Bad slur shape

• May 23, 2019 - 17:53
Reported version
3.0
Priority
P2 - Medium
Type
Functional
Frequency
Once
Severity
S4 - Minor
Reproducibility
Always
Status
active
Regression
No
Workaround
Yes
Project

Enter measure 224 for the second violin of the attached score.

This is an ugly slur that could be improved. It has too much arch.

Not in this score, but in some scores this causes staves to be farther apart than necessary. Add a dim. line to the first violin's 224 measure to see an example.

Attachment Size
SINFONIA VII.mscz 135.54 KB

Comments

Priority P2 - Medium

So, this was a deliberate change, in 3.0.5 what we got was a slur whose endpoints are too far from the note. Is that better? To me, no. There are some cases the old algorithm was clearly bad, no doubt cases where it the current might be seen as not as good, but I still think the new is better overall. Neither is necessarily ideal, and there is no getting around the need to manually adjust in some cases.

One assumption we've been going on thus far is that the slur should be symmetric, we can play with the endpoints and the "tilt" but it's always symmetric by default. Playing with non-symmetric shapes could produce a shape some would find better, but even so others would probably object.

For the record, here's the specific PR that changed this, complete with images showing the cases I was trying to optimize:

https://github.com/musescore/MuseScore/pull/4750

The second commit of the PR in particular - and the second measure in the image comparison - is the one relevant here. You may recall discussing this on Telegram. It's certainly possible to tweak parameters of the algorithm further. It works by trying a simple flattish slur first as per 3.0.5, seeing how far the endpoints are from the notes, and if it's too far, curving the slur more and trying again, but eventually giving up before it gets stupid. So the parameters are, how far is too far from the endpoints, how much more to curve at each iteration, and when to give up.

Also a regression in 3.1RC vs 3.0.5?
I wouldn't say this is a must fix for the final 3.1. It's different than 3.0.5 I'm sure but it's not worthy of delaying 3.1's release.

I do remember a discussion on telegram about slurs. This one caught my eye as a specific case of a slur that needs some work. It really caught my eye before I put the system break in that measure and there were about 3 more empty measures at the end of the system. So I continued to the end of the page and saw it hadn't changed to a reasonable shape.

The basic problem with this one is that the middle looks like it's unnecessarily growing up. Perhaps in reality, it's symmetric with one end raised much higher giving the illusion that the center has risen.

I did the following to make a decent looking slur:

Move the left anchor down 3 times (using ctrl+arrow)
Move the right anchor down 1 time
Move the center top anchor down 3 times

It now looks like this
adjusted slur.PNG

I'll admit this is not perfect, but it looks far more reasonable than before I adjusted it.

Perhaps this will give someone an idea as to how they can approach an improvement.

I agree this looks better. What you have done here is basically rotated the slur so it's a bit more horizontal than the default, and then allowed the end point of the slur to be higher above the end note than we'd normally prefer.

How to implement such an approach - and to predict when it will be effective - is an interesting question.