// file: $isip/class/pr/LanguageModelXML/lmxml_08.cc
// version: $Id: lmxml_08.cc 10429 2006-02-14 22:35:30Z wholland $

// ISIP include files
//
#include "LanguageModelXML.h"

// method: setRuleModel
//
// arguments:
//   RuleModel& rm_a: rule model to convert
// 
// return: a bool8 value indicating status
//
// this method initiates the BNF->XML conversion.  the resulting
// xml grammar is stored in memory
//
bool8 LanguageModelXML::setRuleModel(const RuleModel& rm_a) {

  if (debug_level_d>Integral::BRIEF) {
    Console::put(L"performing BNF->XML conversion");
    Console::increaseIndention();
  }
  
  // clear the ihd and grammars
  //
  hg_d.clear(Integral::RESET);
  grammars_d.clear(Integral::RESET);

  // first thing's first, save the ihd
  //
  hg_d=rm_a.second();

  int32 i, j;

  // loop through each level
  //
  for (i=0; i<rm_a.first().length(); i++) {
    Vector < Vector<XMLToken> > this_level;

    if (debug_level_d>Integral::BRIEF) {
      String out_string(L"level ");
      out_string.concat((Long)i);
      Console::put(out_string);
      Console::increaseIndention();
    }
    
    // loop through each graph and perform conversion
    //
    for (j=0; j<rm_a.first()(i).length(); j++) {

      if (debug_level_d>Integral::BRIEF) {
	String out_string(L"graph ");
	out_string.concat((Long)j);
	Console::put(out_string);
	Console::increaseIndention();
      }
      
      this_level.concat(convertBNFtoXML(rm_a.first()(i)(j)));

      if (debug_level_d>Integral::BRIEF) {
	  Console::put(L"XML equivalent:");
	  Console::increaseIndention();
	  Console::put(tokensToString(this_level(j)));
	  Console::decreaseIndention();
      }
      
    }

    // concat converted grammars onto storage vector
    //
    grammars_d.concat(this_level);
  }
  
  return true;
}

// method: convertBNFtoXML
//
// arguments:
//   ProductionRuleSet prset_a: production rules to convert
// 
// return: Vector<XMLToken> containing equivalent XML grammar
//
// this method performs the BNF->XML conversion.  this method assumes
// that the root rule is marked appropriately.
// the root rule is defined as the rule that defines the entire
// set (with the aid of other rules).  typically, the root rule has
// a rule name of "S" (for set).
//
Vector<XMLToken> LanguageModelXML::convertBNFtoXML(ProductionRuleSet prset_a) {
  ProductionRuleSet working_set;
  working_set.assign(prset_a);

  Vector<XMLToken> ret_xml;
  // if we have an empty grammar, add an
  // empty grammar to the xml and return
  //
  if (prset_a.length()==0 ||
      (prset_a.length()==1 &&
       prset_a(0).getRuleType()==ProductionRule::NORMAL &&
       prset_a(0).getRuleName().eq(L""))) {
    XMLToken empty_grammar;
    empty_grammar.init(XMLToken::START_AND_END_TAG, GRAMMAR);
    ret_xml.concat(empty_grammar);
    return ret_xml;
  }

  // figure out the root rule 
  //
  String root_rule_uri;
  int32 i, j;
  for (i=0; i<prset_a.length(); i++) {
    if (prset_a(i).getRuleType()==ProductionRule::START) {
      root_rule_uri=prset_a(i).getRuleName();
      break;
    }
  }

  bool multiple_starts=false;
  
  if (i>=prset_a.length()) {
    root_rule_uri=prset_a(0).getRuleName();
  }
  else {
    for (; i<prset_a.length(); i++) {
      if (prset_a(i).getRuleType()==ProductionRule::START) {
	multiple_starts=true;
	break;
      }
    }
  }

  // construct the start grammar tag
  //
  XMLToken start_grammar_tag;
  Vector< Pair<String, String> > attrib_array;
  Pair<String, String> root_attrib;
  root_attrib.first().assign(ROOT);
  root_attrib.second().assign(root_rule_uri);
  attrib_array.concat(root_attrib);
  start_grammar_tag.init(XMLToken::START_TAG,
			 GRAMMAR,
			 attrib_array);

