path: root/grammer.cpp

#include "grammer.h"

#include <algorithm>
#include <limits>

using namespace Gram;

bool debug{false};

void Debug(std::string s)
{
 if (debug)
  std::cout << s << std::endl;
}

void Compiler::clear()
{
 nodes.clear();
 root_node_id = 0;
}

std::string Compiler::GetTypeOfNode(index_t node_id) const
{
 if (node_id >= nodes.size())
  throw std::runtime_error("GetTypeOfNode(): node_id="s + std::to_string(node_id) + ", nodes.size()="s + std::to_string(nodes.size()));
 return nodes[node_id].type;
}

bool Compiler::IsRootNode(index_t node_id) const
{
 auto& node {nodes[node_id]};
 return node.parent_node_id == node.node_id;
}

void Compiler::Validate() const
{
 // A program is non empty
 if (nodes.size() == 0)
  return;
  //throw std::runtime_error("Program is empty");

 // Consistency check for nodes
 if (root_node_id >= nodes.size())
  throw std::runtime_error("Bad root node: "s + std::to_string(root_node_id) + " vs. "s + std::to_string(nodes.size()));
 
 // Start symbol on top
 if (GetTypeOfNode(root_node_id) != m_top)
  throw std::runtime_error("Root node not start symbol!");

 // All nodes filled
 for (const auto& node: nodes) {
  if (node.child_ids.size() != bnf[node.type][node.variant].size())
   throw std::runtime_error("Node not filled: "s + node.type + "["s + std::to_string(node.variant) + "]"s);
 }
}

bool Compiler::rootIsStartSymbol() const
{
 return GetTypeOfNode(root_node_id) == m_top;
}

bool Gram::ChildIdIsToken(int32_t child_id)
{
 return child_id < 0;
}

index_t Gram::TokenIdFromChildId(int32_t child_id)
{
 return index_t(-child_id) - 1;
}

int32_t Gram::ChildIdFromTokenId(index_t token_id)
{
 return -1 - int32_t(token_id);
}

void Compiler::DumpTree()
{
 Debug("= Dump =======================================");
 Debug("nodes.size()="s + std::to_string(nodes.size()));
 if (nodes.size() > 0) {
  if (0) {
   Debug("--- Nodes ------------------------------------");
   Debug("root_node_id="s + std::to_string(root_node_id));
   for (const auto& node : nodes) {
    std::string line{"("s + std::to_string(node.node_id) + "):"s};
    for (const auto& child : node.child_ids) {
     line += " "s;
     if (ChildIdIsToken(child))
      line += "t"s + std::to_string(TokenIdFromChildId(child));
     else
      line += std::to_string(child);
    }
    Debug(line);
   }
  }

  Debug("--- Tree -------------------------------------");
  std::deque<std::pair<int32_t, size_t>> todo{std::pair<int32_t, size_t>{static_cast<int32_t>(root_node_id), 0}}; // id, indent
  while (!todo.empty()) {
   auto [current_index, indent] {todo.front()};
   todo.pop_front();

   std::string line(indent, ' ');
   if (ChildIdIsToken(current_index)) {
    index_t token_id {TokenIdFromChildId(current_index)};
    line += "Token("s + std::to_string(token_id) + "): "s + tokens[token_id].type + "("s + tokens[token_id].value + ")"s;
   } else {
    auto& node {nodes[current_index]};
    line += "Node("s + std::to_string(current_index) + "): "s + node.type + "/" + std::to_string(node.variant);
   
    auto child_ids{node.child_ids};
    for (int i = 0; i < child_ids.size(); i++) {
     todo.insert(todo.begin() + i, std::pair<int32_t, size_t>{child_ids[i], indent + 1});
    }
   }
   Debug(line);
  }
 }
 Debug("==============================================");
}

bool Compiler::treeIsComplete() const
{
 return nodes.size() > 0 && rootIsStartSymbol();
}

index_t Compiler::AddNode(const std::string& child_type, index_t parent_index)
{
 TreeNode& parent {nodes[parent_index]};
 index_t index = nodes.size();
 parent.child_ids.push_back(index);

 index_t variant{};
 std::deque<std::pair<std::string, index_t>> alternatives;

 const auto& variants { bnf[child_type] };
 bool found{false};
 for (int i = 0; i < variants.size(); i++) { // variants
  if (!found) { // use first match
   variant = i;
   found = true;
  } else { // defer all others
   alternatives.emplace_back(child_type, i);
  }
 }

 nodes.emplace_back(TreeNode{parent_index, index, child_type, variant, std::vector<int32_t>{}});
 //root stays

