private void popCurrentState()
{
ParseState ps = parseStateStack.Pop() as ParseState;
candidateSymbols = ps.candidateSymbols;
currentSymbol = ps.currentSymbol;
unusedInputStrokes = ps.unusedInputStrokes;
unusedStrokes = ps.unusedStrokes;
minRequiredStrokes = ps.minRequiredStrokes;
}
// TODO: update to construct LBT here?
public void parse(LBT tree, ParseTreeNode n, List <ParseTreeNode> additional_nodes)
{
candidateSymbols = lexer.Start(tree, MAX_NEIGHBORS);
currentSymbol = null;
LBT oldtree = initLBT;
unusedStrokes = tree.strokes.Count;
List <ParseTreeNode> nodes = additional_nodes == null ? new List <ParseTreeNode>() : new List <ParseTreeNode>(additional_nodes);
nodes.Insert(0, n);
initLBT = tree;
applyRules(nodes);
initLBT = oldtree;
}
public ParserMain(Grammar grammar, InkML inkml_file, int topn, int neighbors, string classify_url, double prob_threshold, string stats_file)
{
TOP_N = topn;
MAX_NEIGHBORS = neighbors;
// YUCK!! COPY AND PASTE BELOW.
// Set symbol tables and input file. Instantiate the lexer. Create empty valid parse list.
//initializeTables(symbolDefFile);
lexer = new LexerMain(stats_file, classify_url, prob_threshold);
validParses = new List <ValidParseTree>();
this.grammar = grammar;
layoutClasses = grammar.GetLayoutClasses();
// Obtain LBT, stroke information.
inputInkML = inkml_file;
initLBT = new LBT(inputInkML, LBT.DefaultAdjacentCriterion);
treeRoot = new ParseTreeNode();
treeRoot.nodeType = "S";
treeRoot.lbt = initLBT;
// treeRoot.lexResult ???
//strokeList = lbt.strokes; // unused?
currentSymbol = null;
minRequiredStrokes = 1;
unusedInputStrokes = initLBT.strokes.Count;
parseStateStack = new Stack();
//Console.WriteLine(tree.ToDOT());
// Create root node.
//rootAll = new ParseTreeNode("S", tree);
// HACK: note that the ParseState constructor will initialize the parser data as well.
//ParseState initState = new ParseState(rootAll, tree); //, 5); // MAGIC
accesses = 0;
hits = 0;
// Invoke the parse
#if DEBUG
//Console.WriteLine("Starting parse...");
#endif
//Parse(initState, true, MAX_NEIGHBORS);
//validParses.Sort();
}
// END ADDED JULY 7, 12
public bool legalPartition(PartitionResultWrapper presult, PreviousSymbol p)
{
if (p == null)
{
return(true); // ???
}
// Compile non-empty regions.
List <string> nonEmptyRegions = new List <string>();
if (presult.result.ABOVE != null && //presult.result.ABOVE.strokes != null &&
presult.result.ABOVE.strokes.Count > 0)
{
nonEmptyRegions.Add("ABOVE");
}
if (presult.result.BELOW != null && // presult.result.BELOW.strokes != null &&
presult.result.BELOW.strokes.Count > 0)
{
nonEmptyRegions.Add("BELOW");
}
if (presult.result.SUPER != null && //presult.result.SUPER.strokes != null &&
presult.result.SUPER.strokes.Count > 0)
{
nonEmptyRegions.Add("SUPER");
}
if (presult.result.SUBSC != null && //presult.result.SUBSC.strokes != null &&
presult.result.SUBSC.strokes.Count > 0)
{
nonEmptyRegions.Add("SUBSC");
}
if (presult.result.CONTAINS != null && // presult.result.CONTAINS.strokes != null &&
presult.result.CONTAINS.strokes.Count > 0)
{
nonEmptyRegions.Add("CONTAINS");
}
// account for regions currently assigned
foreach (string reg in p.regions)
{
if (!nonEmptyRegions.Contains(reg))
{
nonEmptyRegions.Add(reg);
}
}
// TODO: LAYOUT CLASS?
