Re: Feature request

[Werner LEMBERG]

>>>   \beaming {8[ 8] 8 8 8[ 8 8 8] }
>>
>> Uh no.  You need the rest, or you need to write something like
>> \beaming {4 8[ 8] 8[ 8 8 8]}
>
> Don't think so.  With this command, it would seem that the rule is
> "delete all previous autobeaming rules and only autobeam in the
> pattern set with manual beams here".

Exactly.  To be more precise, the entries with the largest duration
*must* have beaming information, this is, either starting or ending a
beam, or be enclosed in a beam.  In other words, this

  \beaming 4/4 {4 8[ 8] 8[ 8 8 8]}

would be invalid because there is no information how to beam the first
quarter, but this

  \beaming 4/4 {8[ 4 8] 8[ 8 8 8]}

would be valid.

>> And where the overrides can be used, the meter might not be known
>> yet, so one needs to write something like
>> \beaming 4/4 { ... }
>> (somewhat redundant) and/or require that the argument consists of
>> whole bars separated by bar checks.
>
> Again, with the suggestion above, this is not needed.  This part of
> the rule is that the bracketed note lengths must form a complete bar
> in the current time signature.  If not, an error is thrown.

Here, I go with David.  Just think of notation used sometimes by
Richard Strauss:

  5 4
  4 4

which means that the first bar of a two-bar phrase holds 5/4, and the
second one 4/4 (cf. the `aria' of Octavian in the first act of
`Rosenkavalier').  To handle this, we need two beaming instructions:

  \beaming 5/4 {...}
  \beaming 4/4 {...}

to redefine the beaming structure for the two bar types.  I think it
looks quite good in general to be able to define beaming structures
like this.

>> It is probably easiest to require the time specification always (it
>> is needed for setting baseMoment and figuring out measureLength):
>> that way the chance for users to mess up once they have LilyPond
>> stop complaining is minimized.

I fully agree.


    Werner