Question

Suppose I have a “database” defined as:

// Baked goods vendors
   var vendor = new[] {
    new { ID = 1, Baker = "Pies R Us", StMnemon = "NY", Items = 8, Rating = 9 },
    new { ID = 2, Baker = "Mikes Muffins", StMnemon = "CA", Items = 5, Rating = 9 },
    new { ID = 3, Baker = "Best Bakers", StMnemon = "FL", Items = 2, Rating = 5 },
    new { ID = 4, Baker = "Marys Baked Treats", StMnemon = "NY", Items = 8, Rating = 7 },
    new { ID = 5, Baker = "Cool Cakes", StMnemon = "NY", Items = 4, Rating = 9 },
    new { ID = 6, Baker = "Pie Heaven", StMnemon = "CA", Items = 12, Rating = 9 },
    new { ID = 7, Baker = "Cakes N More", StMnemon = "GA", Items = 6, Rating = 8 },
    new { ID = 8, Baker = "Dream Desserts", StMnemon = "FL", Items = 2, Rating = 7 }
};

// Locations
var location = new[] {
    new {ID= 1, State = "New York", Mnemonic = "NY"},
    new {ID= 2, State = "Massachusetts", Mnemonic = "MA"},
    new {ID= 3, State = "Ohio", Mnemonic = "OH"},
    new {ID= 4, State = "California", Mnemonic = "CA"},
    new {ID= 5, State = "Florida", Mnemonic = "FL"},
    new {ID= 6, State = "Texas", Mnemonic = "TX"},
    new {ID= 7, State = "Georgia", Mnemonic = "GA" }
};

I want to build a query that would be the equivalent of the SQL query:

SELECT   State, Rating, SUM(Items) AS 'Kinds'
FROM     vendor, location
WHERE    vendor.StMnemon = location.Mnemonic
GROUP BY State, Rating

Two things of interest in this query are:

1. The GROUP BY involves multiple tables, and
2. The result contains a summation of a column not appearing in the grouping criteria.

I’ve seen the solutions in the posts on grouping by multiple tables and summing columns not in the group-by. The problem is that combining both doesn’t really duplicate the relational query.

I try to duplicate it in LINQ with the following code:

var query = from v in vendor
           join l in location
           on v.StMnemon equals l.Mnemonic
           orderby v.Rating ascending, l.State
           select new { v, l };

var result = from q in query
           group q by new { 
               s = q.l.State, 
               r = q.v.Rating
/* ==> */    , i = q.v.Items
           } into grp
        select new 
        {
            State = grp.Key.s,
            Rating = grp.Key.r
/* ==> */ , Kinds = grp.Sum(k => grp.Key.i)
        };

This results in:

=================================
State           Rating  Kinds
Florida         5       2
Florida         7       2
New York        7       8
Georgia         8       6
California      9       5
California      9       12
New York        9       8
New York        9       4
=================================

Whereas, the SQL query given above gives this result:

=========================
State       Rating  Kinds
Florida     5       2
Florida     7       2
New York    7       8
Georgia     8       6
California  9       17
New York    9       12
=========================

The discrepancy is because there seems to be no place to put additional columns, other than in the grouping criteria, which of course changes the grouped result. Commenting out the two lines indicated by the /* ==> */ comment in the code above will give the same grouping as the SQL result, but of course that removes the summation field that I want to include.

How do we group multiple tables in LINQ and include additional criteria without changing the grouped result?

Answer 1

something like this seems to return the same as the SQL query:

var result = from v in vendor
             from l in location
             where l.Mnemonic == v.StMnemon
             group v by new { l.State, v.Rating } into grp
             orderby grp.Key.Rating ascending, grp.Key.State
             select new {State = grp.Key.State, Rating = grp.Key.Rating, Kinds = grp.Sum(p=>p.Items)};

foreach (var item in result)
        Console.WriteLine("{0}\t{1}\t{2}", item.State, item.Rating, item.Kinds);