In Lisp (Clojure, Emacs Lisp), what is the difference between list and quote?
From reading introductory m开发者_运维知识库aterial on Lisp, I now consider the following to be identical:
(list 1 2 3)
'(1 2 3)
However, judging from problems I face when using the quoted form in both Clojure and Emacs Lisp, they are not the same. Can you tell me what the difference is?
The primary difference is that quote
prevents evaluation of the elements, whereas list
does not:
user=> '(1 2 (+ 1 2)) (1 2 (+ 1 2)) user=> (list 1 2 (+ 1 2)) (1 2 3)
For this reason (among others), it is idiomatic clojure to use a vector when describing a literal collection:
user=> [1 2 (+ 1 2)] [1 2 3]
In Common Lisp, quoted objects are constant literal data. The data is not evaluated. You should not modify this data, as the consequences are undefined. Possible consequences are: modification of shared data, attempt to modify read-only data, an error might be signalled, it might just work, shared data, ...
Literal lists:
'(1 2 3)
Above is a constant list, which will be constructed by the reader and evaluating to itself, because it is quoted. If it appears in Lisp code, a compiler will embed this data somehow in the FASL code.
(quote (1 2 3))
is another way to write it.
(list 1 2 3)
this is a call of the Common Lisp function LIST
with three arguments 1
, 2
and 3
. Each of the arguments will be evaluated. Since they are numbers, they evaluate to themselves. When evaluated the result is a fresh new list (1 2 3)
.
Data sharing in compiled code
Imagine in a Lisp file the following four definitions:
(defparameter *list1* (list 1 2 3))
(defparameter *list2* (list 1 2 3))
(defparameter *list3* '(1 2 3))
(defparameter *list4* '(1 2 3))
Then we compile and load the file.
! (eq *list3* *list4*)
now may evaluate to either T
or NIL
depending on the implementation and compiler settings !
Reason: in Common Lisp the Lisp compiler may share structure of literal lists (!) if they are similar. The compiler detects that here the lists are similar and will allocate only one list. Both variables *list1*
and *list2*
the will point to this one list.
All other EQ
(object equality) comparisons of two of the above lists will return NIL
.
Notations for some other data structures:
'(1 . 2) and (cons 1 2) ; cons cell
'#(1 2 3) and (vector 1 2 3) ; vector
'#S(FOO :a 1 :b 2) and (make-foo :a 1 :b 2) ; structure
One is the literal data and the other is a function call that constructs such a data structure.
Quoted lists (e.g. '(1 2 3)
) should be treated carefully (generally as read-only). (see SO answers When to use 'quote in Lisp and When to use 'quote in Lisp).
(list 1 2 3)
will "cons" up a fresh list, independent of all others.
You can see an example of a pitfall of using quoted lists in the manual for nconc
.
And, as you probably know, when you call 'list
- the arguments will obviously be evaluated versus the contents of a quoted list. And 'quote
takes a single argument, versus 'list
s variable number of arguments.
(list (+ 1 2) 3) --> (3 3)
(quote ((+ 1 2) 3)) --> ((+ 1 2) 3)
Their relation can be analogous to function invocation with 'function name' and funcall
.
When you have no idea what function you'll get at runtime, you use funcall
.
When you have no idea what element you may get at runtime, your use list
.
For people like me who get confused because of existence of backquote
and count it as quote tacitly.
backquote
is not quote
It's a reader-macro that expands into quote
, list
or others:
(macroexpand ''(1 2));=> '(1 2)
(macroexpand '`(1 2));=> '(1 2)
(macroexpand '`(1 ,2));=> (list 1 2)
(macroexpand '`(1 ,@foo));=> (cons 1 foo)
(macroexpand '`(1 ,@foo 2));=> (cons 1 (append foo '(2)))
精彩评论