#if DEBUG
Console.WriteLine("[legalPartition] p.parent.nodeType = <{0}>", p.parent.nodeType);
#endif
// get layout (regions) for symbol as specified by the grammar
// FIRST, try looking up parent NT directly as the class
// IF THAT FAILS, look up the class(es) based on the symbol
List <string[]> grammarLayout;
try {
grammarLayout = grammar.GetRegionLayouts(p.parent.nodeType);
} catch (Exception) {
try {
grammarLayout = grammar.GetRegionLayouts(p.symbol.segment.classification[0].symbol);
} catch (Exception) {
return(false); // no, just no
}
}
foreach (string[] regions in grammarLayout)
{
int matchedRegions = 0;
bool satisfies = true;
for (int i = 0; i < regions.Length; i += 2)
{
bool optional = regions[i][0] == '@'; // ow, assume required
string region = regions[i].Substring(1);
// string nt = regions[ i + 1 ];
if (nonEmptyRegions.Contains(region))
{
matchedRegions++;
}
else if (!optional)
{
satisfies = false;
}
}
if (satisfies && matchedRegions == nonEmptyRegions.Count)
{
// add starting NTs to the regions
for (int i = 0; i < regions.Length; i += 2)
{
string region = regions[i].Substring(1);
selectRegion(region, presult, regions[i + 1]);
}
return(true); // at least one satisfies
}
}
return(false);
}
public void applyRules(List <ParseTreeNode> arg_nlist)
{
#if DEBUG
Console.WriteLine("[applyRules] entered.");
treeRoot.ShowTree(4, null);
Console.WriteLine("Min Required Strokes: " + minRequiredStrokes);
Console.WriteLine("Call: " + apply_rule_counter);
#else
//Console.Write(apply_rule_counter);
//Console.Write('\r');
#endif
// increment counter
apply_rule_counter++;
if (arg_nlist.Count == 0)
{
return;
}
if (unusedStrokes < 1 && arg_nlist[0] != ParseTreeNode.EndOfBaseline && !(arg_nlist[0] is RelationTreeNode))
{
return;
}
// if ( unusedStrokes > 0 && !( arg_nlist[ 0 ].GetType().Equals( typeof( ParseTreeNode ) ) ) ) return;
List <ParseTreeNode> nlist = new List <ParseTreeNode>(arg_nlist);
ParseTreeNode n = nlist[0];
nlist.RemoveAt(0);
if (n == ParseTreeNode.EndOfBaseline)
{
// END-OF-BASELINE CASE
PartitionResultWrapper pr = attachSymbol(currentSymbol.symbol.lbt, null);
// continue only if valid partition made
if (pr != null)
{
if (nlist.Count == 0 && unusedStrokes == 0 && unusedInputStrokes == 0)
{
#if DEBUG
Console.WriteLine("***ACCEPT***");
#endif
acceptCurrentParseTree();
}
int nodes_added = nlist.Count;
if (pr.result != null && pr.result.ABOVE.lbt.strokes.Count != 0)
{
nlist.Add(pr.result.ABOVE);
}
if (pr.result != null && pr.result.BELOW.lbt.strokes.Count != 0)
{
nlist.Add(pr.result.BELOW);
}
if (pr.result != null && pr.result.CONTAINS.lbt.strokes.Count != 0)
{
nlist.Add(pr.result.CONTAINS);
}
if (pr.result != null && pr.result.SUBSC.lbt.strokes.Count != 0)
{
nlist.Add(pr.result.SUBSC);
}
if (pr.result != null && pr.result.SUPER.lbt.strokes.Count != 0)
{
nlist.Add(pr.result.SUPER);
}
// add BLEFT/TLEFT
//if ( pr.result != null && pr.result.BLEFT.lbt.strokes.Count != 0 ) nlist.Add( pr.result.BLEFT );
//if ( pr.result != null && pr.result.TLEFT.lbt.strokes.Count != 0 ) nlist.Add( pr.result.TLEFT );
nodes_added = nlist.Count - nodes_added;
minRequiredStrokes += nodes_added;
applyRules(nlist);
}
else
{
#if DEBUG
Console.WriteLine("**BACKTRACK: end-of-baseline, invalid partition");
#endif
}
}
else if (n is RelationTreeNode)
{
// handle relation nodes by adding new parse tree node
RelationTreeNode rtn = n as RelationTreeNode;
ParseTreeNode ptn = new ParseTreeNode();
ptn.strokes = rtn.strokes;
ptn.nodeType = rtn.nodeType;
// increment here for new production
ptn.lbt = rtn.lbt;
rtn.children.Clear(); // remove all before
rtn.children.Add(ptn);
parse(ptn.lbt, ptn, nlist);
}
else if (n.nodeType.StartsWith("*"))
{ // if n generates terminal symbols
List <LexerResult> C = SelectCandidateSymbols(n.nodeType);
//candidateSymbols = C;
foreach (LexerResult c in C)
{
PartitionResultWrapper pr = attachSymbol(currentSymbol == null ? initLBT : currentSymbol.symbol.lbt, c);
if (pr != null)
{
pushCurrentState();
// update current state
currentSymbol = new PreviousSymbol(c, n, null, null, true);
unusedStrokes -= currentSymbol.symbol.segment.strokes.Count;
unusedInputStrokes -= currentSymbol.symbol.segment.strokes.Count;
// remove one of the min required strokes for the current token
minRequiredStrokes -= 1;
// prune by number of strokes left
if (unusedInputStrokes < minRequiredStrokes)
{
popCurrentState();
return; // continue;
}
