So I have a many to many relationship between something known as Specialism and SpecialismCombo. What I'm trying to do is take an int[] of ids and check if there is a combo already that contains the specialisms with those ids.
I was close but not quite right.
Say I have specialisms with Ids 1 and 3 and I create a combo with those specialisms.
If I pass in 3 & 1 then it returns the expected combo id.
If I pass in 1 then it returns the combo id that has both 1 and 3.
I can't just rely on total number of specialisms associated with the combo. Because if a combo has two items, 1 and 4 and the items being matched on are 1 and 3 I don't want this coming back as a matched combo.
So it's like I do need the count of this result, and match the count of total specialisms associated to the combo. I don't quite get whether I'm after a subquery or detatchedcriteria or how to get the result I want using nhibernate criteria. Thanks for your help!
int[] SpecialismIds = ArrayExtensions.ConvertArray<int>(idCollection.Split(new char[] { '|' }));
ICriteria query = m_SpecialismComboRepository.QueryAlias("sc");
query.CreateAlias("sc.Specialisms", "s", NHibernate.SqlCommand.JoinType.InnerJoin);
ICriterion lastCriteria = null;
foreach(int i in SpecialismIds)
{
ICriterion currentCriteria = Restrictions.Eq("s.SpecialismId", i);
if (lastCriteria != null)
lastCriteria = Restrictions.Or(lastCriteria, currentCriteria);
else
lastCriteria = currentCriteria;
}
if (lastCriteria != null)
query.Add(lastCriteria);
IProjection IdCount = Projections.Count("s.SpecialismId").As("IdCount");
query.SetProjection(
Projections.GroupProperty("sc.SpecialismComboId"),
IdCount
);
query.Add(Restrictions.Eq(IdCount, SpecialismIds.Count()));
var comboId = query.List();
The sql being generated is:
SELECT this_.SpecialismComboId as y0_, count(s1_.SpecialismId) as y1_
FROM dbo.SpecialismCombo this_
inner join SpecialismComboSpecialism specialism3_ on this_.SpecialismComboId=specialism3_.SpecialismId
inner join dbo.Specialism s1_ on specialism3_.SpecialismComboId=s1_.SpecialismId WHERE s1_.SpecialismId = #p0
GROUP BY this_.SpecialismComboId HAVING count(s1_.SpecialismId) = #p1',N'#p0 int,#p1 int',#p0=3,#p1=1
EDIT - It seems like I either need the having to be something like...
HAVING count(s1_.SpecialismId) = (select count(SpecialismId)
from specialismComboSpecialism
where SpecialismComboId = y0
group by SpecialismComboId) == #p2
Or maybe it's simpler than that and I need to exclude SpecalismCombos where the combo.specialisms are not in the collection of ids.
Ie. if the combo has specialisms 1 and 3 but the collection only has 1.. then we could exclude this combo based on 3 not being in the collection…
Edit 8/8/2011
Went back to focusing on how to get the result I needed in SQL - and I believe this query works.
WITH CustomQuery AS
(
SELECT sc.SpecialismComboId,
count(s.SpecialismId) AS ItemCount
FROM SpecialismCombo sc
inner join SpecialismComboSpecialism scs on sc.SpecialismComboId = scs.SpecialismComboId
inner join Specialism s on s.SpecialismId = scs.SpecialismId
GROUP BY sc.SpecialismComboId
HAVING count(s.SpecialismId) = 2
)
SELECT CustomQuery.SpecialismComboId FROM CustomQuery
INNER JOIN SpecialismComboSpecialism scs on CustomQuery.SpecialismComboId = scs.SpecialismComboId
WHERE scs.SpecialismId in (1,4)
GROUP BY CustomQuery.SpecialismComboId
HAVING count(scs.SpecialismId) = 2
So now I just need to figure out how to call this procedure from my nhibernate code passing in the appropriate values :)
I also discovered in the process that my mapping class was wrong - as it was putting the wrong values in the mapping table (ie. the specialismid was ending up in the specialismcomboid field !)
Your solution should actually work well. The specialisms are filtered by id and there shouldn't be anything left that is not searched for, so count should work. Unless you have the same specialism joined more the once. This currentCriteria lastCriteria stuff looks a bit strange, may be there is an error. Just use Expression.In or Conjunction.
