Protect DBIC as best we can from the failure mode in 7cb35852

[dbsrgits/DBIx-Class.git] / lib / DBIx / Class / Storage / DBI / Sybase / ASE.pm

package DBIx::Class::Storage::DBI::Sybase::ASE;

use strict;
use warnings;

use base qw/
  DBIx::Class::Storage::DBI::Sybase
  DBIx::Class::Storage::DBI::AutoCast
  DBIx::Class::Storage::DBI::IdentityInsert
/;
use mro 'c3';
use DBIx::Class::Carp;
use Scalar::Util qw/blessed weaken/;
use List::Util 'first';
use Sub::Name();
use Data::Dumper::Concise 'Dumper';
use Try::Tiny;
use Context::Preserve 'preserve_context';
use DBIx::Class::_Util qw( sigwarn_silencer dbic_internal_try );
use namespace::clean;

__PACKAGE__->sql_limit_dialect ('GenericSubQ');
__PACKAGE__->sql_quote_char ([qw/[ ]/]);
__PACKAGE__->datetime_parser_type(
  'DBIx::Class::Storage::DBI::Sybase::ASE::DateTime::Format'
);

__PACKAGE__->mk_group_accessors('simple' =>
    qw/_identity _identity_method _blob_log_on_update _parent_storage
       _writer_storage _is_writer_storage
       _bulk_storage _is_bulk_storage _began_bulk_work
    /
);


my @also_proxy_to_extra_storages = qw/
  connect_call_set_auto_cast auto_cast connect_call_blob_setup
  connect_call_datetime_setup

  disconnect _connect_info _sql_maker _sql_maker_opts disable_sth_caching
  auto_savepoint unsafe cursor_class debug debugobj schema
/;

=head1 NAME

DBIx::Class::Storage::DBI::Sybase::ASE - Sybase ASE SQL Server support for
DBIx::Class

=head1 SYNOPSIS

This subclass supports L<DBD::Sybase> for real (non-Microsoft) Sybase databases.

=head1 DESCRIPTION

If your version of Sybase does not support placeholders, then your storage will
be reblessed to L<DBIx::Class::Storage::DBI::Sybase::ASE::NoBindVars>.
You can also enable that driver explicitly, see the documentation for more
details.

With this driver there is unfortunately no way to get the C<last_insert_id>
without doing a C<SELECT MAX(col)>. This is done safely in a transaction
(locking the table.) See L</INSERTS WITH PLACEHOLDERS>.

A recommended L<connect_info|DBIx::Class::Storage::DBI/connect_info> setting:

  on_connect_call => [['datetime_setup'], ['blob_setup', log_on_update => 0]]

=head1 METHODS

=cut

sub _rebless {
  my $self = shift;

  my $no_bind_vars = __PACKAGE__ . '::NoBindVars';

  if ($self->_using_freetds) {
    carp_once <<'EOF' unless $ENV{DBIC_SYBASE_FREETDS_NOWARN};

You are using FreeTDS with Sybase.

We will do our best to support this configuration, but please consider this
support experimental.

TEXT/IMAGE columns will definitely not work.

You are encouraged to recompile DBD::Sybase with the Sybase Open Client libraries
instead.

See perldoc DBIx::Class::Storage::DBI::Sybase::ASE for more details.

To turn off this warning set the DBIC_SYBASE_FREETDS_NOWARN environment
variable.
EOF

    if (not $self->_use_typeless_placeholders) {
      if ($self->_use_placeholders) {
        $self->auto_cast(1);
      }
      else {
        $self->ensure_class_loaded($no_bind_vars);
        bless $self, $no_bind_vars;
        $self->_rebless;
      }
    }
  }

  elsif (not $self->_get_dbh->{syb_dynamic_supported}) {
    # not necessarily FreeTDS, but no placeholders nevertheless
    $self->ensure_class_loaded($no_bind_vars);
    bless $self, $no_bind_vars;
    $self->_rebless;
  }
  # this is highly unlikely, but we check just in case
  elsif (not $self->_use_typeless_placeholders) {
    $self->auto_cast(1);
  }
}

sub _init {
  my $self = shift;

  $self->next::method(@_);

  if ($self->_using_freetds && (my $ver = $self->_using_freetds_version||999) > 0.82) {
    carp_once(
      "Buggy FreeTDS version $ver detected, statement caching will not work and "
    . 'will be disabled.'
    );
    $self->disable_sth_caching(1);
  }

  $self->_set_max_connect(256);

# create storage for insert/(update blob) transactions,
# unless this is that storage
  return if $self->_parent_storage;

  my $writer_storage = (ref $self)->new;

