Linq projection of flattened table into parent and child object graph

Question

I have a enumerable list that contains a flattened parent-child relationship:

ParentGuid1, ParentName1, ChildGuid1, ChildName1
ParentGuid1, ParentName1, ChildGuid2, ChildName2
ParentGuid2, ParentName2, ChildGuid3, ChildName3
ParentGuid2, ParentName2, ChildGuid4, ChildName4

I have defined a Child class and a Parent class that includes a List<Child> property called Children.

Can I use linq to create on object graph with one instance of the Parent class per unique ParentGuid, referencing a List populated by the children associated with that parent.

Something along the lines of this (note, this code doesn't compile):

myFlattenedHierarchy.Select(p => new Parent
   {Guid = p.ParentGuid, 
    Name = p.ParentName, 
    Children = myFlattenedHierarchy.Where(c => c.ParentGuid == p.ParentGuid).Select(c => new Child{Guid = c.ChildGuid, Name = c.ChildName})
   });

Answer 1

myFlattenedHierarchy.Select(p => new Parent
   {Guid = p.ParentGuid, 
    Name = p.ParentName, 
    Children = myFlattenedHierarchy.Where(c => c.ParentGuid == p.ParentGuid).Select(c => new Child{Guid = c.ChildGuid, Name = c.ChildName})
   });

You should be able to do that, but the Children can not be a List, it has to be IEnumerable .

Answer 2

Here's the pre-Linq way to do it with a simple loop.

Dictionary<Guid, Parent> parents = new Dictionary<Guid, Parent>();
foreach(RowType row in myFlattenedHierarchy) //just enumerate once
{
  if (!parents.ContainsKey(row.ParentGuid)
  {
    Parent newParent = new Parent(row);
    parents[row.ParentGuid] = newParent;
  }

  Child newChild = new Child(row);

  Parent theParent = parents[row.ParentGuid];
  theParent.Children.Add(newChild);  
}

List<Parent> result = parents.Values.ToList();

---

Or you could use GroupBy to get a similar result.

from row in myFlattenedHierarchy
group row by row.ParentGuid into g
select new Parent()
{
  Guid = g.Key,
  Name = g.First().ParentName,
  Children =
  (
    from childRow in g
    select new Child()
    {
      Guid = childrow.ChildGuid,
      Name = childrow.ChildName
    }
  ).ToList()
}

It's a toss-up which is more maintainable. Either way, don't re-enumerate myFlattenedHierarchy inside the loop/query.

Answer 3

I believe you can use GroupBy() (full disclosure: not compiled):

myFlattenedHierarchy.GroupBy(row => row.ParentGuid)
    .Select(group => new Parent
        {
            Guid = group.Key.ParentGuid,
            Name = group.Key.ParentName,
            Children = myFlattenedHierarchy.Where(c => c.ParentGuid == group.Key.ParentGuid)
                .Select(c => new Child{ Guid = c.ChildGuid, Name = c.ChildName })
                .ToList()
        });

Answer 4

This should work, very similar to David B's second example but I couldn't get his to work without a bit of fixing (grouping by multiple column) so I've added it here for the record.

from row in myFlattenedHierarchy
group row by new { row.ParentGuid, row.ParentName } into g
select new Parent()
{
  Guid = g.Key.ParentGuid,
  Name = g.Key.ParentName,
  Children =
  (
    from childRow in g
    select new Child()
    {
      Guid = childRow.ChildGuid,
      Name = childRow.ChildName
    }
  ).ToList()
};

Answer 5

You need to tackle with recursivity and infinite recursivity if you have a loop from within your flat collection. Linq can't be used for the complete problem but can help returning child of a specific node.