IProjection IdCount = Projections.Count("s.SpecialismId").As("IdCount");
IQuery query = session
.CreateCriteria<SpecialismCombo>("sc")
.CreateCriteria("Specialism", "s");
.Add(Expression.In("s.SpecialismId", SpecialismIds));
.SetProjection(
Projections.GroupProperty("sc.SpecialismComboId"),
IdCount);
.Add(Restrictions.Eq(IdCount, SpecialismIds.Count()));
Should result in a query like this:
select ...
from
SpecialismCombo sc
inner join -- linktable ...
inner join Specialism s on ...
where
s.SpecialismId in (1, 4)
Group By sc.SpecialismComboId
having count(*) = 2
The same in HQL
from SpecialismCombo sc
join sc.Specialism s
where s.id in (:ids)
group by sc
having count(*) = :numberOfIds
You could also join the specialism as many times as you have ids to find:
IQuery query = session.CreateCriteria<SpecialismCombo>("sc")
int counter = 0;
foreach(int id in ids)
{
string alias = "s" + counter++;
query
.CreateCriteria("Specialism", alias);
.Add(Expression.Eq(alias + ".SpecialismId", id));
}
should create a query like this:
select ...
from
SpecialismCombo sc
inner join -- linktable ...
inner join Specialism s0 on ...
inner join -- linktable ...
inner join Specialism s1 on ...
where
s0.SpecialismId = 1
and s1.SpecialismId = 4
So I ended up creating a stored proc and using SQL CTE in order to get only the specialism combos with the correct count of specialisms. Posting this in case someone else comes across a similar issue.
Rediscovered after 8 months of using nhibernate that I'd forgotten a lot of SQL stuff :)
DECLARE #IdCollectionCount INT
, #IdCollection VARCHAR(250)
, #CollectionDelimiter NVARCHAR
SET #IdCollectionCount = 2;
SET #IdCollection = '1,4';
SET #CollectionDelimiter= ',';
WITH CustomQuery AS
(
SELECT sc.SpecialismComboId,
count(s.SpecialismId) AS ItemCount
FROM SpecialismCombo sc
inner join SpecialismComboSpecialism scs on sc.SpecialismComboId = scs.SpecialismComboId
inner join Specialism s on s.SpecialismId = scs.SpecialismId
GROUP BY sc.SpecialismComboId
HAVING count(s.SpecialismId) = #IdCollectionCount
)
SELECT Top 1 CustomQuery.SpecialismComboId FROM CustomQuery
INNER JOIN SpecialismComboSpecialism scs on CustomQuery.SpecialismComboId = scs.SpecialismComboId
INNER JOIN dbo.fn_SplitDelimited(#IdCollection,#CollectionDelimiter) AS ids
ON scs.SpecialismId = CAST(ids.ListValue AS INT)
GROUP BY CustomQuery.SpecialismComboId
HAVING count(scs.SpecialismId) = #IdCollectionCount
Related
I have the following SQL request for a report.
select customers."AppId", second_dep "SecondDeps", first_dep "FirstDeps",
customers_count "Customers", registrations "Registrations"
From (select Count("AppId") as customers_count, "AppId"
FROM "Customers"
join "Advertisers" A on "Customers"."AdvertiserId" = A."AdvertiserId"
join "Categories" C2 on "Customers"."CategoryId" = C2."CategoryId"
where A."Name" in (:AdvertiserNames)
AND C2."Name" = :CategoryName
GROUP BY "AppId"
) as customers
left join
(select C."AppId", count(CE.*) as second_dep
from "CustomerEvents" as CE
inner join "Customers" C on CE."CustomerId" = C."CustomerId"
WHERE "EventType" = 'deposit'
and "Again" = TRUE
GROUP BY C."AppId") as dep2 on customers."AppId" = dep2."AppId"
left join
(select C."AppId", count(CE.*) as first_dep
from "CustomerEvents" as CE
inner join "Customers" C on CE."CustomerId" = C."CustomerId"
WHERE "EventType" = 'deposit'
and "Again" = false
GROUP BY C."AppId") as dep on customers."AppId" = dep."AppId"
left join
(select C."AppId", count(CE.*) as registrations
from "CustomerEvents" as CE
inner join "Customers" C on CE."CustomerId" = C."CustomerId"
WHERE "EventType" = 'registration'
GROUP BY C."AppId") as regs on regs."AppId" = customers."AppId";
The string with problem is
where A."Name" in (:AdvertiserNames)
I would like to skip it if AdvertiserNames is empty. Is it possible? Ok, I can check it on the code side, but this way will leads me to copy whole request with some small difference (I mean if AdvertiserNames is empty run SQL without where A."Name" in (:AdvertiserNames)). Or I can use concatenation to get suitable SQL. I dont like this way too.
