Re: [Lilypond-auto] Issue 3902 in lilypond: the position of the initial \voiceOne rest is suboptimal

[lilypond]

Comment #19 on issue 3902 by address@hidden: the position of  
the initial \voiceOne rest is suboptimal
http://code.google.com/p/lilypond/issues/detail?id=3902

So here's 2 further sets of comparison rests.  We start with the music used  
by Gould on page 36.  In bar 1, the old version is clearly better.   
However, this is Issue 3043: if the note was an e rather than an a, both  
would look awful, so the fact that the new one looks worse is simply a  
quirk: moving the default position outside the staff would then fix both,  
but be awful elsewhere.  Looking at bar 2: the new version conforms to  
Gould's recommendation.  The old one doesn't and looks much worse, to my  
eye.  The same is true of bar 3 and one of the rests in bar 4: the other is  
positioned using melody following.

The second line of rests shows 2 things: r8 and r16 don't actually collide  
with the new version, but are close.  To my eye they're fine but I  
appreciate others may disagree.  It also illustrates another issue: rests  
move readily to avoid close proximity to notes, but less readily for other  
rests.  I believe the correct fix here would be to improve rest collision  
avoidance whilst maintaining the new default positioning from this fix.   
The same is true for the longas in bar 7.

While experimenting with these, I looked at 2 minimum-distance overrides:  
Rest.minimum-distance and RestCollision.minimum-distance.  AFAICS from both  
code inspection and experiment, the first has no effect whatever - it's  
documented but not implemented.  Seems like a bug.  The second affects the  
distance between rests and notes as shown in bar 5.

Finally the midi-notes retest.  As David says, it's quite difficult to  
follow the voice line as shown in the new version in 8 and 9.  Howeever,  
with a slightly different set of notes, it's no better with the old  
version: it again is simply seredipity that the old version is slightly  
better.  It's likely that extreme crossing of notes and rests like this  
will either need an automatic melody following solution (hard) or pitched  
rests (easy).

My conclusion:

* my proposed fix provides correct default positioning, as shown in Gould  
page 36 and in the blog post identified earlier.
* There is a bug in the way rest vertical spacing is handled: they are  
allowed to get too close anyway: look at the 3rd image, from currect  
2.19.6: this has the same issue over closeness of rests.
* There is a bug in how minimum-distance is handled.
* There is a bug over ledger line poisioning for rests >= minim outside the  
staff.

Attachments:
        3902Old.png  21.3 KB
        3902New.png  21.2 KB
        RestPos4.png  1.5 KB

--
You received this message because this project is configured to send all  
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings