Geometry coming out incorrect on first attempt

[Tristan Croll]

This is something new (code has been working perfectly for months). The applied code that is coming out wrong is my rotamer annotation (the whole drawing gets hidden inside the "dot" of the exclamation mark), but the below is a minimal example.

def exclamation_mark(radius = 0.1, height = 2, nc=8, color = [255,0,0,255]):
    '''
    An exclamation mark for annotating rotamer outliers
    '''
    from chimerax.surface.shapes import cone_geometry, sphere_geometry2
    from chimerax.core.geometry import translation, scale
    import numpy
    stem = cone_geometry(radius=radius, height=height, nc=nc, caps=True)
    spheres = list(sphere_geometry2(nc*4))
    spheres[0] = scale(radius*0.7).moved(spheres[0])

    v, n, t = stem


    vbottom = translation((0,0, height/2+radius*1.5)).moved(spheres[0])
    t = numpy.concatenate((t, spheres[2]+len(v)))
    v = numpy.concatenate((v, vbottom))
    n = numpy.concatenate((n, spheres[1]))

    return v, n, t

from chimerax.core.models import Model, Drawing
m = Model('test', session)
d = Drawing('test')
m.add_drawing(d)
d.set_geometry(*exclamation_mark())
session.models.add([m])

The sphere will be much too large (looks like it's still unit radius, as if the scale was not applied).

d.set_geometry(*exclamation_mark())

... and the drawing now looks as expected. Similarly, if I repeat the whole of the above script (creating a new Model), then the geometry comes out correct this time around. Loading and displaying an AtomicStructure *first* (with atoms displayed - ribbon only isn't sufficient) also seems to correct it.

[goddard@…]

Probably caused by last week’s matrix optimizations.  Will investigate today.

[Tom Goddard]

Fixed.

My matrix optimizations reuse Place instances and their numpy arrays when they are no longer referenced. There was a bug where a the scale() function produced a Place that was flagged as being the identity matrix even though it was not an identity matrix.