hi,

it's me again:
while solving my other problem, i came accross the question if it's always necessary to have an id-section for PK in the mapping files?

let's say, i have a SQL-view which is just used for reading and i'm not interested in having a primary key, how can i achieve this?

thanks.

In your case, the id would be "assigned", without generator.

hi,
thank you for your input.

what i mean is, are there any chance of getting rid of the <id>-block within the mapping files?
in my case, i don't need the <id>-block since i'm just looking up some information in that view - but having the <id>-block enforces me to load also the PK/ID column within the view and i'm trying to avoid this.

(i tried also "mapping" the <id>-block onto an existing column which is loaded within the view, but this causes bugs.)

the problem is: if i throw out the <id>-block, the application won't run...

as far as I know, you have to always specify the id.

hi,

thanks for your reply.

ok, if i can't get rid off it ... then i can't get rid off it.

perhaps someone from the NHibernate team can confirm this?

i'm afraid there is no way to eliminate the id as it acts as an object identifier. We declare an artificial primary key on each view by adding a guid value via



to each view definition.

Luedi,

is it possible to map the <id> tag on a distinct coloumn which is included in the views output or may this interfere with the result of the view (and its validity) ?

the reason is simple:
in my case, i don't need a PK in the result of my view, so i need to map the <id>-tag to another column, if the <id>-tag is necessary...

definetly yes. But you must ensure that the column values are unique.