Trying to LINQ it all together (Pt. I)

Yeah, yeah, we’ve all done our from thingy in stuff where thingy.Age > 26, etc., etc.

I wanted to try something else for starters. I fired up my March CTP Orcas and got myself a Person (as usual, I can relate to them). The list I create for testing looks as follows:

public static Person[] CreatePeopleList() {
return new Person[] {
new Person { Name = "Arthur", Posn = Position.Boss, Salary=10000},
new Person { Name = "Selma", Posn = Position.Middleman, Salary=6000},
new Person { Name = "Buttocks", Posn = Position.Middleman, Salary=5500},
new Person { Name = "Shawn", Posn = Position.Peasant, Salary=4000},
new Person { Name = "Burgess", Posn = Position.Peasant, Salary=3800},
};
}

Bog standard hierarchy in any company. My aim now was to get a total of the salaries grouped by the available positions. The new capabilities indeed allow to quickly express such a thing, albeit I suppose I am becoming a lambda fan, since that was my first result:

Person[] pees = Person.CreatePeopleList();
var salaryDistribution =
pees.GroupBy(p => p.Posn)
.Select(group =>
new {
Position = group.Key,
TotalSalary = group.Sum(p => p.Salary)
});
salaryDistribution.ToList()
.ForEach(entry =>
Console.WriteLine("Position {0} earns {1} bucks",entry.Position,entry.TotalSalary));

Those funny Select() and GroupBy() all come from the System.Linq namespace and are extension methods to your well-known IEnumerable<> interface. In other words, all arrays, collections, etc. can benefit from LINQ functionality.

First question: Is the code understandable? I hope so. In this specific case

So far, so nice. I spared myself of writing the same code with C#2.0 capabilities, but I would expect it to be roughly 5-8 lines more. We don’t really need funny keywords to get the power of LINQ, …[snip]

Update from 29.04.07 - The following lines lead to wrong conclusions. For a better picture read the next post

In fact in this case a bug stopped me in my attempt to express this query with the available keywords select, group, etc.

That is as far as I got:

var salarySums = from p in pees group p by p.Posn into groups from grp in groups select grp.Key;

The error you’ll get:

‘Popular.Person’ does not contain a definition for ‘Key’ and no extension method ‘Key’ accepting a first argument of type ‘Popular.Person’ could be found (are you missing a using directive or an assembly reference?)

Oops, did I miss something? Heck, no, you would expect groups to be an IEnumerable containing IGrouping<Position,Person>, just like what I did with the lambdas further up. That is in fact what the current documentation also expects and when I copy the doc’s group by example into my source code it fails by the same reason: The compiler thinks that groups is an IEnumerable containing…Persons. This is plain wrong and I will see if I get it to Microsoft somehow, if it hasn’t already happened.

To sum it up for today: I think the LINQ methods are very practical but I have the feeling that I will favour the lambda-based usage over the keyword-based one. In this case the keyword-based query would have gotten pretty convoluted. Time will tell what is more readable / maintainable/ powerful!