//------------------------------------------------------------------------
//
// printRanges A debugging function.
// dump out all of the range definitions.
//
//------------------------------------------------------------------------
///CLOVER:OFF
internal virtual void PrintRanges()
{
RangeDescriptor rlRange;
int i;
Console.Out.Write("\n\n Nonoverlapping Ranges ...\n");
for (rlRange = fRangeList; rlRange != null; rlRange = rlRange.fNext)
{
Console.Out.Write(" " + rlRange.fNum + " " + rlRange.fStartChar + "-" + rlRange.fEndChar);
for (i = 0; i < rlRange.fIncludesSets.Count; i++)
{
RBBINode usetNode = rlRange.fIncludesSets[i];
String setName = "anon";
RBBINode setRef = usetNode.fParent;
if (setRef != null)
{
RBBINode varRef = setRef.fParent;
if (varRef != null && varRef.fType == RBBINode.varRef)
{
setName = varRef.fText;
}
}
Console.Out.Write(setName); Console.Out.Write(" ");
}
Console.Out.WriteLine("");
}
}
//-----------------------------------------------------------------------------
//
// calcFollowPos. Impossible to explain succinctly. See Aho, section 3.9
//
//-----------------------------------------------------------------------------
internal virtual void CalcFollowPos(RBBINode n)
{
if (n == null ||
n.fType == RBBINode.leafChar ||
n.fType == RBBINode.endMark)
{
return;
}
CalcFollowPos(n.fLeftChild);
CalcFollowPos(n.fRightChild);
// Aho rule #1
if (n.fType == RBBINode.opCat)
{
foreach (RBBINode i /* is 'i' in Aho's description */ in n.fLeftChild.fLastPosSet)
{
i.fFollowPos.UnionWith(n.fRightChild.fFirstPosSet);
}
}
// Aho rule #2
if (n.fType == RBBINode.opStar ||
n.fType == RBBINode.opPlus)
{
foreach (RBBINode i /* again, n and i are the names from Aho's description */ in n.fLastPosSet)
{
i.fFollowPos.UnionWith(n.fFirstPosSet);
}
}
}
//CLOVER:ON
//------------------------------------------------------------------------
//
// printRangeGroups A debugging function.
// dump out all of the range groups.
//
//------------------------------------------------------------------------
///CLOVER:OFF
internal virtual void PrintRangeGroups()
{
RangeDescriptor rlRange;
RangeDescriptor tRange;
int i;
int lastPrintedGroupNum = 0;
Console.Out.Write("\nRanges grouped by Unicode Set Membership...\n");
for (rlRange = fRangeList; rlRange != null; rlRange = rlRange.fNext)
{
int groupNum = rlRange.fNum & 0xbfff;
if (groupNum > lastPrintedGroupNum)
{
lastPrintedGroupNum = groupNum;
if (groupNum < 10)
{
Console.Out.Write(" ");
}
Console.Out.Write(groupNum + " ");
if ((rlRange.fNum & 0x4000) != 0)
{
Console.Out.Write(" <DICT> ");
}
for (i = 0; i < rlRange.fIncludesSets.Count; i++)
{
RBBINode usetNode = rlRange.fIncludesSets[i];
String setName = "anon";
RBBINode setRef = usetNode.fParent;
if (setRef != null)
{
RBBINode varRef = setRef.fParent;
if (varRef != null && varRef.fType == RBBINode.varRef)
{
setName = varRef.fText;
}
}
Console.Out.Write(setName); Console.Out.Write(" ");
}
i = 0;
for (tRange = rlRange; tRange != null; tRange = tRange.fNext)
{
if (tRange.fNum == rlRange.fNum)
{
if (i++ % 5 == 0)
{
Console.Out.Write("\n ");
}
RBBINode.PrintHex(tRange.fStartChar, -1);
Console.Out.Write("-");
RBBINode.PrintHex(tRange.fEndChar, 0);
}
}
Console.Out.Write("\n");
}
}
Console.Out.Write("\n");
}
//-------------------------------------------------------------------------
//
// cloneTree Make a copy of the subtree rooted at this node.
