Semantics of 'C-c -' and massively-indented lists

Semantics of 'C-c -' and massively-indented lists

[Huy]

Hello,
I'm a beginner, so please let me know if I'm completely off the mark.

The effects of converting from headline to list with 'C-c -' and back to 'C-c *' don't behave as I'd expect.  The indentation levels seem all wrong, no matter how you look at it.

My Setup
~~~~~~~~
I'm using org 7.5 with Cocoa emacs 23.3 on OSX
Note that I do not have the "odd" option on, I don't do visual-line-mode, and I rely on org-adapt-indentation to be on.

Converting from headline to list
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
If I have a LEVEL=5 headline and I convert it to a list with 'C-c -', I get too much indentation
***** Headline
***** Converted Line
becomes
***** Headline
        - Converted Line

After looking at the code, I have the impression that the intent is that the LEVEL of the line should remain 5; thus, the list is indented by 2 spaces for each level under 5, i.e. 8 spaces.
But who needs a LEVEL=5 list at this point?

Expected behavior
~~~~~~~~~~~~~~~~~
What I would like is a LEVEL=1 list, properly indented along with the rest of the body text.

If org-adapt-indentation is t, then I would expect to see:
***** Headline
      - Converted Line
where the '-' is aligned with the 'H' of the above headline, at the same column where any body text would start if I hit <TAB> in a line after Headline (if there were no list).

And if org-adapt-indentation is nil, then I would expect to see:
***** Headline
- Converted Line

Workaround
~~~~~~~~~~
Right now, what I have to do is manuallay remove the extra spaces to do the alignment myself.

Converting from list to headline
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In fact, if I take the list and I try to convert it back to a headline with 'C-c *', it doesn't even matter how many spaces there are in front of the '-': it's always converted to a LEVEL=5 headline because the parent is LEVEL=4:

***** Headline
        - Converted Line
becomes
***** Headline
****** Converted Line

***** Headline
                   - Converted Line
also becomes
***** Headline
****** Converted Line

***** Headline
- Converted Line
also becomes
***** Headline
****** Converted Line

Well, at least 'C-c *' acts the way I'd expect it to: the line becomes a sub-heading of Headline.

Other modes
~~~~~~~~~~~
Things look especially weird when org-indent-mode is turned on, since you have a mixture of "visual" indentation and hard space indentation.  I would expect a 'C-c -' to insert no hard spaces at all and leave everything in column 0 and let the visual-line-mode adjust the display.

If you have 'odd' mode on, however, everything seems to look perfectly.  Why?  Sheer luck and maybe an off-by-1 bug.  The headlines go up by 2 columns, and '- ' also happens to be 2 characters long.  But also, if 'odd' mode is on, for some reason 1 is added to the resulting list level.  So after the conversion, the list just happens to lign up perfectly with the 'H' of Headline.

***** Headline LEVEL=3
***** Converted Line LEVEL=3
becomes
***** Headline LEVEL=3
      - Converted Line LEVEL=4

This might be the reason that some people haven't noticed this problem.  Maybe most users use 'odd' mode?

Acting relative to the parent
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Now, what happens when we convert the previous example back by using 'C-c *'?  We get a LEVEL-4 headline.
***** Headline LEVEL=3
      - Converted Line LEVEL=4
becomes
***** Headline LEVEL=3
******* Converted Line LEVEL=4

The operations are not symmetric because somehow the LEVEL was incremented.
Well, this behavior actually makes sense.  When we converted from headline to list, we were saying that the headline should now be part of the body text of Headline, so it should be considered a child of Headline, i.e. LEVEL=4.  And going back, well, we might as well preserve LEVEL=4, keeping it as a child of Headline, instead of guessing the user's intent to be LEVEL=3, a sibling of Headline.

What doesn't make sense is what happens when do 'C-c -' once again.
***** Headline LEVEL=3
******* Converted Line LEVEL=4
becomes
***** Headline LEVEL=3
        - Converted Line LEVEL=5

Whoa, all we did was 'C-c -' then 'C-c *' then 'C-c -' and the line got demoted.
I bet at this time, you'd like to know what hapepns when you hit 'C-c *'.  Don't worry, I did it for you.  It comes back to LEVEL=4.  It actually stabilizes, thank god.
***** Headline LEVEL=3
        - Converted Line LEVEL=5
becomes
***** Headline LEVEL=3
******* Converted Line LEVEL=4

Expected behavior
~~~~~~~~~~~~~~~~~
For a given parent headline, there are several cases when applying 'C-c -'.
1)
***** Headline LEVEL=3
***** Converted Line LEVEL=3
2)
***** Headline LEVEL=3
******* Converted Line LEVEL=4
3)
***** Headline LEVEL=3
********* Converted Line LEVEL=5
4)
***** Headline LEVEL=3
*** Converted Line LEVEL=2

For any case, the resulting list should just be a LEVEL=1 list regardless of the level that it started, and it should be properly indented according to the parent headline.  That's because a list should become part of the body of the above headline, no matter what.

And when we go back to 'C-c *', what makes sense is for the line to become a sub-heading of Headline, as is the current behavior.

'C-c *' fringe case
~~~~~~~~~~~~~~~~~~~
In most cases, 'C-c *' acts sanely.
It acts weird when there is no Headline above a list item and you try to convert to a Headline.  Again, in 'odd' mode.
- Converted Line LEVEL=1 (alone without parent)
becomes
** Converted Line LEVEL=1.5 (alone without parent)

Yes, that's a decimal. It's trying to hide in the cracks.

