determination of update error

[Kamp, Rolf F, ALCNS]

Hi:
        I have a question dealing with the detection of an error condition. I am 
updating and 
inserting many rows into the database. Most of the time the record exists, so I just 
want to 
do an update. If the update fails, I assume the record does not exist and I must do an 
insert.
        What is the most positive way to discern an update failure due to a 
record not being there from another type of error (such as bad SQL syntax, listener 
gone, or 
some other database error that I cannot recover from). Note that RaiseError is ON.

$dbh = DBI->connect(undef, 'i', 'i3', {AutoCommit => 0, RaiseError => 1,}) 
or die ("Cannot connect to the database " . $DBI::errstr);
# assume usg_update and usg_insert are prepared already
eval {
        # must this line be in an eval block?
        $uval = $usg_update->execute(...);
};
if($@ || $uval == 0) {
        # might I get here for some other reason?
        # update failed...do an insert
        $ival = $usg_insert->execute(...);
        if($ival <= 0) {
                print "Error: insert usage failed\n";
                print "MSG = $@, DBI::err = $DBI::err\n";
                return -1;
        }
}