Linq and Lambda expression for a complex sql query involving joins - linq

Question

Using Linq to Entity (Entity Framework) in MVC 3 project.
My model:
Table - Users
UserID (PK)
...
Table - Clients
ClientID (PK)
Table - PropertyItems
PropertyItemID (PK)
Table - MemberContactPreference (Contains PropertyItems selected by Users- many to many)
UserID(FK)
PropertyItemID(FK)
Table ClientProperties (Contains PropertyItems that belong to Clients - many to many)
ClientID (FK)
PropertyItemID (FK)
I want to list all the distinct users that have selected all the properties selected by clients.
My Approach :
I got a list of all properties for a particular client in
Iqueryable<ClientProperty> clientProperties = GetClientProperties(ClientID)
Iqueryable<User> UsersMatchingClientProperties = GetAllUsers();
foreach (ClientProperty property in clientproperties)
{
UsersMatchingClientProperties = (from uem in UsersMatchingClientProperties
join ucp in GetAllMemberContactPreferences on
ucp.UserID == uem.UserID
where uem.MemberContactPreferences.SelectMany(
mcp => mcp.PropertyItemID == property.PropertyItemID)
select uem).Distinct;
}
It gives the right result only first time. As it doesn't reduce the number of items in UsersMatchingClientProperties with each iteration. actually it replaces the collection with new resultset. I want to filter out this collection with each iteration.
Also, any suggestions to do this in Lambda expression without using Linq.
Thanks

Answer 1

That generation of an iqueryable in a for loop seems like a dangerous thing, which could end up in a monster sql join being executed at once.
Anyway, I don't think you need that. How about something like this?
// for a given client, find all users
// that selected ALL properties this client also selected
Iqueryable<ClientProperty> clientProperties = GetClientProperties(ClientID)
Iqueryable<User> allUsers= GetAllUsers();
Iqueryable<MemberContactPreference> allMemberContactProperties = GetAllMemberContactPreferences();
Iqueryable<User> UsersMatchingClientProperties = allUsers
.Where(user => allMemberContactProperties
.Where(membP => membP.UserID==user.UserID)
.All(membP => clientProperties
.Select(clientP => clientP.PropertyID)
.Contains(membP.PropertyID)
)
);
Here is an alternative query in case you want the users that selected ANY property for a given client
// for a given client, find all users
// that selected ANY properties this client also selected
Iqueryable<ClientProperty> clientProperties = GetClientProperties(ClientID)
Iqueryable<User> allUsers= GetAllUsers();
Iqueryable<MemberContactPreference> allMemberContactProperties = GetAllMemberContactPreferences();
Iqueryable<User> UsersMatchingClientProperties = clientproperties
.Join(allMembersContactProperties, // join clientproperties with memberproperties
clientP => clientP.PropertyItemID,
membP => membP.PropertyItemID,
(clientP, membP) => membP)) // after the join, ignore the clientproperties, keeping only memberproperties
.Distinct() // distinct is optional here. but perhaps faster with it?
.Join(allUsers, //join memberproperties with users
membP => membP.UserID,
user => user.UserID,
(membP, user) => user)) // after the join, ignore the member properties, keeping only users
.Distinct();

Answer 2

I trust Hugo did a good job suggesting ways to improve your query (+1). But that does not yet explain the cause of your problem, which is the modified closure pitfall.
I think that after your loop there is some code that actually executes the query in UsersMatchingClientProperties. At that moment the query is executed with the last value of the loop variable property! (The loop variable is the closure in each query delegate that is created in an iteration, and it is modified by each iteration).
Change the loop like this:
foreach (ClientProperty property in clientproperties)
{
var property1 = property;
...
and use property1 in the query. That should solve the cause of the problem. But as said, it looks like the whole process can be improved.