// Discard any variable references encountered along the way,
// and replace with copies of the variable's definitions.
// Used to replicate the expression underneath variable
// references in preparation for generating the DFA tables.
//
//-------------------------------------------------------------------------
internal virtual RBBINode CloneTree()
{
RBBINode n;
if (fType == RBBINode.varRef)
{
// If the current node is a variable reference, skip over it
// and clone the definition of the variable instead.
n = fLeftChild.CloneTree();
}
else if (fType == RBBINode.uset)
{
n = this;
}
else
{
n = new RBBINode(this);
if (fLeftChild != null)
{
n.fLeftChild = fLeftChild.CloneTree();
n.fLeftChild.fParent = n;
}
if (fRightChild != null)
{
n.fRightChild = fRightChild.CloneTree();
n.fRightChild.fParent = n;
}
}
return(n);
}
//-----------------------------------------------------------------------------
//
// printSet Debug function. Print the contents of a set of Nodes
//
//-----------------------------------------------------------------------------
internal virtual void PrintSet(ICollection <RBBINode> s)
{
foreach (RBBINode n in s)
{
RBBINode.PrintInt32(n.fSerialNum, 8);
}
Console.Out.WriteLine();
}
//
// RBBISymbolTable::lookupNode Given a key (a variable name), return the
// corresponding RBBI Node. If there is no entry
// in the table for this name, return NULL.
//
internal virtual RBBINode LookupNode(String key)
{
RBBINode retNode = null;
RBBISymbolTableEntry el;
el = fHashTable.Get(key);
if (el != null)
{
retNode = el.Val;
}
return(retNode);
}
//
// RBBISymbolTable::addEntry Add a new entry to the symbol table.
// Indicate an error if the name already exists -
// this will only occur in the case of duplicate
// variable assignments.
//
internal virtual void AddEntry(string key, RBBINode val)
{
if (fHashTable.TryGetValue(key, out RBBISymbolTableEntry e) && e != null)
{
fRuleScanner.Error(RBBIRuleBuilder.U_BRK_VARIABLE_REDFINITION);
return;
}
fHashTable[key] = new RBBISymbolTableEntry {
Key = key, Val = val
};
}
//-----------------------------------------------------------------------------
//
// addRuleRootNodes Recursively walk a parse tree, adding all nodes flagged
// as roots of a rule to a destination vector.
//
//-----------------------------------------------------------------------------
internal virtual void AddRuleRootNodes(IList <RBBINode> dest, RBBINode node)
{
if (node == null)
{
return;
}
if (node.fRuleRoot)
{
dest.Add(node);
// Note: rules cannot nest. If we found a rule start node,
// no child node can also be a start node.
return;
}
AddRuleRootNodes(dest, node.fLeftChild);
AddRuleRootNodes(dest, node.fRightChild);
}
//
// RBBISymbolTable::addEntry Add a new entry to the symbol table.
// Indicate an error if the name already exists -
// this will only occur in the case of duplicate
// variable assignments.
//
internal virtual void AddEntry(string key, RBBINode val)
{
RBBISymbolTableEntry e;
e = fHashTable.Get(key);
if (e != null)
{
fRuleScanner.Error(RBBIRuleBuilder.U_BRK_VARIABLE_REDFINITION);
return;
}
e = new RBBISymbolTableEntry();
e.Key = key;
e.Val = val;
fHashTable[e.Key] = e;
}
//-----------------------------------------------------------------------------
//
// calcNullable. Impossible to explain succinctly. See Aho, section 3.9
//
//-----------------------------------------------------------------------------
internal virtual void CalcNullable(RBBINode n)
{
if (n == null)
{
return;
}
if (n.fType == RBBINode.setRef ||
n.fType == RBBINode.endMark)
{
// These are non-empty leaf node types.
n.fNullable = false;
return;
}
if (n.fType == RBBINode.lookAhead || n.fType == RBBINode.tag)
{
// Lookahead marker node. It's a leaf, so no recursion on children.
// It's nullable because it does not match any literal text from the input stream.
n.fNullable = true;
return;
}
// The node is not a leaf.
// Calculate nullable on its children.