  // add the start grammar tag
  //
  ret_xml.concat(start_grammar_tag);

  // if we have multiple starts, go ahead and
  // take care of the start rule.  otherwise, the
  // code below this will handle it
  //
  if (multiple_starts) {
    ProductionRuleSet finished_set;
    ProductionRuleSet return_set;
    
    int32 i;
    
    for (i=0; i<working_set.length(); i++) {
      if (working_set(i).getRuleType()==ProductionRule::START) {
	return_set.concat(working_set(i));
      }
      else {
	finished_set.concat(working_set(i));
      }
    }
    
    working_set.assign(finished_set);

    int32 temp_depth = 1;    
    
    // make start rule tag and add it
    //
    addStartRuleTag(ret_xml, root_rule_uri, temp_depth);
 
    // add one-of
    //
    addStartBranchTag(ret_xml, temp_depth);

    // add items
    //
    for (j=0; j<return_set.length(); j++) {
      return_set(j).gotoFirst();
      String temp_rule_name = return_set(j).getValue();
      Float temp_rule_weight = return_set(j).getWeight();
      addStartItemTag(ret_xml, temp_depth, temp_rule_weight);
      addRuleRef(ret_xml, temp_rule_name, temp_depth);
      addEndItemTag(ret_xml, temp_depth);
    }

    // add end one-of
    //
    addEndBranchTag(ret_xml, temp_depth);

    // add end rule tag
    //
    addEndRuleTag(ret_xml, temp_depth);
  }
  
  while (working_set.length()!=0) {
    ProductionRuleSet curr_rule_set;

    // get all alternative rules matching the name of the
    // first rule left in the set.  note that this removes
    // some of the rules from the set, which will eventually
    // lead to the exit condition of the enclosing loop
    //
    curr_rule_set=extractRule(working_set(0).getRuleName(),
			      working_set);

    
    addRule(curr_rule_set, ret_xml);
  }

  
  // construct and add the end grammar tag
  //
  XMLToken end_grammar_tag;
  end_grammar_tag.init(XMLToken::END_TAG, GRAMMAR);
  ret_xml.concat(end_grammar_tag);

  
  return ret_xml;
}

// method: extractRule
//
// arguments:
//   const String& rule_name_a: string containg name of rules to extract
//   ProductionRuleSet prset_a: production rules from which to extract rule
// 
// return: Vector<ProductionRule> containing all rules named rule_name_a
//
// this method extracts all rules named rule_name_a from the given set
// of production rules.  note that these rules are removed from the
// input set
//
ProductionRuleSet LanguageModelXML::extractRule(const String& rule_name_a,
						ProductionRuleSet& prset_a) const {
  ProductionRuleSet finished_set;
  ProductionRuleSet return_set;

  int32 i;

  for (i=0; i<prset_a.length(); i++) {
    if (prset_a(i).getRuleName().eq(rule_name_a)) {
      return_set.concat(prset_a(i));
    }
    else {
      finished_set.concat(prset_a(i));
    }
  }

  
  prset_a.assign(finished_set);

  return return_set;
}

// method: addRule
//
// arguments:
//   ProductionRuleSet prset_a: production rules to convert to xml
//   Vector<XMLToken>& token_vector_a: vector to which to add the xml rule
// 
// return: bool8 value indicating status
//
// this method converts a set of production rule alternatives (production
// rules having the same left-hand side) into a single xml format rule
//
bool8 LanguageModelXML::addRule(ProductionRuleSet& prset_a,
				  Vector<XMLToken>& token_vector_a) const {
  if (prset_a.length()==0) {
    return true;
  }

  
  int32 i;
  int32 depth=1;
  String rule_name;
  rule_name.assign(prset_a(0).getRuleName());

  // add start rule tag
  //
  addStartRuleTag(token_vector_a, rule_name, depth);

  
  // if there's only one possible expansion, we don't need
  // a one-of
  //
  if (prset_a.length()==1) {
    addAlternative(prset_a(0), token_vector_a, depth);
    
