//-----------------------------------------------------------------------------
//
// RBBITableBuilder::build - This is the main function for building the DFA state transtion
// table from the RBBI rules parse tree.
//
//-----------------------------------------------------------------------------
void RBBITableBuilder::build() {
if (U_FAILURE(*fStatus)) {
return;
}
// If there were no rules, just return. This situation can easily arise
// for the reverse rules.
if (fTree==NULL) {
return;
}
//
// Walk through the tree, replacing any references to $variables with a copy of the
// parse tree for the substition expression.
//
fTree = fTree->flattenVariables();
#ifdef RBBI_DEBUG
if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ftree")) {
RBBIDebugPuts("Parse tree after flattening variable references.");
fTree->printTree(TRUE);
}
#endif
//
// 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);
// Delete and exit if memory allocation failed.
if (bofTop == NULL || bofLeaf == NULL) {
*fStatus = U_MEMORY_ALLOCATION_ERROR;
delete bofTop;
delete bofLeaf;
return;
}
bofTop->fLeftChild = bofLeaf;
bofTop->fRightChild = fTree;
bofLeaf->fParent = bofTop;
bofLeaf->fVal = 2; // Reserved value for {bof}.
fTree = 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);
// Exit if memory allocation failed.
if (cn == NULL) {
*fStatus = U_MEMORY_ALLOCATION_ERROR;
return;
}
cn->fLeftChild = fTree;
fTree->fParent = cn;
cn->fRightChild = new RBBINode(RBBINode::endMark);
// Delete and exit if memory allocation failed.
if (cn->fRightChild == NULL) {
*fStatus = U_MEMORY_ALLOCATION_ERROR;
delete cn;
return;
}
cn->fRightChild->fParent = cn;
fTree = cn;
//
// Replace all references to UnicodeSets with the tree for the equivalent
// expression.
//
fTree->flattenSets();
#ifdef RBBI_DEBUG
if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "stree")) {
RBBIDebugPuts("Parse tree after flattening Unicode Set references.");
fTree->printTree(TRUE);
}
#endif
//
// 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(fTree);
calcFirstPos(fTree);
calcLastPos(fTree);
calcFollowPos(fTree);
if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "pos")) {
RBBIDebugPuts("\n");
printPosSets(fTree);
}
//
// For "chained" rules, modify the followPos sets
//
if (fRB->fChainRules) {
calcChainedFollowPos(fTree);
}
//
// 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 && uprv_strstr(fRB->fDebugEnv, "states")) {printStates();};
}
//-----------------------------------------------------------------------------
//
// RBBITableBuilder::build - This is the main function for building the DFA state transtion
// table from the RBBI rules parse tree.
//
//-----------------------------------------------------------------------------
void RBBITableBuilder::build() {
if (U_FAILURE(*fStatus)) {
return;
}
// If there were no rules, just return. This situation can easily arise
// for the reverse rules.
if (fTree==NULL) {
return;
}
//
// Walk through the tree, replacing any references to $variables with a copy of the
// parse tree for the substition expression.
//
fTree = fTree->flattenVariables();
if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ftree")) {
RBBIDebugPrintf("Parse tree after flattening variable references.\n");
fTree->printTree(TRUE);
}
//
// 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 = fTree;
fTree->fParent = cn;
cn->fRightChild = new RBBINode(RBBINode::endMark);
cn->fRightChild->fParent = cn;
fTree = cn;
//
// Replace all references to UnicodeSets with the tree for the equivalent
// expression.
//
fTree->flattenSets();
if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "stree")) {
RBBIDebugPrintf("Parse tree after flattening Unicode Set references.\n");
fTree->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(fTree);
calcFirstPos(fTree);
calcLastPos(fTree);
calcFollowPos(fTree);
if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "pos")) {
RBBIDebugPrintf("\n\n");
printPosSets(fTree);
}
//
// For "chained" rules, modify the followPos sets
//
if (fRB->fChainRules) {
calcChainedFollowPos(fTree);
}
//
// 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 && uprv_strstr(fRB->fDebugEnv, "states")) {printStates();};
}
