Re: Proposed new available and recommended behavior of \relative

[David Kastrup]

Jim Long <address@hidden> writes:

> I think 'recommended' is going too far.  At least, I can't see
> that one is always or nearly always better than the other.  David
> made some examples of when the proposal could be better,

Not really.  I just went into the Learning manual (intended for
beginners) without looking much and picked out several examples where
the choice of reference pitches was not helpful for the user.

Now one could probably go over all uses of \relative manually and try
improving this situation by picking sets of reference pitches that
better complement one another.  There are over 6000 instances in our
documentation.  Who is going to do that?

> but I don't know that anyone has tried to make counter-examples of
> when an explicit reference pitch would be "better".

The thing is that the non-explicit reference pitch is a consistent and
non-arbitrary straightforward choice.  Consistency is an advantage in
itself since it saves you thinking about unexpected situations.

Can you think of a situation where an explicit reference pitch could be
better, short of the special case \relative x''' { x where one does not
need to think about the relation between reference pitch and starting
pitch in the scale?

Is it better enough to make it worth the discipline for using it
consistently?  It is, obviously, not hard to change the convert-ly rule
now converting to \relative f to one which does that instead.  So if one
can agree that this is always the best, we can go there automatically.

But if it is always the best, why have to write it so complex?

>> 1) possibly change the octave of the first pitch inside { ... } 2)
>> possibly add an explicit reference pitch.
>> 
>> I'm not sure which is better, and maybe this is getting ahead of
>> ourselves.
>
> You seem to imply or at least I'm infering that you believe
> having convert-ly convert:
>
> \relative d'' { c d e f g a b c }
>
> to:
>
> \relative { c'' d e f g a b c }
>
> ought to be eliminated from consideration.  I agree with that,
> but I'm not sure that such a convert-ly rule is beyond
> possibility.

The discussion might be running less in circles if people actually
bothered looking at the proposed patch for issue 2329.

It presents where an automated conversion can go.  So please take it
unto yourself to compare, say, some passages of the Learning manual
before and after conversion and see whether you consider the net amount
of changes rather helpful or rather detrimental.

> I would rather not have convert-ly change any use of \relative with an
> explicit reference pitch.

For user code without asking explicitly, maybe.

-- 
David Kastrup