#!/usr/bin/perl 

use warnings;

#
# This program takes a file called ./allocine.html and parses it
# to generate three tab-separated text files: 
#  - ./allocine-theater.txt
#  - ./allocine-movie.txt
#  - ./allocine-perf.txt
# Those files can then be used almost as-is to inject in an SQL database.
#
# $Id: parse.perl,v 1.29 2011-06-20 21:37:42 audran Exp $
#
# Copyright (c) Audran Le Baron 2003-2005
#

##########################
# Program initialization #
##########################

open (FILE_IN, '<allocine.html');
open (THEATER, '>allocine-theater.txt'); # Movie theaters (id, name, address, 
                                         #                 district, metro, multiplexity)
open (MOVIE,   '>allocine-movie.txt');   # Movies (id, title)
open (PERF,    '>allocine-perf.txt');    # Performances (theater, movie, day, hour, version, ads)
    
%state = ("Start",        0, 
	  "Theater",      1,
	  "Address",      2, 
          "Metro",        3,
	  "Week_day",     4,
	  "Movie",        5,
	  "Perf_version", 6,
	  "Perf_details", 7);

my $current_state = $state {"Start"};

my $theater;      # Movie theater name
my $theater_id;   # Movie theater allocine id
my $multiplexity; # 2 = UGC, Gaumont and the like
                  # 1 = MK2
                  # 0 = Others
my $address;      # Movie theater address
my $district;     # Movie theater district (1 to 20)
my $metro;        # Movie theater metro station
my $title;        # Movie title
my $movie_id;     # Movie allocine id
my $week_day;     # A week day's id (Wed = 0, ..., Tue = 6) 
my %hours;        # Table of perf hours => ads durations
my $version;      # VO or VF

my $DEBUG = 0;    # Display debug messages (No = 0, Yes = 1)

#######################
# Program subroutines #
#######################

####
# parse_theater ($line)
# -> tries to parse $line for a new movie theater
# -> returns 1 upon success, 0 otherwise
####
sub parse_theater {

    my $line = $_[0];

    if ($line =~ /<a href=\"\/seance\/salle_gen_csalle=([A-Z])(\d{4}).html\">([^<]+)<\/a>/) {
        if ($1 eq "C") {
            # Theater in Paris
            $theater = $3;
            $theater_id = "C$2";
            $multiplexity = 0;
            $address = "";
            $district = "0";
            $metro = "";
            if ($theater =~ /^(Gaumont|UGC|Bienvenue Montparnasse|Bretagne)/) {
                $multiplexity = 2;
            } elsif ($theater =~ /^MK2/) {
                $multiplexity = 1;
            }
            $current_state = $state {"Theater"};
            if ($DEBUG) {
                print "[$theater_id: $theater] -> Multiplexity = $multiplexity\n";
            }        
        } else {
            # Theater not in Paris => skip
            $current_state = $state {"Start"};
        }
        return 1;
    } else {
        return 0;
    }
}

####
# parse_address ($line)
# -> tries to parse $line for a movie theater address
# -> returns 1 upon success, 0 otherwise
####
sub parse_address {

    my $line = $_[0];

    if ($line =~ / +(.+? 75[01]([0-2][0-9]) Paris)/) {
        $address = $1;
	$district = $2;
        if ($DEBUG) {
            print " $address [$district]\n";
        }
        $current_state = $state {"Address"};
        return 1;
    } else {
        return 0;
    }
}

####
# parse_metro ($line)
# -> tries to parse $line for a movie theater metro station
# -> returns 1 upon success, 0 otherwise
####
sub parse_metro {

    my $line = $_[0];

    if ($line =~ / +Métro /) {
        $line =~ s/.*Métro( et RER)? //;
        $line =~ s/\r//;
        $line =~ s/\..*//;
        $line =~ s/<\/.*//;
        $line =~ s/, bus .*//;
        $line =~ s/Ligne [^,]+, //;
        chomp ($line);
        $metro = $line;
        if ($DEBUG) {
            print " Métro: $metro\n";
        }
        $current_state = $state {"Metro"};
        return 1;
    } else {
        return 0;
    }
}

####
# parse_week_day ($line)
# -> tries to parse $line for a new week day tab
# -> returns 1 upon success, 0 otherwise
####
sub parse_week_day {

    my $line = $_[0];

    if ($line =~ /<div class=\"pane[ hide]*\" rel=\"([0-6])\">/) {
        $week_day = $1;
        if ($DEBUG) {
            print "  Day $week_day\n";
        }
        $current_state = $state {"Week_day"};
        return 1;
    } else {
        return 0;
    }
}

####
# parse_movie ($line)
# -> tries to parse $line for a new movie
# -> returns 1 upon success, 0 otherwise
####
sub parse_movie {

    my $line = $_[0];

    if ($line =~ /<a class=\"underline\" href=\"\/film\/fichefilm_gen_cfilm=(\d+)\.html\">([^<]+)<\/a>/) {
        $movie_id = $1;
        $title = $2;
	$version = "";
	%hours = (); # array of (hour:min, ads)
	if ($DEBUG) {
            print "   [$movie_id] $title\n";
        }
        $current_state = $state {"Movie"};
        return 1;
    } else {
        return 0;
    }
}

