Just to clarify what I meant. Functions in pl/sql only allow input arguments. You cannot write:

FUNCTION fred (arg1 OUT INTEGER) RETURN integer;
[download]

but you can write

FUNCTION fred (arg1 IN INTEGER) RETURN integer;
[download]

and the IN is implied anyway.

Also, I was not saying bind_param_inout would not work on an input only parameter but it is pointless using it and may require an unnecessary overhead on input only arguments.

You cannot write:
FUNCTION fred (arg1 OUT INTEGER) RETURN integer;

Yes you can... (and if you do use such output parameters, bind_param_inout is the way to bind/access them).

You are right you can do that - I didn't know that. I use functions mostly in SQL and procedures if there is more than one parameter to return but as you say the following does work:

use DBI;
use strict;
my $h = DBI->connect('dbi:Oracle:xxx','xxx','xxx');
eval {$h->do(q{drop function fred});};
my $sql = <<'EOT';
create function fred(a OUT integer) return integer as
begin
  a := 1;
  return (a + 1);
end;
EOT

$h->do($sql);
my $s = $h->prepare(q{begin ? := fred(?); end;});
my $in = 1;
$s->bind_param_inout(2, \$in, 100);
my $res;
$s->bind_param_inout(1, \$res, 100);
$s->execute;
print "fred=$res, a=$in\n";
[download]

outputs fred=2, a=1

However, I still stand by the my comment that if the parameter is an input parameter it is probably better to bind it with bind_param and not bind_param_inout.