Batch mode produces voiceless notes

[olin256]

The following file was processed with -batch argument 505-3_small.png with the resulting file 505-3_small.zip

Now the extra quarter note in measure 7 somehow got "voiceless" and is not present in the export. When this image is imported in the GUI, it is assigned to voice 2.

Perhaps this is not a bug and there is some logic behind this behaviour, but I couldn't find this in the documentation…

(Why there are five quarters at all: There was a triplet sign, but it got "accidentally" deleted in the preprocessing of the images.)

[hbitteur]

In the MusicXML export, notes within the same measure are exported voice after voice. So, if a note has no related voice, it won't get exported.

The question is why the fifth quarter wasn't assigned any voice. This is certainly related to the 4/4 time signature (four quarters). Note the measure rhythm is flagged as abnormal (painted with a pink background).

Here is the strip for this measure:

MeasureStack#7
    |0       |1/4     |1/2     |3/4     |1
--- P1
V 1 |Ch#480  |Ch#481  |Ch#482  |Ch#483  |1
V 2 |........|........|........|........|0

Note that voice N°2 exists but is empty! Further investigation is needed...

[hbitteur]

Displayed measure N°7:

Re-processing rhythm on this measure gives:

Measure{#7P1} No timeOffset for HeadChordInter#484{(0.947/0.947) stf:2 slot#5 dur:1/4}

[hbitteur]

Here is what happened during the processing of measure rhythm: 5 slots were detected, each containing a quarter note. For slot 1 (first in measure), the time offset was set to 0 (start of the measure) and the voice N°1 was created. For slots 2, 3 and 4, the quarter note of the slot was time-compatible with the previous one, still in voice N°1.

Fot slot 5 however, the quarter note was not found time-compatible with the previous one, because this previous note ended at measure end. Hence there was no time left in voice N°1. A new voice (N°2) was created for this quarter note. But its time-offset since measure start could not be computed. Hence the warning message and the measure being set as abnormal.

From the engine point of view, that's all it could do! [And this has nothing to do with the fact that the engine runs in batch or interactive mode].

[hbitteur]

To be precise, the last chord (N°484) was assigned to voice N°2 (and indeed displayed in green, the color for voice N°2). However its time-offset could not be set (thus impeding any export). And the slot N°5 itself was discarded because its time-offset could not be computed.

[olin256]

Thanks for the investigation, but I'm afraid I'm simply witnessing something else in the interactive mode—see this: audi.gif This is the resulting file: 505-3_small_gui.zip

EDIT: Sorry, I just re-opened the OMR file above and now I see that actually the note is "voiceless" there as well. Then perhaps we're talking about a different bug: The note is viewed as green (i.e. but belonging voice 2), but this is in fact not the case, as dumping measure voices shows.

EDIT2: Let me rephrase this from a very "high-level user" point of view: I've got a green note, but if I save the file and reopen it, it's no longer green…