We Keep Coding

Performing a join on a group by linq

I'm trying to perform a group by and then a join after (based on what other posts here suggested) and i'm getting the following error in .NET 5
'RelationalProjectionBindingExpressionVisitor' failed. This may
indicate either a bug or a limitation in Entity Framework.
code:
var test = (from ra in _db.ResourceAllocation
group ra by new { ra.Date, ra.ResourceId } into g
join resource in _db.Resource on g.Key.ResourceId equals resource.Id
select new
{
ResourceName = resource.Name,
ResourceDayOfWeek = resource.DayOfWeek,
g.Key.Date,
userSum = g.Sum(x => x.userCount)
}).OrderByDescending(e => e.Date).ToList();

Try to simplify life for LINQ Translator and separate queries by parts. This query should be translatable:
var grouped =
from ra in _db.ResourceAllocation
group ra by new { ra.Date, ra.ResourceId } into g
select new
{
g.Key.Date,
g.Key.ResourceId,
userSum = g.Sum(x => x.userCount)
}
var query =
from g in grouped
join resource in _db.Resource on g.ResourceId equals resource.Id
select new
{
ResourceName = resource.Name,
ResourceDayOfWeek = resource.DayOfWeek,
g.Date,
g.userSum
);
var result = query.OrderByDescending(e => e.Date).ToList();

SUM and COUNT in single LINQ to SQL query

I'm trying to create the following query in LINQ-TO-SQL.
select count(*), sum( o.CostInCents ) from Orders o
where Flag = true;
I came up with the following LINQ query:
var q = db.Orders
.Where(o => o.Flag )
var result = q
.GroupBy(o => 1)
.Select(g => new MyDTO
{
NoOfOrders = g.Count(),
TotalInCents = g.Sum(o => o.CostInCents )
})
.SingleOrDefaultAsync();
Is there a better way?
Is .GroupBy(o => 1) even OK?
The other option would be to do two queries, like below.
var q = db.Orders
.Where(o => o.Flag );
//No groupBy
var result2 = new MyDTO
{
NoOfCostedOrders = q.Count(),//hit the db
TotalInCents = q.Sum(o => o.CostInCents )//hit the db 2nd time
};
How should I judge which approach is better?
Thanks in advance!

This query can be rewritten in sql format as follows
var orderList = db.Orders.Where(o => o.Flag );
var orderSummary = from o in orderList
group o by 1 into p
select new
{
Items = p.Count(),
Total = p.Sum( x => x.CostInCents)
}

I think what you are searching for is the following slight adjustment:
var q = db.Orders
.Where(o => o.Flag).Select(o => o.CostInCents).ToList(); // hit the db here once
//No groupBy
var result2 = new MyDTO
{
NoOfCostedOrders = q.Count(), // don't hit the db
TotalInCents = q.Sum() // don't hit the db a 2nd time
};
If you have a question to my adjustment feel free to comment.

Linq Dynamic Query With Group By

I need extra where clause for my Linq query. For example if customer choose a date filter so i need to date filter to my query etc... When i try to myQuery.Where predicate there is visible just group by's field.
How can i append new where condition to my query.
//for example i need dynamically append o.OrderDate==Datetime.Now or another where clause
var myQuery =(from o in _db.Orders
join l in _db.OrderLines.Where(x => x.ParaBirimi == model.ParaBirimi) on o.orderId equals
l.OrderId
where o.OrderDate.Value.Year == year1
group o by new {o.OrderDate.Value.Month}
into g
select
new
{
Month = g.Key.Month,
Total = g.Select(t => t.OrderLines.Sum(s => s.OrderTotal)).FirstOrDefault()
});

You are too late at the end of the query to add new Where. You have already grouped the data, and projected it, removing nearly all the fields.
Try:
var baseQuery = from o in _db.Orders
join l in _db.OrderLines.Where(x => x.ParaBirimi == model.ParaBirimi) on o.orderId equals l.OrderId
where o.OrderDate.Value.Year == year1
select new { Order = o, OrderLine = l };
if (something)
{
baseQuery = baseQuery.Where(x => x.Order.Foo == "Bar");
}
var myQuery = (from o in baseQuery
group o by new { o.Order.OrderDate.Value.Month }
into g
select
new
{
Month = g.Key.Month,
Total = g.Sum(t => t.OrderLine.OrderTotal)
});
Clearly you can have multiple if. Each .Where() is in && (AND) with the other conditions.
Note how the result of the join is projected in an anonymous class that has two properties: Order and OrderLine

