#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#include <iostream>
#include <string>
#include <boost/program_options.hpp>

#include "profile.h"
#include "mrsReader.h"
#include "unicode.h"

using namespace std;

void parse_options(int argc, char **argv, string *gprof, string *tprof,
	bool *ignore, bool *stats, bool *quiet, int *itemno, int *parseno,
	bool *verbose, bool *veryverbose, bool *score_p);

void addCounts(char index, map<string,map<char,int> > &totals, 
	map<string,int> &counts);

//prints raw stats in tab-delimited form for use in statistical significance
//testing
void printStats(tIid item, map<string,int> &ccounts, map<string,int> &gcounts,
	map<string,int> &tcounts);

void printScores(map<string,map<char,int> > &totals);

//tests specified or (default) top reading against (all) gold and adds counts to
//total
pair<pair<tIid,int>,string> finishItem(Profile &test, tIid item, int parse_id, 
	map<string, map<char,int> > &totals, vector<tMRS *> &goldmrs, bool stats,
	pair<pair<tIid,int>,string> lastline, bool verbose, bool veryverbose, 
	bool ignore);

//tests all readings against (all) gold and outputs individual fscores in an
//@-delimited file format
pair<pair<tIid,int>,string> finishFullItem(Profile &test, tIid item, 
	vector<tMRS *> &goldmrs, pair<pair<tIid,int>,string> lastline);

string subspan(string itemstr, string span); 

int main (int argc, char **argv)
{
	string gprofile, tprofile;
	bool ignore, stats, quiet, verbose, veryverbose, score_p;
	int itemno, parseno;
	parse_options(argc, argv, &gprofile, &tprofile, &ignore, &stats, &quiet,
		&itemno, &parseno, &verbose, &veryverbose, &score_p); 
	initialize_encoding_converter("utf-8"); 

	Profile gold(gprofile);
	Profile test(tprofile);

	vector<tMRS *> goldmrs;
	pair<pair<tIid,int>, string> gres = gold.getResult(MRS);
	pair<pair<tIid,int>, string> lastline 
		= pair<pair<tIid,int>, string>(pair<tIid,int>(-1,-1),string());
	tIid item = -1;
	map<string, map<char,int> > totals;

	while (gres.first.first >= 0 //still returning results
		&& (itemno == -1 || gres.first.first <= itemno)) { 
		if (itemno > -1 && gres.first.first < itemno) {
			//looking for a specific item
			gres = gold.getResult(MRS);
			continue;	// and not there yet
		}
		if (gres.first.first != item) { //new item
			if (item != -1 && goldmrs.size() > 0) { //finish last item
				if (score_p)
					lastline = finishFullItem(test, item, goldmrs, lastline);
				else
					lastline = finishItem(test, item, parseno, totals, goldmrs, 
						stats, lastline, verbose, veryverbose, ignore);
			}
		}
		item = gres.first.first;
		if (!ignore || test.getReadings(item) > 0) { 
		//if we are ignoring items that have a gold reading but didn't parse in
		//test, skip this if there are no test readings
			if (gold.isGold(item, gres.first.second) && !gres.second.empty()) {
				//make sure it is annotated as gold (maybe an option to skip this
				//check?)
				tMRS *gmrs = new tMRS(gres.second);
				goldmrs.push_back(gmrs);
			}
		}
		gres = gold.getResult(MRS);
	}
	//finish last item
	if ((itemno == -1 || item == itemno) && goldmrs.size() > 0) {
		if (score_p)
			lastline = finishFullItem(test, item, goldmrs, lastline);
		else
			lastline = finishItem(test, item, parseno, totals, goldmrs, 
				stats, lastline, verbose, veryverbose, ignore);
		if (!quiet)
			printScores(totals);
	}
	else
		cerr << "Item " << itemno << " was not found in the gold profile."
			<< endl;
}

void parse_options(int argc, char **argv, string *gprof, string *tprof,
	bool *ignore, bool *stats, bool *quiet, int *itemno, int *parseno, 
	bool *verbose, bool *veryverbose, bool *score_p)
{
	namespace po = boost::program_options;
	po::options_description visible("Options");
	visible.add_options()
		("help,h", "This usage information.")
		("ignore,i", "Ignore gold triples where the test parse failed.")
		("stats,s", "Print raw counts per item (for significance testing).")
		("quiet,q", "Don't print totals. (Only makes sense with stats option).")
		("verbose,v", "Print unmatched triples.")
		("debug,d", "Print all triples and counts.")
		("num,n", po::value<int>(itemno)->default_value(-1),
			"Select a single item to evaluate (-1 for all items).")
		("parse,p", po::value<int>(parseno)->default_value(0),
			"Reading to evaluate, defaults to 0 for the top reading")
		("full,f", "Score every reading and output an fscore file")
	;

