emacs-orgmode@gnu.org archives
 help / color / mirror / code / Atom feed
From: Bithov Vinu Student <vinub@calday.co.uk>
To: emacs-orgmode@gnu.org
Subject: Org-learn failing to update calendar scheduling
Date: Sun, 19 Dec 2021 18:31:10 +0000	[thread overview]
Message-ID: <CADr1GP-235bwf4Gv3uZUQPWQ8WjuU+ybFQModiKELAExedffSg@mail.gmail.com> (raw)


[-- Attachment #1.1: Type: text/plain, Size: 1529 bytes --]

I installed the plugin org-learn for my Emacs (latest version, latest
org-mode as well). The plugins no longer actively maintained nor does it
need to be, it does a select few things and doesn't need to be updated. The
org-learn module I used is attached to this email, here
<https://lists.gnu.org/archive/html/emacs-orgmode/2009-10/msg00535.html> is
the original email about org-learn that describes how to use it, etc.

Anyway, I created an example org heading as follows:

* TODO Save the world

I hovered my cursor over the heading, M-x org-smart-reschedule, and gave a
grade of 5. It then switched to the calendar minibuffer, where the
preselected date is the current date. It shouldn't be this, it should be
the current date + n days (n increases as the heading is scheduled again
and again).

If I repeatedly org-smart-reschedule, it'll work out an interval and update
the E-factor, but nothing will be done for the next scheduling date. This
is frustrating because the more correct gradings I've give (>= 4 when
prompted) the larger the time until the next scheduling, and the rate of
increase in intervals increases. As a result after a couple of repetitions
it may be scheduled for several years later, which is impractical to do
manually.

How would I fix this?

Thanks,
~

-- 
Student Account

-- 
Calday Grange Grammar School is a charitable company limited by guarantee 
and registered in England and Wales with company number 8332696.
The 
Registered Office is at Grammar School Lane, West Kirby, Wirral, CH48 8GG

[-- Attachment #1.2: Type: text/html, Size: 2104 bytes --]

[-- Attachment #2: org-learn.el --]
[-- Type: text/x-emacs-lisp, Size: 6055 bytes --]

;;; org-learn.el --- Implements SuperMemo's incremental learning algorithm

;; Copyright (C) 2009-2012 Free Software Foundation, Inc.

;; Author: John Wiegley <johnw at gnu dot org>
;; Keywords: outlines, hypermedia, calendar, wp
;; Homepage: http://orgmode.org
;; Version: 6.32trans
;;
;; This file is not part of GNU Emacs.
;;
;; GNU Emacs is free software: you can redistribute it and/or modify
;; it under the terms of the GNU General Public License as published by
;; the Free Software Foundation, either version 3 of the License, or
;; (at your option) any later version.

;; GNU Emacs is distributed in the hope that it will be useful,
;; but WITHOUT ANY WARRANTY; without even the implied warranty of
;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;; GNU General Public License for more details.

;; You should have received a copy of the GNU General Public License
;; along with GNU Emacs.  If not, see <http://www.gnu.org/licenses/>.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;
;;; Commentary:

;; The file implements the learning algorithm described at
;; http://supermemo.com/english/ol/sm5.htm, which is a system for reading
;; material according to "spaced repetition".  See
;; http://en.wikipedia.org/wiki/Spaced_repetition for more details.
;;
;; To use, turn on state logging and schedule some piece of information you
;; want to read.  Then in the agenda buffer type

(require 'org)
(eval-when-compile
  (require 'cl))

(defgroup org-learn nil
  "Options concerning the learning code in Org-mode."
  :tag "Org Learn"
  :group 'org-progress)

(defcustom org-learn-always-reschedule nil
  "If non-nil, always reschedule items, even if retention was \"perfect\"."
  :type 'boolean
  :group 'org-learn)

(defcustom org-learn-fraction 0.5
  "Controls the rate at which EF is increased or decreased.
Must be a number between 0 and 1 (the greater it is the faster
the changes of the OF matrix)."
  :type 'float
  :group 'org-learn)

(defun initial-optimal-factor (n ef)
  (if (= 1 n)
      4
    ef))

(defun get-optimal-factor (n ef of-matrix)
  (let ((factors (assoc n of-matrix)))
    (or (and factors
	     (let ((ef-of (assoc ef (cdr factors))))
	       (and ef-of (cdr ef-of))))
	(initial-optimal-factor n ef))))

(defun set-optimal-factor (n ef of-matrix of)
  (let ((factors (assoc n of-matrix)))
    (if factors
	(let ((ef-of (assoc ef (cdr factors))))
	  (if ef-of
	      (setcdr ef-of of)
	    (push (cons ef of) (cdr factors))))
      (push (cons n (list (cons ef of))) of-matrix)))
  of-matrix)

