Re: Stem direction with ties

[Per Starbäck]

I wrote:
>> With
>>
>> \version "2.16.2"
>>
>> \relative c'' {
>>   b~ b8 a b4~ b8 c
>>   a8 b~ b4 c8 b~ b4
>> }
>>
>> the stems of every b goes downwards. I think it would be better if tied
>> notes had the same direction, to make it clearer visually that they belong
>> together. Then two of these stems should go up instead.

David Kastrup replied:
> 
> Well, how would you typeset
> 
> \relative c'' {
>   e8[ c]~ c a~ a[ f]~ f e
> }
> 
> with that rule?

With all stems up. That looks better to me at least.

> This kind of "look at a larger context and choose a global optimum"
> decision is pretty hard to do using LilyPond's current architecture.

I know almost nothing about the architecture, but if beam handling doesn't
do at least some of that it is faking it well. I'd like your examples to
have all stems up because

\relative c'' {
  d8[ b b g g e e d]
}

would have all stems up. Maybe this doesn't fit in nicely with the order
things are done in now, but somewhat more strictly my suggestion is to first
treat ties as beams when determining stem orientation, so no big new
backtrack for global optimum feature.

Can anyone think of an example where that would give worse results?

(To Phil Holmes: Thanks for the tip! I had only read about up and down as
possible values for neutral-direction.)