Re: Issue 687 in lilypond: Enhancement: inequal MIDI quantization of equal durations (swing, rubato)

[lilypond]

Comment #9 on issue 687 by chicagogrooves: Enhancement: inequal MIDI  
quantization of equal durations (swing, rubato)
http://code.google.com/p/lilypond/issues/detail?id=687

adam.spiers - thank you for your valuable thoughts. I like the  
displacement-list concept you presented, but I think the note boundary  
issues are a red herring, as I think
of swing as a belt-driven turntable with a misshapen cam =)  Still I think  
with a couple tweaks, I can accomodate those scenarios and have an  
expansion point for user tweaks. I would:

1. Define a displacement map, of - say, a quarter note duration. A  
displacement function accepts the position of an event, and the map, and  
returns a new position (which may or may not modify the event, as explained  
by swing_factor section below)

    Example 1: map ( (0 0) (99 99) ) - this is interpreted as a straight  
linear function across 100 (arbitrary) subdivisions of a quarter note. 25  
gets mapped to 25, etc..
    Example 2: map ( (0 0) (50 66) (99 99) ) - this maps 25 to 33, 50 to  
66, and 99 to 99 - its eigth note swing in a 'continuous distortion' model.
    Example 3: map ( (0 -5)  (50 61) (99 95) ) - this is swing, but behind  
the beat by 5 units out of a hundred.

    Maps accomodate arbitrary distortions quite well, you could get as  
crazy as you like with yours !

2. A displacement list like yours adam.spiers can be used to set up the  
map. Common maps can be given names, and invoked as \swing_8 or similar  
variables.

3. Sometimes you will want the output of the displacement map to modify the  
position less than all the way, or none at all. The key to this is the  
choice of various swing_factor functions,
    which for each event, return how much to listen to the map as a number  
from 0 to 1.

    Example 1: a simple implementation always returns 1 - the map  
continuously distorts every event like the misshapen cam. This would be  
enough for me :)
    Example 2: a swing factor function sees that swing is explictly turned  
off in the context during some events - the swing factor returned is 0 for  
each of those events.
    Example 3: a swing factor function returns 0 unless NoteOn and NoteOff  
are both aligned to 'kink' points in the map, 1 otherwise.
    Example 4: a swing factor function cancels swing for events of a  
certain type
    Example 5: a swing factor function scales between 1 and 0 between two  
tempo points.

    etc.. it'd be fun to combine these too - I'd have to think about how  
the user would specify this.