package DBIx::Class::InflateColumn::DateTime;

use strict;
use warnings;
use base qw/DBIx::Class/;
use DBIx::Class::Carp;
use Try::Tiny;
use namespace::clean;

=head1 NAME

DBIx::Class::InflateColumn::DateTime - Auto-create DateTime objects from date and datetime columns.

=head1 SYNOPSIS

Load this component and then declare one or more
columns to be of the datetime, timestamp or date datatype.

  package Event;
  use base 'DBIx::Class::Core';

  __PACKAGE__->load_components(qw/InflateColumn::DateTime/);
  __PACKAGE__->add_columns(
    starts_when => { data_type => 'datetime' }
    create_date => { data_type => 'date' }
  );

Then you can treat the specified column as a L<DateTime> object.

  print "This event starts the month of ".
    $event->starts_when->month_name();

If you want to set a specific timezone and locale for that field, use:

  __PACKAGE__->add_columns(
    starts_when => { data_type => 'datetime', timezone => "America/Chicago", locale => "de_DE" }
  );

If you want to inflate no matter what data_type your column is,
use inflate_datetime or inflate_date:

  __PACKAGE__->add_columns(
    starts_when => { data_type => 'varchar', inflate_datetime => 1 }
  );

  __PACKAGE__->add_columns(
    starts_when => { data_type => 'varchar', inflate_date => 1 }
  );

It's also possible to explicitly skip inflation:

  __PACKAGE__->add_columns(
    starts_when => { data_type => 'datetime', inflate_datetime => 0 }
  );

NOTE: Don't rely on C<InflateColumn::DateTime> to parse date strings for you.
The column is set directly for any non-references and C<InflateColumn::DateTime>
is completely bypassed.  Instead, use an input parser to create a DateTime
object. For instance, if your user input comes as a 'YYYY-MM-DD' string, you can
use C<DateTime::Format::ISO8601> thusly:

  use DateTime::Format::ISO8601;
  my $dt = DateTime::Format::ISO8601->parse_datetime('YYYY-MM-DD');

=head1 DESCRIPTION

This module figures out the type of DateTime::Format::* class to
inflate/deflate with based on the type of DBIx::Class::Storage::DBI::*
that you are using.  If you switch from one database to a different
one your code should continue to work without modification (though note
that this feature is new as of 0.07, so it may not be perfect yet - bug
reports to the list very much welcome).

If the data_type of a field is C<date>, C<datetime> or C<timestamp> (or
a derivative of these datatypes, e.g. C<timestamp with timezone>), this
module will automatically call the appropriate parse/format method for
deflation/inflation as defined in the storage class. For instance, for
a C<datetime> field the methods C<parse_datetime> and C<format_datetime>
would be called on deflation/inflation. If the storage class does not
provide a specialized inflator/deflator, C<[parse|format]_datetime> will
be used as a fallback. See L<DateTime/Formatters And Stringification>
for more information on date formatting.

For more help with using components, see L<DBIx::Class::Manual::Component/USING>.

=cut

__PACKAGE__->load_components(qw/InflateColumn/);

=head2 register_column

Chains with the L<DBIx::Class::Row/register_column> method, and sets
up datetime columns appropriately.  This would not normally be
directly called by end users.

In the case of an invalid date, L<DateTime> will throw an exception.  To
bypass these exceptions and just have the inflation return undef, use
the C<datetime_undef_if_invalid> option in the column info:

    "broken_date",
    {
        data_type => "datetime",
        default_value => '0000-00-00',
        is_nullable => 1,
        datetime_undef_if_invalid => 1
    }

=cut

sub register_column {
  my ($self, $column, $info, @rest) = @_;

  $self->next::method($column, $info, @rest);

  my $requested_type;
  for (qw/datetime timestamp date/) {
    my $key = "inflate_${_}";
    if (exists $info->{$key}) {

      # this bailout is intentional
      return unless $info->{$key};

      $requested_type = $_;
      last;
    }
  }

  return if (!$requested_type and !$info->{data_type});

  my $data_type = lc( $info->{data_type} || '' );

  # _ic_dt_method will follow whatever the registration requests
  # thus = instead of ||=
  if ($data_type eq 'timestamp with time zone' || $data_type eq 'timestamptz') {
    $info->{_ic_dt_method} = 'timestamp_with_timezone';
  }
  elsif ($data_type eq 'timestamp without time zone') {
    $info->{_ic_dt_method} = 'timestamp_without_timezone';
  }
  elsif ($data_type eq 'smalldatetime') {
    $info->{_ic_dt_method} = 'smalldatetime';
  }
  elsif ($data_type =~ /^ (?: date | datetime | timestamp ) $/x) {
    $info->{_ic_dt_method} = $data_type;
  }
  elsif ($requested_type) {
    $info->{_ic_dt_method} = $requested_type;
  }
  else {
    return;
  }

  if ($info->{extra}) {
    for my $slot (qw/timezone locale floating_tz_ok/) {
      if ( defined $info->{extra}{$slot} ) {
        carp "Putting $slot into extra => { $slot => '...' } has been deprecated, ".
             "please put it directly into the '$column' column definition.";
        $info->{$slot} = $info->{extra}{$slot} unless defined $info->{$slot};
      }
    }
  }

