Count nulls in properties with one LINQ query - linq

Question

Paste into LINQPad:
void Main()
{
List<Data> list = new List<Data>();
list.Add(new Data());
list.Add(new Data{a="a", b="b"});
list.Add(new Data{a=null, b="b"});
var queryA = from data in list where data.a == null select data;
var queryB = from data in list where data.b == null select data;
var countNulls = new {a = queryA.Count(),b = queryB.Count()};
countNulls.Dump();
}
class Data
{
public string a {get;set;}
public string b {get;set;}
}
Instead of using queryA and queryB is it possible to do this in one query?
Answer:
All queries below generate exactly same SQL, so it's just a preference of coder what to choose.
var countNulls = new
{
a = queryA.Count(),
b = queryB.Count()
};
var countNulls2 = new
{
a = list.Count(d => d.a == null),
b = list.Count(d => d.b == null)
};
var countNulls3 = list.Aggregate(
new { a = 0, b = 0 },
(acc, data) => new
{
a = acc.a + (data.a == null ? 1 : 0),
b = acc.b + (data.b == null ? 1 : 0),
});
Update: apparently (thanks to Evan Stoev) this task can be done 20x faster on EF's DbSet and it creates one SQL query.
var countNulls4 =
(from data in db.Data
group data by 1 into g
select new
{
a = g.Sum(data => data.a == null ? 1 : 0),
b = g.Sum(data => data.b == null ? 1 : 0)
}).First();

Answer 1

For completeness, you can use Aggregate to get the result in a single pass over the input sequence:
var countNulls = list.Aggregate(
new { a = 0, b = 0 },
(acc, data) => new
{
a = acc.a + (data.a == null ? 1 : 0),
b = acc.b + (data.b == null ? 1 : 0),
});
But I'm not sure it would be more efficient compared to 2 separate Count calls due to the need of anonymous object allocation on each step.
UPDATE: It turns out that you are asking for a single SQL query (so list is not actually a List<Data> but DbSet<Data> I guess). In LINQ to Entities, you can use group by constant technique, which combined with replacing Count(condition) with Sum(condition ? 1 : 0) will produce a nice single SQL query pretty similar to what you would write manually:
var countNulls =
(from data in db.Data
group data by 1 into g
select new
{
a = g.Sum(data => data.a == null ? 1 : 0),
b = g.Sum(data => data.b == null ? 1 : 0)
}).First();

Answer 2

You can simply use the Count() overload that takes a predicate:
var countNulls = new
{
a = list.Count(d => d.a == null),
b = list.Count(d => d.b == null)
};
Output is:

Answer 3

As far as I understand, you want to count nulls for two different object properties regardless if the other one is null. In other words:
a=null, b="b"
a="a", b=null
a=null, b=null
would return count for 2 and 2.
In that case you can do it this way in one query:
var queryAB = from data in list where data.a == null || data.b == null select data;