current-column and display (space :align-to) properties

current-column and display (space :align-to) properties

[Michael Mauger]

I'm not sure if this is a bug or a feature...

Along the lines of a question that arose a couple of weeks ago re: counting 
before and after properties when calculating current-column, should the 
space :align-to and :width properties be reflected in current-column?

For example, in *scratch*, try the following:

  (progn
  (insert "["
	  (propertize "01" 
		      'display '(space :align-to 10))
	  "]XX\n")
  (insert "[" 
	  (propertize "0123456789012"
		      'display '(space :align-to 5)) 
	  "]YY\n")
  ) `C-j'

You should see something like the following in *scratch*:

  [         ]XX
  [    ]YY
  nil

If you place the cursor on the second of the two XXs (actually point is between 
the two XXs), the current-column is 5 (the two square brackets, the two 
characters in the propertized string and the first X.)

If you place the cursor on the second of the two YYs, the current-column is 16 
(the two square brackets, the twelve characters in the propertized string and 
the first Y.)

Even stranger, if you again place the cursor between the two XXs and hit `C-n', 
the cursor will go to before the closing bracket on the second line in an 
effort to preserve the column position.

Is this desired behavior?  Should current-column look at the display property 
to determine how to calculate the column position?

-- Michael

Re: current-column and display (space :align-to) properties

[Stefan Monnier]

> Is this desired behavior?

No.

> Should current-column look at the display property to determine how to
> calculate the column position?

Yes.  It should also return a float value reflecting the real display size
(just like :align-to can be fractional, or if you use proportional fonts,
...).

IIRC someone had started working on that, but I haven't heard any news
from it.


        Stefan

Re: current-column and display (space :align-to) properties

[Richard M. Stallman]

    Along the lines of a question that arose a couple of weeks ago re: counting 
    before and after properties when calculating current-column, should the 
    space :align-to and :width properties be reflected in current-column?

I think this is part of the to-do item to write a version
of current-column that takes account of pixel widths.

Re: current-column and display (space :align-to) properties

[Michael Mauger]

Richard M. Stallman writes:

> 
>     Along the lines of a question that arose a couple of weeks ago re: 
counting 
>     before and after properties when calculating current-column, should the 
>     space :align-to and :width properties be reflected in current-column?
> 
> I think this is part of the to-do item to write a version
> of current-column that takes account of pixel widths.
> 

Actually, on Dec 22nd, during a discussion with Drew titled "current-column and 
invisible text", you said:

> I think that is a bug.  I see two possible meaningful ways to handle
> an overlay before-string: either ignore it, or count its contents.
> The latter seems more correct.
>
> Would someone like to look at this bug, and ack?

I think a basic solution to this (report integer column number with no 
accounting for font widths) is needed now (before the release) since many of 
the display properties are new and cause inconsistent behavior in some basic 
functions.  

I've looked at the code, and understand where the changes are needed, but am 
not very comfortable with Emacs internals.  I'll take a look and see if I can 
puzzle my way thru but my time for it is limited.

-- Michael

Re: current-column and display (space :align-to) properties

[Kim F. Storm]

Michael Mauger <mmaug@yahoo.com> writes:

> I think a basic solution to this (report integer column number with no 
> accounting for font widths) is needed now (before the release) since many of 
> the display properties are new and cause inconsistent behavior in some basic 
> functions.  

The mentioned display properties are not new (they were introduced in
21.1), so the problem is not new either.

> I've looked at the code, and understand where the changes are needed, but am 
> not very comfortable with Emacs internals.  I'll take a look and see if I can 
> puzzle my way thru but my time for it is limited.

If you manage to fix current-column (actually current_column_1),
please remember to fix move-to-column in a similar way!

My best guess is that it must be fixed in a way which calls redisplay
internally to calculate the proper column.  Otherwise, there are just
too many things which can go wrong -- e.g. newlines with display
properties, etc.

Try this:

  (insert "["
	  (propertize "01\n23" 
		      'display '(space :align-to 10))
	  "]ZZ\n")


A good test would be 
M-: (move-to-column (current-column))

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: current-column and display (space :align-to) properties

[Richard M. Stallman]

    I think a basic solution to this (report integer column number with no 
    accounting for font widths) is needed now (before the release) since many of 
    the display properties are new and cause inconsistent behavior in some basic 
    functions.  

I think that failing to handle :align-to is less clearly a bug than
failing to handle a before-string.  But it is a matter of judgment.
Both are clearly flaws.

I think they are more or less part of a bigger flaw which is really
the lack of a feature.  So we want to fix this, but not necessarily
now.

    I've looked at the code, and understand where the changes are
    needed, but am not very comfortable with Emacs internals.  I'll
    take a look and see if I can puzzle my way thru but my time for it
    is limited.

Would someone who knows this code better like to help Michael?