Repeated sim.data[probe] accesses are very slow

[arvoelke]

import time
import nengo

with nengo.Network() as model:
    x = nengo.Ensemble(1000, 1)
    p_neurons = nengo.Probe(x.neurons)

with nengo.Simulator(model) as sim:
    sim.run(10)

# Repeated probe accesses
start = time.time()
for m in range(x.n_neurons):
    sim.data[p_neurons][0, m]
print time.time() - start

# Only one probe access
start = time.time()
u = sim.data[p_neurons][0, :]
for m in range(x.n_neurons):
    u[m]
print time.time() - start

It takes 37 seconds on my computer to access a mere 1000 spikes. It looks like every time sim.data[p_neurons] is accessed the entire array is re-copied. And of course the same effect is seen when accessing various time points.

Perhaps we should be caching the read-only object?