Unhiding edited areas

Unhiding edited areas

[Martin Pohlack]

Hi there,

I'm using undo from time to time in org-mode buffers and sometimes the
area covered by the undo step is hidden (folded away) at the undo
time.  I feel somewhat uncomfortable in this situation as I,
apparently, rely on the visual appearance of the buffer when undoing
and redoing several steps.

For interactive search, the area around point is unfolded incrementally.

Is there a way to always show the area where edits are done?

Cheers,
Martin Pohlack

Re: Unhiding edited areas

[Eric S Fraga]

At Thu, 30 Jul 2009 17:42:08 +0200,
Martin Pohlack wrote:
> 
> Hi there,
> 
> I'm using undo from time to time in org-mode buffers and sometimes the
> area covered by the undo step is hidden (folded away) at the undo
> time.  I feel somewhat uncomfortable in this situation as I,
> apparently, rely on the visual appearance of the buffer when undoing
> and redoing several steps.
> 
> For interactive search, the area around point is unfolded incrementally.
> 
> Is there a way to always show the area where edits are done?

+1

Of course, there may be solutions already available, such as
UndoBrowse [1].  But I've not tried anything.

eric

Footnotes: 
[1]  http://www.emacswiki.org/emacs/UndoBrowse

Re: Unhiding edited areas

[Samuel Wales]

Hi Martin,

This is a huge issue.[1]

Here is what I do to try to work around it.

I use git, to limit the damage from confusion.

I expand the entire buffer if I think I am about to be confused.

;;i like the idea of clustering undo but find it disconcerting
(setf org-self-insert-cluster-for-undo nil)
;;somebody, I think Carsten, suggested this, and it might work for
you, but for some reason I commented it out.  I don't remember what
the reason was.  Maybe speed.
'(defadvice undo (after org-undo-reveal activate)
 "Make point and context visible after an undo command in Org-mode."
 (and (org-mode-p) (org-reveal)))
;;(ad-unadvise 'undo)


[1]  It is even more important when combined with what is IMO Emacs's
greatest need for improvement, which is that you can undo, and undo an
undo, and this is considered to be sufficient since you can get
anywhere in the timeline in principle -- but many users, myself
included, prefer a true redo command, both because undoing an undo
does not let you do commands (such as copying) in the middle of an
undo sequence without going the other direction, and because it feels
more intuitive to tell emacs where in the timeline we want to go, and
go forward or backward if we overshoot, thus making it possible to
view the timeline the same way as we go backward and forward in any
linear sequence.  (redo.el provides the functionality, but it corrupts
the buffer.)  Of course, many are comfortable with the traditional
undo-the-undo mechanism, so that should stay possible, but there are
many who are not, and a redo mechanism would satisfy them.  It is
possible to get more fancy with a tree.

-- 
Myalgic encephalomyelitis makes you die decades early (Jason
et al. 2006) and suffer severely.  Conflicts of interest are
destroying research. What people "know" is wrong. Silence = death.
http://www.meactionuk.org.uk/What_Is_ME_What_Is_CFS.htm

Re: Unhiding edited areas

[Martin Pohlack]

Hi Samuel,

Samuel Wales wrote:
> Hi Martin,
> 
> This is a huge issue.[1]
> 
> Here is what I do to try to work around it.
> 
> I use git, to limit the damage from confusion.

Yes, this or a versioning filesystem is probably advisable.

> I expand the entire buffer if I think I am about to be confused.
> 
> ;;i like the idea of clustering undo but find it disconcerting
> (setf org-self-insert-cluster-for-undo nil)
> ;;somebody, I think Carsten, suggested this, and it might work for
> you, but for some reason I commented it out.  I don't remember what
> the reason was.  Maybe speed.
> '(defadvice undo (after org-undo-reveal activate)
>  "Make point and context visible after an undo command in Org-mode."
>  (and (org-mode-p) (org-reveal)))
> ;;(ad-unadvise 'undo)

Awesome, this is exactly what I was looking for!

> [1]  It is even more important when combined with what is IMO Emacs's
> greatest need for improvement, which is that you can undo, and undo an
> undo, and this is considered to be sufficient since you can get
> anywhere in the timeline in principle -- but many users, myself
> included, prefer a true redo command, both because undoing an undo
> does not let you do commands (such as copying) in the middle of an
> undo sequence without going the other direction, and because it feels
> more intuitive to tell emacs where in the timeline we want to go, and
> go forward or backward if we overshoot, thus making it possible to
> view the timeline the same way as we go backward and forward in any
> linear sequence.  (redo.el provides the functionality, but it corrupts
> the buffer.)  Of course, many are comfortable with the traditional
> undo-the-undo mechanism, so that should stay possible, but there are
> many who are not, and a redo mechanism would satisfy them.  It is
> possible to get more fancy with a tree.

The current undo system is very powerful as it doesn't lose history
(unless you hit a quota limit).  With undo-redo systems you usually can
lose history if you edit things in an old state.  Suddenly redo is not
available anymore.  You can only access the most recent branch in the
history tree.

In emacs this will not happen as you can reach all nodes in the buffer
history, but these states are not easily accessible, especially, if you
went back and forth some times.  I cannot track the list of states in my
mind or imagine the current structure of the undo tree, I can only go
step by step and look at the situation in the buffer and decide whether
this is what I wanted or not.

