Multiple implementations of low-level functions

Multiple implementations of low-level functions

[Michael J. Barillier]

I've been hacking on a system for which I would like to have multiple
implementations of low-level functions, which would be called by
higher-level code.  (e.g. A test version with static data, v. a test
version with data stored in a database, v. a production version
running against an Oracle database and communicating with other
processes.)

As I see it, I've got two options, but I'm trolling for others:

  (1) Something SICP-like, with a closure that's passed around,
      returning a function based on a selector, as in:

        (use-module (test-implementation))
        (define impl (make-test-implementation args...))
        ((impl 'get-records) "first key" "second key")

  (2) Load functions at the start of program execution, before the
      high-level stuff is loaded/called:

        (load "test-implementation")
        (get-records "first key" "second key")

There also might be a module-based solution, but I haven't ventured
down that route.

I've written apps using the first solution, and it worked fine, other
than some nasty nesting of functions and having to pass a parameter
around.  The second option is nice for its simplicity, and from the
viewpoint of building up the Scheme (or Lisp) environment/library of
available functions, loading higher-level stuff onto lower-level code.

Anyone care to weigh in on either method, or propose a sexier
solution?

-- 
Michael J. Barillier mbarilli(at)midsouth.rr.com  /"\ ASCII Ribbon
Reg Linux user #125310 <http://counter.li.org/>   \ / Campaign against
  A. Because that's not how people read.           X  HTML and RTF
  Q. Why is top-posting email replies bad?        / \ email


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-user

Re: Multiple implementations of low-level functions

[Keith Wright]

> From: "Michael J. Barillier" <mbarilli@midsouth.rr.com>
> 
> I've been hacking on a system for which I would like to have multiple
> implementations of low-level functions, which would be called by
> higher-level code.  (e.g. A test version with static data, v. a test
> version with data stored in a database, v. a production version
> running against an Oracle database and communicating with other
> processes.)
>       ...
> Anyone care to weigh in on either method, or propose a sexier
> solution?

Why not just have a variable and assign it the value you want?

(set! compute
      (case version
            ((test1) (lambda(n) 42))
            ((test2) (lambda(n) (+ 1 n)))
            (else (lambda(n) (call-oracle "select answer from air"))) ))

-- 
     -- Keith Wright  <kwright@free-comp-shop.com>

Programmer in Chief, Free Computer Shop <http://www.free-comp-shop.com>
         ---  Food, Shelter, Source code.  ---


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-user

Re: Multiple implementations of low-level functions

[Thien-Thi Nguyen]

   From: "Michael J. Barillier" <mbarilli@midsouth.rr.com>
   Date: Fri, 31 Oct 2003 19:33:23 -0600

   Anyone care to weigh in on either method, or propose a sexier
   solution?

closure dispatch is fine.  personally, i like to use keywords instead of
symbols since emacs displays them in a different face, but ymmv.

thi


_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://mail.gnu.org/mailman/listinfo/guile-user