You can do something like this to get just the TODO headlines in the current buffer. If you make the helm-todo-candidates map over all the files in (org-agenda-files) you can make it give all the TODO headings. You can change the match criteria in org-map-entries to be more selective.
#+BEGIN_SRC emacs-lisp :results raw (defun helm-todo-candidates () (let ((results '())) (org-map-entries (lambda () (add-to-list 'results (cons (concat (make-string (nth 1 (org-heading-components)) ?*) " TODO " (nth 4 (org-heading-components))) (point-marker)))) "TODO=\"TODO\"") results)) #+END_SRC #+RESULTS: ((** post it . #<marker at 977 in blog.org>) (** work it out . #<marker at 941 in blog.org>)) Now to run helm, there is a subtle point. We need to map the current buffer /before/ running helm, otherwise we will map an empty helm buffer. #+BEGIN_SRC emacs-lisp (defun helm-todo () "Helm interface to headlines with TODO status in current buffer." (interactive) (let ((candidates (helm-todo-candidates))) (setq helm-todo-source '((name . "TODO headlines") (candidates . candidates) (action . (("open" . goto-char))))) (helm :sources '(helm-todo-source)))) (helm-todo) Simon Thum writes: > Hi all, > > I recently updated my helm install so it includes > helm-org-agenda-headings which is just AWESOME (to me at least). A bit > like org-goto but across all agenda files at once, with goto, refile, > linking built in. If you haven't tried it, I definitely recommend to do so. > > > Yet I'm missing a few things so far, I would like to have different > datasources differentiated by tags, in particular the ARCHIVE tag, and > the infamous FILETAGS so I cannot just regex my way through as the > current approach does. > > This requires making more use of org-ode when filling helm's buffers. My > elisp isn't great but I might be able to get there if the approach is sane. > > Any pointers are welcome! If you might help me please read on. > > I would like to ask what would be the best approach for better utilising > org infrastructure so I may have separate helm sources for > live/archived, private/work, the clocking history, stuff like that. > > The helm-org definition looks deceptively simple: > > https://github.com/emacs-helm/helm/blob/master/helm-org.el > > (defun helm-org-agenda-files-headings () > (interactive) > (helm :sources (helm-source-org-headings-for-files (org-agenda-files)) > :candidate-number-limit 99999 > :buffer "*helm org headings*")) > > > FWICT, in effect helm-org is chewing itself through the buffers: > > (defun helm-get-org-candidates-in-file (filename min-depth max-depth > &optional fontify) > (with-current-buffer (find-file-noselect filename) > (and fontify (jit-lock-fontify-now)) > (let ((match-fn (if fontify 'match-string 'match-string-no-properties))) > (save-excursion > (goto-char (point-min)) > (cl-loop while (re-search-forward org-complex-heading-regexp nil t) > if (let ((num-stars (length (match-string-no-properties 1)))) > (and (>= num-stars min-depth) (<= num-stars max-depth))) > collect `(,(funcall match-fn 0) . ,(point-marker))))))) > > I don't really get what it does but I have a hunch that org-element or > other org-mode functions could be used to achieve the same with more > precision. That's what I would need to do. FWIW I'd be happy to take a > performance hit. > > Thanks in advance, > > Simon -- Professor John Kitchin Doherty Hall A207F Department of Chemical Engineering Carnegie Mellon University Pittsburgh, PA 15213 412-268-7803 @johnkitchin http://kitchingroup.cheme.cmu.edu