Question

A while ago I found a handy query for mysql to get the top X per group by.
This is what I mean:

if this is the table:

rid id1 id2 id3 value
1   1   1   1   10
2   1   1   2   11
3   1   1   3   9
4   1   2   1   20
5   1   2   2   18
6   1   2   3   23
7   1   3   1   30
8   1   3   2   34
9   1   3   3   31
10  1   3   4   27
11  1   3   5   32
12  1   4   1   41
13  1   4   2   40
14  1   4   3   43
15  1   5   1   53
16  1   5   2   51
17  1   5   3   50
18  2   1   1   11
19  2   1   2   9
20  2   1   3   12

I want to get this result:

rid id1 id2 id3 value
2   1   1   2   11
6   1   2   3   23
8   1   3   2   34
14  1   4   3   43
15  1   5   1   53

I can get this by running the following mysql query:

SELECT * FROM
  (SELECT * FROM idsandvalue 
   WHERE id1=1 AND 
     (SELECT COUNT(*) FROM idsandvalue AS helper 
      WHERE helper.id1 = idsandvalue.id1 
      AND helper.id2= idsandvalue.id2 
      AND helper.value > idsandvalue.value
     ) < 1
  )a;

if I change < 1 to lets say 2, 3 or x I can get the top x per id2 where id1=1 (so, two of the same id2’s with different id3’s) like this:

rid id1 id2 id3 value
1   1   1   1   10
2   1   1   2   11
4   1   2   1   20
6   1   2   3   23
8   1   3   2   34
11  1   3   5   32
12  1   4   1   41
14  1   4   3   43
15  1   5   1   53
16  1   5   2   51

two questions.
A) the query is not really fast in MySQL. Takes a while (runs a table with 3207394 rows). Can I get the same result with the use of a different query (I was not able to get it).
B) How can I translate this to linq? Due to the strange where statement, I have no clue how to translate this into linq.

(later I added this extra question as well)
in MySQL I use this query:

SELECT *,COUNT(*) AS Counter FROM idsandvalue GROUP BY id1,id2;

to get this result:

rid id1 id2 id3 value   Counter
1   1   1   1   10       3
4   1   2   1   20       3
7   1   3   1   30       5
12  1   4   1   41       3
15  1   5   1   53       3
18  2   1   1   11       3

I’m also having difficulties translating this to Linq.

(extra info was too big for comment)

Hi John (thanks for the quick respond).
with this mysql query

SELECT * FROM 
  (SELECT * FROM idsandvalue 
   WHERE id1=1 AND 
     (SELECT COUNT(*) FROM idsandvalue AS helper 
      WHERE helper.id1 = idsandvalue.id1 
      AND helper.id2= idsandvalue.id2 
      AND helper.value > idsandvalue.value
     ) < 1
  )a 

I try to get the rows for each grouped id1 and id2 with it’s biggest value. That’s why in this case I get for instance row with id 2. 11 is the biggest of 10,11 and 9 where id1=1 and id2=1. and that’s why I get the row with id 8, because where id1=1 and id2=3 the biggest value for column value is 34. If I change the query to < 2, I get the top two. for id2=1 and id2=3 this would give the rows with id 8 and 11. Is this better explained?

Answer 1

Recreated your table in SQL Server and ran your query against it, than converted the query via linqer:

from a in ((from idsandvalue in db.idsandvalue whereidsandvalue.id1 == 1 &&

    (from helper in db.idsandvalue
    where
      helper.id1 == idsandvalue.id1 &&
      helper.id2 == idsandvalue.id2 &&
      helper.value > idsandvalue.value
    select new {
      helper
    }).Count() < 1
select new {
  idsandvalue
}))

select new {
a.idsandvalue.rid,
a.idsandvalue.id1,
a.idsandvalue.id2,
a.idsandvalue.id3,
a.idsandvalue.value
}