About my technology stack. It's .NET Core 2.2 with PostgreSQL. Here is the code of whole report method:
public IQueryable<ByApplicationsReportModel> ByApplications(string category, List<string> advertisers)
{
var rawSql = new RawSqlString(#"
select customers.""AppId"", second_dep ""SecondDeps"", first_dep ""FirstDeps"",
customers_count ""Customers"", registrations ""Registrations""
From (select Count(""AppId"") as customers_count, ""AppId""
FROM ""Customers""
join ""Advertisers"" A on ""Customers"".""AdvertiserId"" = A.""AdvertiserId""
join ""Categories"" C2 on ""Customers"".""CategoryId"" = C2.""CategoryId""
where A.""Name"" in (#AdvertiserNames)
AND C2.""Name"" = #CategoryName
GROUP BY ""AppId""
) as customers
left join
(select C.""AppId"", count(CE.*) as second_dep
from ""CustomerEvents"" as CE
inner join ""Customers"" C on CE.""CustomerId"" = C.""CustomerId""
WHERE ""EventType"" = 'deposit'
and ""Again"" = TRUE
GROUP BY C.""AppId"") as dep2 on customers.""AppId"" = dep2.""AppId""
left join
(select C.""AppId"", count(CE.*) as first_dep
from ""CustomerEvents"" as CE
inner join ""Customers"" C on CE.""CustomerId"" = C.""CustomerId""
WHERE ""EventType"" = 'deposit'
and ""Again"" = false
GROUP BY C.""AppId"") as dep on customers.""AppId"" = dep.""AppId""
left join
(select C.""AppId"", count(CE.*) as registrations
from ""CustomerEvents"" as CE
inner join ""Customers"" C on CE.""CustomerId"" = C.""CustomerId""
WHERE ""EventType"" = 'registration'
GROUP BY C.""AppId"") as regs on regs.""AppId"" = customers.""AppId""");
var advertisersParam = new NpgsqlParameter("AdvertiserNames",
string.Join(",", advertisers) );
var categoryParam = new NpgsqlParameter("CategoryName", category);
return _context.ByApplicationsReportModels
.FromSql(rawSql, categoryParam, advertisersParam);
}
Any ideas?
You might try changing where A.""Name"" in (#AdvertiserNames) to where (A.""Name"" in (#AdvertiserNames) or #AdvertiserNames = '').
Instead of concatenating your advertisers into a string, you could just pass an array of strings directly to your query:
var advertisersParam = new NpgsqlParameter("AdvertiserNames", advertisers));
In SQL, instead of using the x IN (#advertisers) construct, you would need to change to x = ANY (#advertisers).
Note: you would still need an additional clause if you want the check to pass when #advertisers is empty.
I'm trying to convert the following MySQL statement in to LINQ query format
SELECT * FROM table1 WHERE table1.id IN (SELECT c_id FROM table2 WHERE a_id IN (1, 49) GROUP BY c_id HAVING COUNT(*) = 2) ORDER BY name
Got as far as this, but I'm drawing a blank on how to handle the IN and 2nd SELECT statement
myItems = from c in table1
let id = c.id
where ????
orderby c.name
select c;
Would appreciate some guidance with this please
Try this:
var ids=new[]{1,49};
var innerquery=table2.Where(e=>ids.Contains(e.a_id))
.GroupBy(e=>e.c_id)
.Where(g=>g.Count()==2)
.Select(g=>g.Key);
var myItems = from c in table1
where innerquery.Contains(c.id)
orderby c.name
select c;
First define your inner query,after the group by you will get a collection of IGrouping<TKey, TElement>> that represent a collection of objects that have a common key, filter the groups choosing only those where count==2, and select the keys of those groups. The second part is really easy to understand. I split the process in two queries to do it more readable, but you can merge both query in one.