Query Users who have all requested records in two tables

Data structure looks like:
User(id)
UserApp(user_id, app_id)
UserSkill(user_id, skill_id)
Using linq-to-sql or EF, how would I construct a query to elegantly return only users who possess every requested app and skill?
In addition, how would I adjust the query to return any user who possesses at least one of the requested apps or skills? Essentially an OR vs AND (above).
UPDATE 1:
So I think we're close. Basically I want to only return users who have ALL the requested apps and skills. If we have two arrays of requested ids for skills and apps:
int[] requestedAppIDs // [1, 2, 3]
int[] requestedSkillIDs // [4, 5, 6]
I would only want to return a user if they have apps 1,2,3 AND skills 4,5,6.
var usersWithAllSelectedAppsAndSkills =
context.Users
.GroupJoin(context.UserApp,
k => k.id,
k => k.user_id,
(o, i) => new { User = o, UserApps = i })
.GroupJoin(context.UserSkill,
k => k.User.id,
k => k.user_id,
(o, i) => new { User = o.User, o.UserApps, UserSkills = i })
.Where(w => !requestedAppIDs.Except(w.UserApps.Select(x => x.app_id).ToArray()).Any() && !requestedSkillIDs.Except(w.UserSkills.Select(x => x.skill_id).ToArray()).Any())
.Select(s => s.User)
.ToList();
Obviously, LINQ does not know how to translate the UserSkills.Select().ToArray()'s in my Where() to SQL. How can I accomplish this?
And, secondarily the OR solution as well (user has any one of the requested apps or skills).

This will do the job as long as the user_id – app_id and user_id – skill_id values in the UserApp and UserSkill tables are unique.
var requestedSkillIDs = new[] { 4, 5, 6 };
var skillCount = requestedSkillIDs.Length;
var requestedAppIDs = new[] { 1, 2, 3 };
var appCount = requestedAppIDs.Length;
using (var context = new TestContext()) {
context.Database.CreateIfNotExists();
var appQuery = context.UserApp.Where(p => requestedAppIDs.Contains(p.AppId))
.GroupBy(p => p.UserId)
.Where(p => p.Count() == appCount);
var skillQuery = context.UserSkill.Where(p => requestedSkillIDs.Contains(p.SkillId))
.GroupBy(p => p.UserId)
.Where(p => p.Count() == skillCount);
var result = from a in appQuery
join s in skillQuery on a.Key equals s.Key
join u in context.Users on s.Key equals u.Id
select u;
var users = result.ToList();
}

Here's one way to do it, I hope i got all the syntax right :)
using (var context = new YourContext())
{
var usersWithAllSkills = context.User
.Where(w => w.id == yourId)
.Join(context.UserApp,
k => k.id,
k => k.user_id,
(o,i) => o)
.Join(context.UserSkill,
k => k.id,
k => k.user_id,
(o,i) => o)
.ToList();
var usersWithAnySkill = context.User
.Where(w => w.id == yourId)
.GroupJoin(context.UserSkill,
k => k.id,
k => k.user_id,
(o,i) => new { User = o, UserSkills = i })
.GroupJoin(context.UserApp,
k => k.User.id,
k => k.user_id,
(o,i) => new { User = o.User, o.UserSkills ,UserApps = i })
.Where(w => w.UserSkills != null || w.UserApps != null)
.Select(s => s.User)
.ToList();
}

For the first case (AND) You just need to make inner join like below:
from t1 in db.UserApp
join t2 in db.UserSkill on t1.user_id equals t2.user_id
where t1.app_id == "someID" && t2.skill_id == "someID"
select new { t1.user_id,t1.user_app_id, t2.user_skill}
For the second case just swap &&(AND) with ||(OR).

There is a more direct way to write the required queries using L2E. To write these queries you have to forget thinking in SQL and start thinking in LINQ.
For the first case, look for users which have all the skills:
var usersWithAll = ctx.Users2.Where(u =>
appIds.All(aid => u.Apps.Any(a => a.AppId == aid))
&& skillIds.All(sid => u.Skills.Any(s => s.SkillId == sid))
);
Translated as: get the users where, for all the appIds the user has al leat an app with that application id and for all skillIds the user has at least one skill with that id
And, for the second case, users which have any of the apps and any of the skills:
var usersWithAny = ctx.Users2.Where(u =>
appIds.Any(aid => u.Apps.Any(a => a.AppId == aid))
&& skillIds.Any(sid => u.Skills.Any(s => s.SkillId == sid))
).ToList();
Translated as: get the users where, for at least one appId the user has an app with that application id and for any skillIds the user has at least one skill with that id
If you run this test class, you'll also see the executed query (please, note that, to do so, I'm using the Log property of Database. I think it's only available from EF6 on).
namespace Tests
{
[TestClass]
public class CheckSeveralRelationsAtOnce
{
[TestMethod]
public void HasAllAppsAndSkills()
{
int[] appIds = {1, 2, 3};
int[] skillIds = {6, 7, 8};
using (var ctx = new MyDbContext())
{
ctx.Database.Log = Console.Write;
var usersWithAll = ctx.Users2.Where(u =>
appIds.All(aid => u.Apps.Any(a => a.AppId == aid))
&& skillIds.All(sid => u.Skills.Any(s => s.SkillId == sid))
).ToList();
Assert.IsNotNull(usersWithAll);
}
}
[TestMethod]
public void HasAnyAppsOrSkill()
{
int[] appIds = { 1, 2, 3 };
int[] skillIds = { 6, 7, 8 };
using (var ctx = new MyDbContext())
{
ctx.Database.Log = Console.Write;
var usersWithAny = ctx.Users2.Where(u =>
appIds.Any(aid => u.Apps.Any(a => a.AppId == aid))
&& skillIds.Any(sid => u.Skills.Any(s => s.SkillId == sid))
).ToList();
Assert.IsNotNull(usersWithAny);
}
}
}
}

