Re: Basic Emacs Lisp question

[Giorgos Keramidas <keramida@ceid.upatras.gr>]

On Tue, 29 Apr 2008 20:56:17 +0300, Giorgos Keramidas <keramida@ceid.upatras.gr> wrote:
> On Tue, 29 Apr 2008 17:49:19 +0200, Matthias Pfeifer <pfemat@web.de> wrote:
>> Hello,
>>
>> What is the difference between
>>
>> (list 0 nil -1)
>>
>> and
>>
>> '(0 nil -1)
>
> In Common Lisp (list 0 nil -1) is required to 'cons' a new list every
> time it is called.  Quoting the list as in '(0 nil -1) is not required
> to build a new list.  In fact, in compiled code it may reuse the same
> static object over and over again.

Reading my own post reveals that I may have been too terse.  To clarify
the point I was trying to make, here's a small test in Common Lisp, and
the equivalent test in Emacs Lisp.

1. Common Lisp test
-------------------

* Save the following Lisp code to a file called "foo.lisp":

  (defun foo-quoted ()
    '(0 nil -1))

  (defun foo-list ()
    (list 0 nil -1))

* Then compile the file, and load it.  Here's the output from loading
  the compiled file in SBCL:

  CL-USER> (compile-file "foo")

  ; compiling file "/home/keramida/foo.lisp" (written 30 APR 2008 01:48:02 AM):
  ; compiling (DEFUN FOO-QUOTED ...)
  ; compiling (DEFUN FOO-LIST ...)

  ; /home/keramida/foo.fasl written
  ; compilation finished in 0:00:00
  #P"/home/keramida/foo.fasl"
  NIL
  NIL
  CL-USER> (load "foo")          ;; This actually loads "foo.fasl" in SBCL.
  T
  CL-USER>

* Every time the `foo-quoted' function runs it returns exactly the same
  compiled object.  The object returned by separate calls to
  `foo-quoted' is all of EQ, EQL and EQUAL to any previous call, as you
  can see in:

  CL-USER> (let ((one-list (foo-quoted))
                 (another-list (foo-quoted)))
             (mapcar (lambda (test)
                       (funcall test one-list another-list))
                     (list #'eq #'eql #'equal)))
  (T T T)
  CL-USER>

* In contrast, the object returned by the `foo-list' function is a newly
  CONS-ed list every time the function runs:

  CL-USER> (let ((one-list (foo-list))
                 (another-list (foo-list)))
             (mapcar (lambda (test)
                       (funcall test one-list another-list))
                     (list #'eq #'eql #'equal)))
  (NIL NIL T)
  CL-USER>

The lists returned by `foo-list' are EQUAL, but they are neither EQ nor
EQL to each other.  They are created from scratch by allocating new
storage for the value of the expression every time the `foo-list'
function is called.

2. Emacs Lisp test
------------------

* Save the same two functions in a file called "foo.el".

* Fire up Emacs, and byte-compile the file by typing

  M-x byte-compile-file RET foo.el RET

* Load the byte-compiled file by typing

  M-x load-file RET foo.elc RET

* Now evaluate the same two LET forms in your scratch buffer, by pasting
  them in the buffer and typing `C-x C-e' after each expression.

  Emacs Lisp should also evaluate them as:

  (let ((one-list (foo-quoted))
        (another-list (foo-quoted)))
    (mapcar (lambda (test)
              (funcall test one-list another-list))
            (list #'eq #'eql #'equal)))
  => (t t t)

  (let ((one-list (foo-list))
        (another-list (foo-list)))
    (mapcar (lambda (test)
              (funcall test one-list another-list))
            (list #'eq #'eql #'equal)))
  => (nil nil t)

I hope this makes what I initially wrote a bit easier to grasp :-)

Giorgos