// file: $isip/class/pr/LanguageModelIHD/lmihd_07.cc
//

// system include files
//
#include <typeinfo>

// isip include files
//
#include "LanguageModelIHD.h"

// method: getRuleModel
//
// arguments: none
//
// return: a RuleModel object containing BNF rules representing the IHD
//
// this method converts an IHD object to a set of BNF production rules and
// returns a RuleModel object
//
RuleModel LanguageModelIHD::getRuleModel() {

  Vector< DiGraph<Ulong> > digraphs;
  Vector<SearchSymbol> symbol_table;
  SingleLinkedList<Ulong> vertices;
  SingleLinkedList< Triple< Pair<Long, Long>, Float, Boolean> > arcs;
  Vector< Vector< Vector<ProductionRule> > > prod_rule_model;
  HierarchicalDigraph hg(hg_d);
  
  // loop over the levels in the HierarchicalDigraph0
  //
  for (int i=0; i<hg_d.length(); i++) {
    
    // get the digraphs from this level of the HierarchicalDigraph
    //
    digraphs = hg_d(i).convertSubgraphs();

    // get the symbol table from this level of the HierarchicalDigraph
    //
    symbol_table = hg_d(i).getSymbolTable();
    
    Vector< Vector<ProductionRule> > level_prod_rules;

    // loop over each graph at this level
    //
    for (int j=0; j<digraphs.length(); j++) {
      
      vertices.clear();
      arcs.clear();
      digraphs(j).get(vertices, arcs);
  
      Vector<ProductionRule> prod_rules;

      // for every arc, create a rule:
      //
      // (node number)::=(node value), (branch number)
      //
      // where:
      //
      // '(node number)' = rule_name_d
      // (node value) = 
      // ',' =
      // (branch number) =
      //
      //vertices.getCurr()->debug(L"current vertex");
      //
      arcs.gotoFirst();
      do {

	ProductionRule rule;
	
	if (arcs.getCurr() != NULL) {	  

	  // create rule name (in the form Rxx where xx is the symbol index)
	  //
	  String rule_name;
	  String symbol_name;
	  int32 rule_index;
	  int32 symbol_index;
	  
	  rule_index = (arcs.getCurr()->first()).first();
	  symbol_index = (arcs.getCurr()->first()).second();
	  
	  if (rule_index == DiGraph<Ulong>::START_INDEX) {

	    if (i <= 0) {
	      rule.setRuleName(ProductionRule::START_RULE_NAME);
	    }
	    else {
	      if (hg_d(i-1).getContextMap().length() > 0) {
		String temp_name;
		temp_name.concat((Long)j);
		rule.setRuleName(temp_name);
	      }
	      else {
		rule.setRuleName(hg_d(i-1).getSymbolTable()(j));
	      }
	    }
 	    rule.setRuleType(ProductionRule::START);
	    
	    symbol_name.assign(ProductionRule::RULE_NAME_BASE);
	    symbol_name.concat((Long)symbol_index);
	    rule.append(ProductionRuleTokenType::NON_TERMINAL, symbol_name,
			arcs.getCurr()->second());
	  }
	  else {
	    
	    // need to convert rule_index to a string Rxx where xx is the
	    // symbol index
	    //
	    rule_name.assign(ProductionRule::RULE_NAME_BASE);
	    rule_name.concat((Long)rule_index);
	    rule.setRuleName(rule_name);
	    
	    vertices.gotoPosition(rule_index);

	    symbol_name.assign(symbol_table(*vertices.getCurr()));
	    
	    rule.append(ProductionRuleTokenType::TERMINAL, symbol_name);
	    
	    
	    // append a concatenation symbol to the rule
	    //
	    rule.append(ProductionRuleTokenType::CONCATENATION,
			ProductionRule::DEF_RULE_NAME,
			arcs.getCurr()->second());
	    

	    if (symbol_index == DiGraph<Ulong>::TERM_INDEX) {

	      // add a non_terminal that references a rule to an
	      // epsilon symbol
	      //
	      symbol_name.assign(ProductionRule::TERM_RULE_NAME);
	      rule.append(ProductionRuleTokenType::NON_TERMINAL, symbol_name);
	      prod_rules.concat(rule);
	      
