sqlobject-discuss

[SQLObject] FloatCol, distinct+count and others

[Marcin W. <wo...@un...>]

Hi,
I will write about a few things, that make me not 100% happy
with SQLObject, in one mail.

First about MultipleJoin: as Robert wrote in
http://sourceforge.net/mailarchive/forum.php?thread_id=3D5435228&forum_id=
=3D30269
person.addresses will generate n+1 queries to DB.

About FloatCol, what do you think should be a type of FloatCol?
I think double precision in every database (but I'm not sure if other
people agree), as I wrote in
http://sourceforge.net/mailarchive/forum.php?thread_id=3D5604389&forum_id=
=3D30269

About count() in distinct'ed select: now it silently returns not-distinct=
ed
count. As Jeremy wrote, SQLite doesn't support count(distinct ...),
but I think the patch from
http://sourceforge.net/mailarchive/forum.php?thread_id=3D5545866&forum_id=
=3D30269
would work fine - it works as expected in dbms that support count(distinc=
t),
and probably raises exception in SQLite, what is better than returning
wrong value.

And when I found SQLObject, I first downloaded "Mostly Live CVS Tarball".
It took me a while to realize, that it is outdated. Now it seems to be
older than 0.5.2 tarball.

Anyway, I'm quite happy with SQLObject :)

Marcin

--=20
Marcin Wojdyr  |  http://www.unipress.waw.pl/~wojdyr

Re: [SQLObject] FloatCol, distinct+count and others

[Ian B. <ia...@co...>]

Marcin Wojdyr wrote:
> Hi,
> I will write about a few things, that make me not 100% happy
> with SQLObject, in one mail.
> 
> First about MultipleJoin: as Robert wrote in
> http://sourceforge.net/mailarchive/forum.php?thread_id=5435228&forum_id=30269
> person.addresses will generate n+1 queries to DB.

I don't think that is what happens...?  It's certainly not intended to 
cause that many selects.

> About FloatCol, what do you think should be a type of FloatCol?
> I think double precision in every database (but I'm not sure if other
> people agree), as I wrote in
> http://sourceforge.net/mailarchive/forum.php?thread_id=5604389&forum_id=30269

Sure, I guess that's fine.

> About count() in distinct'ed select: now it silently returns not-distincted
> count. 

Right now it's supposed to fail with an assertion error.  If it's not, 
I'll have to look into it.  Or...

> As Jeremy wrote, SQLite doesn't support count(distinct ...),
> but I think the patch from
> http://sourceforge.net/mailarchive/forum.php?thread_id=5545866&forum_id=30269
> would work fine - it works as expected in dbms that support count(distinct),
> and probably raises exception in SQLite, what is better than returning
> wrong value.

I can just apply that patch.  It's probably fine even without 
cross-database support.

> And when I found SQLObject, I first downloaded "Mostly Live CVS Tarball".
> It took me a while to realize, that it is outdated. Now it seems to be
> older than 0.5.2 tarball.

Oops, I have to get rid of that link, thanks.

-- 
Ian Bicking  /  ia...@co...  /  http://blog.ianbicking.org

Re: [SQLObject] FloatCol, distinct+count and others

[Robert L. <ro...@le...>]

(Oops, replied directly to Ian - hey, it's 4:45am what else do you expect!)

Ian Bicking wrote:
> Marcin Wojdyr wrote:
> 
>> Hi,
>> I will write about a few things, that make me not 100% happy
>> with SQLObject, in one mail.
>>
>> First about MultipleJoin: as Robert wrote in
>> http://sourceforge.net/mailarchive/forum.php?thread_id=5435228&forum_id=30269 
>>
>> person.addresses will generate n+1 queries to DB.
> 
> 
> I don't think that is what happens...?  It's certainly not intended to 
> cause that many selects.

Unfortunately, it does, at least it does when using Postgres, a fresh install of
SQLObject 0.6, Python 2.3.3, W2000 :-(

The first select is :

SELECT id FROM address WHERE person_id = ###

which returns 'n' id's

then for every id in the returned list there is :

SELECT column(s)... FROM address WHERE id = ####

i.e. n+1 queries to DB.

Robert