I have the following SQL statement I'm trying to convert to Entity Framework.
SELECT S_NUMBER,A_NUMBER,FIRST_NAME,LAST_NAME
FROM EMPLOYEE WHERE S_NUMBER IN (
SELECT S_NUMBER
FROM EMPLOYEE
WHERE CO='ABC'
GROUP BY S_NUMBER
HAVING COUNT(*) > 1)
I've done some searching on using Group By in LINQ as well as sub-queries. I'm using LinqPad with a "C# Statement" and I came up with the following which based on some examples I found looks like it should work. However, I'm getting errors when trying to assign esn.S_NUMBER to sNumber in the anonymous object. The message says 'IGrouping' does not contain a definition for 'S_NUMBER'.
var result = from e in EMPLOYEE
where e.CO=="ABC"
group e by e.S_NUMBER into esn
select new
{
sNumber = esn.S_NUMBER
};
result.Dump();
I was under the impression that all the records would basically get put into a temp table called esn and I could be able to call the temptable.column name to assign it to my object that I will eventually return as a list.
You want to use Key instead of S_NUMBER. When grouping, the results get put into a IEnumerable<IGrouping>>. The grouping has a Key property which holds the key for that group, which in this case it's your S_NUMBER.
select new
{
sNumber = esn.Key
};
The following query should be a translation of the original SQL query. Instead of using a subquery, we're grouping and doing another from...in to "flatten" the sequence, and also checking that each grouping has a count > 1 like the original query.
var result = from e in EMPLOYEE
where e.CO=="ABC"
group e by e.S_NUMBER into esn
from e2 in esn
where esn.Count() > 1
select new
{
e.S_NUMBER,
e.A_NUMBER,
e.FIRST_NAME,
e.LAST_NAME
};
Since you're using the results of one query to filter another we can do a fairly direct transliteration of the query like so:
var result =
from e in EMPLOYEE
join f in (
from fe in EMPLOYEE
where fe.CO == 'ABC'
group null by S_NUMBER into grp
where grp.Count() > 1
select grp.Key
)
on e.S_NUMBER equals f
select new { e.S_NUMBER, e.A_NUMBER, e.FIRST_NAME, e.LAST_NAME };
Not only does this look a lot more like the original query but it should perform a bit faster (on MS SQL at least, can't speak for others) than the other form that might be simpler in LINQ but is much more complex when converted to SQL... four selects and a cross join, in my test version, vs two selects and an inner join for this one.
Of course if you prefer you can pull the inner query out as a separate IQueryable for clarity:
var filter =
from e in EMPLOYEE
where e.CO == 'ABC'
group null by S_NUMBER into grp
where grp.Count() > 1
select grp.Key;
var result =
from e in EMPLOYEE
join f in filter
on e.S_NUMBER equals f
select new { e.S_NUMBER, e.A_NUMBER, e.FIRST_NAME, e.LAST_NAME };
I have to retrieve all the columns from table1 and matching columns from table2. I have a stored procedure as :
alter Procedure [dbo].[usp_Property]
#UserId bigint =null
As
Begin
select P.PID, P.PropertyName, P.SBUArea, P.ListedOn,
P.Availability, P.Price, F.UserID, F.PID as FavProjId
from dbo.Property P left outer join dbo.Favorite F
on (F.PID=P.PID And F.UserID=#UserId)
I want to get Linq query for the same. So far I tried with something like
//User Id comes from session..
//var userId
var result=(from p in Properties
join f in Favorites
on p.PID equals f.PID into r
from r1 in r.DefaultIfEmpty()
where r1.UserID==userId
select new
{
p.PID,
p.PropertyName,
p.SBUArea, p.ListedOn,
r1.UserId
});
Can anyone please correct me. I want to use left outer join or any other alternate thing here.
If I beautify your SP's code, I get this:
DECLARE #UserId int
SET #UserId = 12435
SELECT
P.PID
,P.PropertyName
,P.SBUArea
,P.ListedOn
,P.Availability
,P.Price
,F.UserID
,F.PID AS FavProjId
FROM Property AS P
LEFT JOIN Favorite AS F
ON (F.PID=P.PID AND F.UserID = #UserId)
Now I wonder if you need that UserId in the WHERE clause of the SQL, or really in the join.