List <string> layoutClasses = grammar.GetLayoutClassesFromTerminal(c.segment.classification[0].symbol);
if (layoutClasses.Count == 0)
{
continue;
}
candidateSymbols = new List <LexerResult>();
foreach (string layoutClass in layoutClasses)
{
List <LexerResult> res = lexer.Next(c.lbt, c.segment, layoutClass, MAX_NEIGHBORS);
foreach (LexerResult r in res)
{
if (!candidateSymbols.Contains(r))
{
candidateSymbols.Add(r);
}
}
}
if (candidateSymbols.Count == 0 && !nlist.Contains(ParseTreeNode.EndOfBaseline))
{
nlist.Insert(0, ParseTreeNode.EndOfBaseline);
}
// not end of baseline, so prune and backtrack if necessary
else if (nlist.Count > 0 && (nlist[0] is RelationTreeNode == false))
{
// prune candidates based on the current node type
foreach (LexerResult lr in candidateSymbols)
{
for (int k = 0; k < lr.segment.classification.Count; k++)
{
if (grammar.NonTerminalCanGenerateTerminal(nlist[0].nodeType, lr.segment.classification[k].symbol) == false)
{
lr.segment.classification.RemoveAt(k--);
}
}
}
// remove candidate symbols which contain no symbol alternatives
for (int k = 0; k < candidateSymbols.Count; k++)
{
if (candidateSymbols[k].segment.classification.Count == 0)
{
candidateSymbols.RemoveAt(k--);
}
}
// no valid symbols given the grammar, so bbreak out early
if (candidateSymbols.Count == 0)
{
popCurrentState();
return; // continue;
}
}
SymbolTreeNode nc = new SymbolTreeNode(c);
n.children.Add(nc);
/*
* if ( nlist.Count == 0 && unusedStrokes == 0 ) {
* if ( unusedInputStrokes == 0 ) acceptCurrentParseTree();
* } else {
*/
// append relation nodes to the END
List <RelationTreeNode> nlist_rels = new List <RelationTreeNode>();
if (pr.result != null && pr.result.ABOVE.lbt.strokes.Count != 0)
{
nlist_rels.Add(pr.result.ABOVE);
}
if (pr.result != null && pr.result.BELOW.lbt.strokes.Count != 0)
{
nlist_rels.Add(pr.result.BELOW);
}
if (pr.result != null && pr.result.CONTAINS.lbt.strokes.Count != 0)
{
nlist_rels.Add(pr.result.CONTAINS);
}
if (pr.result != null && pr.result.SUBSC.lbt.strokes.Count != 0)
{
nlist_rels.Add(pr.result.SUBSC);
}
if (pr.result != null && pr.result.SUPER.lbt.strokes.Count != 0)
{
nlist_rels.Add(pr.result.SUPER);
}
//if ( pr.result != null && pr.result.BLEFT.lbt.strokes.Count != 0 ) nlist_rels.Add( pr.result.BLEFT );
//if ( pr.result != null && pr.result.TLEFT.lbt.strokes.Count != 0 ) nlist_rels.Add( pr.result.TLEFT );
foreach (RelationTreeNode rtn in nlist_rels)
{
nlist.Add(rtn);
}
minRequiredStrokes += nlist_rels.Count;
applyRules(nlist);
//}
if (nlist.Count > 0 && nlist[0] == ParseTreeNode.EndOfBaseline)
{
nlist.RemoveAt(0); // !
}
// remove any leftover relation nodes
n.children.Remove(nc);
List <ParseTreeNode> n_children_tmp = new List <ParseTreeNode>(n.children);
for (int i = 0; i < n_children_tmp.Count; i++)
{
if (n_children_tmp[i] is RelationTreeNode)
{
n.children.Remove(n_children_tmp[i]);
}
}
foreach (RelationTreeNode rtn in nlist_rels)
{
nlist.Remove(rtn);
}
popCurrentState();
}
}
}
else
{
// NONTERMINALS
//n.lexResult.segment.classification[0].symbol;
List <string[]> productions = grammar.GetProductions(n.nodeType);
if (productions == null)
{
#if DEBUG
Console.Error.WriteLine("Error: invalid nonterminal ({0}).", n.nodeType);
#endif
return;
}
// remove one for the token we are replacing with productions
minRequiredStrokes--;
foreach (string[] production in productions)
{
minRequiredStrokes += production.Length;
// prune by number of strokes left
if (unusedInputStrokes < minRequiredStrokes)
{
minRequiredStrokes -= production.Length;
continue;
}
// prune based on the candidate symbols and the first rule in production
bool candidate_can_be_generated = false;
foreach (LexerResult lr in candidateSymbols)
{
foreach (Classification csf in lr.segment.classification)
{
if (grammar.NonTerminalCanGenerateTerminal(production[0], csf.symbol))
{
candidate_can_be_generated = true;
break;
}
}
if (candidate_can_be_generated)
{
break;
}
}
if (!candidate_can_be_generated)
{
minRequiredStrokes -= production.Length;
continue;
}
List <ParseTreeNode> nodes = new List <ParseTreeNode>();
foreach (string p in production)
{
ParseTreeNode n0 = new ParseTreeNode();
n0.nodeType = p;
n0.lexResult = null;
nodes.Add(n0);
n.children.Add(n0);
}
for (int i = nodes.Count - 1; i >= 0; i--)
{
nlist.Insert(0, nodes[i]);
}
applyRules(nlist);
// restore min required strokes
minRequiredStrokes -= production.Length;
foreach (ParseTreeNode node in nodes)
{
n.children.Remove(node);
nlist.Remove(node);
}
}
minRequiredStrokes++;
}
}
