bug#2086: 23.0.60; Todo mode bug with completing-read + patch

[Stephen Berman]

GNU Emacs 23.0.60.30 (i686-pc-linux-gnu, GTK+ Version 2.14.4) of
2009-01-21 on escher

1. emacs -Q
2. Type `M-x todo-show', and in the Todo mode buffer type `s' to save
~/.todo-do, then type the following sequence of Todo mode commands and input:
`I item1 RET i item2 RET Newcat RET'.  Widen the buffer and you see the
following (modulo the time-stamp headers):

-*- mode: todo; todo-categories: ("Newcat" "Todo"); -*-
*/* --- Newcat
*/* 2009-01-14 13:48 steve: item2
--- End
*/* ---------------------------------------------------------------------------
*/* --- Todo
*/* 2009-01-14 13:48 steve: item1
--- End
*/* ---------------------------------------------------------------------------
   
3. Now type `M-x customize-option RET history-delete-duplicates RET', toggle
   the value to `on (non-nil)' and save for the current session.
4. Type `M-x todo-show'.
5. In the Todo mode buffer, with category "Newcat" displayed, type `j' and at
   the prompt, type `Todo', to jump the the category "Todo".
=> Now category "Todo" is displayed, but it is empty.
6. Widen the Todo mode buffer, and this is what it now looks like:

-*- mode: todo; todo-categories: ("Todo" "Newcat"); -*-
*/* --- Todo
--- End
*/* ---------------------------------------------------------------------------
*/* --- Newcat
*/* 2009-01-14 13:48 steve: item2
--- End
*/* ---------------------------------------------------------------------------
*/* --- Todo
*/* 2009-01-14 13:48 steve: item1
--- End
*/* ---------------------------------------------------------------------------

As a second example, after step 4 above do the following:
5'. While in category "Newcat", type `i item3 RET Todo RET' to insert "item3".
=> Now category "Todo" is displayed, but it contains only item3, not also
item1.
6. Widen the Todo mode buffer, and this is what it now looks like:

-*- mode: todo; todo-categories: ("Todo" "Newcat"); -*-
*/* --- Todo
*/* 2009-01-14 13:48 steve: item3
--- End
*/* ---------------------------------------------------------------------------
*/* --- Newcat
*/* 2009-01-14 13:48 steve: item2
--- End
*/* ---------------------------------------------------------------------------
*/* --- Todo
*/* 2009-01-14 13:48 steve: item1
--- End
*/* ---------------------------------------------------------------------------


The problem in both cases is due, I believe, to the failure to make a copy of
todo-categories before calling completing-read and to restore it afterwards.
This is because, when history-delete-duplicates is non-nil, completing-read
(via read_minibuf) destructively deletes the list element used as the
completion and conses it on to the front, yielding a reordered list.  The fix
I propose, in the attached patch, is a wrapper for completing-read that copies
and restores todo-categories as required.  The wrapper is used instead of
completing-read by todo-insert-item and todo-jump-to-category.  (The resulting
modularization also implements one to the "things to do" mentioned in the
todo-mode.el commentary.)

I also think the completing-read code in todo-mode.el should be simplified by
replacing todo-category-alist.  The doc string of the latter says: "Generate
an alist for use in `completing-read' from `todo-categories'."  AFAICS there
is no need for this structure.  The members of the alist, which constitutes
the COLLECTION argument of completing-read, are just the single element lists
of each category name.  This provides no more information than the value of
todo-categories itself, which as a list of strings is suitable for the
COLLECTION argument.  (Was there a time when completing-read took an alist but
not a list of strings as the COLLECTION argument?)  The wrapper uses
todo-categories instead of todo-category-alist, so the patch eliminates the
latter function from todo-mode.el.

Finally, the patch makes the code of todo-insert-item conform to its doc
string.  The latter says: "With a prefix argument solicit the category,
otherwise use the current category."  But the code does just the opposite.  I
assume the desired behavior is as stated in the doc string.  (In
http://lists.gnu.org/archive/html/bug-gnu-emacs/2002-01/msg00543.html there
was a patch that contained a fix for this, but it was not committed to CVS.)


2009-01-19  Stephen Berman  <stephen.berman@gmx.net>

        * calendar/todo-mode.el (todo-category-alist): Delete.
        (todo-completing-read): New function.
        (todo-insert-item, todo-jump-to-category): Use it.
        (todo-insert-item): Make the use of the prefix argument
        conform to the doc string. 
        

*** emacs/lisp/calendar/todo-mode.el.~1.72.~    2009-01-09 11:48:58.000000000 
+0100
--- emacs/lisp/calendar/todo-mode.el    2009-01-21 10:36:15.000000000 +0100
***************
*** 596,610 ****
                              "New TODO entry: "
                              (if todo-entry-prefix-function
                                  (funcall todo-entry-prefix-function)))))
-          (categories todo-categories)
-          (history (cons 'categories (1+ todo-category-number)))
           (current-category (nth todo-category-number todo-categories))
!          (category
!           (if arg
!               current-category
!             (completing-read (concat "Category [" current-category "]: ")
!                              (todo-category-alist) nil nil nil
!                              history current-category))))
        (todo-add-item-non-interactively new-item category))))
  
  (defalias 'todo-cmd-inst 'todo-insert-item)
--- 596,603 ----
                              "New TODO entry: "
                              (if todo-entry-prefix-function
                                  (funcall todo-entry-prefix-function)))))
           (current-category (nth todo-category-number todo-categories))
!          (category (if arg (todo-completing-read) current-category)))
        (todo-add-item-non-interactively new-item category))))
  
  (defalias 'todo-cmd-inst 'todo-insert-item)
***************
*** 801,812 ****
  (defun todo-jump-to-category ()
    "Jump to a category.  Default is previous category."
    (interactive)
!   (let* ((categories todo-categories)
!          (history (cons 'categories (1+ todo-category-number)))
!        (default (nth todo-category-number todo-categories))
!        (category (completing-read
!                     (concat "Category [" default "]: ")
!                     (todo-category-alist) nil nil nil history default)))
      (if (string= "" category)
          (setq category (nth todo-category-number todo-categories)))
      (setq todo-category-number
--- 794,800 ----
  (defun todo-jump-to-category ()
    "Jump to a category.  Default is previous category."
    (interactive)
!   (let ((category (todo-completing-read)))
      (if (string= "" category)
          (setq category (nth todo-category-number todo-categories)))
      (setq todo-category-number
***************
*** 861,869 ****
    "Return non-nil if STRING spans several lines."
    (> (todo-string-count-lines string) 1))
  
! (defun todo-category-alist ()
!   "Generate an alist for use in `completing-read' from `todo-categories'."
!   (mapcar #'list todo-categories))
  
  ;; ---------------------------------------------------------------------------
  
--- 849,867 ----
    "Return non-nil if STRING spans several lines."
    (> (todo-string-count-lines string) 1))
  
! (defun todo-completing-read ()
!   "Return a category name, with completion, for use in Todo mode."
!   ;; make a copy of todo-categories in case history-delete-duplicates is
!   ;; non-nil, which makes completing-read alter todo-categories
!   (let* ((categories (copy-sequence todo-categories))
!          (history (cons 'todo-categories (1+ todo-category-number)))
!        (default (nth todo-category-number todo-categories))
!        (category (completing-read
!                     (concat "Category [" default "]: ")
!                     todo-categories nil nil nil history default)))
!     ;; restore the original value of todo-categories
!     (setq todo-categories categories)
!     category))
  
  ;; ---------------------------------------------------------------------------