Add dead-is-alive boolean property to Hara_kiri_group_spanner

Add dead-is-alive boolean property to Hara_kiri_group_spanner

This reverses the decisions that the Hara_kiri_engraver makes.
Particularly in connection with the Keep_alive_together_engraver, this
allows for per-line content-based switching between alternative staff
displays.

[lemzwerg, 2012-12-15 10:48:12]

LGTM.  Do we have a regtest (or rather, an example) which demonstrates how to
use it?

[dak, 2012-12-15 11:40:17]

On 2012/12/15 10:48:12, lemzwerg wrote:
> LGTM.  Do we have a regtest (or rather, an example) which demonstrates how to
> use it?

I don't actually have one.  Since this was inspired by a discussion on the
mailing list, I was sort of hoping that Keith would be able to make use of this
for improving his example.

[Keith, 2012-12-15 18:22:35]

https://codereview.appspot.com/6948058/diff/1/lily/hara-kiri-group-spanner.cc
File lily/hara-kiri-group-spanner.cc (right):

https://codereview.appspot.com/6948058/diff/1/lily/hara-kiri-group-spanner.cc...
lily/hara-kiri-group-spanner.cc:80: bool in_Hades = to_boolean (me->get_property
("dead-is-alive"));
The flag, used this way, could be named "censor"

[Keith, 2012-12-15 18:22:46]

On 2012/12/15 11:40:17, dak wrote:
> 
> I was sort of hoping that Keith would be able to make use of this

While the negative logic makes it not negligibly difficult to avoid errors in
understanding, I see no good way to use this.

If the usual method is:
"If this Staff has anything interesting to say on this line, or if his friends
in his keep-alive-together group do, then print him."

this gives an option to reverse the action
"If a Staff with the 'censor' flag set has anything interesting to say on this
line, or if his friends in his keep-alive-together group do, then omit him."

So while this would give users a way to have a combined staff disappear on
exactly those lines when divisi staves are printed, it interferes with the usual
function of removeEmptyStaves on the combined staff.   We would need to declare
that nothing the combined staff says is ever 'interesting', so there is no
longer any way to LilyPond to remove the combined staff when the part is truly
empty.

A re-arrangement in logic could give a useful option 
"If a Staff with the 'alive_together_alternative' flag set has anything
interesting to say on this line, but /none/ of his (un-flagged) friends in his
keep-alive-together group do, then print him."

[dak, 2012-12-15 21:32:41]

On 2012/12/15 18:22:46, Keith wrote:

[...]

> A re-arrangement in logic could give a useful option 
> "If a Staff with the 'alive_together_alternative' flag set has anything
> interesting to say on this line, but /none/ of his (un-flagged) friends in his
> keep-alive-together group do, then print him."

Keith, this is all pretty much over my head.  The "dead-is-alive"-thing had the
advantage that even though I did not understand what I am working with, it was
easy to make it behave as advertised.

What you are saying now is that it is not really useful.  Bummer.  Of course, if
that is the case, we don't want the added complexity, even though it is not
overly complex.

Can you try coming up with some flag or flag combination that would be more
useful without being considerably more complex?  You have a better feeling for
the mechanisms and the original problem, I guess.

[Keith, 2012-12-16 04:15:41]

On 2012/12/15 21:32:41, dak wrote:
> The "dead-is-alive"-thing had the
> advantage that even though I did not understand what I am 
> working with, it was easy to make it behave as advertised.

"This reverses the decisions that the Hara_kiri_engraver makes." 

This engraver removes empty staves, so reversing its decisions results in
printing staves only when empty.

> What you are saying now is that it is not really useful.  

Maybe some perverse re-definition of 'empty' in user input could profitably use
the dead-is-alive behavior, but I could not find a way to do so.

> Can you try coming up with some flag or flag combination that 
> would be more useful without being considerably more complex?  

I was thinking of something like
  https://codereview.appspot.com/6945064/
but I do not think it the added property.

[dak, 2012-12-16 08:09:38]

k-ohara5a5a@oco.net writes:

> On 2012/12/15 21:32:41, dak wrote:
>> The "dead-is-alive"-thing had the
>> advantage that even though I did not understand what I am
>> working with, it was easy to make it behave as advertised.
>
> "This reverses the decisions that the Hara_kiri_engraver makes."
>
> This engraver removes empty staves, so reversing its decisions results
> in printing staves only when empty.
>
>> What you are saying now is that it is not really useful.
>
> Maybe some perverse re-definition of 'empty' in user input could
> profitably use the dead-is-alive behavior, but I could not find a way
> to do so.