	      // create the epsilon rule
	      //
	      rule.clear();
	      rule.setRuleName(ProductionRule::TERM_RULE_NAME);		
	      rule.append(ProductionRuleTokenType::EPSILON,
			  ProductionRule::DEF_RULE_NAME);
	    }
	    else {

	      symbol_name.assign(ProductionRule::RULE_NAME_BASE);
	      symbol_name.concat((Long)symbol_index);
	      
	      // set symbol name based on symbol index	  
	      //
	      rule.append(ProductionRuleTokenType::NON_TERMINAL, symbol_name); 
	    }
	  }
	  
	  prod_rules.concat(rule);
	  
	}
	else {	  
	  rule.setRuleType(ProductionRule::NORMAL);
	  
	  if (i <= 0) {
	    rule.setRuleName(ProductionRule::START_RULE_NAME);
	  }
	  else {
	    if (hg_d(i-1).getContextMap().length() > 0) {
	      rule.setRuleName((String)((Long)j));
	    }
	    else {
	      rule.setRuleName(hg_d(i-1).getSymbolTable()(j));
	    }
	  }

	  prod_rules.concat(rule);
	}
      } while (arcs.gotoNext());
      
      level_prod_rules.concat(prod_rules);
      // after rules have been created for all vertices in the
      // graph, and the Vector<ProductionRule> object to a
      // Vector<Vector<ProductionRule>> object.
    }
    
    // add the Vector<Vector<ProductionRule>> object to a
    // Vector< Vector < Vector<ProductionRule> > > object.
    prod_rule_model.concat(level_prod_rules);
  }
  
  // after iterative over all levels, there will be a
  // Vector< Vector < Vector<ProductionRule> > > with the dimensions
  // (levels) x (graphs in level) x (rules in graph)
  //
  // add this object to the typedefed Pair object RuleModel as the
  // first component, and add the HierarchicalDigraph object as the
  // second component.  the digraphs in the hierarchical digraph
  // object will be cleared before adding to the RuleModel object
  //
  Vector<DiGraph<SearchNode> > graphs;

  // loop over each level and clear the subgraphs
  //
  for (int i = 0; i < hg.length(); i++) {
  
    hg(i).setSubGraphs(graphs);
  }
  
  // finally, create the RuleModel object
  //
  RuleModel rm(prod_rule_model, hg);

  return rm;
}

// method: setRuleModel
//
// arguments: RuleModel rm_a
//
// return: a bool8 indicating status
//
// this method converts an IHD object to a set of BNF production rules and
// returns a RuleModel object
//
bool8 LanguageModelIHD::setRuleModel(const RuleModel& rm_a) {
  
  typedef Triple< Pair<Long, Long>, Float, Boolean> TopoTriple;
  Vector< DiGraph<Ulong> > digraphs;
  SingleLinkedList<Ulong> vertices;
  SingleLinkedList<TopoTriple> arcs;

  // assign all HierarchicalDigraph information from RuleModel
  // to this class's hg_d object
  //
  hg_d.assign(rm_a.second());

  // loop over all levels
  //
  for (int32 i = 0; i < rm_a.first().length(); i++) {

    // retrieve graphs at this level
    //
    Vector<ProductionRuleSet> bnf_graphs(rm_a.first()(i));
    
    Vector<SearchSymbol> search_symbols = hg_d(i).getSymbolTable();
    
    // loop over all graphs at this level
    //
    for (int32 j = 0; j < bnf_graphs.length(); j++) {

      Vector< Pair<String, SearchSymbol> > nodes;
      DiGraph<Ulong> digraph;

      // loop over all rules in this graph
      //
      for (int32 k = 0; k < bnf_graphs(j).length(); k++) {

	bnf_graphs(j)(k).gotoFirst();

	do {

	  if (bnf_graphs(j)(k).length() > 0) {

	    SearchSymbol symbol = bnf_graphs(j)(k).getValue();
	    	    
	    // try to identify unique nodes by finding unique pairs of rule
	    // names and terminal tokens. this is an attempt to minimize
	    // the graph as much as possible.
	    //
	    Pair<String, SearchSymbol> node(bnf_graphs(j)(k).getRuleName(),
					    symbol);

	    if (bnf_graphs(j)(k).getType() == ProductionRuleTokenType::TERMINAL
		&& !nodes.contains(&node)) {
	      nodes.concat(node);
	    }
	  }
	} while (bnf_graphs(j)(k).gotoNext());
      }

