How to convert Sum over partition from T-SQL

Question

I have the following T-SQL Table, data, and query.

create table sampledata 
(
    name nvarchar(50),
    sampletime datetime,
    samplevalue decimal
);

insert into sampledata (name, sampletime, samplevalue) values
('ABC1235', cast('2016/01/01 10:00:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:05:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:10:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:15:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 10:20:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 10:25:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 10:30:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:35:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:40:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:45:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:50:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 10:55:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:00:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:05:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 11:10:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 11:15:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:20:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:25:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:30:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 11:35:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:40:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:45:00 AM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 11:50:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 11:55:00 AM' as datetime), null),
('ABC1235', cast('2016/01/01 12:00:00 PM' as datetime), 50.00),
('ABC1235', cast('2016/01/01 12:05:00 PM' as datetime), null),
('ABC1235', cast('2016/01/01 12:10:00 PM' as datetime), null),
('ABC1235', cast('2016/01/01 12:15:00 PM' as datetime), null),
('ABC1235', cast('2016/01/01 12:20:00 PM' as datetime), null),
('ABC1235', cast('2016/01/01 12:25:00 PM' as datetime), null),
('ABC1235', cast('2016/01/01 12:30:00 PM' as datetime), null)
('ZYA4567', cast('2016/01/01 10:00:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:05:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:10:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:15:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 10:20:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 10:25:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 10:30:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 10:35:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:40:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:45:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:50:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 10:55:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:00:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:05:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 11:10:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 11:15:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:20:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:25:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:30:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 11:35:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:40:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:45:00 AM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 11:50:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 11:55:00 AM' as datetime), null),
('ZYA4567', cast('2016/01/01 12:00:00 PM' as datetime), 50.00),
('ZYA4567', cast('2016/01/01 12:05:00 PM' as datetime), null),
('ZYA4567', cast('2016/01/01 12:10:00 PM' as datetime), null),
('ZYA4567', cast('2016/01/01 12:15:00 PM' as datetime), null),
('ZYA4567', cast('2016/01/01 12:20:00 PM' as datetime), null),
('ZYA4567', cast('2016/01/01 12:25:00 PM' as datetime), 40.00),
('ZYA4567', cast('2016/01/01 12:30:00 PM' as datetime), 50.00)

The query:

select 
    name, 
    sampletime,
    samplevalue,
sum(case when samplevalue is null then 0 else 1 end) 
over(partition by name order by sampletime) 
* case when samplevalue is null then 1 else 0 end as block
from sampledata

I need to convert it into LINQ.

I figured I'd break it into two steps (My actual query is more involved contain joins and where clauses than this example)

var list = (from s in db.sampledata).ToList();

Now I am not sure how to do the

sum(case when samplevalue is null then 0 else 1 end) 
over(partition by name order by sampletime) 
* case when samplevalue is null then 1 else 0 end as block

score 2 · Accepted Answer · answered Aug 25 '16 at 05:54

You can't do it using LinQToSQL, becouse there is no straight convertion to SQL over partition constraction from Linq.

But you can do it on client side i suggest you this way:

int i = 0; //we need this for closure
var list = db.sampledata
    .OrderBy(x => x.name) 
    .ThenBy(x => x.sampletime) //order your data like partition function do
    .ToList() //get all from DB to server memory
    .Select(x =>
    {
        int block = 0; //logic of your block calculation
        if (x.samplevalue.HasValue)
            i++;
        else
            block = i;
        //return your full collection
        return new { 
                      name = x.name, 
                      sampletime = x.sampletime, 
                      samplevalue = x.samplevalue, 
                      block = block 
                   };
    });

As you can see sorting and selectiong logic will be on DB side, but calculated field block will be on LinQ side.

Another way that may intrest you is to wrap your T-SQL query with stored procedure or stored function and map it to your LinQToSQL DataContext so you can call it.

How to convert Sum over partition from T-SQL

1 Answers1