On Mar 20, 2006, at 5:48 PM, Perrin Harkins wrote:
Here's my guess -- are you calling you DBI->connect() on every
request?
yep.
If this is all making you annoyed with Apache::DBI, just make your own
cleanup handler that does a rollback. It's easy to do, as long as you
have a simple way to get the database handle.
well, i already have a cleanup handler that does some stuff with my
db handle , so i'll just stuck a rollback in there. problem solved.
and in solving it, I also found the issue. and ( of course ) it was
entirely my fault .
i connect as AutoCommit = 0 so I get the rollback
then i have a db abstraction level that flips AutoCommit to 1 (shown
below) and replaces the begin_work / rollback / commit -- because I
need to explicitly declare those commands
i *think* a bad error trap missed the rollback, which would have
flipped my autocommit.
anyways, thanks a ton. this was killing me. i've been porting 160+
tables of a mysql-centric db layout to postgres, and having to redo a
significant amount of sql statements to standard sql language. so my
db handles were corrupting nonstop.
eval
{
$db->{'dbh'}->do($SQL) or die "Can not insert";
};
if ( $@ )
{
$self->set_error( "Error: $@" )
$db-> _dbwriter_endwork_rollback();
}
sub _dbwriter_beginwork
{
eval
{
$_[0]->{'DBH'}{AutoCommit} = 1;
$_[0]->{'DBH'}->begin_work;
};
if ( $@ )
{
return 0;
}
return 1;
}
sub _dbwriter_endwork_rollback
{
$_[0]->{'DBH'}->rollback;
$_[0]->{'DBH'}{AutoCommit} = 0;
}
sub _dbwriter_endwork_commit
{
$_[0]->{'DBH'}->commit;
$_[0]->{'DBH'}{AutoCommit} = 0;
}