But anyway, here the LINQ-equivalent of exactly that SQL:
System.Int64 __UserId = 12435;
var query = (
from P in Repo.Property
from F in Repo.Favorite
.Where(fav=> fav.PID == P.PID && fav.UserID == __UserId)
.DefaultIfEmpty() // <== makes join left join
select new
{
PID = P.PID
,PropertyName = P.PropertyName
,SBUArea = P.SBUArea
,ListenOn = P.ListedOn
,Availabiity = P.Availability
,Price = P.Price
,UserId = F.UserID
,FavProjId = F.PID
}
);
var data = (query).ToList();
Use anonymous objects in your selection
var result = from t in table1
join x in table2
on t.id equals x.id
select new { id = t.id, col1 = t.col1, col2 = x.col2 }
If you will put the where clause after join you may get null reference exception because DefaultIfEmpty returns default value for non matching rows. You can filter the records before joining itself like this:-
var result=(from p in Properties
join f in Favorites.Where(x => x.UserID == userId)
on p.PID equals f.PID into r
from r1 in r.DefaultIfEmpty()
select new
{
p.PID,
p.PropertyName,
p.SBUArea,
p.ListedOn,
r1.UserId
});
Please note you need to access properties of Favorites using r1.
Update:
As far as I have understood you need all records from Property table and only matching rows from Favorite table. But you have a filter on your Favorite table so the ultimate data source will differ. Let me make my point clear by this example:-
Suppose you have following data in Property table:-
PID PropertyName Availability Price
1 aaa true 20
2 bbb false 10
3 ccc true 50
4 ddd false 80
5 eee true 55
6 fff false 70
and Favorite table like this:-
FID PID UserId
1 4 1001
2 2 1005
3 5 1007
And let's say you want all records for UserId 1005, then the result should contain all the property Id's from 1 till 6 even if UserId 1005 doesn't match for property Id's 4 & 2 right? So the query above is as per this understanding. Check this Fiddle with same example and output.
results = (from r in results
where r.Buildings.Any(x=>x.StructuralElements.Any(s=>s.VALUE == Model.Bedrooms.ToString() && s.CATEGORY=="RM"))
select r);
I think I'm missing joins here. But maybe they are implied? The execution runs so long I can't do a watch to evaluate the generated query expression
The biggest problem in this query is this:
--#p1 = Models.Bedrooms.ToString()
--#p2 = "RM"
SELECT * FROM Results r WHERE EXISTS
(SELECT x.* FROM Results tr JOIN Buildings x ON tr.SomeID=x.SomeID WHERE tr.ID = r.ID AND EXISTS
(SELECT s.* FROM StructuralElements s JOIN Buildings tx ON tx.OtherID = s.OtherID WHERE tx.ID=x.ID AND s.VALUE = #p1 AND s.Category = #p2))
Do you see why this would be bad? For every Result, you're running a subquery (which in itself is running a subquery). This is going to be an exponential increase in time/processing as you start adding things at the root levels (Results and Buildings) because of these nested subqueries. Your best bet is to use joins and get distinct r values after you're done. The SQL would like like this:
SELECT DISTINCT
r.*
FROM
Results r
INNER JOIN Buildings x ON x.SomeID = r.SomeID
INNER JOIN StructuralElements s ON s.OtherID = r.OtherID
WHERE
s.VALUE = #p1 AND s.CATEGORY = #p2
The reason this will work is that when you join, if there are more than one to join back, it will duplicate the original row. The following illustration shows
IDs
R X S
1 - -
Join X
1 1 -
1 2 -
1 3 -
Join S
1 1 1
1 1 2
1 2 5
1 2 6
Assuming S=2 and S=6 meet your criteria, then it will return (in R,X,S form) rows 1,1,2 and 1,2,6. Getting just the distinct r in this case will only return R=1, which is what you're trying to accomplish. Using EF, the relationships already exist, so you don't need to do anything extra, just reference the columns you're trying to filter by:
results = (from r in results
from x in r.Buildings
from s in x.StructuralElements
where s.VALUE == Model.Bedrooms.ToString() && s.CATEGORY=="RM"
select r).Distinct();
This is the SelectMany operator at play (which takes a collection and flattens out subcollections into a single collection).