CalcNullable(n.fLeftChild);
CalcNullable(n.fRightChild);
// Apply functions from table 3.40 in Aho
if (n.fType == RBBINode.opOr)
{
n.fNullable = n.fLeftChild.fNullable || n.fRightChild.fNullable;
}
else if (n.fType == RBBINode.opCat)
{
n.fNullable = n.fLeftChild.fNullable && n.fRightChild.fNullable;
}
else if (n.fType == RBBINode.opStar || n.fType == RBBINode.opQuestion)
{
n.fNullable = true;
}
else
{
n.fNullable = false;
}
}
internal RBBINode(RBBINode other)
{
fSerialNum = ++gLastSerial;
fType = other.fType;
fInputSet = other.fInputSet;
fPrecedence = other.fPrecedence;
fText = other.fText;
fFirstPos = other.fFirstPos;
fLastPos = other.fLastPos;
fNullable = other.fNullable;
fVal = other.fVal;
fRuleRoot = false;
fChainIn = other.fChainIn;
fFirstPosSet = new HashSet <RBBINode>(other.fFirstPosSet);
fLastPosSet = new HashSet <RBBINode>(other.fLastPosSet);
fFollowPos = new HashSet <RBBINode>(other.fFollowPos);
}
//-----------------------------------------------------------------------------
//
// bofFixup. Fixup for state tables that include {bof} beginning of input testing.
// Do an swizzle similar to chaining, modifying the followPos set of
// the bofNode to include the followPos nodes from other {bot} nodes
// scattered through the tree.
//
// This function has much in common with calcChainedFollowPos().
//
//-----------------------------------------------------------------------------
internal virtual void BofFixup()
{
//
// The parse tree looks like this ...
// fTree root --. <cat>
// / \
// <cat> <#end node>
// / \
// <bofNode> rest
// of tree
//
// We will be adding things to the followPos set of the <bofNode>
//
RBBINode bofNode = fRB.fTreeRoots[fRootIx].fLeftChild.fLeftChild;
Assert.Assrt(bofNode.fType == RBBINode.leafChar);
Assert.Assrt(bofNode.fVal == 2);
// Get all nodes that can be the start a match of the user-written rules
// (excluding the fake bofNode)
// We want the nodes that can start a match in the
// part labeled "rest of tree"
//
ISet <RBBINode> matchStartNodes = fRB.fTreeRoots[fRootIx].fLeftChild.fRightChild.fFirstPosSet;
foreach (RBBINode startNode in matchStartNodes)
{
if (startNode.fType != RBBINode.leafChar)
{
continue;
}
if (startNode.fVal == bofNode.fVal)
{
// We found a leaf node corresponding to a {bof} that was
// explicitly written into a rule.
// Add everything from the followPos set of this node to the
// followPos set of the fake bofNode at the start of the tree.
//
bofNode.fFollowPos.UnionWith(startNode.fFollowPos);
}
}
}
//-----------------------------------------------------------------------------
//
// calcLastPos. Impossible to explain succinctly. See Aho, section 3.9
//
//-----------------------------------------------------------------------------
internal virtual void CalcLastPos(RBBINode n)
{
if (n == null)
{
return;
}
if (n.fType == RBBINode.leafChar ||
n.fType == RBBINode.endMark ||
n.fType == RBBINode.lookAhead ||
n.fType == RBBINode.tag)
{
// These are non-empty leaf node types.
n.fLastPosSet.Add(n);
return;
}
// The node is not a leaf.
// Calculate lastPos on its children.
CalcLastPos(n.fLeftChild);
CalcLastPos(n.fRightChild);
// Apply functions from table 3.40 in Aho
if (n.fType == RBBINode.opOr)
{
n.fLastPosSet.UnionWith(n.fLeftChild.fLastPosSet);
n.fLastPosSet.UnionWith(n.fRightChild.fLastPosSet);
}
else if (n.fType == RBBINode.opCat)
{
n.fLastPosSet.UnionWith(n.fRightChild.fLastPosSet);
if (n.fRightChild.fNullable)
{
n.fLastPosSet.UnionWith(n.fLeftChild.fLastPosSet);
}
}
else if (n.fType == RBBINode.opStar ||
n.fType == RBBINode.opQuestion ||
n.fType == RBBINode.opPlus)
{
n.fLastPosSet.UnionWith(n.fLeftChild.fLastPosSet);
}
}
//-----------------------------------------------------------------------------
//
// printPosSets Debug function. Dump Nullable, firstpos, lastpos and followpos
// for each node in the tree.