	po::options_description hidden("Hidden options");
	hidden.add_options()
		("goldprof", po::value<string>(gprof), "gold profile")
		("testprof", po::value<string>(tprof), "test profile")
	;

	po::options_description cmd_line ("Command line options");
	cmd_line.add(visible).add(hidden);

	po::positional_options_description p;
	p.add("goldprof", 1).add("testprof", 1);

	po::variables_map vm;
	try {
		po::store(po::command_line_parser(argc, argv).
			options(cmd_line).positional(p).run(), vm);
		notify(vm);

		if (vm.count("help")) {
			cout << "Usage: " << argv[0] << " [options] "
				<< "gold-profile test-profile" << endl;
			cout << visible << endl;
			exit(0);
		}

		if (!vm.count("goldprof") || !vm.count("testprof")) {
			cerr << "Insufficient arguments given." << endl;
			cerr << "Usage: " << argv[0] << " [options] "
				<< "gold-profile test-profile" << endl;
			cerr << visible << endl;
			exit(1);
		}
		
		if (vm.count("ignore"))
			*ignore = true;
		else
			*ignore = false;
		if (vm.count("verbose"))
			*verbose = true;
		else
			*verbose = false;
		if (vm.count("debug"))
			*veryverbose = true;
		else
			*veryverbose = false;
		if (vm.count("stats"))
			*stats = true;
		else
			*stats = false;
		if (vm.count("quiet")) {
			if (!vm.count("stats"))
				cerr << "WARNING: using -q without -s means no output will be produced."
					<< " Continuing anyway." << endl;
			*quiet = true;
		}
		else
			*quiet = false;
		if (vm.count("full")) {
			*score_p = true;
			*quiet = true;
			*stats = false;
		}
		else
			*score_p = false;
	} catch (po::error& e ) {
		cerr << "Error: " << e.what() << endl;
		cerr << "Usage: " << argv[0] << " [options] "
			<< "gold-profile test-profile" << endl;
		cerr << visible << endl;
		exit(1);
	}
}

//add counts for a single comparison to the total, index specifies whether the
//counts are for gold (G), test (T) or correct (C)
void addCounts(char index, map<string,map<char,int> > &totals, map<string,int> &counts)
{
	for (map<string,int>::iterator it = counts.begin();
		it != counts.end(); ++it) {
		if (totals.count(it->first) == 0)
			totals[it->first] = map<char,int>();
		if (totals[it->first].count(index) == 0)
			totals[it->first][index] = 0;
		totals[it->first][index]+= it->second;
	}
}

void printStats(tIid item, map<string,int> &ccounts, map<string,int> &gcounts,
	map<string,int> &tcounts)
{	
	cout << item << ".gz\t" 
		<< ccounts["ALL"] << "\t"
		<< gcounts["ALL"] << "\t"
		<< tcounts["ALL"] << "\t"
		<< ccounts["ARGS"] << "\t"
		<< gcounts["ARGS"] << "\t"
		<< tcounts["ARGS"] << "\t"
		<< ccounts["NAMES"] << "\t"
		<< gcounts["NAMES"] << "\t"
		<< tcounts["NAMES"] << "\t"
		<< ccounts["NAMES"]+ccounts["ARGS"] << "\t"
		<< gcounts["NAMES"]+gcounts["ARGS"] 
			<< "\t"
		<< tcounts["NAMES"]+tcounts["ARGS"] << "\t"
		<< endl;
}