      // insert nodes into digraph
      //
      for (int k = 0; k < nodes.length(); k++) {

	Ulong vertex = search_symbols.first(nodes(k).second());

	digraph.insertVertex(&vertex);
      }

      // now that we have a list of unique nodes in this graph and a list of
      // symbols, we can construct the digraph object for this graph.  the
      // list of symbols doesn't contain all symbols at this level yet,
      // but it does contain the only symbols used in this graph.
      // a complete symbol list isn't necessary at this point.  the
      // indices of the symbols in the symbol list will remain constant
      // as we iterate over the graphs, and after all graphs have been
      // processed, we will have a complete list of symbols
      //
      
      // loop over all graphs and search for concatenation (',') symbols.
      // all rules are assumed to be
      // in normalized BNF form.  the following rule formats are permitted
      //
      // (rule_name) -> (terminal) , (non_terminal)
      // (rule_name) -> (non_terminal)
      // (rule_name) -> (terminal) , (epsilon)
      // (rule_name) -> (epsilon)
      //
      for (int k = 0; k < bnf_graphs(j).length(); k++) {

	bnf_graphs(j)(k).gotoFirst();

	if (bnf_graphs(j)(k).getRuleType() == ProductionRule::START) {
	  
	  if (bnf_graphs(j)(k).getType() ==
	      ProductionRuleTokenType::NON_TERMINAL) {

	    float32 weight = bnf_graphs(j)(k).getWeight();

	    if (isEpsilon(bnf_graphs(j), bnf_graphs(j)(k).getValue())) {
	      digraph.insertArc(digraph.getStart(),
				digraph.getTerm(),
				GraphArc<Ulong>::DEF_EPSILON, weight);
	    }

	    Vector<Pair<String, SearchSymbol> > terminals =
	      findTerminals(bnf_graphs(j), bnf_graphs(j)(k).getValue());
	    
	    for (int l = 0; l < terminals.length(); l++) {
	      digraph.insertArc(digraph.getStart(),
				digraph.getVertex(nodes.first(terminals(l))),
				GraphArc<Ulong>::DEF_EPSILON, weight);
	    }
	  }
	  else {
	    return Error::handle(L"LanguageModelIHD",
				 L"setRuleModel: invalid start rule",
				 Error::READ, __FILE__, __LINE__,
				 Error::ERROR);
	  }
	}
	else if (bnf_graphs(j)(k).getRuleType() == ProductionRule::NORMAL) {
	  
	  // if the first token is a terminal symbol, the following token must
	  // be a concatenation token. following the concatenation will be
	  // either a non_terminal symbol or an epsilon symbol.
	  //
	  if (bnf_graphs(j)(k).getType() ==
	      ProductionRuleTokenType::TERMINAL) {
	    
	    // determine which node this is
	    //
	    Pair<String, SearchSymbol> vertex(bnf_graphs(j)(k).getRuleName(),
					      bnf_graphs(j)(k).getValue());
	    int32 start_index = nodes.first(vertex);
	    
	    // move forward 1 tokens to the concatenation token
	    //
	    bnf_graphs(j)(k).gotoNext();
	    
	    // get the weight for this concatenation token.  this will be the
	    // weight of the arc
	    //
	    float32 weight = bnf_graphs(j)(k).getWeight();
	    
	    
	    // move forward again to either a non_terminal or an epsilon
	    //
	    bnf_graphs(j)(k).gotoNext();
	    
	    if (bnf_graphs(j)(k).getType() ==
		ProductionRuleTokenType::NON_TERMINAL) {
	      
	      // check to see whether or not the non_terminal symbol results
	      // in an epsilon symbol before a terminal symbol.  for example:
	      //
	      // (non_terminal1) -> (terminal),(non_terminal2)
	      // (non_terminal2) -> (epsilon)
	      //
	      if (isEpsilon(bnf_graphs(j), bnf_graphs(j)(k).getValue())) {
		digraph.insertArc(digraph.getVertex(start_index),
				  digraph.getTerm(),
				  GraphArc<Ulong>::DEF_EPSILON, weight);
	      }
	      // if it does not result in an epsilon, find all of the
	      // connecting terminal nodes.
	      // for example:
	      //
	      // (non_terminal1) -> (terminal1),(non_terminal2)
	      // (non_terminal2) -> (terminal2), (non_terminaln)
	      //
	      else {
		Vector<Pair<String, SearchSymbol> > terminals =
		  findTerminals(bnf_graphs(j), bnf_graphs(j)(k).getValue());
		
