Re: [CDBI] Odd behaviour under RedHat vs Debian

On 17 Nov 2006, at 10:48, Mike Whitaker wrote:

> I have a chunk of code which works fine under Debian etch (CDBI  
> 3.0.14, DBI 1.52, CDBI::Loader 0.33, CDBI::L::mysql 0.30,  
> DBIx::ContextualFetch 1.03), and fails under Redhat (CDBI 3.0.16,  
> DBI 1.40 OR 1.53, CDBI::Loader 0.32 being the only differences)...
>
> Basically, it's screwing up an add_to_xxx on a has_many, viz:
>
> Can't insert new MCD::SQL::PointAward: DBD::mysql::st execute  
> failed: You have an error in your SQL syntax; check the manual that  
> corresponds to your MySQL server version for the right syntax to  
> use near '1210220 (table_id, reason, points, player_id, table_name)
> VALUES ('1210220', 'So' at line 1 [for Statement "INSERT INTO  
> 1210220 (table_id, reason, points, player_id, table_name)
> VALUES (?, ?, ?, ?, ?)
> " with ParamValues: 4='uid_solve', 1='Solved card 40', 3='46151',  
> 0='1210220', 2=10
>
> Note that the first param value (1210220) seems to have got somehow  
> inserted as the TABLE name in the generated SQL.

Well. I found it. If someone can explain why it works under 3.0.14  
and fails under 3.0.16, that would be shiny:

Oversimplifying a bit: deep down in the bowels of Class::DBI, it  
takes a template piece of SQL that looks like this:
    INSERT INTO __TABLE__ (%s) VALUES (%s)
and runs a quick substitution over it that replaces
    __TABLE__
with $class->table.

I've been overriding accessor names with:

sub accessor_name
{
     my ($class, $column) = @_;
     $column =~ s/_id$//;
     return $column;
}

which basically strips trailing '_id' off any accessor name, so that  
for example $class->player will work just as well as $class->player_id

Which is all fine and shiny. Except that the table referenced by  
MCD::SQL::PointAward has a column called table_id...
-- 
Mike Whitaker - [email protected]
Tel: +44 (0) 20 7501 1920 - Mob: +44 (0) 7971 977375
Skype: FleetfootMike; AIM: FleetfootMike