  $writer_storage->_is_writer_storage(1); # just info
  $writer_storage->connect_info($self->connect_info);
  $writer_storage->auto_cast($self->auto_cast);

  weaken ($writer_storage->{_parent_storage} = $self);
  $self->_writer_storage($writer_storage);

# create a bulk storage unless connect_info is a coderef
  return if ref($self->_dbi_connect_info->[0]) eq 'CODE';

  my $bulk_storage = (ref $self)->new;

  $bulk_storage->_is_bulk_storage(1); # for special ->disconnect acrobatics
  $bulk_storage->connect_info($self->connect_info);

# this is why
  $bulk_storage->_dbi_connect_info->[0] .= ';bulkLogin=1';

  weaken ($bulk_storage->{_parent_storage} = $self);
  $self->_bulk_storage($bulk_storage);
}

for my $method (@also_proxy_to_extra_storages) {
  no strict 'refs';
  no warnings 'redefine';

  my $replaced = __PACKAGE__->can($method);

  *{$method} = Sub::Name::subname $method => sub {
    my $self = shift;
    $self->_writer_storage->$replaced(@_) if $self->_writer_storage;
    $self->_bulk_storage->$replaced(@_)   if $self->_bulk_storage;
    return $self->$replaced(@_);
  };
}

sub disconnect {
  my $self = shift;

# Even though we call $sth->finish for uses off the bulk API, there's still an
# "active statement" warning on disconnect, which we throw away here.
# This is due to the bug described in _insert_bulk.
# Currently a noop because 'prepare' is used instead of 'prepare_cached'.
  local $SIG{__WARN__} = sigwarn_silencer(qr/active statement/i)
    if $self->_is_bulk_storage;

# so that next transaction gets a dbh
  $self->_began_bulk_work(0) if $self->_is_bulk_storage;

  $self->next::method;
}

# This is only invoked for FreeTDS drivers by ::Storage::DBI::Sybase::FreeTDS
sub _set_autocommit_stmt {
  my ($self, $on) = @_;

  return 'SET CHAINED ' . ($on ? 'OFF' : 'ON');
}

# Set up session settings for Sybase databases for the connection.
#
# Make sure we have CHAINED mode turned on if AutoCommit is off in non-FreeTDS
# DBD::Sybase (since we don't know how DBD::Sybase was compiled.) If however
# we're using FreeTDS, CHAINED mode turns on an implicit transaction which we
# only want when AutoCommit is off.
sub _run_connection_actions {
  my $self = shift;

  if ($self->_is_bulk_storage) {
    # this should be cleared on every reconnect
    $self->_began_bulk_work(0);
    return;
  }

  $self->_dbh->{syb_chained_txn} = 1
    unless $self->_using_freetds;

  $self->next::method(@_);
}

=head2 connect_call_blob_setup

Used as:

  on_connect_call => [ [ 'blob_setup', log_on_update => 0 ] ]

Does C<< $dbh->{syb_binary_images} = 1; >> to return C<IMAGE> data as raw binary
instead of as a hex string.

Recommended.

Also sets the C<log_on_update> value for blob write operations. The default is
C<1>, but C<0> is better if your database is configured for it.

See
L<DBD::Sybase/Handling IMAGE/TEXT data with syb_ct_get_data()/syb_ct_send_data()>.

=cut

sub connect_call_blob_setup {
  my $self = shift;
  my %args = @_;
  my $dbh = $self->_dbh;
  $dbh->{syb_binary_images} = 1;

  $self->_blob_log_on_update($args{log_on_update})
    if exists $args{log_on_update};
}

sub _is_lob_column {
  my ($self, $source, $column) = @_;

  return $self->_is_lob_type($source->column_info($column)->{data_type});
}

sub _prep_for_execute {
  my ($self, $op, $ident, $args) = @_;

  my $limit;  # extract and use shortcut on limit without offset
  if ($op eq 'select' and ! $args->[4] and $limit = $args->[3]) {
    $args = [ @$args ];
    $args->[3] = undef;
  }

  my ($sql, $bind) = $self->next::method($op, $ident, $args);

  # $limit is already sanitized by now
  $sql = join( "\n",
    "SET ROWCOUNT $limit",
    $sql,
    "SET ROWCOUNT 0",
  ) if $limit;

  if (my $identity_col = $self->_perform_autoinc_retrieval) {
    $sql .= "\n" . $self->_fetch_identity_sql($ident, $identity_col)
  }

  return ($sql, $bind);
}

sub _fetch_identity_sql {
  my ($self, $source, $col) = @_;

  return sprintf ("SELECT MAX(%s) FROM %s",
    map { $self->sql_maker->_quote ($_) } ($col, $source->from)
  );
}