I believe the answer by codeworx is correct for getting users with all the skills / apps
As an aside - I answered pretty much the same question recently with a pure SQL solution (SQL Server) - that could be turned into a stored procedure (with a table valued parameter) - See here if interested. This will perform better for a large number of values. Entity framework will turn every skill/app in the list into its own SQL parameter, which is much slower.
Unfortunately Entity framework doesn't really support table valued parameters yet - although you can use the entity framework connection to directly call a stored procedure with a table valued parameter (as per this article
Back to the question at hand...
I'll add the (easier) query to select a user with ANY of the skills & ANY of the apps:
var result = from u in context.Users
join _a in (
from a in context.UserApp
where requestedAppIDs.Contains(a.AppId)
select a.UserId;
) on u.Id equals _a
into aGrp
join _s in (
from s in context.UserSkill
where requestedSkillIDs.Contains(s.SkillId)
select s.UserId;
) on u.Id equals _s
into sGrp
where aGrp.Any()
&& sGrp.Any()
select u;
And just for completeness - the ALL solution again:
var skillCount = requestedSkillIDs.Length;
var appCount = requestedAppIDs.Length;
var result = from u in context.Users
join _a in (
from a in context.UserApp
where requestedAppIDs.Contains(a.AppId)
select a.UserId;
) on u.Id equals _a
into aGrp
join _s in (
from s in context.UserSkill
where requestedSkillIDs.Contains(s.SkillId)
select s.UserId;
) on u.Id equals _s
into sGrp
where aGrp.Count() == appCount
&& sGrp.Count() == skillCount
select u;
and finally - an example where the main query body is fixed, but you can add differing where clauses depending upon the AND/OR requirement
bool onlyReturnWhereAllAreMatched = false;
var skillCount = requestedSkillIDs.Length;
var appCount = requestedAppIDs.Length;
IQueryable<User> result;
var query = from u in context.Users
join _a in (
from a in context.UserApp
where requestedAppIDs.Contains(a.AppId)
select a.UserId;
) on u.Id equals _a
into aGrp
join _s in (
from s in context.UserSkill
where requestedSkillIDs.Contains(s.SkillId)
select s.UserId;
) on u.Id equals _s
into sGrp
select new {u, aCount = aGrp.Count(), sCount = sGrp.Count()};
if (onlyReturnWhereAllAreMatched)
{
result = from x in query
where x.aCount == appCount
&& x.sCount == skillCount
select x.u;
} else {
result = from x in query
where x.aCount > 0
&& x.sCount > 0
select x.u;
}

Linq to SQL in C# Unique data

I have a long list with ships that I get from a Linq to SQL query, but I only want one row per ImoNo. Today i have about 4 rows per ImoNo. I just need the one row that has been last updated (so in this example I need 2013-01-27).
This is my Linq To SQL query:
var res = from positions in context.Lloyds_ETAs
join vessels in context.Lloyds_Vessels on positions.ImoNumber equals vessels.imo_no
select new PositionData {
ImoNo = positions.ImoNumber,
PositionCordinates = positions.AIS_Latest_Position,
CompassOverGround = positions.Compass_over_Ground_Heading,
VesselId = positions.Vessel_ID,
Equipment = vessels.vessel_type,
Updated = positions.Last_Place_Location
};
return res.ToList();

var res = (from positions in context.Lloyds_ETAs
join vessels in context.Lloyds_Vessels on positions.ImoNumber equals vessels.imo_no
select new PositionData {
ImoNo = positions.ImoNumber,
PositionCordinates = positions.AIS_Latest_Position,
CompassOverGround = positions.Compass_over_Ground_Heading,
VesselId = positions.Vessel_ID,
Equipment = vessels.vessel_type,
Updated = positions.Last_Place_Location
})
.GroupBy(x => x.ImoNo)
.Select(g => g.OrderByDescending(pd => pd.Updated).First());

If you want the last one, all you have to do is append .OrderBy(pd => pd.Updated).Last() after your select.
var res = (from positions in context.Lloyds_ETAs
join vessels in context.Lloyds_Vessels on positions.ImoNumber equals vessels.imo_no
select new PositionData {
ImoNo = positions.ImoNumber,
PositionCordinates = positions.AIS_Latest_Position,
CompassOverGround = positions.Compass_over_Ground_Heading,
VesselId = positions.Vessel_ID,
Equipment = vessels.vessel_type,
Updated = positions.Last_Place_Location
}).OrderBy(pd => pd.Updated).Last();
return res.ToList();

(yourQuery).OrderByDescending(pd=>pd.Updated).First()

There are several ways to get just one "row" as a result:
res.OrderByDescending(x => x.Updated).Take(1);
res.OrderByDescending(x => x.Updated).First();
res.Order(x => x.Updated).Last();
It seems that you have some duplication though, so maybe doing a group by would be more appropriate.