You should care....

No, I shouldn't :-), I could!
because, at least in the code I tested, I created the database with SQL myself, knowing about primaray keys, foreign keys etc.

and after that I used a Perl-Module, namely Class::DBI, to access this DB, according to it's documentation; the documentation stated clearly that many-to-many relations aren't supported directly, but how to circumvent this fact using the mappings available via Class::DBI

So I should care about my data-definition, you are right with that, but shouldn't care about the internals of a module I use, if this works as stated in the documentation, and nothing else was my intention to express in my first post.

The code you tested works because the problem was circumvented without actually solving it. Try retrieving one key from that table instead of accessing it through relationships and the problem will become apparent. With specifying extra params with Primary, you no longer have a Class::DBI object that matches up with your data definition. The definition says that a primary key for this table is made up of a single column, so that's what Class::DBI needs to think, too.

This doesn't have anything to do with how Class::DBI is handling it internally, as all this is plainly documented. I'm not sure what the real solution is (I don't have time to make a test database and run some code against it), but I'm sure this isn't the correct solution.

----
: () { :|:& };:

Note: All code is untested, unless otherwise stated

No, the code worked because the code (as well as the exmple data) is more restrictive than the data-definition. You can define the n-m table as

mysql> create table user_role (
    -> user_id int(11) not null default '0',
    -> role_id int(11) not null default '0',
    -> primary key (user_id, role_id),
    -> foreign key user_fk (user_id) references user (id),
    -> foreign key role_fk (role_id) references role (id) 
    -> );
[download]

the data fits, the program works, and the Class::DBI definition actualy meets the data-definition as close as possible (AFAI understand it at a quick glance), but given the example-data, nothing really changed. If the data-definition allows for more than the program actualy does, thats undesirable and bad style and a big problem in the future, but not a problem now.

may well be that I don't understand you at all, I'm really tired and the finer points of english conversation escape me all the time....

regards,
tomte

---

Hlade's Law:

If you have a difficult task, give it to a lazy person --
they will find an easier way to do it.