Context

I was working on a project that involved preforming deep searches for propagation thresholds in cells with randomly generated geometries (via python).
As such, tiny changes in cell geometry could completely throw off the data.

Overview of the issue

After losing my marbles over some extremely inconsistent simulation behaviors, I realized that using ShapePlot (in order to view channel activity) was causing small changes in the diameters of axon sections near branch points, and was causing smaller changes everywhere else.
Eventually, I accepted that I would no longer the GUI to observe my simulations, since I couldn’t find any information on how to remedy this problem.

Obviously, I am now aware that ShapePlot is intended to adjust diameters near branch points via tapering (though it appears to cast float64 diameters down to float32 somewhere along the way--maybe this is unintended), but I think this should be documented, especially considering that this was the recommended method of viewing simulations when I attended the NEURON course.

I am also aware that pt3d data must be generated for this to be possible, but there's no reason this should destroy previously defined diameter information. One simply does not expect their model to literally change when they try to view it in the GUI.

NEURON setup

• Version: 8.2.2 HEAD (93d41fa) 2022-12-15
• Installation method: pip install
• OS + Version: Linux Mint 21.2 x86_64
• Compiler + Version: probably gcc

Minimal working example - MWE

from neuron import h

# We begin by creating two sections with unequal, fractional diameters:

sec_a, sec_b = h.Section("a"), h.Section("b")
sec_a.diam = 0.6
sec_b.diam = 0.2
sec_a.nseg = sec_b.nseg = 9

# connect them to make a 'T'
sec_b.connect(sec_a(0.5))
h.topology()

# To see the effects of PlotShape, we can easily scrape diameters using an h.RangeVarPlot, along
# with h.Vector-s:
rvp = h.RangeVarPlot("diam", sec_a(0), sec_a(1))
x, diams1, diams2, diams3 = [h.Vector() for i in range(4)]
# save the initial diameters, then compare to new diameters:
rvp.to_vector(diams1,x)

h.define_shape()    # define_shape, despite the fact that it generates pt3d data, does not affect
                    # diameter.

rvp.to_vector(diams2,x)
diff = list(h.Vector().copy(diams1).sub(diams2))
print(f"\n{diff=}")

h.PlotShape(False)  # PlotShape does affect diameter.

rvp.to_vector(diams3,x)
diff2 = list(h.Vector().copy(diams3).sub(diams2))
print(f"\n{diff2=}")