Re: elisp help: Getting the current word position. (closure?)

[Tassilo Horn]

cinsk <cinsky@gmail.com> writes:

Hi!

> (defun current-word-markers (&optional STRICT REALLY-WORD)
>   "Return the position of the current word"
>   (let ((old (symbol-function
>                 'buffer-substring-no-properties))
>         (saved-start (make-marker))
>         (saved-end (make-marker)) ret)
>
>     (fset 'buffer-substring-no-properties
>           (lambda (start end)
>             (let (ret)
>               (set-marker saved-start start)
>               (set-marker saved-end end)
>               (setq ret (funcall old start end))
>               ret)))
>     (setq ret (current-word STRICT REALLY-WORD))
>     (fset 'buffer-substring-no-properties old)
>     (if ret
>         (cons saved-start saved-end)
>       nil)))
>
>
> This function works.  With my short knowledge of LISP, I think the
> (lambda ...) form in `current-word-markers' is a closure, since it
> refers to the unbound variables (saved-start and saved-end).

Due to emacs' dynamic scoping they're bound when this function is
called.  Dynamic scoping basically means that a function can see all
variables its caller defined (recursively).  Here's an example:

--8<---------------cut here---------------start------------->8---
(defun foo ()
  (let ((foo-val "I'm a local variable of foo!"))
    (bar)))

(defun bar ()
  foo-val)

(foo) ;; C-x C-e ==> "I'm a local variable of foo!"
--8<---------------cut here---------------end--------------->8---

> Recently, I found that Emacs Lisp does not have "closures" according
> to "GNU Emacs Lisp Reference Manual 2.8".

Yes, that's true.

> Here are a few questions:
>
> 1. Does the (lambda ...) in `current-word-markers' is a closure???

No.

> 3. If not, enlighten me what is a closure.

This real closure won't work in elisp:

--8<---------------cut here---------------start------------->8---
(defun make-adder (n)
  (lambda (m)
    (+ n m)))

(fset 'adder (make-adder 3))
(adder 3) ;; C-x C-e
==>
Debugger entered--Lisp error: (void-variable n)
  (+ n m)
  adder(3)
  eval((adder 3))
  eval-last-sexp-1(nil)
  eval-last-sexp(nil)
  call-interactively(eval-last-sexp nil nil)
--8<---------------cut here---------------end--------------->8---

Bye,
Tassilo
-- 
Richard Stallman can coerce meaningful data from /dev/null.