  # shallow copy to avoid unfounded(?) Devel::Cycle complaints
  my $infcopy = {%$info};

  $self->inflate_column(
    $column =>
      {
        inflate => sub {
          my ($value, $obj) = @_;

          # propagate for error reporting
          $infcopy->{__dbic_colname} = $column;

          my $dt = $obj->_inflate_to_datetime( $value, $infcopy );

          return (defined $dt)
            ? $obj->_post_inflate_datetime( $dt, $infcopy )
            : undef
          ;
        },
        deflate => sub {
          my ($value, $obj) = @_;

          $value = $obj->_pre_deflate_datetime( $value, $infcopy );
          $obj->_deflate_from_datetime( $value, $infcopy );
        },
      }
  );
}

sub _flate_or_fallback
{
  my( $self, $value, $info, $method_fmt ) = @_;

  my $parser = $self->_datetime_parser;
  my $preferred_method = sprintf($method_fmt, $info->{ _ic_dt_method });
  my $method = $parser->can($preferred_method) || sprintf($method_fmt, 'datetime');

  return try {
    $parser->$method($value);
  }
  catch {
    $self->throw_exception ("Error while inflating '$value' for $info->{__dbic_colname} on ${self}: $_")
      unless $info->{datetime_undef_if_invalid};
    undef;  # rv
  };
}

sub _inflate_to_datetime {
  my( $self, $value, $info ) = @_;
  return $self->_flate_or_fallback( $value, $info, 'parse_%s' );
}

sub _deflate_from_datetime {
  my( $self, $value, $info ) = @_;
  return $self->_flate_or_fallback( $value, $info, 'format_%s' );
}

sub _datetime_parser {
  shift->result_source->storage->datetime_parser (@_);
}

sub _post_inflate_datetime {
  my( $self, $dt, $info ) = @_;

  $dt->set_time_zone($info->{timezone}) if defined $info->{timezone};
  $dt->set_locale($info->{locale}) if defined $info->{locale};

  return $dt;
}

sub _pre_deflate_datetime {
  my( $self, $dt, $info ) = @_;

  if (defined $info->{timezone}) {
    carp "You're using a floating timezone, please see the documentation of"
      . " DBIx::Class::InflateColumn::DateTime for an explanation"
      if ref( $dt->time_zone ) eq 'DateTime::TimeZone::Floating'
          and not $info->{floating_tz_ok}
          and not $ENV{DBIC_FLOATING_TZ_OK};

    $dt->set_time_zone($info->{timezone});
  }

  $dt->set_locale($info->{locale}) if defined $info->{locale};

  return $dt;
}

1;
__END__

=head1 USAGE NOTES

If you have a datetime column with an associated C<timezone>, and subsequently
create/update this column with a DateTime object in the L<DateTime::TimeZone::Floating>
timezone, you will get a warning (as there is a very good chance this will not have the
result you expect). For example:

  __PACKAGE__->add_columns(
    starts_when => { data_type => 'datetime', timezone => "America/Chicago" }
  );

  my $event = $schema->resultset('EventTZ')->create({
    starts_at => DateTime->new(year=>2007, month=>12, day=>31, ),
  });

The warning can be avoided in several ways:

=over

=item Fix your broken code

When calling C<set_time_zone> on a Floating DateTime object, the timezone is simply
set to the requested value, and B<no time conversion takes place>. It is always a good idea
to be supply explicit times to the database:

  my $event = $schema->resultset('EventTZ')->create({
    starts_at => DateTime->new(year=>2007, month=>12, day=>31, time_zone => "America/Chicago" ),
  });

=item Suppress the check on per-column basis

  __PACKAGE__->add_columns(
    starts_when => { data_type => 'datetime', timezone => "America/Chicago", floating_tz_ok => 1 }
  );

=item Suppress the check globally

Set the environment variable DBIC_FLOATING_TZ_OK to some true value.

=back

Putting extra attributes like timezone, locale or floating_tz_ok into extra => {} has been
B<DEPRECATED> because this gets you into trouble using L<DBIx::Class::Schema::Versioned>.
Instead put it directly into the columns definition like in the examples above. If you still
use the old way you'll see a warning - please fix your code then!

=head1 SEE ALSO

=over 4

=item More information about the add_columns method, and column metadata,
      can be found in the documentation for L<DBIx::Class::ResultSource>.

=item Further discussion of problems inherent to the Floating timezone:
      L<Floating DateTimes|DateTime/Floating DateTimes>
      and L<< $dt->set_time_zone|DateTime/"Set" Methods >>

=back

=head1 FURTHER QUESTIONS?

Check the list of L<additional DBIC resources|DBIx::Class/GETTING HELP/SUPPORT>.

=head1 COPYRIGHT AND LICENSE

This module is free software L<copyright|DBIx::Class/COPYRIGHT AND LICENSE>
by the L<DBIx::Class (DBIC) authors|DBIx::Class/AUTHORS>. You can
redistribute it and/or modify it under the same terms as the
L<DBIx::Class library|DBIx::Class/COPYRIGHT AND LICENSE>.