  }
  
  // otherwise, each rule is nested inside a one-of
  //
  else {
    // add the start one-of
    //
    addStartBranchTag(token_vector_a, depth);
    
    for (i=0; i<prset_a.length(); i++) {
      // add alternative between item tags
      //
      addStartItemTag(token_vector_a, depth);
      addAlternative(prset_a(i), token_vector_a, depth);
      addEndItemTag(token_vector_a, depth);      
    }

    // add the end one-of
    //
    addEndBranchTag(token_vector_a, depth);
  }

  // add end rule tag
  //
  addEndRuleTag(token_vector_a, depth);

  return true;
}

// method: addAlternative
//
// arguments:
//   ProductionRule pr_a: production rule to convert to xml
//   Vector<XMLToken>& token_vector_a: vector to which to add the xml rule
// 
// return: bool8 value indicating status
//
// this method converts a single production rule into xml format
//
bool8 LanguageModelXML::addAlternative(ProductionRule& pr_a,
					 Vector<XMLToken>& token_vector_a,
					 int32 depth_a) const {
  // if we can go to the first token, the rule is not empty
  // and we can proceed.  if we can't go to the first token,
  // the rule is empty and we add an epsilon
  //
  if (pr_a.gotoFirst()) {
    // check the type of the first token (it should be one
    // of these three) and handle it accordingly
    //
    if (pr_a.getType()==ProductionRuleTokenType::TERMINAL) {
      addCDATA(token_vector_a, pr_a.getValue(), depth_a);
    }
    else if (pr_a.getType()==ProductionRuleTokenType::NON_TERMINAL) {
      addRuleRef(token_vector_a, pr_a.getValue(), depth_a);
    }
    else if (pr_a.getType()==ProductionRuleTokenType::EPSILON) {
      addDummyItem(token_vector_a, depth_a);
    }

    // even though in theory we should only have at most one
    // concatenation and one other symbol, we'll allow them
    // to concat as many as they like
    //
    while (pr_a.gotoNext()) {
      // if the symbol isn't concatenation, skip it
      //
      if (pr_a.getType()==ProductionRuleTokenType::CONCATENATION) {
	// make sure there's a next symbol before proceeding
	//
	if (pr_a.gotoNext()) {
	  // if it's a terminal symbol, handle it appropriately
	  //
	  if (pr_a.getType()==ProductionRuleTokenType::TERMINAL) {
	    pr_a.gotoPrev();
	    
	    // if it has no weight, make it plain ol' cdata
	    //
	    if (pr_a.getWeight().eq(0)) {
	      pr_a.gotoNext();
	      addCDATA(token_vector_a, pr_a.getValue(), depth_a);
	    }

	    // otherwise, enclose it in one-of and item tags
	    //
	    else {
	      // because of the screwy XML rules, we have to add
	      // a one-of around this item to give it a weight
	      //
	      addStartBranchTag(token_vector_a, depth_a);
	      addStartItemTag(token_vector_a, depth_a, pr_a.getWeight());
	      pr_a.gotoNext();
	      addCDATA(token_vector_a, pr_a.getValue(), depth_a);
	      addEndItemTag(token_vector_a, depth_a);
	      addEndBranchTag(token_vector_a, depth_a);
	      
	    }
	  }

	  // if it's a non-terminal, handle it appropriately
	  //
	  else if (pr_a.getType()==ProductionRuleTokenType::NON_TERMINAL) {
	    pr_a.gotoPrev();
	    
	    // if it has no weight, make it plain ol' cdata
	    //
	    if (pr_a.getWeight().eq(0)) {
	      pr_a.gotoNext();
	      addRuleRef(token_vector_a, pr_a.getValue(), depth_a);
	    }

	    // otherwise, enclose it in one-of and item tags
	    //
	    else {
	      // because of the screwy XML rules, we have to add
	      // a one-of around this item to give it a weight
	      //
	      addStartBranchTag(token_vector_a, depth_a);
	      addStartItemTag(token_vector_a, depth_a, pr_a.getWeight());
	      pr_a.gotoNext();
	      addRuleRef(token_vector_a, pr_a.getValue(), depth_a);
	      addEndItemTag(token_vector_a, depth_a);
	      addEndBranchTag(token_vector_a, depth_a);
	    }
	  }
	}
      }
    }
  }
  else {
    addDummyItem(token_vector_a, depth_a);
  }

  return true;
}

// method: addRuleRef
//
// arguments:
//  Vector<XMLToken>& token_vector_a: vector to store tag in
//  String rule_name_a: rule name
//  int32& depth_a: depth to add token at
//
// return: bool8 value indicating status
//
// this method adds a ruleref with name rule_name_a to the token
// vector
//
bool8 LanguageModelXML::addRuleRef(Vector<XMLToken>& token_vector_a,
				     String rule_name_a,
				     int32& depth_a) const {
  XMLToken temp_token;

  String temp_string;
  temp_string.assign(L"#");

  temp_string.concat(rule_name_a);
  
  // attribute vector for start rule tag
  //
  Vector< Pair<String, String> > attrib_array;
  Pair<String, String> attrib_pair;
  attrib_pair.assign(URI, temp_string);
  attrib_array.concat(attrib_pair);
  temp_token.init(XMLToken::START_AND_END_TAG, RULEREF, attrib_array, depth_a);

  token_vector_a.concat(temp_token);

  return true;
}

// method: addStartRuleTag
//
// arguments:
//  Vector<XMLToken>& token_vector_a: vector to store tag in
//  String rule_name_a: rule name
//  int32& depth_a: depth to add token at
//
// return: bool8 value indicating status
//
// this method adds a start rule tag to the token vector with name
// rule_name_a.
//
bool8 LanguageModelXML::addStartRuleTag(Vector<XMLToken>& token_vector_a,
				     String rule_name_a,
				     int32& depth_a) const {
  XMLToken temp_token;

  String temp_string;

  temp_string.assign(rule_name_a);
  
  // attribute vector for start rule tag
  //
  Vector< Pair<String, String> > attrib_array;
  Pair<String, String> attrib_pair;
  attrib_pair.assign(ID, temp_string);
  attrib_array.concat(attrib_pair);
  temp_token.init(XMLToken::START_TAG, RULE, attrib_array, depth_a++);

  token_vector_a.concat(temp_token);

  return true;
}

// method: addEndRuleTag
//
// arguments:
//  Vector<XMLToken>& token_vector_a: vector to store tag in
//  int32& depth_a: depth to add token at
//
// return: bool8 value indicating status
//
// this method adds an end rule tag to the token vector
//
bool8 LanguageModelXML::addEndRuleTag(Vector<XMLToken>& token_vector_a,
					int32& depth_a) const {
  XMLToken temp_token;

  temp_token.init(XMLToken::END_TAG, RULE, --depth_a);

  token_vector_a.concat(temp_token);

  return true;
}

// method: addDummyItem
//
// arguments:
//  Vector<XMLToken>& token_vector_a: vector to store tag in
//  int32& depth_a: depth to add token at
//
// return: bool8 value indicating status
//
// this method adds a NULL ruleref to token_vector_a
//
bool8 LanguageModelXML::addDummyItem(Vector<XMLToken>& token_vector_a,
		     int32& depth_a) const {
  XMLToken temp_token;

  Vector< Pair<String, String> > attrib_array;
  Pair <String, String> special_attrib(SPECIAL, SPECIAL_NULL);

  attrib_array.concat(special_attrib);
  
  // create and add token
  //
  temp_token.init(XMLToken::START_AND_END_TAG, RULEREF,
		  attrib_array, depth_a);
  token_vector_a.concat(temp_token);
  
  return true;
}

// method: addCDATA
//
// arguments:
//  Vector<XMLToken>& token_vector_a: vector to store tag in
//  int32& depth_a: depth to add token at
//
// return: bool8 value indicating status
//
// this method adds CDATA to token_vector_a
//
bool8 LanguageModelXML::addCDATA(Vector<XMLToken>& token_vector_a,
				   const String& cdata_a,
				   int32& depth_a) const {
  XMLToken temp_token;
  
