[O] org-map-entries and org-map-continue-from

[Richard Lawrence]

Hi all,

Bastien had advised me [1] to use

(setq org-map-continue-from (outline-next-heading))

in a function called by org-map-entries in order to map that function
across just the /children/ of the current entry (i.e., to exclude the
current `parent' entry itself).  This works great, but I have now found
that it has a weird side-effect: it calls the function twice on the last
child.

For a simple example, suppose I write:
#+BEGIN_SRC emacs-lisp
(defun get-export-filenames ()
  (interactive)
  (setq export-files '())
  (progn
    (org-map-entries
     (lambda ()
       (setq org-map-continue-from (outline-next-heading))
       (let ((org-trust-scanner-tags t))
         (push (org-entry-get (point) "EXPORT_FILE_NAME") export-files)))
     nil 'tree)
    (message export-files))) ; errors, but lets me see the list of collected 
values
#+END_SRC

And I call this function from a buffer that looks like:

* Top
<point is here when I call get-export-filenames>
** One
   :PROPERTIES:
   :EXPORT_FILE_NAME: one
   :END:
** Two
   :PROPERTIES:
   :EXPORT_FILE_NAME: two
   :END:
** Three
   :PROPERTIES:
   :EXPORT_FILE_NAME: three
   :END:
** Four
   :PROPERTIES:
   :EXPORT_FILE_NAME: four
   :END:

Then the list that I get back (the value of export-files) looks like:

("four" "four" "three" "two" "one")

Whereas I would like it to be just:

("four" "three" "two" "one")

Can anyone see what I need to do to achieve that? [Apart from just using
(cdr export-files), I mean -- I'd like to know the /right/ way.]  I'm
puzzled because outline-next-heading, if called interactively from the
last child, does indeed put point at the end of that child or at the
next (parent-level) heading, so it doesn't seem that the problem is that
it somehow loops back when there is no next child-level entry.

Many thanks if you catch something I've missed!

Best,
Richard

[1] http://article.gmane.org/gmane.emacs.orgmode/37244/