		for (int l = 0; l < terminals.length(); l++) {
		  digraph.insertArc(digraph.getVertex(start_index),
				    digraph.getVertex(nodes.first(terminals(l))),
				    GraphArc<Ulong>::DEF_EPSILON, weight);
		}
	      }
	    }
	    // otherwise, there is an epsilon after the terminal.
	    //
	    else if (bnf_graphs(j)(k).getType() ==
		     ProductionRuleTokenType::EPSILON) {
	      
	      digraph.insertArc(digraph.getVertex(start_index),
				digraph.getTerm(),
				GraphArc<Ulong>::DEF_EPSILON, weight);
	    }
	    // if none of the cases above, this rule is invalid
	    //
	    else {
	      return Error::handle(L"LanguageModelIHD",
				   L"setRuleModel: invalid rule",
				   Error::READ, __FILE__, __LINE__,
				   Error::ERROR);	      
	    }
	  }
	}
      }
      digraphs.concat(digraph);      
    }
    
    hg_d(i).setSymbolTable(search_symbols);    
    hg_d(i).setTempGraphs(digraphs);
    hg_d(i).convertDigraphs();
    hg_d(i).convertContexts();
    digraphs.clear();
        
  }

  // exit gracefully
  //
  return true;    
}

// method: isEpsilon
//
// arguments: Vector<ProductionRule> rules
//            String rule_name
//
// return: a bool8 indicating whether or not the non_terminal results
//         in an epsilon token
//
// determines whether a non_terminal token results in an epsilon token
//
bool8 LanguageModelIHD::isEpsilon(Vector<ProductionRule> rules_a,
				    String rule_name_a) {
    
  // loop over all rules and look for rule names matching rule_name_a
  //
  for (int i = 0; i < rules_a.length(); i++) {
    
    if (rules_a(i).getRuleName().eq(rule_name_a)){
      
      rules_a(i).gotoFirst();
      
      // check to see if the first token is an epsilon.  if so, return true
      //
      if (rules_a(i).getType() == ProductionRuleTokenType::EPSILON) {

	return true;
      }
      // check to see if the first token is a non_terminal symbol.  if so,
      // recursivley check to see if this rule results in an epsilon;
      //
      else if (rules_a(i).getType() == ProductionRuleTokenType::NON_TERMINAL) {
	
	return isEpsilon(rules_a, rules_a(i).getRuleName());
      }
    }
  }
  
  // this rule does not result in an epsilon
  //     
  return false;
}

// method: findTerminals
//
// arguments: Vector<ProductionRule> rules
//            String rule_name
//
// return: a Vector<Pair<String, SearchSymbol> > containing all
//         rule_name + terminal token pairs.  these correspond
//         to unique terminal nodes
//
// finds all unique terminal symbols given a non_terminal symbol
//
Vector<Pair<String, SearchSymbol> > LanguageModelIHD::findTerminals(Vector<ProductionRule> rules_a, String rule_name_a) {

  Vector<Pair<String, SearchSymbol> > terminal_nodes;
  Pair<String, SearchSymbol> terminal_node;
  
  for (int i = 0; i < rules_a.length(); i++) {
    
    if (rules_a(i).getRuleName().eq(rule_name_a)){
      
      rules_a(i).gotoFirst();
      
      if (rules_a(i).getType() == ProductionRuleTokenType::TERMINAL) {
	
	terminal_node.assign(rule_name_a, rules_a(i).getValue());
	if (!terminal_nodes.contains(&terminal_node)) {
	  terminal_nodes.concat(terminal_node);
	}
      }
      else if (rules_a(i).getType() == ProductionRuleTokenType::NON_TERMINAL) {
	
	terminal_nodes.concat(findTerminals(rules_a, rules_a(i).getValue()));
      }
    }
  }
  
  return terminal_nodes;
}