ibatis-user-java mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brad Handy" <brad.ha...@gmail.com>
Subject Re: Avoiding N+1 Selects
Date Wed, 28 Feb 2007 14:10:02 GMT
I think I could use a row handler to arrange the objects in a situation like
this.  One question I have:  Do the results of a query with a custom
RowHandler still get cached when using OSCache?


Brad

On 2/22/07, Brad Handy <brad.handy@gmail.com> wrote:
>
> iBatis came back with an error message indicating too many objects were
> being returned for executeQueryForObject.
>
>
> Brad
>
> On 2/20/07, Brad Handy <brad.handy@gmail.com> wrote:
> >
> > Wouldn't I have to give unique column names for those identifying the
> > child data?
> >
> > For example
> >
> > table parentChild
> >    some_foreign_key int
> >    id int
> >    description varchar(64)
> >    parent_id int
> >
> >
> > rows
> >
> >   id          description
> > parent_id
> >   0           grandparent                                       null
> >   1           parent 1                                             0
> >   2           child 1                                               1
> >   3           child 2                                               1
> >   4           parent 2                                             0
> >   5           child 3                                               4
> >   6           parent 3                                             0
> >   7           child 4                                               6
> >   8           child 5                                               6
> >
> >
> > select
> >     p.id as parentId
> >     , p.description as parentDesc
> >     , p.parent_id as parentAncestorId
> >     , c.id as childId
> >     , c.description as childDesc
> >     , c.parent_id as childAncestorId
> > from
> >    parentChild p
> >     , parentChild c
> > where
> >    p.some_foreign_key = c.some_foreign_key
> >    and p.id = c.parent_id
> >    and p.id <> c.id
> > order by
> >    p.id
> >    , c.id
> >
> >
> > should return
> >
> > parentId  parentDesc               parentAncestorId  childId
> > childDesc                childAncestorId
> > 0           grandparent               null                      1
> > parent 1                    0
> > 0           grandparent               null                      4
> > parent 2                    0
> > 0           grandparent               null                      6
> > parent 3                    0
> > 1           parent 1                    0
> > 2        child 1                       1
> > 1           parent 1                    0
> > 3        child 2                       1
> > 4           parent 2                    0                         5
> >        child 3                       4
> > 6           parent 3                    0                         7
> >        child 4                       6
> > 6           parent 3                    0                         8
> >        child 5                       6
> >
> >
> >
> >
> >
> > On 2/20/07, Clinton Begin < clinton.begin@gmail.com > wrote:
> > >
> > >
> > > i.e. is it a join with a bridge table or not (1:M or M:N)?
> > >
> > > Regardless, I wonder if a recursive result map would work....It
> > > might.
> > >
> > > <resultMap id="Node" ... >
> > >   <result name="children" ... resultMap="Node"/>
> > > ...
> > >
> > > I don't see any reason why that would cause any problems...perhaps try
> > > it and let us know.
> > >
> > > Cheers,
> > > Clinton
> > >
> > > On 2/19/07, Clinton Begin <clinton.begin@gmail.com> wrote:
> > > >
> > > >
> > > > Oh...is it a self join?
> > > >
> > > > Clinton
> > > >
> > > > On 2/19/07, Brad Handy < brad.handy@gmail.com> wrote:
> > > > >
> > > > > I guess I'll have to take a different approach.  The depth isn't
> > > > > set to be a defined level; so if I go more than two levels below
the
> > > > > grandparent, it won't work.
> > > > >
> > > > > On 2/19/07, Clinton Begin <clinton.begin@gmail.com> wrote:
> > > > > >
> > > > > > You'll need one query with 3 result maps.  The result maps will
> > > > > > be chained together with collection properties using the resultMap
> > > > > > attribute.
> > > > > >
> > > > > >
> > > > > > <resultMap id="Child">
> > > > > >   ...
> > > > > > <resultMap id="Parent">
> > > > > >   <result ... resultMap="Child"/>
> > > > > >   ...
> > > > > > <resultMap id="GrandParent">
> > > > > >   <result ... resultMap="Parent"/>
> > > > > >   ...
> > > > > > <select ... resultMap="GrandParent">
> > > > > > ...
> > > > > >
> > > > > > The select statement should join the tables together and you
may
> > > > > > need to be very explicit with the column names.
> > > > > >
> > > > > > Clinton
> > > > > >
> > > > > > On 2/19/07, Brad Handy <brad.handy@gmail.com> wrote:
> > > > > > >
> > > > > > > I have a table which has all of the parent/child relationships
> > > > > > > in the same table.  I would like to avoid the N+1 selects
with this
> > > > > > > construct, but it's unclear from the documentation if this
can be done.
> > > > > > >
> > > > > > > Let's say I have the following relationships defined in
the
> > > > > > > table:
> > > > > > >
> > > > > > > Grand Parent
> > > > > > >    Parent 1
> > > > > > >       Child 1
> > > > > > >       Child 2
> > > > > > >    Parent 2
> > > > > > >    Parent 3
> > > > > > >       Child 3
> > > > > > >       Child 4
> > > > > > >
> > > > > > >
> > > > > > > When creating the child objects for "Grand Parent", will
the
> > > > > > > same "Parent*" objects be used to add the children "Child*"
objects to the
> > > > > > > appropriate parents?
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > Brad
> > > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

Mime
View raw message