Well, the idea was to use this in pairs of contexts with otherwise
identical harakiri conditions, and give only one of them the
dead-is-alive flag.  So one context pops up only when the other doesn't.

I had the idea of using this with the Keep_alive_together engraver in
order to make the conditions identical by force (and that is mentioned
in the doc strings), but that is not going to work since the
Keep_alive_together engraver will, if it decides to kill, just kill both
contexts without asking again.

But if contents and keepAliveInterfaces are essentially identical, the scheme
should be able to work, shouldn't it?

[Keith, 2012-12-17 06:13:22]

On 2012/12/16 08:09:38, dak wrote:
> Well, the idea was to use this in pairs of contexts with otherwise
> identical harakiri conditions, and give only one of them the
> dead-is-alive flag.  So one context pops up only when the other 
> doesn't.
> 
> I had the idea of using this with the Keep_alive_together engraver > in order
to make the conditions identical by force (and that is 
> mentioned in the doc strings), but [...]

That works fine.  The input below will show a single staff, except for the one
line containing the voice-crossing section, regardless of where the line-breaks
fall. (We can do this with current LilyPond using two Staff contexts; the new
feature enables the measure that could be printed either as chords or divided
staves, depending on linebreaks.)

The problems with 'dead-is-alive' are the difficulty of keeping track of the
reversed logic, and loss of the usual action of removeEmptyStaves when all
affected staves are truly empty.

\paper {
  #(set-paper-size "a6")
  indent = 1.0\cm
  short-indent = 1.2\cm
}

violinII = \relative c' {
  c4 d e f g f e d
  <c e>-"divisi" <d f> <e g> <f a>
  s1  % this measure is too complex to show on a single staff
  c4-"unison" d e f g f e d
  c d e f c1
  \bar "|."}

letStaffVanish = \set Staff.keepAliveInterfaces = #'()
showStaff = \unset Staff.keepAliveInterfaces

violinIIa = \relative c' {
  \letStaffVanish \quoteDuring "violinII" s1*2
  e4 f g a
  \showStaff b4 g d g
  \letStaffVanish \quoteDuring "violinII" s1*4
}

violinIIb = \relative c' {
  \letStaffVanish \quoteDuring #"violinII" s1*2
  c4 d e f
  \showStaff g4 b d b
  \letStaffVanish \quoteDuring #"violinII" s1*4 }

\addQuote "violinII" \violinII

\new StaffGroup <<
  \new Staff \with {
    instrumentName = "Vn 1"
    shortInstrumentName = "Vn1"
  } \repeat unfold 8 c''1
  \new GrandStaff \with {
    \consists Keep_alive_together_engraver
    instrumentName = "Vn 2"
    shortInstrumentName = "Vn2"
  } <<
    \new Staff \with {
      \override VerticalAxisGroup #'remove-first = ##t
      shortInstrumentName="outer"
    } \new Voice \violinIIa
    \new Staff \with {
      \override VerticalAxisGroup #'dead-is-alive = ##t
      \override VerticalAxisGroup #'remove-first = ##t
      keepAliveInterfaces = #'()
    } \violinII
    \new Staff \with {
      \override VerticalAxisGroup #'remove-first = ##t
      shortInstrumentName="inner"
    } \new Voice \violinIIb >> >>
\layout { \context {
        \Staff \RemoveEmptyStaves }}

[dak, 2012-12-17 08:22:51]

On 2012/12/17 06:13:22, Keith wrote:
> On 2012/12/16 08:09:38, dak wrote:

> That works fine.  The input below will show a single staff, except for the one
> line containing the voice-crossing section, regardless of where the
line-breaks
> fall. (We can do this with current LilyPond using two Staff contexts; the new
> feature enables the measure that could be printed either as chords or divided
> staves, depending on linebreaks.)
> 
> The problems with 'dead-is-alive' are the difficulty of keeping track of the
> reversed logic, and loss of the usual action of removeEmptyStaves when all
> affected staves are truly empty.

Ok, so we should likely not be abusing the keepAliveInterfaces as
keepDeadInterfaces as it may be separately needed.  Instead we likely want
something like a killContexts list which names a number of contexts to kill if
this context is not suicided (or already killed) itself.