# Stolen from SQLT, with some modifications. This is a makeshift
# solution before a sane type-mapping library is available, thus
# the 'our' for easy overrides.
our %TYPE_MAPPING  = (
    number    => 'numeric',
    money     => 'money',
    varchar   => 'varchar',
    varchar2  => 'varchar',
    timestamp => 'datetime',
    text      => 'varchar',
    real      => 'double precision',
    comment   => 'text',
    bit       => 'bit',
    tinyint   => 'smallint',
    float     => 'double precision',
    serial    => 'numeric',
    bigserial => 'numeric',
    boolean   => 'varchar',
    long      => 'varchar',
);

sub _native_data_type {
  my ($self, $type) = @_;

  $type = lc $type;
  $type =~ s/\s* identity//x;

  return uc($TYPE_MAPPING{$type} || $type);
}


sub _execute {
  my $self = shift;
  my ($rv, $sth, @bind) = $self->next::method(@_);

  $self->_identity( ($sth->fetchall_arrayref)->[0][0] )
    if $self->_perform_autoinc_retrieval;

  return wantarray ? ($rv, $sth, @bind) : $rv;
}

sub last_insert_id { shift->_identity }

# handles TEXT/IMAGE and transaction for last_insert_id
sub insert {
  my $self = shift;
  my ($source, $to_insert) = @_;

  my $columns_info = $source->columns_info;

  my ($identity_col) = grep
    { $columns_info->{$_}{is_auto_increment} }
    keys %$columns_info
  ;

  $identity_col = '' if ! defined $identity_col;

  # FIXME - this is duplication from DBI.pm. When refactored towards
  # the LobWriter this can be folded back where it belongs.
  local $self->{_autoinc_supplied_for_op} = exists $to_insert->{$identity_col}
    ? 1
    : 0
  ;

  local $self->{_perform_autoinc_retrieval} = $self->{_autoinc_supplied_for_op}
    ? undef
    : $identity_col
  ;

  # check for empty insert
  # INSERT INTO foo DEFAULT VALUES -- does not work with Sybase
  # try to insert explicit 'DEFAULT's instead (except for identity, timestamp
  # and computed columns)
  if (not %$to_insert) {
    for my $col ($source->columns) {
      next if $col eq $identity_col;

      my $info = $source->column_info($col);

      next if ref $info->{default_value} eq 'SCALAR'
        || (exists $info->{data_type} && (not defined $info->{data_type}));

      next if $info->{data_type} && $info->{data_type} =~ /^timestamp\z/i;

      $to_insert->{$col} = \'DEFAULT';
    }
  }

  my $blob_cols = $self->_remove_blob_cols($source, $to_insert);

  # if a new txn is needed - it must happen on the _writer/new connection (for now)
  my $guard;
  if (
    ! $self->transaction_depth
      and
    (
      $blob_cols
        or
      # do we need the horrific SELECT MAX(COL) hack?
      (
        $self->_perform_autoinc_retrieval
          and
        ( ($self->_identity_method||'') ne '@@IDENTITY' )
      )
    )
  ) {
    $self = $self->_writer_storage;
    $guard = $self->txn_scope_guard;
  }

  my $updated_cols = $self->next::method ($source, $to_insert);

  $self->_insert_blobs (
    $source,
    $blob_cols,
    {
      ( $identity_col
        ? ( $identity_col => $self->last_insert_id($source, $identity_col) )
        : ()
      ),
      %$to_insert,
      %$updated_cols,
    },
  ) if $blob_cols;

  $guard->commit if $guard;

  return $updated_cols;
}

sub update {
  my $self = shift;
  my ($source, $fields, $where, @rest) = @_;

  #
  # When *updating* identities, ASE requires SET IDENTITY_UPDATE called
  #
  if (my $blob_cols = $self->_remove_blob_cols($source, $fields)) {

    # If there are any blobs in $where, Sybase will return a descriptive error
    # message.
    # XXX blobs can still be used with a LIKE query, and this should be handled.