//
//-----------------------------------------------------------------------------
internal virtual void PrintPosSets(RBBINode n)
{
if (n == null)
{
return;
}
RBBINode.PrintNode(n);
Console.Out.Write(" Nullable: " + n.fNullable);
Console.Out.Write(" firstpos: ");
PrintSet(n.fFirstPosSet);
Console.Out.Write(" lastpos: ");
PrintSet(n.fLastPosSet);
Console.Out.Write(" followpos: ");
PrintSet(n.fFollowPos);
PrintPosSets(n.fLeftChild);
PrintPosSets(n.fRightChild);
}
//-------------------------------------------------------------------------
//
// flattenVariables Walk a parse tree, replacing any variable
// references with a copy of the variable's definition.
// Aside from variables, the tree is not changed.
//
// Return the root of the tree. If the root was not a variable
// reference, it remains unchanged - the root we started with
// is the root we return. If, however, the root was a variable
// reference, the root of the newly cloned replacement tree will
// be returned, and the original tree deleted.
//
// This function works by recursively walking the tree
// without doing anything until a variable reference is
// found, then calling cloneTree() at that point. Any
// nested references are handled by cloneTree(), not here.
//
//-------------------------------------------------------------------------
internal virtual RBBINode FlattenVariables()
{
if (fType == varRef)
{
RBBINode retNode = fLeftChild.CloneTree();
retNode.fRuleRoot = this.fRuleRoot;
retNode.fChainIn = this.fChainIn;
return(retNode);
}
if (fLeftChild != null)
{
fLeftChild = fLeftChild.FlattenVariables();
fLeftChild.fParent = this;
}
if (fRightChild != null)
{
fRightChild = fRightChild.FlattenVariables();
fRightChild.fParent = this;
}
return(this);
}
//-------------------------------------------------------------------------
//
// print. Print out a single node, for debugging.
//
//-------------------------------------------------------------------------
////CLOVER:OFF
internal static void PrintNode(RBBINode n)
{
if (n == null)
{
Console.Out.Write(" -- null --\n");
}
else
{
RBBINode.PrintInt32(n.fSerialNum, 10);
RBBINode.PrintString(nodeTypeNames[n.fType], 11);
RBBINode.PrintInt32(n.fParent == null ? 0 : n.fParent.fSerialNum, 11);
RBBINode.PrintInt32(n.fLeftChild == null ? 0 : n.fLeftChild.fSerialNum, 11);
RBBINode.PrintInt32(n.fRightChild == null ? 0 : n.fRightChild.fSerialNum, 12);
RBBINode.PrintInt32(n.fFirstPos, 12);
RBBINode.PrintInt32(n.fVal, 7);
if (n.fType == varRef)
{
Console.Out.Write(" " + n.fText);
}
}
Console.Out.WriteLine("");
}
//CLOVER:ON
//------------------------------------------------------------------------
//
// printSets A debugging function.
// dump out all of the set definitions.
//
//------------------------------------------------------------------------
///CLOVER:OFF
internal virtual void PrintSets()
{
int i;
Console.Out.Write("\n\nUnicode Sets List\n------------------\n");
for (i = 0; i < fRB.fUSetNodes.Count; i++)
{
RBBINode usetNode;
RBBINode setRef;
RBBINode varRef;
String setName;
usetNode = fRB.fUSetNodes[i];
//System.out.print(" " + i + " ");
RBBINode.PrintInt32(2, i);
setName = "anonymous";
setRef = usetNode.fParent;
if (setRef != null)
{
varRef = setRef.fParent;
if (varRef != null && varRef.fType == RBBINode.varRef)
{
setName = varRef.fText;
}
}
Console.Out.Write(" " + setName);
Console.Out.Write(" ");
Console.Out.Write(usetNode.fText);
Console.Out.Write("\n");
if (usetNode.fLeftChild != null)
{
usetNode.fLeftChild.PrintTree(true);
}
}
Console.Out.Write("\n");
}
//-------------------------------------------------------------------------
//
// flattenSets Walk the parse tree, replacing any nodes of type setRef
// with a copy of the expression tree for the set. A set's
// equivalent expression tree is precomputed and saved as
// the left child of the uset node.