void printScores(map<string,map<char,int> > &totals)
{
	cout << setw(10) << " " << setw(7) << "    PREC  " << setw(7) << "  REC  "
		<< setw(7) << "F" << setw(7) << "   TOTAL" << endl;
	cout << setw(10) << "EDM" << " " << fixed << setprecision(3)
		<< setw(7) << (totals["ALL"]['T']>0?
			static_cast<double>(totals["ALL"]['C'])/totals["ALL"]['T']:0)
		<< " " << setw(7) << (totals["ALL"]['G']>0?
			static_cast<double>(totals["ALL"]['C'])/totals["ALL"]['G']:0)
		<< " " << setw(7) << (totals["ALL"]['G']>0?
			(2.0*totals["ALL"]['C'])/(totals["ALL"]['T']+totals["ALL"]['G']):0)
		<< " " << setw(7) << totals["ALL"]['G']
		<< endl;
	cout << setw(10) << "EDM_NA" << " " << fixed << setprecision(3)
		<< setw(7) << (totals["NAMES"]['T']>0?
			(static_cast<double>(totals["ARGS"]['C']) + totals["NAMES"]['C'])
			/(totals["ARGS"]['T']+totals["NAMES"]['T']):0)
		<< " " << setw(7) << (totals["NAMES"]['G']>0?
			(static_cast<double>(totals["ARGS"]['C']) + totals["NAMES"]['C'])
			/(totals["ARGS"]['G']+totals["NAMES"]['G']):0)
		<< " " << setw(7) << (totals["NAMES"]['G']>0?
			(2.0*(totals["ARGS"]['C']+totals["NAMES"]['C']))
			/(totals["ARGS"]['T']+totals["NAMES"]['T']+totals["ARGS"]['G']
			+totals["NAMES"]['G']):0)
		<< " " << setw(7) << totals["ARGS"]['G']+totals["NAMES"]['G']
		<< endl;
	cout << setw(10) << "EDM_A" << " " << fixed << setprecision(3)
		<< setw(7) << (totals["ARGS"]['T']>0?
			static_cast<double>(totals["ARGS"]['C'])/totals["ARGS"]['T']:0)
		<< " " << setw(7) << (totals["ARGS"]['G']>0?
			static_cast<double>(totals["ARGS"]['C'])/totals["ARGS"]['G']:0)
		<< " " << setw(7) << (totals["ARGS"]['G']>0?
			(2.0*totals["ARGS"]['C'])/(totals["ARGS"]['T']+totals["ARGS"]['G']):0)
		<< " " << setw(7) << totals["ARGS"]['G']
		<< endl;
	cout << setw(10) << "EDM_N" << " " << fixed << setprecision(3)
		<< setw(7) << (totals["NAMES"]['T']>0?
			static_cast<double>(totals["NAMES"]['C'])/totals["NAMES"]['T']:0)
		<< " " << setw(7) << (totals["NAMES"]['G']>0?
			static_cast<double>(totals["NAMES"]['C'])/totals["NAMES"]['G']:0)
		<< " " << setw(7) << (totals["NAMES"]['G']>0?
			(2.0*totals["NAMES"]['C'])/(totals["NAMES"]['T']+totals["NAMES"]['G']):0)
		<< " " << setw(7) << totals["NAMES"]['G']
		<< endl;
	cout << setw(10) << "EDM_P" << " " << fixed << setprecision(3)
		<< setw(7) << (totals["PROPS"]['T']>0?
			static_cast<double>(totals["PROPS"]['C'])/totals["PROPS"]['T']:0)
		<< " " << setw(7) << (totals["PROPS"]['G']>0?
			static_cast<double>(totals["PROPS"]['C'])/totals["PROPS"]['G']:0)
		<< " " << setw(7) << (totals["PROPS"]['G']>0?
			(2.0*totals["PROPS"]['C'])/(totals["PROPS"]['T']+totals["PROPS"]['G']):0)
		<< " " << setw(7) << totals["PROPS"]['G']
		<< endl << endl;
	for (map<string,map<char,int> >::iterator it = totals.begin();
		it != totals.end(); ++it) {
		if (it->first != "ALL" && it->first != "ARGS" &&
			it->first != "NAMES" && it->first != "PROPS") {
			cout << setw(10) << it->first << " " << fixed << setprecision(3)
				<< setw(7) << (totals[it->first]['T']>0?
					static_cast<double>(totals[it->first]['C'])/
					totals[it->first]['T']:0)
				<< " " << setw(7) << (totals[it->first]['G']>0?
					static_cast<double>(totals[it->first]['C'])/
					totals[it->first]['G']:0)
				<< " " << setw(7) << (totals[it->first]['G']>0?
					(2.0*totals[it->first]['C'])/
					(totals[it->first]['T']+totals[it->first]['G']):0)
				<< " " << setw(7) << totals[it->first]['G']
				<< endl;
		}
	}
}