    # update+blob update(s) done atomically on separate connection
    $self = $self->_writer_storage;

    my $guard = $self->txn_scope_guard;

    # First update the blob columns to be updated to '' (taken from $fields, where
    # it is originally put by _remove_blob_cols .)
    my %blobs_to_empty = map { ($_ => delete $fields->{$_}) } keys %$blob_cols;

    # We can't only update NULL blobs, because blobs cannot be in the WHERE clause.
    $self->next::method($source, \%blobs_to_empty, $where, @rest);

    # Now update the blobs before the other columns in case the update of other
    # columns makes the search condition invalid.
    my $rv = $self->_update_blobs($source, $blob_cols, $where);

    if (keys %$fields) {

      # Now set the identity update flags for the actual update
      local $self->{_autoinc_supplied_for_op} = (first
        { $_->{is_auto_increment} }
        values %{ $source->columns_info([ keys %$fields ]) }
      ) ? 1 : 0;

      my $next = $self->next::can;
      my $args = \@_;
      return preserve_context {
        $self->$next(@$args);
      } after => sub { $guard->commit };
    }
    else {
      $guard->commit;
      return $rv;
    }
  }
  else {
    # Set the identity update flags for the actual update
    local $self->{_autoinc_supplied_for_op} = (first
      { $_->{is_auto_increment} }
      values %{ $source->columns_info([ keys %$fields ]) }
    ) ? 1 : 0;

    return $self->next::method(@_);
  }
}

sub _insert_bulk {
  my $self = shift;
  my ($source, $cols, $data) = @_;

  my $columns_info = $source->columns_info;

  my $identity_col =
    first { $columns_info->{$_}{is_auto_increment} }
      keys %$columns_info;

  # FIXME - this is duplication from DBI.pm. When refactored towards
  # the LobWriter this can be folded back where it belongs.
  local $self->{_autoinc_supplied_for_op} =
    (first { $_ eq $identity_col } @$cols)
      ? 1
      : 0
  ;

  my $use_bulk_api =
    $self->_bulk_storage &&
    $self->_get_dbh->{syb_has_blk};

  if (! $use_bulk_api and ref($self->_dbi_connect_info->[0]) eq 'CODE') {
    carp_unique( join ' ',
      'Bulk API support disabled due to use of a CODEREF connect_info.',
      'Reverting to regular array inserts.',
    );
  }

  if (not $use_bulk_api) {
    my $blob_cols = $self->_remove_blob_cols_array($source, $cols, $data);

# next::method uses a txn anyway, but it ends too early in case we need to
# select max(col) to get the identity for inserting blobs.
    ($self, my $guard) = $self->transaction_depth
      ? ($self, undef)
      : ($self->_writer_storage, $self->_writer_storage->txn_scope_guard)
    ;

    $self->next::method(@_);

    if ($blob_cols) {
      if ($self->_autoinc_supplied_for_op) {
        $self->_insert_blobs_array ($source, $blob_cols, $cols, $data);
      }
      else {
        my @cols_with_identities = (@$cols, $identity_col);

        ## calculate identities
        # XXX This assumes identities always increase by 1, which may or may not
        # be true.
        my ($last_identity) =
          $self->_dbh->selectrow_array (
            $self->_fetch_identity_sql($source, $identity_col)
          );
        my @identities = (($last_identity - @$data + 1) .. $last_identity);

        my @data_with_identities = map [@$_, shift @identities], @$data;

        $self->_insert_blobs_array (
          $source, $blob_cols, \@cols_with_identities, \@data_with_identities
        );
      }
    }

    $guard->commit if $guard;

    return;
  }

# otherwise, use the bulk API

# rearrange @$data so that columns are in database order
# and so we submit a full column list
  my %orig_order = map { $cols->[$_] => $_ } 0..$#$cols;

  my @source_columns = $source->columns;

  # bcp identity index is 1-based
  my $identity_idx = first { $source_columns[$_] eq $identity_col } (0..$#source_columns);
  $identity_idx = defined $identity_idx ? $identity_idx + 1 : 0;

  my @new_data;
  for my $slice_idx (0..$#$data) {
    push @new_data, [map {
      # identity data will be 'undef' if not _autoinc_supplied_for_op()
      # columns with defaults will also be 'undef'
      exists $orig_order{$_}
        ? $data->[$slice_idx][$orig_order{$_}]
        : undef
    } @source_columns];
  }

  my $proto_bind = $self->_resolve_bindattrs(
    $source,
    [map {
      [ { dbic_colname => $source_columns[$_], _bind_data_slice_idx => $_ }
        => $new_data[0][$_] ]
    } (0 ..$#source_columns) ],
    $columns_info
  );

## Set a client-side conversion error handler, straight from DBD::Sybase docs.
# This ignores any data conversion errors detected by the client side libs, as
# they are usually harmless.
  my $orig_cslib_cb = DBD::Sybase::set_cslib_cb(
    Sub::Name::subname _insert_bulk_cslib_errhandler => sub {
      my ($layer, $origin, $severity, $errno, $errmsg, $osmsg, $blkmsg) = @_;

      return 1 if $errno == 36;

      carp
        "Layer: $layer, Origin: $origin, Severity: $severity, Error: $errno" .
        ($errmsg ? "\n$errmsg" : '') .
        ($osmsg  ? "\n$osmsg"  : '')  .
        ($blkmsg ? "\n$blkmsg" : '');

      return 0;
  });

  my $exception = '';
  dbic_internal_try {
    my $bulk = $self->_bulk_storage;

    my $guard = $bulk->txn_scope_guard;

