Re: possible Class::DBI::Oracle [patch]

[prev] [thread] [next] [Date index for 2005/04/22]

Jay Strauss wrote:
> Just a couple of nit picks, why not :)
> 
>> --- Oracle.pm.orig    2003-07-15 19:56:23.000000000 -0500
>> +++ Oracle.pm    2005-04-21 16:09:07.000000000 -0500
>> @@ -16,6 +16,15 @@
>> +
>> +        my $cdbi_table;
>> +
>> +        if ($table =~ /\./ and not $schema) {
> 
> 
> If someone sends in a "schema.tablename", this has to be true otherwise 
> you should raise an error.  (or I suppose you could go on further and 
> make sure that the schemas are == )

should probably be an error.

>>  
>> +        # upper-case here for backward compatibility with 0.51?
>> +        # convenience?
>>      $table = uc $table;
>> +        $schema = uc $schema;
> 
> 
> I put in the, uc $table, just for searching the oracle catalog, since 
> that is obsolete, its probably not needed.

column_info() and primary_key_info() just search the catalog anyway, so in 
the common case they'll need to be uc.  However, if someone did this:

David Naughton wrote:
 > CREATE TABLE "my_table" ...
 > CREATE TABLE "My_Table" ...

then set_up_table will fail if we uc.

So as I see it, if we uc then we're wrong in the general case (but 
convenient for the common case), and if we don't uc then we're not strictly 
backward compatible with 0.51.

Perhaps we should not uc, but _die with a message about case sensitivity if 
column_info() yields no columns.

-thepler