//
//-------------------------------------------------------------------------
internal virtual void FlattenSets()
{
Assert.Assrt(fType != setRef);
if (fLeftChild != null)
{
if (fLeftChild.fType == setRef)
{
RBBINode setRefNode = fLeftChild;
RBBINode usetNode = setRefNode.fLeftChild;
RBBINode replTree = usetNode.fLeftChild;
fLeftChild = replTree.CloneTree();
fLeftChild.fParent = this;
}
else
{
fLeftChild.FlattenSets();
}
}
if (fRightChild != null)
{
if (fRightChild.fType == setRef)
{
RBBINode setRefNode = fRightChild;
RBBINode usetNode = setRefNode.fLeftChild;
RBBINode replTree = usetNode.fLeftChild;
fRightChild = replTree.CloneTree();
fRightChild.fParent = this;
// delete setRefNode;
}
else
{
fRightChild.FlattenSets();
}
}
}
//-----------------------------------------------------------------------------
//
// printStates Debug Function. Dump the fully constructed state transition table.
//
//-----------------------------------------------------------------------------
internal virtual void PrintStates()
{
int c; // input "character"
int n; // state number
Console.Out.Write("state | i n p u t s y m b o l s \n");
Console.Out.Write(" | Acc LA Tag");
for (c = 0; c < fRB.fSetBuilder.NumCharCategories; c++)
{
RBBINode.PrintInt32(c, 3);
}
Console.Out.Write("\n");
Console.Out.Write(" |---------------");
for (c = 0; c < fRB.fSetBuilder.NumCharCategories; c++)
{
Console.Out.Write("---");
}
Console.Out.Write("\n");
for (n = 0; n < fDStates.Count; n++)
{
RBBIStateDescriptor sd = fDStates[n];
RBBINode.PrintInt32(n, 5);
Console.Out.Write(" | ");
RBBINode.PrintInt32(sd.fAccepting, 3);
RBBINode.PrintInt32(sd.fLookAhead, 4);
RBBINode.PrintInt32(sd.fTagsIdx, 6);
Console.Out.Write(" ");
for (c = 0; c < fRB.fSetBuilder.NumCharCategories; c++)
{
RBBINode.PrintInt32(sd.fDtran[c], 3);
}
Console.Out.Write("\n");
}
Console.Out.Write("\n\n");
}
//-------------------------------------------------------------------------------------
//
// RangeDescriptor::setDictionaryFlag
//
// Character Category Numbers that include characters from
// the original Unicode Set named "dictionary" have bit 14
// set to 1. The RBBI runtime engine uses this to trigger
// use of the word dictionary.
//
// This function looks through the Unicode Sets that it
// (the range) includes, and sets the bit in fNum when
// "dictionary" is among them.
//
// TODO: a faster way would be to find the set node for
// "dictionary" just once, rather than looking it
// up by name every time.
//
// -------------------------------------------------------------------------------------
internal virtual void SetDictionaryFlag()
{
int i;
for (i = 0; i < this.fIncludesSets.Count; i++)
{
RBBINode usetNode = fIncludesSets[i];
string setName = "";
RBBINode setRef = usetNode.fParent;
if (setRef != null)
{
RBBINode varRef = setRef.fParent;
if (varRef != null && varRef.fType == RBBINode.varRef)
{
setName = varRef.fText;
}
}
if (setName.Equals("dictionary"))
{
this.fNum |= 0x4000;
break;
}
}
}
internal virtual void AddValToSet(RBBINode usetNode, int val)
{
RBBINode leafNode = new RBBINode(RBBINode.leafChar);
leafNode.fVal = val;
if (usetNode.fLeftChild == null)
{
usetNode.fLeftChild = leafNode;
leafNode.fParent = usetNode;
}
else
{
// There are already input symbols present for this set.