pair<pair<tIid,int>,string> finishItem(Profile &test, tIid item, int parse_id, 
	map<string, map<char,int> > &totals, vector<tMRS *> &goldmrs,
	bool stats, pair<pair<tIid,int>,string> lastline, bool verbose, 
	bool veryverbose, bool ignore)
{
	pair<pair<tIid,int>,string> tres = lastline;
	if (test.getReadings(item) <= parse_id) { //just add counts
		if (parse_id > 0)
			cerr << "No reading " << parse_id << " available for item "
				<< item << "." << endl;
		if (!ignore) {
			addCounts('G', totals, goldmrs[0]->getCounts());
			if (veryverbose) {
				cout << "GOLD TRIPLES: " << item << endl;
				goldmrs[0]->printTriples();
				goldmrs[0]->printCounts();
			}
			if (stats) {
				map<string,int> zeros;
				printStats(item, zeros, goldmrs[0]->getCounts(), zeros);
			}
		}
	} else {
		if (tres.first.first == -1)
			tres = test.getResult(MRS);
		while (tres.first.first != -1 && tres.first.first < item)
			tres = test.getResult(MRS);
		if (tres.first.first != item) {
			cerr << "Item " << item << " is not in the test profile." << endl;
			if (!ignore) {
				addCounts('G', totals, goldmrs[0]->getCounts());
				if (stats) {
					map<string,int> zeros;
					printStats(item, zeros, goldmrs[0]->getCounts(), zeros);
				}
			}
		} else {
			while (tres.first.first != -1 && tres.first.second < parse_id) 
				tres = test.getResult(MRS);
			if (tres.first.first != item || tres.first.second != parse_id
				|| tres.second.empty()) {
				cerr << "No reading " << parse_id << " for item " << item 
					<< " in the test profile." << endl;
				if (!ignore) {
					addCounts('G', totals, goldmrs[0]->getCounts());
					if (stats) {
						map<string,int> zeros;
						printStats(item, zeros, goldmrs[0]->getCounts(), zeros);
					}
				}
			} else {//we have an appropriate test reading to compare

				tMRS *tmrs = new tMRS(tres.second);
				pair<map<string,int>, vector<Unmatched> > maxscores;
				double maxscore = 0;
				tMRS *bestgold = NULL;
				//compare this reading to each gold MRS
				for (vector<tMRS *>::iterator it = goldmrs.begin();
					it != goldmrs.end(); ++it) {
					pair<map<string,int>, vector<Unmatched> > correct 
						= (*it)->compareTriples(*tmrs);
					double fscore = (*it)->getCounts()["ALL"]>0?
						((2.0*correct.first["ALL"])/
						(tmrs->getCounts()["ALL"]+(*it)->getCounts()["ALL"])):0;
					if (bestgold == NULL || fscore > maxscore) {	
						//and use the match with the highest EDM fscore as the final
						//match
						maxscores = correct;
						maxscore = fscore;
						bestgold = *it;
					}
				}
				addCounts('G', totals, bestgold->getCounts());
				addCounts('T', totals, tmrs->getCounts());
				addCounts('C', totals, maxscores.first);
				if (veryverbose) { //print all triples (debugging)
					cout << "GOLD TRIPLES: " << item << endl;
					bestgold->printTriples();
					bestgold->printCounts();
					cout << "TEST TRIPLES: " << item << endl;
					tmrs->printTriples();
					tmrs->printCounts();
				}
				if(verbose && !maxscores.second.empty()) {
					// print unmatched triples
					cout << "Unmatched triples for " << item << 
						":" << test.getItem(item) << endl;
					sort(maxscores.second.begin(), maxscores.second.end(), 
						unmatched_cmp());
					for (vector<Unmatched>::iterator it = maxscores.second.begin();
						it != maxscores.second.end(); ++it) {
							ostringstream relstr1, relstr2;
							string ss(subspan(test.getItem(item), it->span1));
							if (ss.length() > 23 ) 
								ss = string(ss.substr(0,20)+"...");
							relstr1 << ss << " " << it->span1;
							if ((it->span2).empty())
								relstr2 << it->name2.substr(0,23);
							else {
								ss = string(subspan(test.getItem(item), it->span2));
								if (ss.length() > 23 ) 
									ss = string(ss.substr(0,20)+"...");
								relstr2 << ss << " " << it->span2;
							}
							cout << it->dir << "  " 
								<< left << setw(33) << relstr1.str() << " "
								<< left << setw(7) << it->pred << " "
								<< relstr2.str() << endl;
					}
				}
					
