improve explorer for non-time variables

[corakingdon]

using Mimi

@defcomp example begin
    p1 = Parameter(index = [foo])
    p2 = Parameter(index = [foo, bar])
    p3 = Parameter(index = [foo, baz])
    p4 = Parameter(index = [time, bar])
    p5 = Parameter(index = [time, baz])
end

m = Model()
set_dimension!(m, :time, 10)
add_comp!(m, example)
set_dimension!(m, :foo, 1:5)
set_dimension!(m, :bar, 1:3)
set_dimension!(m, :baz, [:A, :B, :C])
set_param!(m, :example, :p1, 1:5)
set_param!(m, :example, :p2, reshape(1:15, 5, 3))
set_param!(m, :example, :p3, reshape(1:15, 5, 3))
set_param!(m, :example, :p4, reshape(1:30, 10, 3))
set_param!(m, :example, :p5, reshape(1:30, 10, 3))
run(m)
explore(m)

I think the plot for p1 should be with points instead of bars. (If the dimension values are numerical, I think they should get plotted as a scatter plot.)

Then the plots for p2 and p3 are currently meaningless, but what's interesting is that they are rendered differently from each other I guess because one of them has symbols for the second dimension and one of them has numbers for the second dimension. I think we should stick with putting the first dimension, foo, as the x axis, and then still produce a scatter plot where the second dimension is represented by different colors, like we do for p4 and p5.

There's probably more corner cases we need to define, but this is a start!

[rjplevin]

set_param!(m, :example, :p1, 1:5)
set_param!(m, :example, :p2, reshape(1:15, 5, 3))

In the current code under development, the first statement won't work because the import will not have happened yet. We might import on demand if the requested param doesn't exist in m.md and it is unbound in one subcomponent (but not in multiple).

Alternatively, you could call import_params!(m) any time after the add_comp! and before the set_param! calls. Alternatively alternatively, we could add a kwarg, e.g., add_comp!(m, import_all=True).

[corakingdon]

I'm pretty sure we want those calls to set_param! to still work. But I think any discussion about that belongs on this issue: #524