Re: [Haskelldb-users] composing relational operations with aggregate expressions

[Bjorn B. <bri...@cs...>]

On Jul 11, 2006, at 2:38 PM, Matthias Radestock wrote:

> I am trying to construct a haskelldb query that returns the count =20
> of the
> ids in one table that are not present in another table. Here's what I
> came up with:
>
> test1Query =3D
>     do t <- table Table1.table1
>        project (idField << t1!Table1.xid)
>
> test2Query =3D
>     do t <- table Table2.table2
>        project (idField << t2!Table2.xid)
>
> test3Query =3D
>     do ids <- minus test1Query test2Query
>        project (countField << count (ids!idField))
>
> Unfortunately this is not producing the right SQL. showSql returns
>
> (SELECT DISTINCT COUNT(id) as count
>  FROM  table1 as T1)
> EXCEPT
> (SELECT DISTINCT COUNT(id) as count
>  FROM  table2 as T1)
>
> The same happens for other combinations of relational operations with
> aggregate expressions, e.g. 'union' combined with '_min'. In all cases
> the aggregate expression gets pushed under the relational operation,
> which is just plain wrong.
>
> Am I not constructing my queries the right way or is this a bug?

That sure seems like a bug. Try changing line 187 in Optimize.hs to

           	=3D not (any (isAggregate.snd) assoc)

Let me know if that helps and I'll fix it in darcs.

> Matthias.
> PS: Is the haskelldb bug tracker at
> http://sourceforge.net/tracker/?group_id=3D101095&atid=3D629040 still =
in
> use? There hasn't been any activity on it for nearly 18 months.

It's still in use in the sense that most of those bugs still haven't =20
been fixed. It's not in use in the sense that noone has reported any =20
new bugs since then. It's basically a result of the lack of users and =20=

releases. Which in turn probably depends on a lack of developer time. =20=

Basically we would only enter stuff there if it couldn't be fixed =20
right away.

/Bj=F6rn