				if (stats) 
					printStats(item, maxscores.first, bestgold->getCounts(), 
						tmrs->getCounts());
			}
		}
	}
	for (vector<tMRS *>::iterator it = goldmrs.begin();
		it != goldmrs.end(); ++it) 
		delete *it;
	goldmrs.clear();
	return tres;
}

pair<pair<tIid,int>,string> finishFullItem(Profile &test, tIid item, 
	vector<tMRS *> &goldmrs, pair<pair<tIid,int>,string> lastline)
{
	//record lastline to save re-reading the entire file every time
	pair<pair<tIid,int>,string> tres = lastline;
	if (tres.first.first == -1)
		tres = test.getResult(MRS);
	while (tres.first.first != -1 && tres.first.first < item)
		tres = test.getResult(MRS);

	while (tres.first.first == item && !tres.second.empty()) {
		tMRS *tmrs = new tMRS(tres.second);
		pair<map<string,int>, vector<Unmatched> > maxscores;
		double maxscore = 0;
		tMRS *bestgold = NULL;
		// compare analysis to each gold
		for (vector<tMRS *>::iterator it = goldmrs.begin();
			it != goldmrs.end(); ++it) {
			pair<map<string,int>, vector<Unmatched> > correct 
				= (*it)->compareTriples(*tmrs);
			double fscore = (*it)->getCounts()["ALL"]>0?
				((2.0*correct.first["ALL"])/
				(tmrs->getCounts()["ALL"]+(*it)->getCounts()["ALL"])):0;
			if (bestgold == NULL || fscore > maxscore) {	
				//and use the match with the highest EDM fscore as the final match
				maxscores = correct;
				maxscore = fscore;
				bestgold = *it;
			}
		}
		//print item no, parse no and fscores
		cout << item << "@" << tres.first.second;
		cout << "@" << fixed << setprecision(3) 
			<< (bestgold->getCounts()["ALL"]>0?
			(2.0*maxscores.first["ALL"])/
			(tmrs->getCounts()["ALL"]+bestgold->getCounts()["ALL"]):0);
		cout << "@" << fixed << setprecision(3) 
			<< (bestgold->getCounts()["NAMES"]>0?
			(2.0*(maxscores.first["ARGS"]+maxscores.first["NAMES"]))
			/(tmrs->getCounts()["ARGS"]+tmrs->getCounts()["NAMES"]
			+bestgold->getCounts()["ARGS"]+bestgold->getCounts()["NAMES"]):0);
		cout << "@" << fixed << setprecision(3)
			<< (bestgold->getCounts()["ARGS"]>0?
			(2.0*maxscores.first["ARGS"])
			/(tmrs->getCounts()["ARGS"] +bestgold->getCounts()["ARGS"]):0);
		cout << "@" << fixed << setprecision(3)
			<< (bestgold->getCounts()["NAMES"]>0?
			(2.0*maxscores.first["NAMES"])
			/(tmrs->getCounts()["NAMES"] +bestgold->getCounts()["NAMES"]):0);
		cout << "@" << fixed << setprecision(3)
			<< (bestgold->getCounts()["PROPS"]>0?
			(2.0*maxscores.first["PROPS"])
			/(tmrs->getCounts()["PROPS"] +bestgold->getCounts()["PROPS"]):0);
		cout << endl;
		tres = test.getResult(MRS);
	}

	for (vector<tMRS *>::iterator it = goldmrs.begin();
		it != goldmrs.end(); ++it) 
		delete *it;
	goldmrs.clear();
	return tres;
}

string subspan(string itemstr, string span) 
{
	string substr;
	if (span.at(0) != '<')
		return substr;

	int mid = span.find(':');
	if (mid == string::npos)
		return substr;
	istringstream startstr(span.substr(1, mid-1));	
	istringstream endstr(span.substr(mid+1, span.find('>', mid)-(mid+1)));
	int start, end;
	startstr >> start;
	endstr >> end;
	if (start >= 0 && end > 0) {
		UnicodeString uitem = Conv->convert(itemstr);
		if (start < uitem.length() && end <= uitem.length()) {
			UnicodeString subuitem(uitem, start, end-start);
			substr = Conv->convert(subuitem);
		} else
			return substr;
	} else 
		return substr;

	return substr;
}