  // create and add token
  //
  temp_token.init(XMLToken::CDATA, cdata_a, depth_a);
  token_vector_a.concat(temp_token);
  
  return true;
}

// method: addStartItemTag
//
// arguments:
//  Vector<XMLToken>& token_vector_a: working token vector
//  int32 depth_a: depth
//  bool8 repeat_a: flag indicating whether to use repeat attribute
//
// return: bool8 value indicating status
//
// this method adds a start item tag to the token vector
//
bool8 LanguageModelXML::addStartItemTag(Vector<XMLToken>& token_vector_a,
					  int32& depth_a,
					  float32 weight_a,
					  bool8 repeat_a,
					  float32 repeat_prob_a) const {
  XMLToken temp_token;

  String weight_str;
  weight_str.concat((Float)weight_a);
  String repeat_prob_str;
  repeat_prob_str.concat((Float)repeat_prob_a);
  
  Vector< Pair<String, String> > attrib_array;
  Pair <String, String> repeat_attrib(REPEAT, REPEAT_LOOP_BACK);
  Pair <String, String> weight_attrib(WEIGHT, weight_str);
  Pair <String, String> repeat_prob_attrib(REPEAT_PROB, repeat_prob_str);

  // set up attributes
  //
  if (repeat_a) {
    attrib_array.concat(repeat_attrib);
  }

  if (!((Float)weight_a).eq(1)) {
    attrib_array.concat(weight_attrib);
  }

  if (repeat_a && !((Float)repeat_prob_a).eq(1)) {
    attrib_array.concat(repeat_prob_attrib);
  }
  
  // create and add start <item> tag
  //
  temp_token.init(XMLToken::START_TAG, ITEM,
		  attrib_array, depth_a++);
  token_vector_a.concat(temp_token);
  
  return true;
}

// method: addEndItemTag
//
// arguments:
//  Vector<XMLToken>& token_vector_a: working token vector
//  int32 depth_a: depth
//
// return: bool8 value indicating status
//
// this method adds an end item tag to the token vector
//
bool8 LanguageModelXML::addEndItemTag(Vector<XMLToken>& token_vector_a,
					int32& depth_a) const {
  XMLToken temp_token;

  // create and add end </item> tag
  //
  temp_token.init(XMLToken::END_TAG, ITEM, --depth_a);
  token_vector_a.concat(temp_token);

  return true;
}

// method: addStartBranchTag
//
// arguments:
//  Vector<XMLToken>& token_vector_a: working token vector
//  int32 depth_a: depth
//
// return: bool8 value indicating status
//
// this method adds a start branch tag to the token vector
//
bool8 LanguageModelXML::addStartBranchTag(Vector<XMLToken>& token_vector_a,
					    int32& depth_a) const {
  XMLToken temp_token;

  // create and add start <one-of> tag
  //
  temp_token.init(XMLToken::START_TAG, ONE_OF, depth_a++);
  token_vector_a.concat(temp_token);

  return true;
}

// method: addEndBranchTag
//
// arguments:
//  Vector<XMLToken>& token_vector_a: working token vector
//  int32 depth_a: depth
//
// return: bool8 value indicating status
//
// this method adds an end branch tag to the token vector
//
bool8 LanguageModelXML::addEndBranchTag(Vector<XMLToken>& token_vector_a,
					  int32& depth_a) const {
  XMLToken temp_token;

  // create and add end </one-of> tag
  //
  temp_token.init(XMLToken::END_TAG, ONE_OF, --depth_a);
  token_vector_a.concat(temp_token);

  return true;
}

// method: getXMLModel
//
// arguments:
//  none
//
// return: an xml rule model
//
// this method returns the XML model in memory
//
Pair< Vector< Vector< Rule > >, HierarchicalDigraph> LanguageModelXML::getXMLModel() {
  Pair< Vector< Vector< Rule > >, HierarchicalDigraph> ret;
  ret.first().assign(grammars_d);
  ret.second().assign(hg_d);
  return ret;
}
  
// method: setXMLModel
//
// arguments:
//  Pair< Vector< Vector< Rule > >, HierarchicalDigraph> xml_model_a: model
//  
// return: bool8 value indicating status
//
// this method sets the XML model in memory to the specified model
//
bool8 LanguageModelXML::setXMLModel(Pair< Vector< Vector< Rule > >, HierarchicalDigraph>
		    xml_model_a) {
  return (grammars_d.assign(xml_model_a.first()) &&
	  hg_d.assign(xml_model_a.second()));
}