## FIXME - once this is done - address the FIXME on finish() below
## XXX get this to work instead of our own $sth
## will require SQLA or *Hacks changes for ordered columns
#    $bulk->next::method($source, \@source_columns, \@new_data, {
#      syb_bcp_attribs => {
#        identity_flag   => $self->_autoinc_supplied_for_op ? 1 : 0,
#        identity_column => $identity_idx,
#      }
#    });
    my $sql = 'INSERT INTO ' .
      $bulk->sql_maker->_quote($source->name) . ' (' .
# colname list is ignored for BCP, but does no harm
      (join ', ', map $bulk->sql_maker->_quote($_), @source_columns) . ') '.
      ' VALUES ('.  (join ', ', ('?') x @source_columns) . ')';

## XXX there's a bug in the DBD::Sybase bulk support that makes $sth->finish for
## a prepare_cached statement ineffective. Replace with ->sth when fixed, or
## better yet the version above. Should be fixed in DBD::Sybase .
    my $sth = $bulk->_get_dbh->prepare($sql,
#      'insert', # op
      {
        syb_bcp_attribs => {
          identity_flag   => $self->_autoinc_supplied_for_op ? 1 : 0,
          identity_column => $identity_idx,
        }
      }
    );

    {
      # FIXME the $sth->finish in _execute_array does a rollback for some
      # reason. Disable it temporarily until we fix the SQLMaker thing above
      no warnings 'redefine';
      no strict 'refs';
      local *{ref($sth).'::finish'} = sub {};

      $self->_dbh_execute_for_fetch(
        $source, $sth, $proto_bind, \@source_columns, \@new_data
      );
    }

    $guard->commit;

    $bulk->_query_end($sql);
  } catch {
    $exception = shift;
  };

  DBD::Sybase::set_cslib_cb($orig_cslib_cb);

  if ($exception =~ /-Y option/) {
    my $w = 'Sybase bulk API operation failed due to character set incompatibility, '
          . 'reverting to regular array inserts. Try unsetting the LC_ALL environment variable'
    ;
    $w .= "\n$exception" if $self->debug;
    carp $w;

    $self->_bulk_storage(undef);
    unshift @_, $self;
    goto \&_insert_bulk;
  }
  elsif ($exception) {
# rollback makes the bulkLogin connection unusable
    $self->_bulk_storage->disconnect;
    $self->throw_exception($exception);
  }
}

# Make sure blobs are not bound as placeholders, and return any non-empty ones
# as a hash.
sub _remove_blob_cols {
  my ($self, $source, $fields) = @_;

  my %blob_cols;

  for my $col (keys %$fields) {
    if ($self->_is_lob_column($source, $col)) {
      my $blob_val = delete $fields->{$col};
      if (not defined $blob_val) {
        $fields->{$col} = \'NULL';
      }
      else {
        $fields->{$col} = \"''";
        $blob_cols{$col} = $blob_val
          if length $blob_val;
      }
    }
  }

  return %blob_cols ? \%blob_cols : undef;
}

# same for _insert_bulk
sub _remove_blob_cols_array {
  my ($self, $source, $cols, $data) = @_;

  my @blob_cols;

  for my $i (0..$#$cols) {
    my $col = $cols->[$i];

    if ($self->_is_lob_column($source, $col)) {
      for my $j (0..$#$data) {
        my $blob_val = delete $data->[$j][$i];
        if (not defined $blob_val) {
          $data->[$j][$i] = \'NULL';
        }
        else {
          $data->[$j][$i] = \"''";
          $blob_cols[$j][$i] = $blob_val
            if length $blob_val;
        }
      }
    }
  }

  return @blob_cols ? \@blob_cols : undef;
}

sub _update_blobs {
  my ($self, $source, $blob_cols, $where) = @_;

  my @primary_cols = dbic_internal_try
    { $source->_pri_cols_or_die }
    catch {
      $self->throw_exception("Cannot update TEXT/IMAGE column(s): $_")
    };

  my @pks_to_update;
  if (
    ref $where eq 'HASH'
      and
    ! grep { ! defined $where->{$_} } @primary_cols
  ) {
    my %row_to_update;
    @row_to_update{@primary_cols} = @{$where}{@primary_cols};
    @pks_to_update = \%row_to_update;
  }
  else {
    my $cursor = $self->select ($source, \@primary_cols, $where, {});
    @pks_to_update = map {
      my %row; @row{@primary_cols} = @$_; \%row
    } $cursor->all;
  }

  for my $ident (@pks_to_update) {
    $self->_insert_blobs($source, $blob_cols, $ident);
  }
}

sub _insert_blobs {
  my ($self, $source, $blob_cols, $row_data) = @_;

  my $table = $source->name;

  my @primary_cols = dbic_internal_try
    { $source->_pri_cols_or_die }
    catch {
      $self->throw_exception("Cannot update TEXT/IMAGE column(s): $_")
    };

