Re: [ggt-devel] another potential usability issue from naming

[Ben S. <bs...@vr...>]

On Thu, 30 May 2002, Kevin Meinert wrote:

> > I think that the translational aspect of a matrix is A PROPERTY OF THAT 
> > MATRIX. 
> 
> maybe this is your problem?  :)  actually it is a property, but how does 
> this matter?  You're using that property of the matrix to initialize a 
> newly created item returned by the "make" function.  right?

It matters because I don't create properties of other objects, I ask 
for/retrieve them. Ignore what a "make" function is in GMTL and think 
about it. Do you really want to create a new translation vector or 
retrieve information about the matrix you have?

> 
> > That property is generally considered as being the fourth column 
> > vector that makes up that matrix. Thus when I wish to retrieve the 
> > translation of a matrix I don't consider it creational, because IT ALREADY
> > EXISTS as a part of the matrix.
> 
> it is creational.  2 tip offs to the fact:
> - a clone operation is creation.  

Agreed.

> - the other tip off is that we're returning a temporary.  
>   by definition, that is creational.

But only as a consequence that the data structure required to pass on that 
information requires a Vec as a vessel for the data. I would say that 
creating a temporary object does make make a function creational since the 
intent is not to create the object and keep it around but to retrieve some 
data from an object which requires a temporary to transfer the data 
required by the language's syntax.

> 
> > Now, consider this; we happened to implement our matrix as 16 floating 
> > point values. Thus to get the translation of a matrix, we must create a 
> > temporary vector to store those values in. What if instead we implemented 
> > the matrix as 4 column vectors? Then retrieving the translation from the 
> > matrix is no longer causes the construction of a new vector since just a 
> > reference to the fourth column vector would be returned. In that case, the 
> > function IS NOT CREATIONAL.
> 
> but we dont implement it that way... and yes it would still be creational 
> (it is a clone, copy constructor, etc...)...

EXACTLY! We call it makeXX because the syntax shows that we are returning 
a temporary rather than noticing that the semantics of the function 
dictate that it is really returning a distinct property of the object.

> 
> > See, I think that the naming conventions used in GMTL are based on 
> > syntactic reasons rather than semantic reasons that may cause some 
> > interesting learning and usability problems since it goes against common 
> > Object Oriented design idioms.
> 
> I don't agree...  what book have you been reading?

Not a book, but experience. When an object has a property, do you write a 
set/make method pair to manipulate it or a set/get pair? What does the 
industry standard tell you? What does any UML tool use while reverse 
engineering a diagram from the code to determine an object's properties?

[snip]

cheers,
-----
Ben Scott
Research Assistant VRAC
bs...@ia...