/*
 * Copyright (c) 2011-2013 Vlad Voicu
 * Licensed under the Simplified BSD license.
 * See Documentation/Licenses/BSD-simplified.txt for more information.
 */

#include <SwifTools/SpellParser.h>

#include <boost/spirit/include/lex_lexertl.hpp>
#include <boost/bind.hpp>
#include <boost/ref.hpp>
#include <boost/numeric/conversion/cast.hpp>

#include <string>

namespace lex = boost::spirit::lex;

namespace Swift {

template <typename Lexer>
struct word_count_tokens : lex::lexer<Lexer>
{
	word_count_tokens()
	{
	// define tokens (regular expresions) to match strings
	// order is important
		this->self.add
			("w{3}.[^ ]+", ID_WWW)
			("http:\\/\\/[^ ]+", ID_HTTP)
			("\\w{1,}['?|\\-?]?\\w{1,}", ID_WORD)
			(".", ID_CHAR);
	}
};

struct counter
{
	typedef bool result_type;
	// the function operator gets called for each of the matched tokens
	template <typename Token>
	bool operator()(Token const& t, PositionPairList& wordPositions, std::size_t& position) const
	{
		switch (t.id()) {
			case ID_WWW:
				position += boost::numeric_cast<size_t>(t.value().size());
				break;
			case ID_HTTP:
				position += boost::numeric_cast<size_t>(t.value().size());
				break;
			case ID_WORD:       // matched a word
				wordPositions.push_back(boost::tuples::make_tuple(position, position + boost::numeric_cast<size_t>(t.value().size())));
				position += boost::numeric_cast<size_t>(t.value().size()); 
				break;
			case ID_CHAR:       // match a simple char
				++position; 
				break;
		}
		return true;        // always continue to tokenize
	}
};

void SpellParser::check(const std::string& fragment, PositionPairList& wordPositions) {
	std::size_t position = 0;
	// create the token definition instance needed to invoke the lexical analyzer
	word_count_tokens<lex::lexertl::lexer<> > word_count_functor;
	char const* first = fragment.c_str();
	char const* last = &first[fragment.size()];
	lex::tokenize(first, last, word_count_functor, boost::bind(counter(), _1, boost::ref(wordPositions), boost::ref(position)));
}

}