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  pushNewNode   create a new RBBINode of the specified type and push it
 //                onto the stack of nodes.
 //
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 RBBINode  *RBBIRuleScanner::pushNewNode(RBBINode::NodeType  t) {
-    fNodeStackPtr++;
-    if (fNodeStackPtr >= kStackSize) {
-        error(U_BRK_INTERNAL_ERROR);
+    if (U_FAILURE(*fRB->fStatus)) {
+        return NULL;
+    }
+    if (fNodeStackPtr >= kStackSize - 1) {
+        error(U_BRK_RULE_SYNTAX);
         RBBIDebugPuts("RBBIRuleScanner::pushNewNode - stack overflow.");
-        *fRB->fStatus = U_BRK_INTERNAL_ERROR;
         return NULL;
     }
+    fNodeStackPtr++;
     fNodeStack[fNodeStackPtr] = new RBBINode(t);
     if (fNodeStack[fNodeStackPtr] == NULL) {
         *fRB->fStatus = U_MEMORY_ALLOCATION_ERROR;
@@ -1105,7 +1194,7 @@ RBBINode  *RBBIRuleScanner::pushNewNode(RBBINode::NodeType  t) {
 
 
 
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 //
 //  scanSet    Construct a UnicodeSet from the text at the current scan
 //             position.  Advance the scan position to the first character
@@ -1118,7 +1207,7 @@ RBBINode  *RBBIRuleScanner::pushNewNode(RBBINode::NodeType  t) {
 //             that controls rule parsing.  UnicodeSets, however, are parsed by
 //             the UnicodeSet constructor, not by the RBBI rule parser.
 //
-//---------------------------------------------------------------------------------
+//------------------------------------------------------------------------------
 void RBBIRuleScanner::scanSet() {
     UnicodeSet    *uset;
     ParsePosition  pos;
@@ -1132,9 +1221,12 @@ void RBBIRuleScanner::scanSet() {
     pos.setIndex(fScanIndex);
     startPos = fScanIndex;
     UErrorCode localStatus = U_ZERO_ERROR;
-    uset = new UnicodeSet(fRB->fRules, pos, USET_IGNORE_SPACE,
-                         fSymbolTable,
-                         localStatus);
+    uset = new UnicodeSet();
+    if (uset == NULL) {
+        localStatus = U_MEMORY_ALLOCATION_ERROR;
+    } else {
+        uset->applyPatternIgnoreSpace(fRB->fRules, pos, fSymbolTable, localStatus);
+    }
     if (U_FAILURE(localStatus)) {
         //  TODO:  Get more accurate position of the error from UnicodeSet's return info.
         //         UnicodeSet appears to not be reporting correctly at this time.
@@ -1148,6 +1240,7 @@ void RBBIRuleScanner::scanSet() {
 
     // Verify that the set contains at least one code point.
     //
+    U_ASSERT(uset!=NULL);
     if (uset->isEmpty()) {
         // This set is empty.
         //  Make it an error, because it almost certainly is not what the user wanted.
@@ -1174,6 +1267,9 @@ void RBBIRuleScanner::scanSet() {
         RBBINode         *n;
 
         n = pushNewNode(RBBINode::setRef);
+        if (U_FAILURE(*fRB->fStatus)) {
+            return;
+        }
         n->fFirstPos = startPos;
         n->fLastPos  = fNextIndex;
         fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText);