RangeVarPlot enhancements

[ramcdougal]

Support for rxd, e.g.

rvp = h.RangeVarPlot(ca, dend(0), dend(1))

or

rvp = h.RangeVarPlot(ca[er], dend(0), dend(1))

and a new method that returns a new Vector (the old rvp.to_vector only manipulates an existing Vector):

vec = rvp.vector()

[ramcdougal]

This is a work in progress because of the following bug:

If you run this:

from neuron import h, rxd

dend = h.Section(name='dend')
dend.nseg = 3

cyt = rxd.Region([dend], name='cyt')
c = rxd.Species(cyt)#, initial=lambda node: 1 if 0.4 < node.x < 0.6 else 0)

rvp = h.RangeVarPlot('v', dend(0), dend(1))

h.finitialize(-65)
print(list(rvp.vector()))

you'll see the 5 values (3 segments + endpoints):

[-65.0, -65.0, -65.0, -65.0, -65.0]

which is correct. If instead, you uncomment the species initialization:

c = rxd.Species(cyt, initial=lambda node: 1 if 0.4 < node.x < 0.6 else 0)

it incorrectly shows 10 values (and with a slight modification, one can show that it generates two copies of the correct values):

[-65.0, -65.0, -65.0, -65.0, -65.0, -65.0, -65.0, -65.0, -65.0, -65.0]

If the RangeVarPlot is defined after the finitialize, it works correctly.

This is mysterious because neuron-nightly doesn't do this, the RangeVarPlot doesn't involve rxd, and the new code is essentially unused in this case. Any thoughts @pramodk , @adamjhn , @nrnhines ?

[ramcdougal]

Ok, so the above bug does NOT appear to be a bug in this commit, as it appears when I build NEURON from sources (d1ba505) using autotools (or cmake) on Ubuntu 20.04 (with Python 3.8.2) but it does not occur if I use the pip install neuron-nightly (dbeb52b) on Google Colab or on Ubuntu 20.04.

The test needs to be slightly revised to run on those versions as they do not have a .vector method:

from neuron import h, rxd

dend = h.Section(name='dend')
dend.nseg = 3

cyt = rxd.Region([dend], name='cyt')
c = rxd.Species(cyt)#, initial=lambda node: 1 if 0.4 < node.x < 0.6 else 0)

rvp = h.RangeVarPlot('v', dend(0), dend(1))

h.finitialize(-65)
y = h.Vector()
rvp.to_vector(y)
print(list(y))

[ramcdougal]

I removed the WIP designation since the bug doesn't seem to be coming from here.

[nrnhines]

LGTM. Have you tested to your satisfaction?

[ramcdougal]

Yes. I think this is ready to merge.

[adamjhn]

Should we include "<class 'neuron.rxd.species.State'>", "<class 'neuron.rxd.species.Parameter'>" and "<class 'neuron.rxd.species.ParameterOnRegion'>"?