  $self->throw_exception('Cannot update TEXT/IMAGE column(s) without primary key values')
    if grep { ! defined $row_data->{$_} } @primary_cols;

  # if we are 2-phase inserting a blob - there is nothing to retrieve anymore,
  # regardless of the previous state of the flag
  local $self->{_perform_autoinc_retrieval}
    if $self->_perform_autoinc_retrieval;

  my %where = map {( $_ => $row_data->{$_} )} @primary_cols;

  for my $col (keys %$blob_cols) {
    my $blob = $blob_cols->{$col};

    my $cursor = $self->select ($source, [$col], \%where, {});
    $cursor->next;
    my $sth = $cursor->sth;

    if (not $sth) {
      $self->throw_exception(
          "Could not find row in table '$table' for blob update:\n"
        . (Dumper \%where)
      );
    }

    dbic_internal_try {
      do {
        $sth->func('CS_GET', 1, 'ct_data_info') or die $sth->errstr;
      } while $sth->fetch;

      $sth->func('ct_prepare_send') or die $sth->errstr;

      my $log_on_update = $self->_blob_log_on_update;
      $log_on_update    = 1 if not defined $log_on_update;

      $sth->func('CS_SET', 1, {
        total_txtlen => length($blob),
        log_on_update => $log_on_update
      }, 'ct_data_info') or die $sth->errstr;

      $sth->func($blob, length($blob), 'ct_send_data') or die $sth->errstr;

      $sth->func('ct_finish_send') or die $sth->errstr;
    }
    catch {
      if ($self->_using_freetds) {
        $self->throw_exception (
          "TEXT/IMAGE operation failed, probably because you are using FreeTDS: $_"
        );
      }
      else {
        $self->throw_exception($_);
      }
    }
    finally {
      $sth->finish if $sth;
    };
  }
}

sub _insert_blobs_array {
  my ($self, $source, $blob_cols, $cols, $data) = @_;

  for my $i (0..$#$data) {
    my $datum = $data->[$i];

    my %row;
    @row{ @$cols } = @$datum;

    my %blob_vals;
    for my $j (0..$#$cols) {
      if (exists $blob_cols->[$i][$j]) {
        $blob_vals{ $cols->[$j] } = $blob_cols->[$i][$j];
      }
    }

    $self->_insert_blobs ($source, \%blob_vals, \%row);
  }
}

=head2 connect_call_datetime_setup

Used as:

  on_connect_call => 'datetime_setup'

In L<connect_info|DBIx::Class::Storage::DBI/connect_info> to set:

  $dbh->syb_date_fmt('ISO_strict'); # output fmt: 2004-08-21T14:36:48.080Z
  $dbh->do('set dateformat mdy');   # input fmt:  08/13/1979 18:08:55.080

This works for both C<DATETIME> and C<SMALLDATETIME> columns, note that
C<SMALLDATETIME> columns only have minute precision.

=cut

sub connect_call_datetime_setup {
  my $self = shift;
  my $dbh = $self->_get_dbh;

  if ($dbh->can('syb_date_fmt')) {
    # amazingly, this works with FreeTDS
    $dbh->syb_date_fmt('ISO_strict');
  }
  else {
    carp_once
      'Your DBD::Sybase is too old to support '
     .'DBIx::Class::InflateColumn::DateTime, please upgrade!';

    # FIXME - in retrospect this is a rather bad US-centric choice
    # of format. Not changing as a bugwards compat, though in reality
    # the only piece that sees the results of $dt object formatting
    # (as opposed to parsing) is the database itself, so theoretically
    # changing both this SET command and the formatter definition of
    # ::S::D::Sybase::ASE::DateTime::Format below should be safe and
    # transparent

    $dbh->do('SET DATEFORMAT mdy');
  }
}


sub _exec_txn_begin {
  my $self = shift;

# bulkLogin=1 connections are always in a transaction, and can only call BEGIN
# TRAN once. However, we need to make sure there's a $dbh.
  return if $self->_is_bulk_storage && $self->_dbh && $self->_began_bulk_work;

  $self->next::method(@_);

