efficient way to implement paging

Should I use LINQ’s Skip() and Take() method for paging, or implement my own paging with a SQL query?

Which is most efficient? Why would I choose one over the other?

I’m using SQL Server 2008, ASP.NET MVC and LINQ.

Answer

Trying to give you a brief answer to your doubt, if you execute the skip(n).take(m) methods on linq (with SQL 2005 / 2008 as database server) your query will be using the Select ROW_NUMBER() Over ... statement, with is somehow direct paging in the SQL engine.

Giving you an example, I have a db table called mtcity and I wrote the following query (work as well with linq to entities):

using (DataClasses1DataContext c = new DataClasses1DataContext())
{
    var query = (from MtCity2 c1 in c.MtCity2s
                select c1).Skip(3).Take(3);
    //Doing something with the query.
}

The resulting query will be:

SELECT [t1].[CodCity], 
    [t1].[CodCountry], 
    [t1].[CodRegion], 
    [t1].[Name],  
    [t1].[Code]
FROM (
    SELECT ROW_NUMBER() OVER (
        ORDER BY [t0].[CodCity], 
        [t0].[CodCountry], 
        [t0].[CodRegion], 
        [t0].[Name],
        [t0].[Code]) AS [ROW_NUMBER], 
        [t0].[CodCity], 
        [t0].[CodCountry], 
        [t0].[CodRegion], 
        [t0].[Name],
        [t0].[Code]
    FROM [dbo].[MtCity] AS [t0]
    ) AS [t1]
WHERE [t1].[ROW_NUMBER] BETWEEN @p0 + 1 AND @p0 + @p1
ORDER BY [t1].[ROW_NUMBER]

Which is a windowed data access (pretty cool, btw cuz will be returning data since the very begining and will access the table as long as the conditions are met). This will be very similar to:

With CityEntities As 
(
    Select ROW_NUMBER() Over (Order By CodCity) As Row,
        CodCity //here is only accessed by the Index as CodCity is the primary
    From dbo.mtcity
)
Select [t0].[CodCity], 
        [t0].[CodCountry], 
        [t0].[CodRegion], 
        [t0].[Name],
        [t0].[Code]
From CityEntities c
Inner Join dbo.MtCity t0 on c.CodCity = t0.CodCity
Where c.Row Between @p0 + 1 AND @p0 + @p1
Order By c.Row Asc

With the exception that, this second query will be executed faster than the linq result because it will be using exclusively the index to create the data access window; this means, if you need some filtering, the filtering should be (or must be) in the Entity listing (where the row is created) and some indexes should be created as well to keep up the good performance.

Now, whats better?

If you have pretty much solid workflow in your logic, implementing the proper SQL way will be complicated. In that case LINQ will be the solution.

If you can lower that part of the logic directly to SQL (in a stored procedure), it will be even better because you can implement the second query I showed you (using indexes) and allow SQL to generate and store the Execution Plan of the query (improving performance).