// Set up an OR node, with the previous stuff as the left child
// and the new value as the right child.
RBBINode orNode = new RBBINode(RBBINode.opOr);
orNode.fLeftChild = usetNode.fLeftChild;
orNode.fRightChild = leafNode;
orNode.fLeftChild.fParent = orNode;
orNode.fRightChild.fParent = orNode;
usetNode.fLeftChild = orNode;
orNode.fParent = usetNode;
}
}
//-----------------------------------------------------------------------------
//
// printRuleStatusTable Debug Function. Dump the common rule status table
//
//-----------------------------------------------------------------------------
internal virtual void PrintRuleStatusTable()
{
int thisRecord = 0;
int nextRecord = 0;
int i;
IList <int> tbl = fRB.fRuleStatusVals;
Console.Out.Write("index | tags \n");
Console.Out.Write("-------------------\n");
while (nextRecord < tbl.Count)
{
thisRecord = nextRecord;
nextRecord = thisRecord + tbl[thisRecord] + 1;
RBBINode.PrintInt32(thisRecord, 7);
for (i = thisRecord + 1; i < nextRecord; i++)
{
int val = tbl[i];
RBBINode.PrintInt32(val, 7);
}
Console.Out.Write("\n");
}
Console.Out.Write("\n\n");
}
//-----------------------------------------------------------------------------
//
// RBBITableBuilder::build - This is the main function for building the DFA state transtion
// table from the RBBI rules parse tree.
//
//-----------------------------------------------------------------------------
internal virtual void Build()
{
// If there were no rules, just return. This situation can easily arise
// for the reverse rules.
if (fRB.fTreeRoots[fRootIx] == null)
{
return;
}
//
// Walk through the tree, replacing any references to $variables with a copy of the
// parse tree for the substition expression.
//
fRB.fTreeRoots[fRootIx] = fRB.fTreeRoots[fRootIx].FlattenVariables();
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("ftree", StringComparison.Ordinal) >= 0)
{
Console.Out.WriteLine("Parse tree after flattening variable references.");
fRB.fTreeRoots[fRootIx].PrintTree(true);
}
//
// If the rules contained any references to {bof}
// add a {bof} <cat> <former root of tree> to the
// tree. Means that all matches must start out with the
// {bof} fake character.
//
if (fRB.fSetBuilder.SawBOF)
{
RBBINode bofTop = new RBBINode(RBBINode.opCat);
RBBINode bofLeaf = new RBBINode(RBBINode.leafChar);
bofTop.fLeftChild = bofLeaf;
bofTop.fRightChild = fRB.fTreeRoots[fRootIx];
bofLeaf.fParent = bofTop;
bofLeaf.fVal = 2; // Reserved value for {bof}.
fRB.fTreeRoots[fRootIx] = bofTop;
}
//
// Add a unique right-end marker to the expression.
// Appears as a cat-node, left child being the original tree,
// right child being the end marker.
//
RBBINode cn = new RBBINode(RBBINode.opCat);
cn.fLeftChild = fRB.fTreeRoots[fRootIx];
fRB.fTreeRoots[fRootIx].fParent = cn;
cn.fRightChild = new RBBINode(RBBINode.endMark);
cn.fRightChild.fParent = cn;
fRB.fTreeRoots[fRootIx] = cn;
//
// Replace all references to UnicodeSets with the tree for the equivalent
// expression.
//
fRB.fTreeRoots[fRootIx].FlattenSets();
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("stree", StringComparison.Ordinal) >= 0)
{
Console.Out.WriteLine("Parse tree after flattening Unicode Set references.");
fRB.fTreeRoots[fRootIx].PrintTree(true);
}
//
// calculate the functions nullable, firstpos, lastpos and followpos on
// nodes in the parse tree.
// See the alogrithm description in Aho.
// Understanding how this works by looking at the code alone will be
// nearly impossible.
