Updated PlasmoPlots to work with v0.6.0+

[dlcole3]

PlasmoPlots had not been updated to work with the new release of Plasmo (v0.6). I updated the source code to handle the new version, including upding some of the function calls (e.g., num_all_variables is replaced by num_variables or replaced getnodes with local_nodes). I also updated the label calls (which need to access label.x instead of label since label.x results in BaseRefValue{Symbol} being included in the label) and the call to the clique projection. The tests and Project.toml files have been updated.

I did not update the PlasmoPlots.jl version number in the Project.toml file, which I think should be updated to 0.2.0. @jalving do you normally do that as a separate PR, or should I have included that in this PR?

[dlcole3]

I did not see that there was already another PR open (#5) for this same thing. I am going to close this one and instead look at that other one. I see the tests here work, but the tests on #5 are failing, so I will take a look and see if I can figure out why that is. Apologies for duplicating the PR.

[jalving]

@dlcole3 I forgot I had opened the other PR. If your implementation works, we can close the other one in favor of this.
The version number can be updated here or afterwards. It just has to get incremented before we register.

[dlcole3]

@dlcole3 I forgot I had opened the other PR. If your implementation works, we can close the other one in favor of this. The version number can be updated here or afterwards. It just has to get incremented before we register.

Sounds good. I copied over the examples and tests from your PR (#5) and followed a couple of the formatting things from your source code. On my local machine, the tests all run, but the overlapping matrix_plot example does not work as I expected. I will try to look into the code more tomorrow. layout_plot works as expected, and matrix_plot seems to work okay for the non-overlapping case, but I have to go deeper in the code to address the overlapping problem I think. I believe the constraints are not tracked properly in the overlapping case because some of the node labels go below the xaxis of the plot. Was this a problem on the old version of PlasmoPlots as well? I see there was not a test for the overlapping case before.

[dlcole3]

@jalving I noticed one difference between #5 and #6 is that #5 uses n_cons_total for n_all_cons_total. This makes sense to me, but when I removed the n_linkcons_total from this line, the resulting spy matrix has node labels that run off the plot, and I don't think the number of constraints on the y axis is being mapped correctly. Consequently, I put the n_linkcons_total back into this equation and the plots look better. I can try to dive into the code mroe to figure out why this is.

[jalving]

I suspect the issue has to do with inconsistencies with how constraints used to be tracked versus how they are now. There are likely a few kinks to work through here.

[jalving]

overlapping used to work. i suspect issues have to do with erroneous tracking due to the v0.6 updates.

[dlcole3]

overlapping used to work. i suspect issues have to do with erroneous tracking due to the v0.6 updates.

It looks like it was not strictly an issue with the overlapping case. The num_constriants(node) was not using the include_variable_constraints argument, so it was counting more node constraints than were initially being counted for the graph. In addition, plotting the link constraints for the non overlapping case was calling all_link_constraints instead of local_link_constraints. For the overlapping case, it also offset the column by 1 (it started at 1 instead of 0), which does not appear to be consistent with the non-overlapping case, and this resulted in there being one node that was being placed outside the plot. The plots now look about like I expected for the three examples. I did plot the three examples with the old version of Plasmo/PlasmoPlots, and the matrix plots look the same.

[dlcole3]

For the overlapping case, it does look like both the old and new (i.e., this PR) versions of PlasmoPlots do not directly overlap the matrix_plot. For instance, the plot from example 3 looks like this

which says it has 33 variables listed on the x axis. The graph technically only has 19 variables on it (the additional 14 come from the overlap). Is that the expected behavior from plotting? It seems like there is an alternative visualization where the x-axis only has 19 variables on it. However, the current approach seems okay because it highlights the variables that are "overlapping"

[jalving]

Is that the expected behavior from plotting? It seems like there is an alternative visualization where the x-axis only has 19 variables on it. However, the current approach seems okay because it highlights the variables that are "overlapping"

Yes, IIRC, I never produced a visual that directly overlaps the matrix plot. I think this version was meant to show the 'size' of the overlapped problems, but a direct representation would also be nice to have.

See issue #7

[jalving]

The num_constriants(node) was not using the include_variable_constraints argument, so it was counting more node constraints than were initially being counted for the graph

I think Plasmo needs to change the default such that include_variable_constraints is always false. I believe JuMP makes that their default as well.