At which musical position starts and ends a part?

[lemzwerg]

In the documentation for <measure> you can find the following for the ‘number’ attribute.

“The attribute that identifies the measure. [...] In partwise files, it should be the same for measures in different parts that share the same left barline.”

Let's assume I have an input file like the following.

<part id="P1">
  <measure number="1"> ... </measure>
</part>
<part id="P2">
  <measure number="2"> ... </measure>
</part>
<part id="P3">
  <measure number="3"> ... </measure>
</part>
<part id="P4">
  <measure number="4"> ... </measure>
</part>

There are four parts, with one measure each. However, the <measure> elements have different IDs. What is expected to happen?

Testing with both MuseScore and Finale, I get something like the following

without any warnings. But is this really the intended output? Shouldn't it be rather looking like this?

Where exactly is it documented that all parts start at the same musical position?

[lemzwerg]

A similar situation arises if one part finishes earlier than another. Let's assume that ‘P1’ has one measure and ‘P2’ consists of two measures. What exactly should be the result? MuseScore emits the error message

“Incomplete measure: Full score, measure 2, staff 1. Found: 0/1. Expected: 4/4.“

for such an input file and shows the following.

On the other hand, Finale accepts it without warning and displays

ignoring the second measure. I could also imagine the following output.

Where exactly is it documented that all parts must end at the same musical position?

[infojunkie]

To me, these 2 scenarios are edge cases that are not handled by applications because they almost never come up in real life. Can you point to actual examples of scores where these scenarios occur? Otherwise, there would be little motivation for software maintainers to handle them.

For what it's worth, I do agree with your expected renderings of both cases on a purely common-sense basis.

[lemzwerg]

I don't have real scores. However, those scenarios aren't purely hyphothetical, at least not the second one – just imagine that you programmatically construct a score from two parts in two different files, and that one part is shorter than the other, probably due to a missed measure during a copy-and-paste operation.

What is relevant, however, is that – as far as I can see – there is no mentioning at all in the MusicXML specification how parts are synchronized time-wise. I think that we should clarify at which musical moment a part starts (always at moment zero), that measures are following sequentially, without any gaps (irrespective of their IDs), and that all parts should have the same length (being padded with one or more invisible measures on an invisible staff, if necessary).

[samuelbradshaw]

Would these songs be examples of parts starting mid-song?
Example 1 – piano accompaniment separates from SATB starting in system 3
Example 2 – obbligato starts in system 5

[lemzwerg]

Would these songs be examples of parts starting mid-song?

I don't think so. Imagine that you write a part for the piano accompaniment in Example 1. It would start with a multi-measure rest for eight bars. This implies that the full score as shown in the image is actually a 'frenched' score, that is, it suppresses the display of staves containing only rests for the piano.

What comes to my mind is rather an 'ossia' staff, as shown in this example from the first aria of 'La clemenza di Tito' (above the word 'indegno'):

It would be nice if this could be encoded in MusicXML as a very short part starting directly at the given measure ID, but...

[I know that MusicXML provides the <staves> element for handling ossias – at least I guess this is the intended way to encode it; I haven't tried.]

[lemzwerg]

Maybe the previous ossia example is not the best one. Here is another one, taken from Rachmaninoff's third piano concerto, which provides an alternative cadenza. This cannot be encoded with <staves>, AFAICS.