####
# parse_perf_version ($line)
# -> tries to parse $line for a new movie performance version (VO|VF)
# -> returns 1 upon success, 0 otherwise
####
sub parse_perf_version {
    
    my $line = $_[0];
    
    if ($line =~ /<span class=\"bold\"> En (VF|VO)<\/span>/) {
	if (defined ($1)) {
	    $version = $1;
	} else {
	    $version = "VO";
	}
	if ($DEBUG) {
	    print "    Version : $version\n";
	}
	$current_state = $state {"Perf_version"};
	return 1;
    } else {
	return 0;
    }
}

####
# parse_perf_details ($line)
# -> tries to parse $line for a new movie performance information (hours, ads)
# -> returns 1 upon success, 0 otherwise
####
sub parse_perf_details {
    
    my $line = $_[0];
    
    if ($line =~ /<div class=\"times\"><ul><li [^>]*>(.*?)<\/li><\/ul>/) {
	@perfs = split /<\/li><li >/, $1;
	%hours = ();
	foreach (@perfs) {
	    /<em data-times=\'\[\"([0-9:]+)\",\"([0-9:]+)\"/;
	    $hour_1 = $1;
	    $hour_2   = $2;

	    $hour_1 =~ /([0-2]?[0-9]):([0-5][0-9])/;
	    $time_hour = $1 % 24;
	    $time_minute = $2;
	    $hour = sprintf ("%02d", $time_hour) . ":" . $time_minute; # 24:mm => 00:mm ; h:mm => 0h:mm

	    $ads = "\\N";
	    if ($hour_2 =~ /([0-2]?[0-9]):([0-5][0-9])/) {
		$ads = (($1 - $time_hour) % 24) * 60 + ($2 - $time_minute);
	    }

	    $hours {$hour} = $ads;
	    if ($DEBUG) {
		print "    [$version] $hour (ads: $ads)\n";
	    }
	}
	$current_state = $state {"Perf_details"};
	return 1;
    } else {
	return 0;
    }
}

#####################
# Program main loop #
#####################

#
# Reads lines one by one
#
while ($line = <FILE_IN>) {
    
    my $match = 0;
    
    $last_state = $current_state;

    if ($current_state == $state {"Start"}) {
        # State: Start => Search for a theater
        $match = &parse_theater ($line);

    } elsif ($current_state == $state {"Theater"}) {
        # State: Theater => Search for an address, 
        # or else a metro, 
	# or else a week day,
        # or else a new theater
        $match = (&parse_address ($line) or &parse_metro ($line) or &parse_week_day ($line) or &parse_theater ($line));
        
    } elsif ($current_state == $state {"Address"}) {
        # State: Address => Search for a metro, 
        # or else a week day,
        # or else a new theater
        $match = (&parse_metro ($line) or &parse_week_day ($line) or &parse_theater ($line));

    } elsif ($current_state == $state {"Metro"}) {
        # State: Metro => Search for a week day,
        # or else a new theater
        $match = (&parse_week_day ($line) or &parse_theater ($line));

    } elsif ($current_state == $state {"Week_day"}) {
        # State: Week_day => Search for a movie,
        # or else a new week day,
        # or else a new theater
        $match = (&parse_movie ($line) or &parse_week_day ($line) or &parse_theater ($line));

    } elsif ($current_state == $state {"Movie"}) {
        # State: Movie => Search for a perf_version
	# or else a perf_details,
        # or else a new movie,
	# or else a new week day
        # or else a new theater
        $match = (&parse_perf_version ($line) or &parse_perf_details ($line) or &parse_movie ($line) or &parse_week_day ($line) or &parse_theater ($line));

    } elsif ($current_state == $state {"Perf_version"}) {
        # State: Perf_version => Search for a perf_details
	# or else a new movie,
        # or else a new week day
        # or else a new theater
        $match = (&parse_perf_details ($line) or &parse_movie ($line) or &parse_week_day ($line) or &parse_theater ($line));

    } elsif ($current_state == $state {"Perf_details"}) {
        # State: Perf_details => Search for a new movie,
        # or else a new week day
        # or else a new theater
        $match = (&parse_perf_details ($line) or &parse_movie ($line) or &parse_week_day ($line) or &parse_theater ($line));

    }
    
    if ($match) {
	if ($DEBUG && 0) {
	    print "         ### MATCH! ### Last state: $last_state ### Current state: $current_state ###\n";
	}

	# Write output files
	if ($last_state < $state {"Week_day"} && $current_state == $state {"Week_day"}) {
	    if ($DEBUG) {
		print " => THEATER: $theater_id\t$theater\t$address\t$district\t$metro\t$multiplexity\n";
	    }
	    print THEATER "$theater_id\t$theater\t$address\t$district\t$metro\t$multiplexity\n";
	}
	if ($current_state == $state {"Movie"}) {
	    if ($DEBUG) {
		print "   => MOVIE: $movie_id\t$title\n";
	    }
	    print MOVIE "$movie_id\t$title\n";
	}
	if ($current_state == $state {"Perf_details"}) {
	    foreach $hour (keys (%hours)) {
		if ($DEBUG) {
		    print "    => PERF: $theater_id\t$movie_id\t$week_day\t$hour\t$version\t", $hours{$hour}, "\n";
		}
		print PERF "$theater_id\t$movie_id\t$week_day\t$hour\t$version\t", $hours{$hour}, "\n";
	    }
	}
	
    }
}

################
# Program exit #
################

close (FILE_IN);
close (THEATER);
close (MOVIE);
close (PERF);