C＃Linq GroupBy，获取每个组中的列的列表

Question

I need to write a query to get all the shipping costs of sales, and compare it against an estimated shipping cost. Here is my query:

var sales = (from sale in db.Sales
             where sale.DateOfSale > startDate && sale.DateOfSale < endDate
             group sale by new {sale.ItemID, sale.EstimatedShipping} into g
             select new
             {
                 ItemID = g.Key.ItemID
                 Estimate = g.Key.EstimatedShipping
                 ActualShipCosts = (from gSales in g select gSales.ActualShipping)
             }).ToList();

It seems that doing anything with the group that isn't getting the group by keys or doing g.Count() makes the query run terribly slow, I can't get this query to finish without timing out. Is there anything I can do to help performance here?

Answer 1

Difficult to tell without all the database for test, but this line could be the problem :

(from gSales in g select gSales.ActualShipping)

Try this instead:

var sales = (from sale in db.Sales
             where sale.DateOfSale > startDate && sale.DateOfSale < endDate
             group sale by new {sale.ItemID, sale.EstimatedShipping} into g
             select new
             {
                 ItemID = g.Key.ItemID,
                 Estimate = g.Key.EstimatedShipping,
                 ActualShipCosts = g.Select(gSales => gSales.ActualShipping).ToList()
             }).ToList();

This prevent a second from . The first query might doing some weird joins, whereas with a simple Select it should not.

Answer 2

I think problem can be caused by (from gSales in g select gSales.ActualShipping) . Linq group by is translated into group by in SQL, so you don't have direct access to records which forms group. Each call of (from gSales in g select gSales.ActualShipping) will do separate select.

First possible solution: create index on ItemID, EstimatedShipping pair.

Second possible solution:

var sales = (from sale in db.Sales
         where sale.DateOfSale > startDate && sale.DateOfSale < endDate
         select new
         {
             ItemID = sale.ItemID
             Estimate = sale.EstimatedShipping
             ActualShipCosts = sale.ActualShipping
         }).ToList();

var groupedSales = (from sale in sales
         group sale by new {sale.ItemID, sale.EstimatedShipping} into g
         select new
         {
             ItemID = g.Key.ItemID,
             Estimate = g.Key.EstimatedShipping,
             ActualShipCosts = g.Select(gSales => gSales.ActualShipping).ToList()
         }).ToList();

This will do grouping offline, after records are downloaded. It will for sure reduce number of queries, but you must check on your db to see if it's quicker.

Answer 3

You can try performing the ActualShipping selection while you are building the result set:

var sales = db.Sales
    .Where(sale => sale.DateOfSale > startDate && sale.DateOfSale < endDate)
    .GroupBy(
        sale => new {sale.ItemID, sale.EstimatedShipping},
        sale => sale.ActualShipping)
    .ToList();

Not sure, but it could prevent additional enumeration.