Summary
~~~~~~~
'C-c *' mostly works right.  'C-c -' has strange behavior in all display styles, especially 'oddeven' and org-indent-mode modes and even in 'odd' mode.  The resulting indentation is not useful.

Thanks for responding,
Huy

Re: Semantics of 'C-c -' and massively-indented lists

[Nicolas Goaziou]

Hello,

Huy <list-orgmode@reml.org> writes:

> The effects of converting from headline to list with 'C-c -' and back
> to 'C-c *' don't behave as I'd expect. The indentation levels seem all
> wrong, no matter how you look at it.

I have pushed a patch that should go in the right direction. See below.

> What I would like is a LEVEL=1 list, properly indented along with the
> rest of the body text.
>
> If org-adapt-indentation is t, then I would expect to see:

> ***** Headline
>       - Converted Line

> where the '-' is aligned with the 'H' of the above headline, at the
> same column where any body text would start if I hit <TAB> in a line
> after Headline (if there were no list).
>
> And if org-adapt-indentation is nil, then I would expect to see:

> ***** Headline
> - Converted Line

Fixed.


> In fact, if I take the list and I try to convert it back to a headline
> with 'C-c *', it doesn't even matter how many spaces there are in
> front of the '-': it's always converted to a LEVEL=5 headline because
> the parent is LEVEL=4:

This is the intended behavior.

> Things look especially weird when org-indent-mode is turned on, since
> you have a mixture of "visual" indentation and hard space indentation.
> I would expect a 'C-c -' to insert no hard spaces at all and leave
> everything in column 0 and let the visual-line-mode adjust the
> display.

Fixed. It's the same as the first case, as org-indent-mode sets
org-adapt-indentation to nil.

> Now, what happens when we convert the previous example back by using
> 'C-c *'? We get a LEVEL-4 headline.

> ***** Headline LEVEL=3
>       - Converted Line LEVEL=4

> becomes

> ***** Headline LEVEL=3
> ******* Converted Line LEVEL=4

> The operations are not symmetric because somehow the LEVEL was
> incremented.

Operations are not symmetric because lists and headlines are not
equivalent objects.


> In most cases, 'C-c *' acts sanely. It acts weird when there is no
> Headline above a list item and you try to convert to a Headline.
> Again, in 'odd' mode.

> - Converted Line LEVEL=1 (alone without parent)

> becomes

> ** Converted Line LEVEL=1.5 (alone without parent)
>
> Yes, that's a decimal. It's trying to hide in the cracks.

Fixed.

Please report back if something is still wrong. Thank you.

Regards,

-- 
Nicolas Goaziou

Re: Semantics of 'C-c -' and massively-indented lists

[Huy]

On Sat, May 28, 2011 at 02:29:07PM +0200, Nicolas Goaziou wrote:
> Hello,
> 
> Huy <list-orgmode@reml.org> writes:
> 
> > The effects of converting from headline to list with 'C-c -' and back
> > to 'C-c *' don't behave as I'd expect. The indentation levels seem all
> > wrong, no matter how you look at it.
> 
> I have pushed a patch that should go in the right direction. See below.
> 
> > What I would like is a LEVEL=1 list, properly indented along with the
> > rest of the body text.
> >
> > If org-adapt-indentation is t, then I would expect to see:
> 
> > ***** Headline
> >       - Converted Line
> 
> > where the '-' is aligned with the 'H' of the above headline, at the
> > same column where any body text would start if I hit <TAB> in a line
> > after Headline (if there were no list).
> >
> > And if org-adapt-indentation is nil, then I would expect to see:
> 
> > ***** Headline
> > - Converted Line
> 
> Fixed.
> 
> 
> > In fact, if I take the list and I try to convert it back to a headline
> > with 'C-c *', it doesn't even matter how many spaces there are in
> > front of the '-': it's always converted to a LEVEL=5 headline because
> > the parent is LEVEL=4:
> 
> This is the intended behavior.
> 
> > Things look especially weird when org-indent-mode is turned on, since
> > you have a mixture of "visual" indentation and hard space indentation.
> > I would expect a 'C-c -' to insert no hard spaces at all and leave
> > everything in column 0 and let the visual-line-mode adjust the
> > display.
> 
> Fixed. It's the same as the first case, as org-indent-mode sets
> org-adapt-indentation to nil.
> 
> > Now, what happens when we convert the previous example back by using
> > 'C-c *'? We get a LEVEL-4 headline.
> 
> > ***** Headline LEVEL=3
> >       - Converted Line LEVEL=4
> 
> > becomes
> 
> > ***** Headline LEVEL=3
> > ******* Converted Line LEVEL=4
> 
> > The operations are not symmetric because somehow the LEVEL was
> > incremented.
> 
> Operations are not symmetric because lists and headlines are not
> equivalent objects.
> 
> 
> > In most cases, 'C-c *' acts sanely. It acts weird when there is no
> > Headline above a list item and you try to convert to a Headline.
> > Again, in 'odd' mode.
> 
> > - Converted Line LEVEL=1 (alone without parent)
> 
> > becomes
> 
> > ** Converted Line LEVEL=1.5 (alone without parent)
> >
> > Yes, that's a decimal. It's trying to hide in the cracks.
> 
> Fixed.
> 
> Please report back if something is still wrong. Thank you.

Perfect behavior! Exactly what I was expecting.

The only problem is a typo 'skip-blank' instead of 'skip-blanks'

Thanks a lot for the quick turnaround!

> 
> Regards,
> 
> -- 
> Nicolas Goaziou
> 

Re: Semantics of 'C-c -' and massively-indented lists

[Nicolas Goaziou]

Huy <list-orgmode@reml.org> writes:

> The only problem is a typo 'skip-blank' instead of 'skip-blanks'

Oops. Fixed that too.

Regards,

-- 
Nicolas Goaziou

Re: Semantics of 'C-c -' and massively-indented lists

[Huy]

On Sat, May 28, 2011 at 07:09:06AM -0700, Huy wrote:
> On Sat, May 28, 2011 at 02:29:07PM +0200, Nicolas Goaziou wrote:
> > Fixed.
> > 
> > Please report back if something is still wrong. Thank you.
> 
> Perfect behavior! Exactly what I was expecting.
> 
> The only problem is a typo 'skip-blank' instead of 'skip-blanks'
> 
> Thanks a lot for the quick turnaround!

Actually, I found a few more problems

Multiple Lines
~~~~~~~~~~~~~~
I had forgotten to test when invoking 'C-c -' on multiple lines.