I recently stumbled upon an article which, I think, quiet nicely
summarized what one wants:

  http://e-texteditor.com/blog/2006/making-undo-usable

But it's not available for emacs ...

Cheers,
Martin

Re: Unhiding edited areas

[Samuel Wales]

Hi Martin,

On 2009-07-30, Martin Pohlack <mp26@os.inf.tu-dresden.de> wrote:
>> '(defadvice undo (after org-undo-reveal activate)
>>  "Make point and context visible after an undo command in Org-mode."
>>  (and (org-mode-p) (org-reveal)))
>> ;;(ad-unadvise 'undo)
>
> Awesome, this is exactly what I was looking for!

Maybe we can improve on it with one or more of these:

  1) Check visibility before revealing.
  2) Speed.
  3) (emacs) /Include visibility in the undo stack/ so that
     visibility while undoing is always what it was when you
     did the editing.
  4) (emacs) Implement undo-redo so that manually revealing
     does not break the chain as it does with undo.

> The current undo system is very powerful as it doesn't lose history
> (unless you hit a quota limit).  With undo-redo systems you usually can
> lose history if you edit things in an old state.  Suddenly redo is not
> available anymore.  You can only access the most recent branch in the
> history tree.

Yes, unless you implement a tree.  But even with that
limitation, I prefer undo-redo.

The cognitive burden is not the only limitation
of undo-the-undo.  With undo-the-undo, you cannot
realistically copy text from different places in the undo
history.

Try to go back 50 edits, copy, go back a few more edits
(you're in trouble already :)), copy, go forward 10, copy,
go forward 15, copy, go back 15 more, copy, go back 15 more,
copy.  With undo-redo, I think that it would be faster.

>   http://e-texteditor.com/blog/2006/making-undo-usable

Yes, I agree that there are some good ideas there.

-- 
Myalgic encephalomyelitis makes you die decades early (Jason
et al. 2006) and suffer severely.  Conflicts of interest are
destroying research. What people "know" is wrong. Silence = death.
http://www.meactionuk.org.uk/What_Is_ME_What_Is_CFS.htm

Re: Unhiding edited areas

[Carsten Dominik]

On Jul 30, 2009, at 8:50 PM, Martin Pohlack wrote:

> Hi Samuel,
>
> Samuel Wales wrote:
>> Hi Martin,
>>
>> This is a huge issue.[1]
>>
>> Here is what I do to try to work around it.
>>
>> I use git, to limit the damage from confusion.
>
> Yes, this or a versioning filesystem is probably advisable.
>
>> I expand the entire buffer if I think I am about to be confused.
>>
>> ;;i like the idea of clustering undo but find it disconcerting
>> (setf org-self-insert-cluster-for-undo nil)
>> ;;somebody, I think Carsten, suggested this, and it might work for
>> you, but for some reason I commented it out.  I don't remember what
>> the reason was.  Maybe speed.
>> '(defadvice undo (after org-undo-reveal activate)
>> "Make point and context visible after an undo command in Org-mode."
>> (and (org-mode-p) (org-reveal)))
>> ;;(ad-unadvise 'undo)

Hi Samuel, Martin,

this could be a good FAQ entry, I think.

If speed was the reason to discard this, maybe try again, because
I believe that recent changes will have solved many of the speed issues
that might plague org-reveal in lists with many siblings.

- Carsten

>
> Awesome, this is exactly what I was looking for!
>
>> [1]  It is even more important when combined with what is IMO Emacs's
>> greatest need for improvement, which is that you can undo, and undo  
>> an
>> undo, and this is considered to be sufficient since you can get
>> anywhere in the timeline in principle -- but many users, myself
>> included, prefer a true redo command, both because undoing an undo
>> does not let you do commands (such as copying) in the middle of an
>> undo sequence without going the other direction, and because it feels
>> more intuitive to tell emacs where in the timeline we want to go, and
>> go forward or backward if we overshoot, thus making it possible to
>> view the timeline the same way as we go backward and forward in any
>> linear sequence.  (redo.el provides the functionality, but it  
>> corrupts
>> the buffer.)  Of course, many are comfortable with the traditional
>> undo-the-undo mechanism, so that should stay possible, but there are
>> many who are not, and a redo mechanism would satisfy them.  It is
>> possible to get more fancy with a tree.
>
> The current undo system is very powerful as it doesn't lose history
> (unless you hit a quota limit).  With undo-redo systems you usually  
> can
> lose history if you edit things in an old state.  Suddenly redo is not
> available anymore.  You can only access the most recent branch in the
> history tree.
>
> In emacs this will not happen as you can reach all nodes in the buffer
> history, but these states are not easily accessible, especially, if  
> you
> went back and forth some times.  I cannot track the list of states  
> in my
> mind or imagine the current structure of the undo tree, I can only go
> step by step and look at the situation in the buffer and decide  
> whether
> this is what I wanted or not.
>
> I recently stumbled upon an article which, I think, quiet nicely
> summarized what one wants:
>
>  http://e-texteditor.com/blog/2006/making-undo-usable
>
> But it's not available for emacs ...
>
> Cheers,
> Martin
>
>
> _______________________________________________
> Emacs-orgmode mailing list
> Remember: use `Reply All' to send replies to the list.
> Emacs-orgmode@gnu.org
> http://lists.gnu.org/mailman/listinfo/emacs-orgmode