Monitor array return type

[thesamovar]

We should decide what M.v should return, and which shape it has.

[thesamovar]

Current thinking suggests the best solution is:

• M[i].v returns a 1D array corresponding to neuron i
• M.v is a 2D array, with first index the recorded neuron index and second index time

In general, we shouldn't encourage people to write M.v[i] as this will not be equal to M[i].v.

Plotting multiple recorded traces will be done with plot(M.t, M.v.T).

[mstimberg]

I implemented the described semantics in a branch (see pull request). The only thing we did not discuss is the possible semantics of using more than one index, e.g. when recording neurons [0, 5, 10], what should mon[[5, 10]] do? I opted for consistent "neuron index first" semantic, so for plotting one would have to transpose:

plot(M.t, M[[5, 10]].v.T)

This means that for fully recorded groups, M[x].v == M.v[x].

I did not bother with slicing support, since this is not really obvious (if [0, 5, 10] is recorded, what should [5:] do -- work as [5, 10] or raise an error about 6, 7, 8, 9?)

[thesamovar]

Nice! I don't think slicing syntax is essential. I'll go ahead and merge this.