If I have (with oddeven and org-adapt-indentation on) the following:
** Headline 2
** Converting 2
*** Converting 3
* Converting 1
** Converting 2
*** Converting 3
and I try to do 'C-c -' on the 5 bottom lines, I get:
** Headline 2
   - Converting 1
     - Converting 2
   - Converting 1
   - Converting 1
     - Converting 2

I would hope to get the following instead:
** Headline 2
   - Converting 1
     - Converting 2
   - Converting 1
     - Converting 2
       - Converting 3

Whenever possible, the converted lines should preserve their relative levels with their parent.
In the above example, we have 2 subtrees that should be preserved separately.
Of course, we can't keep all 5 items with levels relative to each other, otherwise we'd get the nonsensical effect:
  - orphan
- parent
But in other cases, entire subtrees can be preserved.

Strange region behavior and 'C-c *'
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
When highlighting a region and then applying 'C-c -', all the lines that have at least one character highlighted (anywhere on the line) will be converted.  That's good and as expected.

But the same doesn't happen with 'C-c *'.  There, it actually matters where the region starts/ends in the line.

For example, if you have
** Headline 2
   - Converting 1
If you triple-mouse-click "Converting 1" or if you just single-click the cursor on "Converting 1", then 'C-c *' will convert the line fine.
However, if you only select any of the letters on that line, e.g. the letter 'v', and then 'C-c *', then you get this:
** Headline 2
*** Con
verting 1


I have other problems that involve vimpulse and its visual mode, but I won't bother you with that :)

Thanks,
Huy

Re: Semantics of 'C-c -' and massively-indented lists

[Nicolas Goaziou]

Huy <list-orgmode@reml.org> writes:

> Whenever possible, the converted lines should preserve their relative
> levels with their parent. In the above example, we have 2 subtrees
> that should be preserved separately.
> Of course, we can't keep all 5 items with levels relative to each
> other, otherwise we'd get the nonsensical effect:

>   - orphan
> - parent

> But in other cases, entire subtrees can be preserved.

Fixed.

> For example, if you have

> ** Headline 2
>    - Converting 1

> If you triple-mouse-click "Converting 1" or if you just single-click
> the cursor on "Converting 1", then 'C-c *' will convert the line fine.
> However, if you only select any of the letters on that line, e.g. the
> letter 'v', and then 'C-c *', then you get this:

> ** Headline 2
> *** Con
> verting 1

Fixed too.

> I have other problems that involve vimpulse and its visual mode, but
> I won't bother you with that :)

I can't help there. But, if something is still wrong about toggling
headlines and items, please keep telling me.

Regards,

-- 
Nicolas Goaziou