  $self->_began_bulk_work(1) if $self->_is_bulk_storage;
}

# savepoint support using ASE syntax

sub _exec_svp_begin {
  my ($self, $name) = @_;

  $self->_dbh->do("SAVE TRANSACTION $name");
}

# A new SAVE TRANSACTION with the same name releases the previous one.
sub _exec_svp_release { 1 }

sub _exec_svp_rollback {
  my ($self, $name) = @_;

  $self->_dbh->do("ROLLBACK TRANSACTION $name");
}

package # hide from PAUSE
  DBIx::Class::Storage::DBI::Sybase::ASE::DateTime::Format;

my $datetime_parse_format  = '%Y-%m-%dT%H:%M:%S.%3NZ';
my $datetime_format_format = '%m/%d/%Y %H:%M:%S.%3N';

my ($datetime_parser, $datetime_formatter);

sub parse_datetime {
  shift;
  require DateTime::Format::Strptime;
  $datetime_parser ||= DateTime::Format::Strptime->new(
    pattern  => $datetime_parse_format,
    on_error => 'croak',
  );
  return $datetime_parser->parse_datetime(shift);
}

sub format_datetime {
  shift;
  require DateTime::Format::Strptime;
  $datetime_formatter ||= DateTime::Format::Strptime->new(
    pattern  => $datetime_format_format,
    on_error => 'croak',
  );
  return $datetime_formatter->format_datetime(shift);
}

1;

=head1 Schema::Loader Support

As of version C<0.05000>, L<DBIx::Class::Schema::Loader> should work well with
most versions of Sybase ASE.

=head1 FreeTDS

This driver supports L<DBD::Sybase> compiled against FreeTDS
(L<http://www.freetds.org/>) to the best of our ability, however it is
recommended that you recompile L<DBD::Sybase> against the Sybase Open Client
libraries. They are a part of the Sybase ASE distribution:

The Open Client FAQ is here:
L<http://www.isug.com/Sybase_FAQ/ASE/section7.html>.

Sybase ASE for Linux (which comes with the Open Client libraries) may be
downloaded here: L<http://response.sybase.com/forms/ASE_Linux_Download>.

To see if you're using FreeTDS run:

  perl -MDBI -le 'my $dbh = DBI->connect($dsn, $user, $pass); print $dbh->{syb_oc_version}'

It is recommended to set C<tds version> for your ASE server to C<5.0> in
C</etc/freetds/freetds.conf>.

Some versions or configurations of the libraries involved will not support
placeholders, in which case the storage will be reblessed to
L<DBIx::Class::Storage::DBI::Sybase::ASE::NoBindVars>.

In some configurations, placeholders will work but will throw implicit type
conversion errors for anything that's not expecting a string. In such a case,
the C<auto_cast> option from L<DBIx::Class::Storage::DBI::AutoCast> is
automatically set, which you may enable on connection with
L<connect_call_set_auto_cast|DBIx::Class::Storage::DBI::AutoCast/connect_call_set_auto_cast>.
The type info for the C<CAST>s is taken from the
L<DBIx::Class::ResultSource/data_type> definitions in your Result classes, and
are mapped to a Sybase type (if it isn't already) using a mapping based on
L<SQL::Translator>.

In other configurations, placeholders will work just as they do with the Sybase
Open Client libraries.

Inserts or updates of TEXT/IMAGE columns will B<NOT> work with FreeTDS.

=head1 INSERTS WITH PLACEHOLDERS

With placeholders enabled, inserts are done in a transaction so that there are
no concurrency issues with getting the inserted identity value using
C<SELECT MAX(col)>, which is the only way to get the C<IDENTITY> value in this
mode.

In addition, they are done on a separate connection so that it's possible to
have active cursors when doing an insert.

When using C<DBIx::Class::Storage::DBI::Sybase::ASE::NoBindVars> transactions
are unnecessary and not used, as there are no concurrency issues with C<SELECT
@@IDENTITY> which is a session variable.

=head1 TRANSACTIONS

Due to limitations of the TDS protocol and L<DBD::Sybase>, you cannot begin a
transaction while there are active cursors, nor can you use multiple active
cursors within a transaction. An active cursor is, for example, a
L<ResultSet|DBIx::Class::ResultSet> that has been executed using C<next> or
C<first> but has not been exhausted or L<reset|DBIx::Class::ResultSet/reset>.

For example, this will not work:

  $schema->txn_do(sub {
    my $rs = $schema->resultset('Book');
    while (my $result = $rs->next) {
      $schema->resultset('MetaData')->create({
        book_id => $result->id,
        ...
      });
    }
  });

This won't either:

  my $first_row = $large_rs->first;
  $schema->txn_do(sub { ... });

Transactions done for inserts in C<AutoCommit> mode when placeholders are in use
are not affected, as they are done on an extra database handle.

