Hello all

I am confused about the mechanics of 'select.. for update' under DBIx::Class (I'm using MySQL but I suppose that shouldn't matter.) So I have reduced it down to the following test case:

my $guard = $schema->txn_scope_guard;

my $rs = $schema->resultset('TestTable')->search({
    status => 'available'
},
{ 
    rows => 1,
    'for' => 'update'
});

my $id_before_update = $rs->first->id;

my $updated = $rs->update({
    status => 'reserved'
});

$guard->commit;

my $id_at_end = $rs->first->id;

print "id before update: $id_before_update\nid at end: $id_at_end\n";
[download]

The table looks like this initially:

+----+-----------+
| id | status    |
+----+-----------+
|  1 | available |
|  2 | available |
|  3 | available |
+----+-----------+
[download]

and after execution like this:

+----+-----------+
| id | status    |
+----+-----------+
|  1 | reserved  |
|  2 | available |
|  3 | available |
+----+-----------+
[download]

The schema was created by DBIx::Class::Schema::Loader and looks like this:

__PACKAGE__->table("test_table");
__PACKAGE__->add_columns(
  "id",
  { data_type => "bigint", is_nullable => 0 },
  "status",
  { data_type => "varchar", is_nullable => 1, size => 255 },
);
[download]

(with boilerplate removed). When I run the above code I get the following:

id before update: 1
id at end: 2
[download]

To me this seems like odd behaviour, because after $rs is selected with rows=1 it surely should only contain a single row with a single id...?

Any idea what stupid thing have I (not) done this time?