Re: Using eval with environments and R7RS modules

[Jamison Hope <jhope@alum.mit.edu>]

On Mon, Nov 30, 2020 at 1:09 PM Damien MATTEI <damien.mattei@oca.eu> wrote:

>
> strange that the SICP cover book put the couple EVAL / APPLY as
> important notion, APPLY seems more usefull,


FWIW, I would point out that SICP is *not* a book on how to program in
Scheme. It uses Scheme, yes, but much like the pseudo code in CLRS, it’s
just as an easy-to-understand syntax for presenting concepts about
programming. We did have to write and run some programs in 6.001 on actual
computers for homework, but most of the work was done on paper or a
chalkboard. (There were no computers in the room for the final exam.)

The pair of EVAL/APPLY presented the two stages of *Interpreting* a
Computer Program using the environment model: after parsing the Structure
(which for Lisp just means turning text with parentheses into a list of
atoms and lists — I don’t recall SICP mentioning vector literals, but it’s
been a while), the two steps to interpret the program — to decide what the
program’s result is — are to evaluate the car of the list to see what
function it represents, then evaluate each function argument, and then to
apply the function to its arguments (each of those EVAL steps might entail
recursively EVAL/APPLYing an inner function application). There’s a little
more to it to handle special forms like IF which don’t evaluate every
argument first, but that’s the essence of the model.

This was really a mental exercise, not a description of how a real physical
computer does or should work. The chapter on register machines and
compilation addresses the fact that in reality we don’t need to use EVAL to
look up each variable at runtime, because we can precompute where that
variable resides in the stack based on where it was defined (see the
section called “Lexical Addressing”).


>
>
> Damien
>
> -JH

-- 
Jamison Hope