Some workarounds:

=over 4

=item * use L<DBIx::Class::Storage::DBI::Replicated>

=item * L<connect|DBIx::Class::Schema/connect> another L<Schema|DBIx::Class::Schema>

=item * load the data from your cursor with L<DBIx::Class::ResultSet/all>

=back

=head1 MAXIMUM CONNECTIONS

The TDS protocol makes separate connections to the server for active statements
in the background. By default the number of such connections is limited to 25,
on both the client side and the server side.

This is a bit too low for a complex L<DBIx::Class> application, so on connection
the client side setting is set to C<256> (see L<DBD::Sybase/maxConnect>.) You
can override it to whatever setting you like in the DSN.

See
L<http://infocenter.sybase.com/help/index.jsp?topic=/com.sybase.help.ase_15.0.sag1/html/sag1/sag1272.htm>
for information on changing the setting on the server side.

=head1 DATES

See L</connect_call_datetime_setup> to setup date formats
for L<DBIx::Class::InflateColumn::DateTime>.

=head1 LIMITED QUERIES

Because ASE does not have a good way to limit results in SQL that works for
all types of queries, the limit dialect is set to
L<GenericSubQ|DBIx::Class::SQLMaker::LimitDialects/GenericSubQ>.

Fortunately, ASE and L<DBD::Sybase> support cursors properly, so when
L<GenericSubQ|DBIx::Class::SQLMaker::LimitDialects/GenericSubQ> is too slow
you can use the L<software_limit|DBIx::Class::ResultSet/software_limit>
L<DBIx::Class::ResultSet> attribute to simulate limited queries by skipping
over records.

=head1 TEXT/IMAGE COLUMNS

L<DBD::Sybase> compiled with FreeTDS will B<NOT> allow you to insert or update
C<TEXT/IMAGE> columns.

Setting C<< $dbh->{LongReadLen} >> will also not work with FreeTDS use either:

  $schema->storage->dbh->do("SET TEXTSIZE $bytes");

or

  $schema->storage->set_textsize($bytes);

instead.

However, the C<LongReadLen> you pass in
L<connect_info|DBIx::Class::Storage::DBI/connect_info> is used to execute the
equivalent C<SET TEXTSIZE> command on connection.

See L</connect_call_blob_setup> for a
L<connect_info|DBIx::Class::Storage::DBI/connect_info> setting you need to work
with C<IMAGE> columns.

=head1 BULK API

The experimental L<DBD::Sybase> Bulk API support is used for
L<populate|DBIx::Class::ResultSet/populate> in B<void> context, in a transaction
on a separate connection.

To use this feature effectively, use a large number of rows for each
L<populate|DBIx::Class::ResultSet/populate> call, eg.:

  while (my $rows = $data_source->get_100_rows()) {
    $rs->populate($rows);
  }

B<NOTE:> the L<add_columns|DBIx::Class::ResultSource/add_columns>
calls in your C<Result> classes B<must> list columns in database order for this
to work. Also, you may have to unset the C<LC_ALL> environment variable before
loading your app, as C<BCP -Y> is not yet supported in DBD::Sybase .

When inserting IMAGE columns using this method, you'll need to use
L</connect_call_blob_setup> as well.

=head1 COMPUTED COLUMNS

If you have columns such as:

  created_dtm AS getdate()

represent them in your Result classes as:

  created_dtm => {
    data_type => undef,
    default_value => \'getdate()',
    is_nullable => 0,
    inflate_datetime => 1,
  }

The C<data_type> must exist and must be C<undef>. Then empty inserts will work
on tables with such columns.

=head1 TIMESTAMP COLUMNS

C<timestamp> columns in Sybase ASE are not really timestamps, see:
L<http://dba.fyicenter.com/Interview-Questions/SYBASE/The_timestamp_datatype_in_Sybase_.html>.

They should be defined in your Result classes as:

  ts => {
    data_type => 'timestamp',
    is_nullable => 0,
    inflate_datetime => 0,
  }

The C<<inflate_datetime => 0>> is necessary if you use
L<DBIx::Class::InflateColumn::DateTime>, and most people do, and still want to
be able to read these values.

The values will come back as hexadecimal.

=head1 TODO

=over

=item *

Transitions to AutoCommit=0 (starting a transaction) mode by exhausting
any active cursors, using eager cursors.

=item *

Real limits and limited counts using stored procedures deployed on startup.

=item *

Blob update with a LIKE query on a blob, without invalidating the WHERE condition.

=item *

bulk_insert using prepare_cached (see comments.)

=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>.