Re: Start, end, gradient of Glissando

[Aaron Hill]

On 2021-10-24 6:22 am, Thomas Morley wrote:
> Though, how could I have found it myself?
> My lack of C++-knowledge hinders me to understand the "logic from
> Line_spanner::print". I usually learn things from looking at
> scheme-codings.

Not sure it would have been clear without digging into the print 
function.  For reference, here are the relevant lines:

////
  Interval normalized_endpoints = from_scm (get_property (me, 
"normalized-endpoints"), Interval (0, 1));
  Real y_length = span_points[RIGHT][Y_AXIS] - 
span_points[LEFT][Y_AXIS];

  span_points[LEFT][Y_AXIS] += normalized_endpoints[LEFT] * y_length;
  span_points[RIGHT][Y_AXIS] -= (1 - normalized_endpoints[RIGHT]) * 
y_length;
////


> But there is none in our source using 'normalized-endpoints, apart
> from setting it in some grob-definitions.
> The description reads: "Represents left and right placement over the
> _total_ spanner, where the _width_ of the spanner is normalized
> between 0 and 1."
> I think the wording is misleading, at least "total" means the unbroken
> spanner for me, thus I never considered to look at it for broken
> spanners.
> The word "width" means the extent in X-axis direction, again no reason
> to look at it for extents in Y-axis.

You are right that "width" is talking about the X extent, but "total" 
does not just mean an unbroken spanner but the entire spanner, broken or 
otherwise.  The resulting normalized values are in essence percentages 
of the total width if you were to lay each broken element end-to-end.  
For example, if a spanner is broken into two parts of equal size, then 
you should see normalized-endpoints as the intervals [0, 0.5] and [0.5, 
1].


> P.S.
> While the Y-values are fixed now, I found some flaw, if left/right
> padding is not zero. At least for broken or cross-staff Glissando.
> I'll continue research..., unless someone has another hint?

Yup, there's more code in Line_spanner::print that factors in the 
padding values.  It is perhaps harder to locate because the variable was 
named "gaps".  And it turns out there is more than just padding that 
will affect your calculation.  Try setting a stencil in either 
bound-details, as that will also shorten the effective spanner line.  
Arrows, however, seem to overlay on the line without changing its 
length.

Let me take a stab at converting the print routine to Scheme.


-- Aaron Hill