[Feature Request] Resetting the time of clock

[Jasmine969]

I often run the network for 1 second before a real simulation to make the neurons reach a steady state. When I start the real simulation, I want to reset the initial time to be zero. But such time is read-only. So I always need to minus the time by 1 second during post-processing. I hope the users can reset the time of the clock manully.

[mstimberg]

When we created Brian 2, we decided to not allow resetting the time (Brian 1 allowed it), because the semantics of it are not clear. If you change the time, what should Brian do about things like the time of the last spike (used for refractoriness), spikes in the SpikeQueue, etc.? Should monitors add new values with the changed time to the end of the recorded values, or should the monitor discard all values after the current time? We now have the store/restore mechanism to do a well-defined rollback of the full network state, including the time. I don't feel that the minor inconvenience of subtracting 1 from the time during post-processing warrants a fundamental change to this approach.

I don't know if it addresses your use case, but note that you have the flexibility to define and use your own time variable. E.g. you can do something like this:

group = NeuronGroup(10, """
                         dv/dt = -v/(100*ms) : volt
                         t_after_steady = t - steady_time : second (shared)""")
group.v = 'rand()*10*mV'
steady_time = 1*second
run(steady_time)
t_after_steady = StateMonitor(group, 't_after_steady', record=0)
mon = StateMonitor(group, 'v', record=True)
run(500*ms)
plt.plot(t_after_steady.t_after_steady[0]/ms, mon.v.T/mV)

But then, I'd personally rather write t = mon.t - steady_time once before doing the plotting and analysis.