//
CalcNullable(fRB.fTreeRoots[fRootIx]);
CalcFirstPos(fRB.fTreeRoots[fRootIx]);
CalcLastPos(fRB.fTreeRoots[fRootIx]);
CalcFollowPos(fRB.fTreeRoots[fRootIx]);
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("pos", StringComparison.Ordinal) >= 0)
{
Console.Out.Write("\n");
PrintPosSets(fRB.fTreeRoots[fRootIx]);
}
//
// For "chained" rules, modify the followPos sets
//
if (fRB.fChainRules)
{
CalcChainedFollowPos(fRB.fTreeRoots[fRootIx]);
}
//
// BOF (start of input) test fixup.
//
if (fRB.fSetBuilder.SawBOF)
{
BofFixup();
}
//
// Build the DFA state transition tables.
//
BuildStateTable();
FlagAcceptingStates();
FlagLookAheadStates();
FlagTaggedStates();
//
// Update the global table of rule status {tag} values
// The rule builder has a global vector of status values that are common
// for all tables. Merge the ones from this table into the global set.
//
MergeRuleStatusVals();
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("states", StringComparison.Ordinal) >= 0)
{
PrintStates();
}
}
//-----------------------------------------------------------------------------
//
// calcChainedFollowPos. Modify the previously calculated followPos sets
// to implement rule chaining. NOT described by Aho
//
//-----------------------------------------------------------------------------
internal virtual void CalcChainedFollowPos(RBBINode tree)
{
IList <RBBINode> endMarkerNodes = new JCG.List <RBBINode>();
IList <RBBINode> leafNodes = new JCG.List <RBBINode>();
// get a list of all endmarker nodes.
tree.FindNodes(endMarkerNodes, RBBINode.endMark);
// get a list all leaf nodes
tree.FindNodes(leafNodes, RBBINode.leafChar);
// Collect all leaf nodes that can start matches for rules
// with inbound chaining enabled, which is the union of the
// firstPosition sets from each of the rule root nodes.
IList <RBBINode> ruleRootNodes = new JCG.List <RBBINode>();
AddRuleRootNodes(ruleRootNodes, tree);
ISet <RBBINode> matchStartNodes = new JCG.HashSet <RBBINode>();
foreach (RBBINode node in ruleRootNodes)
{
if (node.fChainIn)
{
matchStartNodes.UnionWith(node.fFirstPosSet);
}
}
// Iterate over all leaf nodes,
//
foreach (RBBINode tNode in leafNodes)
{
RBBINode endNode = null;
// Identify leaf nodes that correspond to overall rule match positions.
// These include an endMarkerNode in their followPos sets.
foreach (RBBINode endMarkerNode in endMarkerNodes)
{
if (tNode.fFollowPos.Contains(endMarkerNode))
{
endNode = tNode;
break;
}
}
if (endNode == null)
{
// node wasn't an end node. Try again with the next.
continue;
}
// We've got a node that can end a match.
// Line Break Specific hack: If this node's val correspond to the $CM char class,
// don't chain from it.
// TODO: Add rule syntax for this behavior, get specifics out of here and
// into the rule file.
if (fRB.fLBCMNoChain)
{
int c = this.fRB.fSetBuilder.GetFirstChar(endNode.fVal);
if (c != -1)
{
// c == -1 occurs with sets containing only the {eof} marker string.
int cLBProp = UChar.GetIntPropertyValue(c, UProperty.Line_Break);
if (cLBProp == LineBreak.CombiningMark)
{
continue;
}
}
}
// Now iterate over the nodes that can start a match, looking for ones
// with the same char class as our ending node.
foreach (RBBINode startNode in matchStartNodes)
{
if (startNode.fType != RBBINode.leafChar)
{
continue;
}
if (endNode.fVal == startNode.fVal)
{
// The end val (character class) of one possible match is the
// same as the start of another.
// Add all nodes from the followPos of the start node to the
// followPos set of the end node, which will have the effect of
// letting matches transition from a match state at endNode
// to the second char of a match starting with startNode.
endNode.fFollowPos.UnionWith(startNode.fFollowPos);
}
}
}
}
