Re: iPhone app using Lilypond + potential contributions

[Reinhold Kainhofer]

Hi Daniel,

I'm currently the main developer of musicxml2ly (although I don't have much 
time left for it and mostly work on other stuff in LilyPond and my own music 
publishing house).

Am Freitag, 19. März 2010 01:14:36 schrieb Daniel Grover:
> 2) MusicXML converter improvements
> Another area I would love to help out Lilypond with is the MusicXML
> converter. We're in talks with music publishers to make copyrighted music
> available for purchase in the app. A lot of their music is in Sibelius
> format, but there's a plugin called Dolet that seems to do a great job
> converting to mxml.
> 
> The MusicXML->Lilypond converter, however, has problems. Like it seems to
> make duplicates of any dynamic marks used (or even text). 

Yes, that's a design problem between MusicXML and LilyPond: In MusicXML markup 
is assigned to a staff, while in LilyPond it is part of a voice (i.e. assigned 
to one particular note).
Now imagine a staff with two voices (e.g. two instruments sharing one staff, 
or a different rhythm for two voices in a piano staff, these have a totally 
different meaning, but exactly the same notation!) and a text markup.

In MusicXML it is assigned to measure N at position P. However, in LilyPond we 
need to assign it to a note. In the case of two voices in a piano staff it 
would be most appropriate to simply assign the markup to the next note only.
However, when two instruments (e.g. Oboe 1 and Oboe 2) share a staff, probably 
both instruments should get the markup...

For now, I chose the latter option, since removing things is much easier than 
adding them.
However, of course there is lots of room for improvement (e.g. if the markup 
is in a measure where only one voice really occurs, don't add the markup to 
the other voices).

The main obstacle here is that the assignment of xml nodes to voices is done 
relatively early, when there is not much information about the other contents 
of the staff available.

> And it has
> problems with overlapping slurs marks scores (even if it's just a tie mark
> overlapped with a phrasing slur).

Are you sure the tie is really exported to the xml file as a tie? I could 
imagine that it is wrongly exported as a slur and since lilypond doesn't 
handle overlapping slurs, of course, this will lead to problems.

A small sample file would be best to diagnose the problem.


> And other files I've tried just result
> in invalid .ly markup with no warnings as to why.

Sample XML files would help a lot! What is the invalid .ly markup?


> What I'd like to do is add a bunch of regression tests to the existing test
> suite for musicxml2ly, 

Sure, feel free. However, the test cases should really be stripped down to the 
absolute minimum. You can also send me the files and I'll add them (after 
stripping them).

For now, I chose the MIT license for the test suite (i.e. only attribution via 
license and copyright statement is needed, but it can be used for any 
purpose), but I might change my mind and .

Cheers,
Reinhold

-- 
------------------------------------------------------------------
Reinhold Kainhofer, address@hidden, http://reinhold.kainhofer.com/
 * Financial & Actuarial Math., Vienna Univ. of Technology, Austria
 * http://www.fam.tuwien.ac.at/, DVR: 0005886
 * LilyPond, Music typesetting, http://www.lilypond.org