(defun inter-repetition-interval (n ef &optional of-matrix)
  (let ((of (get-optimal-factor n ef of-matrix)))
    (if (= 1 n)
	of
      (* of (inter-repetition-interval (1- n) ef of-matrix)))))

(defun modify-e-factor (ef quality)
  (if (< ef 1.3)
      1.3
    (+ ef (- 0.1 (* (- 5 quality) (+ 0.08 (* (- 5 quality) 0.02)))))))

(defun modify-of (of q fraction)
  (let ((temp (* of (+ 0.72 (* q 0.07)))))
    (+ (* (- 1 fraction) of) (* fraction temp))))

(defun calculate-new-optimal-factor (interval-used quality used-of
						   old-of fraction)
  "This implements the SM-5 learning algorithm in Lisp.
INTERVAL-USED is the last interval used for the item in question.
QUALITY is the quality of the repetition response.
USED-OF is the optimal factor used in calculation of the last
interval used for the item in question.
OLD-OF is the previous value of the OF entry corresponding to the
relevant repetition number and the E-Factor of the item.
FRACTION is a number belonging to the range (0,1) determining the
rate of modifications (the greater it is the faster the changes
of the OF matrix).

Returns the newly calculated value of the considered entry of the
OF matrix."
  (let (;; the value proposed for the modifier in case of q=5
	(mod5 (/ (1+ interval-used) interval-used))
	;; the value proposed for the modifier in case of q=2
	(mod2 (/ (1- interval-used) interval-used))
	;; the number determining how many times the OF value will
	;; increase or decrease
	modifier)
    (if (< mod5 1.05)
	(setq mod5 1.05))
    (if (< mod2 0.75)
	(setq mod5 0.75))
    (if (> quality 4)
	(setq modifier (1+ (* (- mod5 1) (- quality 4))))
      (setq modifier (- 1 (* (/ (- 1 mod2) 2) (- 4 quality)))))
    (if (< modifier 0.05)
	(setq modifier 0.05))
    (setq new-of (* used-of modifier))
    (if (> quality 4)
	(if (< new-of old-of)
	    (setq new-of old-of)))
    (if (< quality 4)
	(if (> new-of old-of)
	    (setq new-of old-of)))
    (setq new-of (+ (* new-of fraction) (* old-of (- 1 fraction))))
    (if (< new-of 1.2)
	(setq new-of 1.2)
      new-of)))

(defvar initial-repetition-state '(-1 1 2.5 nil))

(defun determine-next-interval (n ef quality of-matrix)
  (assert (> n 0))
  (assert (and (>= quality 0) (<= quality 5)))
  (if (< quality 3)
      (list (inter-repetition-interval n ef) (1+ n) ef nil)
    (let ((next-ef (modify-e-factor ef quality)))
      (setq of-matrix
	    (set-optimal-factor n next-ef of-matrix
				(modify-of (get-optimal-factor n ef of-matrix)
					   quality org-learn-fraction))
	    ef next-ef)
      ;; For a zero-based quality of 4 or 5, don't repeat
      (if (and (>= quality 4)
	       (not org-learn-always-reschedule))
	  (list 0 (1+ n) ef of-matrix)
	(list (inter-repetition-interval n ef of-matrix) (1+ n)
	      ef of-matrix)))))

(defun org-smart-reschedule (quality)
  (interactive "nHow well did you remember the information (on a scale of 0-5)? ")
  (let* ((learn-str (org-entry-get (point) "LEARN_DATA"))
	 (learn-data (or (and learn-str
			      (read learn-str))
			 (copy-list initial-repetition-state)))
	 closed-dates)
    (setq learn-data
	  (determine-next-interval (nth 1 learn-data)
				   (nth 2 learn-data)
				   quality
				   (nth 3 learn-data)))
    (org-entry-put (point) "LEARN_DATA" (prin1-to-string learn-data))
    (if (= 0 (nth 0 learn-data))
	(org-schedule t)
      (org-schedule nil (time-add (current-time)
				  (days-to-time (nth 0 learn-data)))))))

(provide 'org-learn)

;;; org-learn.el ends here

             reply	other threads:[~2021-12-20 18:24 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-12-19 18:31 Bithov Vinu Student [this message]
2022-01-30  8:26 ` Org-learn failing to update calendar scheduling Ihor Radchenko

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://www.orgmode.org/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CADr1GP-235bwf4Gv3uZUQPWQ8WjuU+ybFQModiKELAExedffSg@mail.gmail.com \
    --to=vinub@calday.co.uk \
    --cc=emacs-orgmode@gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://git.savannah.gnu.org/cgit/emacs/org-mode.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).