 Debug("AddNode(): "s + nodes[parent_index].type + "->"s + child_type + ": "s + std::to_string(index));
 DumpTree();

 return index;
}

size_t Compiler::minimumSymbolsNeeded(std::string symbol)
{
 if (isTerminal(bnf, symbol)) {
  return 1;
 } else {
  auto it_min{m_min.find(symbol)};
  if (it_min != m_min.end())
   return it_min->second;
  m_min[symbol] = std::numeric_limits<size_t>::max();

  auto it{bnf.find(symbol)};
  if (it != bnf.end()) {
   size_t minimum{std::numeric_limits<size_t>::max()};
   
   for (const auto& list: it->second) {
    minimum = std::min(minimum, minimumSymbolsNeeded(list));
   }

   m_min[symbol] = minimum;

   return minimum;
  } else
   throw std::runtime_error("ICE: Symbol "s + symbol + " expected in BNF"s);
 }
}

size_t Compiler::minimumSymbolsNeeded(std::vector<std::string> symbol_list)
{
 size_t result{0};

 for (const auto& symbol: symbol_list) {
  size_t min { minimumSymbolsNeeded(symbol) };
  if (min == std::numeric_limits<size_t>::max())
   return min;
  result += min;
 }

 return result;
}

/// begin, end: indexes in tokens list
bool Compiler::match(std::vector<std::string> symbol_list, size_t begin, size_t end)
{
 // TODO: isTerminal() necessary here?
 
 // match terminal symbols at start
 while (begin < end && isTerminal(bnf, tokens[begin].type) && symbol_list.size() > 0 && symbol_list.front() == tokens[begin].type) {
  begin++;
  symbol_list.erase(symbol_list.begin());
 }

 // match terminal symbols at end
 while (begin < end && isTerminal(bnf, tokens[end - 1].type) && symbol_list.size() > 0 && symbol_list.back() == tokens[end - 1].type) {
  end--;
  symbol_list.erase(symbol_list.end() - 1);
 }

 // matching of empty lists
 if (symbol_list.size() == 0) {
  if (begin == end) // only match real empty list
   return true;
  return false;
 }

 // now, symbol_list[begin .. end - 1] has size > 0 and contains non-terminal symbols at start and end
 
 // resolve first symbol
 auto it{bnf.find(symbol_list.front())};
 if (it != bnf.end()) {
  for (std::vector<std::string> list: it->second) { // iterate over alternatives
   list.insert(list.end(), symbol_list.begin() + 1, symbol_list.end());
   if (minimumSymbolsNeeded(list) > end - begin) // stop recursion
    continue;

   // TODO: recurse last?
   if (match(list, begin, end))
    return true;
  }
 } else
  return false; // terminal symbol not found in bnf, non-matching
 
 return false; // no match found
}

bool Compiler::match(std::string symbol, size_t begin, size_t end)
{
 std::vector<std::string> symbol_list{symbol};
 return match(symbol_list, begin, end);
}

Compiler::Compiler(BNF& bnf, const std::string& top): bnf(bnf), m_top(top), ReverseBNF{Reverse(bnf)}, reversedFirst{reverseFirst(bnf)}
{
 //std::cout << "DEBUG: " << m_top << std::endl;
 
 //
 // prepare helper cache (TODO: remove this ugly workaround for remaining bad marker elements)
 //
 minimumSymbolsNeeded("translation-unit");
 // remove bad marker elements
 auto it{m_min.begin()};
 while (it != m_min.end()) {
  if (it->second == std::numeric_limits<size_t>::max()) {
   it = m_min.erase(it);
  } else {
   ++it;
  }
 }
 minimumSymbolsNeeded("translation-unit");
}

std::pair<index_t, std::vector<TreeNode>> Compiler::compile(std::vector<Token> p_tokens)
{
 clear();
 tokens = p_tokens;

 if (tokens.size() == 0)
  throw std::runtime_error("No tokens!");

#if 0
 //
 // bottom-up algorithm
 //
 while (!treeIsComplete()) {
  if (!FillTree()) {
   TrackBack();
  } else if (!AddRootNode()) {
   TrackBack();
  } else if (!FillTree()) {
   TrackBack();
  }
 }
#else
 //
 // top-down algorithm
 //
 if (!match(m_top, 0, tokens.size()))
  throw std::runtime_error("Compile error.");
#endif

 Validate();

 return std::pair<index_t, std::vector<TreeNode>>{root_node_id, nodes};
}