Alternately you can quote the columns as identifiers. Unfortunately 
there isn't a cross-db way to do that, but you can look at Rdo for 
examples.

Cool. I sure am getting a lesson in SQL today :)



Thanks. I'm gonna create this as another ticket and do the changes 
after every thing else I'm in the middle of settles down...



Thanks again!

Aliasing something as desc or default will result in a  sql sintax 
error as DESC and DEFAULT are reserved words.

I understand using the sql alias in the select statement, but why 
would we need to rename the table columns to do this?

Ah, of course. MDB2_Results:: does the loop. Forgot about that...



Thanks.

One step less, some servers tends to access table headers to read 
column list before processing the request.
The first loop is the fetchAll, the second create objects. With while 
you fetch the result and create the object in one loop. If you wish, 
the row variable can be checked for errors, but is not needed as the 
statement is executed and checked before. With a little renaming of 
table columns the parameters replace loop could be replace either. for 
example the gallery_desc becomes gallers_description and you can 
select columns with "AS" sql alias inside select statement. Reducing 
the attribute selection only to one loop. Now the are 3, not needed.

k. Thanks for the info, committed that part of the patch as well.

If not a performance gain, this is a least good practice. And I'm 
pretty sure this *is* a performance difference, at least on some RDBMs.

What's the rationale for this? In this case we actually do want all 
the columns. Is there a performance issue with '*' vs putting all the 
column names in the query explicitly?
I don't see how this is reducing the number of loops, your just 
replacing a for loop for a while loop from what I can tell. Also, with 
your change we're no longer catching any errors that the ->fetchAll() 
(->fetchRow() in your version) call can return.
Agreed, thanks.
Yea, I thought I already took that out after I realised that we very 
rarely need to obtain the count for the same gallery more than once 
per page load. Thanks for catching that.

- Don't use SELECT *

- Combine the fetchAll db driver with the loop that creates galleries objects

- As all fields are prefixed with "gallery_id" don't str_replace but 
use substr(), is faster

- countGalleries caches results but never returns it if recalled, as 
this was not noticed till now I guess that is never recalled so remove 
the parameters serialization