What is the difference between the shape and structure of an object?

Question

Title says it all.

I see it used a lot in context of data. From ScottGu's post:

One of the really powerful capabilities provided by LINQ and query syntax is the ability for you to define new classes that are separate from the data being queried, and to then use them to control the shape and structure of the data being returned by the query.

What does he mean when he refers to shape of the data?

Answer 1

The shape is any spatial attributes (especially as defined by outline) of the object, whereas the structure is the manner of construction of the object and the arrangement of its parts. Of course, that can apply to any type of object. :)

Answer 2

I think these are informal terms, and the definitions are subjective. I would use "shape" to refer to how the object fits in with other objects in the system. (Compare "surface area", which is a rough (no pun intended :-) measure of the complexity of the object's interface.) I'd use "structure" to refer to how the object is designed and implemented internally.

Hence you can have classes that have a good "shape", but a structure like crepe paper. That's probably easier to refactor than the other way around: a poor shape, but good implementation. (I'm sure some folks would question whether the latter is even possible.)

Answer 3

consider shape to be the objects "api" while the structure is it's internal implementation. In a well designed system the shape will remain static while the structure may change significantly.

Answer 4

Generally, I would consider the shape of an a class to be the public methods and properties that the class offers. The structure would be the internal constructs and representation used. In the context of the quoted material, I would take it to mean that by allowing one to define the return type of a query using anonymous or alternate named classes, you can redefine the data returned by query, constraining and transforming its shape from the original data source.

For example, say you have a user table that is related to a contacts table. Using LINQ and anonymous class as the selection, you can return a user with primary contact object without having to define a particular view; using only LINQ.

var userWithContact = from u in db.Users
                      select new
                          { 
                            Name = u.Name,
                            Address = u.Contacts
                                       .Where( c => c.Type = "self" ).First().Address
                          };