SQL 2008 | .NET 4.0 | NHibernate 3.1 | NHibernate.Castle 3.1 | Castle.Core 2.5.2
So I have a linking table with metadata, like the author of this question NHibernate Mapping a Many to Many with Data on Join Table
Initially, I mapped just like the answer to this question as it seemed the most parsimonious way to handle it. However, after turning on show_sql
and observing what was going on, the ID lookups ended up yielding N+1 queries where N is the number of associations.
Observe this example database that is analogous to my actual data, defined in sql-like syntax
CREATE TABLE [User]
(
Id int PRIMARY KEY
)
CREATE TABLE UserPref
(
Id int PRIMARY KEY,
Name varchar(32)
)
CREATE TABLE UserPrefAssociation
(
UserId int,
PrefId int,
Value varchar(32)
)
I hacked the following code together with this User one-to-many object mapping IList<UserPrefAssociation> Preferences { get; set; }
public IDictionary<string, string> GeneratePrefDict()
{
return Preferences
.ToDictionary(i => i.UserPref.Name, i => i.Value);
}
Sure, this works great, but as mentioned before, each i.UserPref.Name
, is an additional query to SQL.
After playing in SQL, I have found the query that accomplishes what I want. My question then becomes how can I do this with NHibernate?
SELECT UserPref.Name, UserPrefAssociation.Value
FROM [User]
INNER JOIN UserPrefAssociation ON [User].Id = UserPrefAssociation.UserId
INNER JOIN UserPref ON UserPrefAssociation.UserPrefId = UserPref.Id
WHERE [User].Id = 1
~~~~SOLVED~~~~~
using NHibernate.Linq;
...
public IDictionary<string, string> GeneratePrefDict(ISession s)
{
return
(from entry in s.Query<User_UserPref>()
where entry.User == this
select new
{
key = entry.UserPref.Name,
value = entry.Value
})
.ToDictionary(i => i.key, i => i.value);
}
Generates this SQL
NHibernate: select userpref1_.Name as col_0_0_, user_userp0_.Value as col_1_0_ f
rom User_UserPref user_userp0_ left outer join UserPref userpref1_ on user_userp
0_.UserPrefId=userpref1_.Id where user_userp0_.UserId=@p0;@p0 = 1 [Type: Int32 (
0)]
Which is better than N+1 queries, and solves my issue.