Consider having these Models:
public class BenchMarks{
public List<Parent> Parents { get; set; }
}
public class Parent {
public List<Child> Children { get; set; }
}
public class Child {
public string Name { get; set; }
public int Value { get; set; }
}
and the class BenchMarks also has these methods:
public void GetWithTwoForeach(){
var dict = new Dictionary<string,int>();
foreach(var parent in Parents){
foreach(var child in parent.Children){
if (dict.ContainsKey(child.Name))
dict[child.Name] += child.Value;
else
dict.Add(child.Name, child.Value);
}
}
}
public void GetWithOneForeach(){
var dict = new Dictionary<string,int>();
foreach(var child in Parents.SelectMany(p=>p.Children)){
if (dict.ContainsKey(child.Name))
dict[child.Name] += child.Value;
else
dict.Add(child.Name, child.Value);
}
}
public void GetWithGroupSum(){
var childGroups = Parents.SelectMany(p=>p.Children).GroupBy(c => c.Name);
var dict = childGroups.ToDictionary(cg => cg.Key, cg => cg.Sum(child => child.Value));
}
Than why is it that the nested foreach loop outperforms the other two methods by more than 100%?
Benchmark fiddle
Edit My benchmark was indeed affected by the ordered nature of my test data.
When I changed it to this (I was not able to get this to work in the .NET fiddle)
public void Setup()
{
Parents = new List<Parent>();
var names = new List<string>{"Tigo","Edo","Kumal","Manu","Guido","Anne","Steff","Mimi","Suzan","Jeff"};
var random = new Random();
for (var i = 0; i < 200; i++)
{
int index = random.Next(names.Count);
Parents.Add(new Parent
{
Children = new List<Child>
{
new Child
{
Name = names[index],
Value = i
}
}
});
}
}
The benchmark came out a bit less dramatic, but still the same conclusion that LINQ adds overhead as mentioned in comments
