package WWW::Search::Pagesjaunes;
use strict;
use Carp qw(carp croak);
use HTML::Form;
use WWW::Mechanize;
use HTML::TokeParser;
use HTTP::Request::Common;
use LWP::UserAgent;

$WWW::Search::Pagesjaunes::VERSION = '0.13';

sub ROOT_URL() { 'http://www.pagesjaunes.fr' }

sub new {
    my $class = shift;
    my $self  = {};
    my $ua    = shift() || WWW::Mechanize->new(
        env_proxy  => 1,
        keep_alive => 1,
        timeout    => 30,
        agent      => "WWW::Search::Pagesjaunes/$WWW::Search::Pagesjaunes::VERSION",
    );

    $self->{ua}    = $ua;
    $self->{limit} = 50;
    $self->{fast}  = 0;
    $self->{error} = 1;
    $self->{lang}  = 'FR';

    bless( $self, $class );
}

sub agent {
    my $self = shift;
    if ( $_[0] ) {
        my $old = $self->{ua};
        $self->{ua} = $_[0];
        return $old;
    }
    else {
        return $self->{ua};
    }
}

sub find {
    my $self = shift;
    my %opt  = @_;

    my $p = $opt{activite} ? 'j' : 'b';

    # Make the first request to pagesjaunes.fr
    $self->{URL} = ROOT_URL . "/p$p.cgi";


    if ( $self->{fast} ) {
        $self->{req} = POST(
            $self->{URL},
            [
                faire           => 'decode_input_image',
                DEFAULT_ACTION  => $p . 'f_inscriptions_req',
                lang            => $self->{lang},
                pays            => 'FR',
                srv             => uc("p$p"),
                TYPE_RECHERCHE  => 'ZZZ',
                input_image     => '',
                FRM_ACTIVITE    => $p eq 'j' ? $opt{activite} : undef,
                FRM_NOM         => $opt{nom},
                FRM_PRENOM      => $p eq 'b' ? $opt{prenom}   : undef,
                FRM_ADRESSE     => $opt{adresse},
                FRM_LOCALITE    => $opt{localite},
                FRM_DEPARTEMENT => $opt{departement},
                #'${p}F_INSCRIPTIONS_REQ.x' => 1,
                #'${p}F_INSCRIPTIONS_REQ.y' => 1,
            ]);
    }
    else {
        my $req = $self->{ua}->get($self->{URL});

        if ( !$req->content || !$req->is_success ) {
            croak('Error while retrieving the HTML page');
        }

        my @forms = HTML::Form->parse( $req->content, $self->{URL} );

        # BooK finds the form by grepping thru all of them, instead
        # of limiting ourselves to the first and second form.
        my ($form) = grep { $_->find_input('lang') } @forms;

        eval {
            # HTML::Form complains when you change hidden fields values.
            local $^W;
            $form->value( 'lang', $self->{lang} );
            
            $form->value( 'FRM_ACTIVITE', $opt{activite} ) if $opt{activite};
            $form->value( 'FRM_NOM',      $opt{nom} );
            $form->value( 'FRM_PRENOM',   $opt{prenom} )   if !$opt{activite};
            $form->value( 'FRM_ADRESSE',  $opt{adresse} );
            $form->value( 'FRM_LOCALITE', $opt{localite} );
            $form->value( 'FRM_DEPARTEMENT', $opt{departement} );
        };
        croak "Cannot fill the pagesjaunes request form. try with the 'fast' option\n" if $@;

        $self->{limit} = $opt{limit} || $self->{limit};

        $self->{req} = $form->click;
    }

    return $self;
}

sub results {
    my $self = shift;

    my $result_page = $self->{ua}->request( $self->{req} )->content;

    my $parser      = HTML::TokeParser->new( \$result_page );

    # All the <br> tags are transformed to '§¤§', to separate
    # multiple phone numbers
    $parser->{textify} = {
        'br' => sub() { '§¤§' }
    };

    my @results;

    if ( $self->{limit} == 0 ) {
        $self->{has_more} = 0;
        return @results;
    }

    # XXX This is a really crude parsing of the data, but it seems to
    # get the job done.
    #
    # <table class="fdcadreinscr">
    #   <tr>
    #     <td>
    #       <table class="fdinscr">
    #         <tr class="fdrsinscr">
    #           <td class="txtrsinscr">Name</td>
    #           <td class="txtrsinscr" align=right>&nbsp;</td>
    #         </tr>
    #         <tr valign="top">
    #           <td class="txtinscr">Address</td>
    #           <td align="right" class=txtinscr nowrap>(télécopie)? Phone</td>
    #         </tr>
    #       </table>
    #     </td>
    #   </tr>
    #  </table>
    #
    $self->{has_more} = 0;

    while ( my $token = $parser->get_tag("table") ) {
        next
          unless $token->[1]
          && $token->[1]{class}
          && $token->[1]{class} eq 'fdinscr';
        {    # We're inside an entry table

            $parser->get_tag("td");    # The first <td> is the name
            my $name = _trim( $parser->get_trimmed_text('/td') );

            $parser->get_tag("td");    # The second <td> is ignored

            $parser->get_tag("td");    # The third <td> is the address
            my $address = _trim( $parser->get_trimmed_text('/td') );
            $address =~ s/\W*\|.*$//g;

            $parser->get_tag("td");    # The fourth <td> is the phone number
            my $phone = _trim( $parser->get_trimmed_text('/td') );
            my @phones = map { _trim($_); s/\.(\s*\d)/$1/; $_ }  split(/§¤§/, $phone);

            # The fifth <td> tag is either the mail or the descr, depending
            # on the class
            my @emails = ('');
            my $tag = $parser->get_tag("td");
            if ( $tag->[1]{class} && $tag->[1]{class} eq 'txtinscr'){
               my $email  = _trim( $parser->get_trimmed_text('/td') );
               @emails = map { _trim($_); s/Mail\s*:\s*//; $_ }  split(/§¤§/, $email);
            }

            push(
                @results,
                WWW::Search::Pagesjaunes::Entry->new(
                    $name, $address, [ @phones ], [ @emails ]
                )
            );

            return @results if --$self->{limit} == 0;
        }
    }

    foreach my $form ( HTML::Form->parse( $result_page, $self->{URL} ) ) {
        if (   $form->find_input('faire')
            && $form->value('faire') eq 'inscriptions_suivant' )
        {
            $self->{has_more} = 1;
            $self->{req}      = $form->click();
        }
    }

    # If there was no result, we look for an error message in the HTML page
    if ( !@results && $self->{error} ) {
        $parser = HTML::TokeParser->new( \$result_page );
        while ( my $token = $parser->get_tag("font") ) {
            next
              unless $token->[1]
              && $token->[1]{color}
              && $token->[1]{color} eq '#ff0000';
            $parser->{textify} = {
                'br' => sub() { " " }
            };
            carp _trim( $parser->get_trimmed_text('/font') ) . "\n";
        }
    }

    wantarray ? @results : $results[0];
}

sub _trim {
    $_[0] =~ s/\xa0/ /g;       # Transform the &nbsp; into whitespace
    $_[0] =~ s/^\s*|\s*$//g;
    $_[0] =~ s/\s+/ /g;
    $_[0];
}

sub limit {
    my $self = shift;
    $self->{limit} = $_[0] || $self->{limit};
}

sub has_more { $_[0]->{has_more} }

package WWW::Search::Pagesjaunes::Entry;

# The entry object is a blessed array with the following indices:
# 0 - Name
# 1 - Address
# 2 - Arrayref of phone numbers
# 3 - E-mail (pj)
# 4 - Notes  (pj)

sub new     {
    my $class = shift;
    bless [ @_ ], $class
}
sub name    { $_[0]->[0] }
sub address { $_[0]->[1] }
sub phone   { $_[0]->[2] }
sub email   { $_[0]->[3] }
sub entry   {
    # Name      Address     First email      Phones
    $_[0]->[0], $_[0]->[1], $_[0]->[3]->[0], @{ @{ $_[0] }[2] },
}

1;

__END__

=pod

=head1 NAME

WWW::Search::Pagesjaunes - Lookup phones numbers from www.pagesjaunes.fr

=head1 SYNOPSIS

 use WWW::Search::Pagesjaunes;

 my $pj = new WWW::Search::Pagesjaunes;
 $pj->find( activite => "Plombier", localite => "Paris" );

 do {
    print $_->entry . "\n" foreach ($pj->results);
 } while $pj->has_more;

=head1 DESCRIPTION

The WWW::Search::Pagesjaunes provides name, phone number and addresses of French
telephone subscribers by using the L<http://www.pagesjaunes.fr>
directory.

=head1 METHODS

Two classes are used in this module, a first one (WWW::Search::Pagesjaunes) to do the
fetching and parsing, and the second one and a second one
(WWW::Search::Pagesjaunes::Entry) holding the entry infos.

Here are the methods for the main WWW::Search::Pagesjaunes module:

=over 4

=item new()

The constructor accept an optional LWP::UserAgent as argument, if you want to
provide your own.

=item find( %request )

Here are the values for the %request hash that are understood. They
each have two name, the first is the french one and the second is the
english one:

=over 4

=item nom / name

Name of the person you're looking for.

=item activite / business

Business type of the company you're looking for. Note that if this
field is filled, the module searches in the yellow pages.

=item localite / town

Name of the town you're searching in.

=item prenom / firstname

First name of the person you're looking for. It is not set if you set the
'activite' field.

=item departement / district

=encoding ISO8859-1

Name or number of the Département or Région you're searching in.

=back

=item results()

Returns an array of WWW::Search::Pagesjaunes::Entry containing the first matches of the
query.

=item limit($max_number_of_entries)

Set the maximum number of entries returned. Default to 50.

=item has_more()

If the query leads to more than a few results, the field has_more is set. You
can then call the results() method again to fetch the datas.

=back

The WWW::Search::Pagesjaunes::Entry class has six methods:

=over 4

=item new($name, $address, $phone, $fax)

Returns a new WWW::Search::Pagesjaunes::Entry.

=item name

Returns the name of the entry.

=item address

Returns the address of the entry.

=item phone

Returns the phone number of the entry.

=item is_fax

Returns true if the phone number is a fax one, false otherwise. Note
that currently, this method always returns 0.

=item entry($separator)

Returns the concatenation of the name and the phone number, separated by
" - ". You can specify your own separator as first argument.

=back

=head1 BUGS

The phone numbers are sometimes not correctly parsed, esp. when one
entry has several phone numbers.

If you found a bug and want to report it or send a patch, you are
encouraged to use the CPAN Request Tracker interface:
L<https://rt.cpan.org/NoAuth/Dists.html?Queue=WWW-Search-Pagesjaunes>

=head1 COPYRIGHT

Please read the Publisher information of L<http://www.pagesjaunes.fr> available at the following URL:
L<http://www.pagesjaunes.fr/pj.cgi?html=commun/avertissement.html&lang=en>

WWW::Search::Pagesjaunes is Copyright (C) 2002, Briac Pilpré

This module is free software; you can redistribute it or modify it under the
same terms as Perl itself.

=head1 AUTHOR

Briac Pilpré <briac@cpan.org>

=cut