[Gllug] wikis

[Pete Ryland]

On Thu, Oct 09, 2003 at 02:07:26PM +0100, Richard Jones wrote:
> On Thu, Oct 09, 2003 at 12:50:26PM +0100, itsbruce at uklinux.net wrote:
> > Off on a slight tangent, I'm really get pissed off by people writing
> > applications that only work with one database backend when there's no
> > bloody excuse.
> 
> Shurely this will all be resolved when everyone moves to the One True
> Database (ie. PostgreSQL) :-?
> 
> But seriously, I did once write a significant sized application which
> worked equally well against Informix and PostgreSQL, and it was a
> significant amount of work. I ended up implementing a whole set of
> PostgreSQL stored procedures which emulated the peculiarities of
> Informix (horrible database BTW). This was using the Perl DBI which
> certainly *doesn't* isolate you from details of the databases.
> 
> That's why I tend nowadays to write stuff which works against
> PostgreSQL only.

I totally agree.  The simple fact that mysql (which I assume is being
referred to) has trouble with all manner of *standard* queries (like
sub-queries) should not hinder one's design IMHO.  If I have to write code
that implements even the relatively simple sql:

  update blah where blah in (select blah)

by wrapping it in code, something like:
  
  r = do("select blah")
  foreach id in r
    do("update blah where id=blah")

then it will be dog slow on *all* databases.  No database abstraction layer
will deal with that.  So calling it "immature database wars" is not
necessarily very accurate - it's more complicated than that.

And besides, if everyone coded their queries to the lowest common
denominator[*] then mysql wins, since it is almost always the LCM and is
very efficient because of that! ;-)

Pete

[*] or, rather, the HCF, for the purists/pedants.

-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug