Re: [Phys-l] modern programming

[John Denker <jsd@av8n.com>]

On 07/05/2011 03:09 AM, Brian Blais wrote:
> Then I saw what was, in my mind, the simplest example that showed a
> clear advantage.  It is a class, from the Python Cookbook, which does
> temperature conversions
> (http://code.activestate.com/recipes/286226-temperature-class/).

I'm not convinced by that example.  In particular, the idea of 
converting everything to a standard form and converting it back 
again (if/when necessary) is a good idea _whether or not_ you are 
using classes, objects, et cetera. 

This is a long-established widely-practiced strategy: 2N converters,
used in pairs, give you the full N^2 capability.

> The advantage is that, in this implementation, you only have to know
> how to convert to and from Kelvin from whatever units you're using.
> In a procedural program (like C), you'd have to have all of the
> different combinations (c to f, c to r, r to f, etc...) which is
> clearly less efficient.

Au contraire, would be as easy as π to implement the converters
in plain old C ... converters that can be used efficiently and 
conveniently:

      result = c_of_k(k_of_f(-40));        /* works fine */

This uses the aforementioned strategy.

The temperature-class-recipe is OK as a pedagogical illustration of
language features.  Pedagogical examples are often a bit contrived.

=============

I added several more sections (and several more code samples) to:
  http://www.av8n.com/computer/htm/modern-programming.htm
including some not-very-detailed remarks about STL container classes, 
overloading, templates, call by reference, and exceptions.