Re: calculating durations

[stk]

> . . . I am currently writing a program that transforms scores expressed
> in a piano-roll-like format into LilyPond scores . . .

>  I'm especially thinking of how to resolve a note like "c3" (where 3
> means 3 quarter note beats) into Lilypond's "c2." . . .  in this case,
> the input data is mostly "clean" . . .

If the data isn't _all_ clean, then that will be your biggest headache.

You have received some suggestions that you write your program to convert
your source files into MIDI format; that's presumably a good idea if you
already know the syntax & semantics of MIDI files.

Is the unit of measurement of note durations in the source _always_
a quarter note?  If it is, that'll save you some trouble.

If you don't go the MIDI route and you decide to write a program to
convert directly to LilyPond, then I have one suggestion.  If you
choose to write a (calculational) algorithm to convert your source
"c3" to LilyPond "c2.", then you will spend a long time writing and
debugging the algorithm.  But if you look over your source files and
find that the actual number of different note-durations is small
(less than a dozen different durations, say), then it will be much easier
to just construct, by brute force, two string vectors --

S(1)="1"    LP(1)="4"
S(2)="2"    LP(2)="2"
S(3)="3"    LP(3)="2."
S(4)="4"    LP(4)="1"
S(5)="1/2"  LP(5)="8"    % I don't know source notation for short notes
S(6)="3/4"  LP(6)="8."   % ditto
etc.

so you could search the S( ) vector for each source-note duration; if S(3)
matched the source duration, then LP(3) would give the LilyPond duration.

This of course does not solve the problem that probably some of your
source notes will have durations exceeding 4 (i.e., will represent notes
longer than a whole note).  If the source duration SD of a note is greater
than 4, you will need to calculate

SD div 4 = the number of whole notes you will need to tie together
SD mod 4 = the source-duration of the additional note you will have
           to tie onto the preceding string of tied whole notes;
           you will have to find n such that S(n)=SD mod 4 to obtain
           LP(n), where LP(n) is the LilyPond duration of this final
           tied note.

And as others have pointed out, you will have to use the
Completion_Heads_Engraver  (user manual section 6.2.6).

There are other complications:
(a) Tuplets:  modifying the procedure I sketched above to handle tuplets
    would be very hard.  *And* the Completion_Head_Engraver apparently (?)
    is not able to handle tuplets anyway.
(b) The Completion_Head_Engraver can't split rests at all, according
    to the user manual; that may be a fatal problem.

Now that I look at it, that's all pretty awful, isn't it?  Probably you
would be better off learning how to construct MIDI files (something I've
never done...).

-- Tom