Allows seconds to merge

Allows seconds to merge

[dak, 2012-03-04 14:15:52]

http://codereview.appspot.com/5729057/diff/1/lily/note-collision.cc
File lily/note-collision.cc (right):

http://codereview.appspot.com/5729057/diff/1/lily/note-collision.cc#newcode131
lily/note-collision.cc:131: min_downs = min (min_ups, dps[i][X_AXIS]);
min (min_ups, ...
Seriously?

And it is nonsensical to compare the mins on either side, since for the
stem-down configuration, the max is the part to merge.

http://codereview.appspot.com/5729057/diff/1/lily/note-collision.cc#newcode139
lily/note-collision.cc:139: }
This assumes that only mins/maxes need to be considered when merging when in
reality it is every pairing that has a chance to overlap.  It is just that the
min/max pairing with the smallest distance determines which stem distance has a
chance of succeeding, and once one has determined that, one needs to do a
merge-like run through both columns to make sure that at this stem distance, all
note heads either merge to identical positions, or don't touch one another.

[Keith, 2012-03-04 19:37:24]

Mike I begin to suspect that you employ an army of monkeys on keyboards.  This
code works by accident.

Also, change the comment on ‘collision-seconds.ly’ to reflect the new behavior.

http://codereview.appspot.com/5729057/diff/1/lily/include/stem.hh
File lily/include/stem.hh (right):

http://codereview.appspot.com/5729057/diff/1/lily/include/stem.hh#newcode32
lily/include/stem.hh:32: static vector<Offset> note_head_x_y_positions (Grob
*me);
Better to return x-positions separately.

So far you do not use the Offset as a 2-D position, but rather use only
individual x- y-components, and probably we should continue to think of the
coordinates separately.

Don't combine x and y into an Offset here, because the y position is in
half-staff-spaces and the x-position is in staff-spaces.

http://codereview.appspot.com/5729057/diff/1/lily/note-collision.cc
File lily/note-collision.cc (right):

http://codereview.appspot.com/5729057/diff/1/lily/note-collision.cc#newcode135
lily/note-collision.cc:135: if (ups[i][X_AXIS] != min_ups)
  if (ups_X[i] != 0.0);
It just happens that min_ups and min_downs always end up being 0.0, which is
what you really wanted anyway.

Probably better would be an array telling which heads are on the unusual side:
  if (ups_is_suspended[i])

http://codereview.appspot.com/5729057/diff/1/lily/note-collision.cc#newcode140
lily/note-collision.cc:140: if (dps[j][X_AXIS] != min_downs)
if (dps_X[i] != 0.0);

http://codereview.appspot.com/5729057/diff/1/lily/stem.cc
File lily/stem.cc (right):

http://codereview.appspot.com/5729057/diff/1/lily/stem.cc#newcode264
lily/stem.cc:264: /* The positions, in ascending order.  */
I see you need the y-component so you can sort these x-components properly, but
it complicates things to make a 2D offset.

Maybe return two separate arrays: the integer staff-positions and boolean flags
marking the heads that are on the unusual side of stem.

[mike_apollinemike.com, 2012-03-04 20:13:10]

On Mar 4, 2012, at 8:37 PM, k-ohara5a5a@oco.net wrote:

> Mike I begin to suspect that you employ an army of monkeys on keyboards.
> This code works by accident.
> 

I obviously don't pay them enough :)

I just took a stab at it to see if I could get it done quickly: David has a few
solid criticisms of the approach's fitness for real music, and if it works by
accident as you say (I was following the TODO), then I think he'll be able to
spend the time to make it work on purpose.  I unfortunately have very little
time over the next couple weeks, and I think he can do a good job bringing it to
push stage.

Cheers,
MS

[Keith, 2012-03-04 21:29:32]

On Sun, 04 Mar 2012 12:12:57 -0800, mike@apollinemike.com
<mike@apollinemike.com> wrote:

> I just took a stab at it to see if I could get it done quickly [...]
> if it works by accident as you say (I was following the TODO),

You misunderstand what you, or your army of monkeys, did.

The TODO suggested filtering out heads not in the main note-column, which would
be those with non-zero x-coordinate relative to the note-column.   Your army
created code to compute min_ups and min_downs, and consider only heads at those
positions.  If min_* were actually the minimums, this would do something
different from what the TODO suggested.
--
Keith