I'm bumping this one,

if the description is not clear, is there anything I can do to help
your understanding of the problem?

On Mar 16, 10:30 am, Gauthier Segay <[email protected]> wrote:
> I'm working on an application which has no ORM and a rather special data 
> model.
>
> The model is made of several entities that fits a relational model on
> the logical side, one entity may be related to another, etc. But the
> physical implementation is relying on a specific mechanism targeting
> issue with copying of all the data in a specific use case.
>
> The application is managing large sets of data composing a case.  Each
> case (or "top case") references a number of subcases.
> The user is allowed to create a new case from an existing one;
> physically, creating a new case is just adding a row to the case
> table, this row referencing each set of entities (i.e. the row holds
> the subcase identifiers), the application starts duplication of
> a subset of data only if any data is changed in subcase data that is
> shared among multiple cases.
>
> So, what is special, is that the model is not using strong foreign
> keys to maintain the relations, it's somewhat handled by the
> application code and what I would call loose coupled relations.
>
> The relation is made of the subcase identfier (coming from the case
> row for a given entity) and an arbitrary identifier.
>
> here is a sample schema script (mssql)
>
> ----------------------
> create table cases (
>        [key] int identity
>        , shop_case_key int
>        , product_case_key int
>        , product_by_shop_case_key int
>        , name nvarchar(100)
> );
>
> create table shop_cases (
>        [key] int identity
>        , subcase_name nvarchar(100)
> );
>
> create table product_cases (
>        [key] int identity
>        , subcase_name nvarchar(100)
> );
>
> create table product_by_shop_cases (
>        [key] int identity
>        , subcase_name nvarchar(100)
> );
>
> create table shops (
>        shop_case_key int
>        , shop_id int
>        , name nvarchar(100)
>        , product_space int
> );
>
> create table products (
>        product_case_key int
>        , product_id int
>        , name nvarchar(100)
>        , price decimal
> );
>
> create table product_by_shop (
>        product_by_shop_case_key int
>        , product_id int
>        , shop_id int
> );
>
> insert into shop_cases (subcase_name) values ('shopcase1');
> insert into shop_cases (subcase_name) values ('shopcase2');
>
> insert into product_cases (subcase_name) values ('productcase1');
> insert into product_cases (subcase_name) values ('productcase2');
>
> insert into product_by_shop_cases (subcase_name) values 
> ('productbyshopcase1');
>
> insert into cases (shop_case_key, product_case_key,
> product_by_shop_case_key, name)
> select 1,1,1,'case 1'
> union all
> select 1,2,1,'case 2'
> union all
> select 2,1,1,'case 3'
>
> insert into shops (shop_case_key, shop_id, name, product_space)
> values(1, 1, 'shop 1', 100);
> insert into shops (shop_case_key, shop_id, name, product_space)
> values(1, 2, 'shop 2', 200);
> insert into shops (shop_case_key, shop_id, name, product_space)
> select 2, shop_id, name, product_space*2 from shops
>
> insert into products (product_case_key, product_id, name, price)
> values (1, 1, 'product 1', 1);
> insert into products (product_case_key, product_id, name, price)
> values (1, 2, 'product 2', 10);
> insert into products (product_case_key, product_id, name, price)
> select 2, product_id, name, price * 3 from products;
>
> insert into product_by_shop (product_by_shop_case_key   , product_id,
> shop_id) values (1, 1, 1);
> insert into product_by_shop (product_by_shop_case_key   , product_id,
> shop_id) values (1, 2, 1);
> insert into product_by_shop (product_by_shop_case_key   , product_id,
> shop_id) values (1, 1, 2);
> ----------------------
>
> for a given case key and shop id, looking at the product list for a
> given shop, the SQL whould look like that
>
> select
>        c.*, s.shop_id, p.*
> from
>        cases c
>        inner join shop_cases sc on c.shop_case_key = sc.[key]
>        inner join product_cases pc on c.product_case_key = pc.[key]
>        inner join product_by_shop_cases pbsc on c.product_by_shop_case_key =
> pbsc.[key]
>        inner join product_by_shop pbs on pbs.product_by_shop_case_key
> = pbsc.[key]
>        inner join shops s on
>                sc.[key] = s.shop_case_key
>                and pbs.shop_id = s.shop_id
>        inner join products p on
>                pc.[key] = p.product_case_key
>                and pbs.product_id = p.product_id
> where
>        c.[key] = @caseKey
>        and s.shop_id = @shopId
>
> as you see this is turning pretty complicated with a simple model, the
> application is quite larger than this.
>
> I'm investigating possibilities to fit this in a normal NH model where
> I could use such things:
>
> using(var session = CraftSessionForCaseKey(1) /* some black magic in
> setting the session or session factory */){
>        var shop = session.GetByKey<Shop>(1);
>        var products = shop.Products; // will issue the previous select
> statement (with product columns)
>
> }
>
> Assuming I can't change the way we put the model, is there something
> that you think is remotely feasible, even putting some amount on
> effort to override some key components involved in the way ISession or
> ISessionFactory works?
>
> I currently haven't come up with lots of ideas on how to overcome this
> issue (session filter aren't enough to manipulate entity
> relationships). Things are done this way to allow the end user to
> create new cases at will and only incurring the performance hit of
> duplicating data to a subset each time a tweak is made in an entity
> set which is used across multiple top cases.
>
> I've thought about the following:
>
> - dynamically crafting views with triggers
> - storing changes as diff
>
> but these seem to be seriously complex to me, NH could be of help with
> the first one, but for the second it involves a lot of
> application / model logic.
>
> Alternatively, did anyone encountered such a situation? what
> alternative to such model would you find applicable?
>
> Thanks a bunch for your insight.

-- 
You received this message because you are subscribed to the Google Groups 
"nhusers" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/nhusers?hl=en.

Reply via email to