Open jonathanfischer97 opened 1 month ago
AayushSabharwal
commented
1 month ago I may be misunderstanding EnsembleSummary here, but looking at the plot recipe it seems that the trajectories keyword argument is doing what you want here. No idea why it's not called idxs, since at the EnsembleSummary level there are no trajectories.
jonathanfischer97
commented
1 month ago Yep, you're right, just tested it. Must've been a typo, I'll make a PR
jonathanfischer97
commented
1 month ago Made a PR, but in looking at the code, had a bunch of other thoughts which I'll repost here for posterity since they are more appropriate for an issue than PR comment:
"
Was the original intention of the EnsembleSummary recipes to allow users to choose/truncate the amount of trajectories that statistics are plotted over, similar to EnsembleSimulation? (Explains the mixup with idxs)
Symbolic indexing does not work any of the recipes for either EnsembleSimulation or EnsembleSummary.
AbstractSolution.Plotting state spaces via giving idxs as a Tuple rather than a Vector does not work.
Makie EnsembleSummary recipe doesn't show transparent ranges:
Shown are 100 trajectories.
No user-exposed way to set the color map of EnsembleSimulation trajectories? Would like an easy way to set it to something other than random, like this:
"
ChrisRackauckas
commented
1 month ago Was the original intention of the EnsembleSummary recipes to allow users to choose/truncate the amount of trajectories that statistics are plotted over, similar to EnsembleSimulation? (Explains the mixup with idxs)
Yes. Both used the idxs kwarg before, so it was split so that idxs is the new vars, but I don't think all got fixed on the ensemble side when that change happened. Now that there's trajectories and idxs, the meaning of the two is more clear and non-overlapping, but we didn't ever complete this.
Symbolic indexing does not work any of the recipes for either EnsembleSimulation or EnsembleSummary.
Agreed, it's a missing part of the interface here.
Plotting state spaces via giving idxs as a Tuple rather than a Vector does not work.
Yeah that's unintentional, it should get fixed.
Makie EnsembleSummary recipe doesn't show transparent ranges:
That's unintentional. I don't use Makie so I'm not 100% how to handle that.
No user-exposed way to set the color map of EnsembleSimulation trajectories? Would like an easy way to set it to something other than random, like this:
That would be nice. I'm not sure how to do that with recipes though.
AayushSabharwal
commented
1 month ago Can this PR be closed now that #700 is merged?
ChrisRackauckas
commented
1 month ago We should get issues on the other points though, those are reasonable asks.
AayushSabharwal
commented
1 month ago Was the original intention of the EnsembleSummary recipes to allow users to choose/truncate the amount of trajectories that statistics are plotted over
Doesn't EnsembleSummary reduce over trajectories? The original trajectories are not retained, so we can't just plot over a subset of them.
jonathanfischer97
commented
1 month ago Was the original intention of the EnsembleSummary recipes to allow users to choose/truncate the amount of trajectories that statistics are plotted over
Doesn't
EnsembleSummaryreduce over trajectories? The original trajectories are not retained, so we can't just plot over a subset of them.
The trajectories aren't maintained in the .u field here?
struct EnsembleSummary{T, N, Tt, S, S2, S3, S4, S5} <: AbstractEnsembleSolution{T, N, S}
t::Tt
u::S
v::S2
med::S3
qlow::S4
qhigh::S5
num_monte::Int
elapsedTime::Float64
converged::Bool
endThough allowing subsetting of the trajectories reduced over is definitely low priority compared to the other issues I think.
Problem
plot(ensemble_summary; idxs = 1)
Output of
using Pkg; Pkg.status(; mode = PKGMODE_MANIFEST)Output of
versioninfo()