[contrib] Use standard repn for incidence graph generation

[Robbybp]

Fixes

Variables with zero coefficients appear in incidence graphs

Summary/Motivation:

Incidence graphs where 0*x appears as an edge can lead to incorrect results when analyzing singularity or generating decompositions.

Changes proposed in this PR:

• get_incident_variables function, which accepts a method argument that can be used to switch between using identify_variables and generate_standard_repn
• IncidenceConfig ConfigBlock to consolidate specification of default options for incidence graph generation
• Tests for this new functionality, including a linear_only option for incidence graph generation

Legal Acknowledgement

By contributing to this software project, I have read the contribution guide and agree to the following terms and conditions for my contribution:

1. I agree my contributions are submitted under the BSD license.
2. I represent I am authorized to make the contributions and grant the license. If my employer has rights to intellectual property that includes these contributions, I represent that I have received permission to make contributions and grant the required license on behalf of that employer.

[jsiirola]

This looks good! Just a couple minor questions that you can address or ignore.

(tests need to pass before merging)

[codecov]

Codecov Report

Patch coverage: 93.33% and project coverage change: -0.01 ⚠️

Comparison is base (dca0388) 87.19% compared to head (eb3d362) 87.19%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2834      +/-   ##
==========================================
- Coverage   87.19%   87.19%   -0.01%     
==========================================
  Files         765      767       +2     
  Lines       88471    88548      +77     
==========================================
+ Hits        77146    77211      +65     
- Misses      11325    11337      +12     

Flag	Coverage Δ
linux	84.26% <93.18%> (+0.01%)	⬆️
osx	73.92% <93.18%> (+0.01%)	⬆️
other	84.43% <93.18%> (+0.01%)	⬆️
win	81.81% <93.18%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pyomo/util/subsystems.py	96.02% <75.00%> (-1.80%)	⬇️
pyomo/contrib/incidence_analysis/incidence.py	94.87% <94.87%> (ø)
pyomo/contrib/incidence_analysis/__init__.py	100.00% <100.00%> (ø)
pyomo/contrib/incidence_analysis/config.py	100.00% <100.00%> (ø)
pyomo/contrib/incidence_analysis/interface.py	96.34% <100.00%> (+0.70%)	⬆️

... and 4 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[jsiirola]

Can we leave a comment as to why this is not being tested? Or better, why is this nondeterministic?

[Robbybp]

Added a comment noting that the unmatched variables/constraints are not mathematically unique, and expanded on this in the text, noting that they depend on (a) how we identify incident variables and (b) the implementation of the matching algorithm.

[Robbybp]

"Regression test" on several examples:

Example	N. edges (6.5.0)	N. edges (main)	N. edges (this PR)	Correct?
DoF	3,921	3,933	3,921	Y
Stuct. sing.	6,922	6,943	6,922	Y
Num. sing.	7,282	7,792	7,588	Y
Init. cond.	14, 560	14,560	14,560	Y
Discretization	3,900	3,916	3,900	Y*
Initialization	897	900	897	Y*

*Leads to a different number of diagonal blocks in the block triangularization (between this PR and main)

Here "Correct" means that, with this PR, the analysis leads us to the correct conclusion about what is wrong with the model. These were all run with IDAES 2.0. The comparison in #2816 uses a recent IDAES main, so the numbers reported in these two tables may not be directly comparable.

[Robbybp]

Pypy test failure appears unrelated

[Robbybp]

@andrewlee94 can you run IDAES tests with this branch?

[andrewlee94]

@Robbybp That fixed the known failures, but it does create two new ones, both related to the following:

self = <pyomo.util.subsystems.ParamSweeper object at 0x7f549e6820d0>

    def __enter__(self):
        to_fix = self._vars_to_fix
        to_deactivate = self._cons_to_deactivate
        to_set = self._comps_to_set
        to_unfix = self._vars_to_unfix
>       self._var_was_fixed = [(var, var.fixed) for var in to_fix + to_unfix]
E       TypeError: unsupported operand type(s) for +: 'ComponentMap' and 'list'

These are coming up in the gas network tests (but I don't think that actually matters here).

[Robbybp]

@andrewlee94 this should be fixed in IDAES/idaes-pse#1188

[mrmundt]

@Robbybp - Okay if we merge?

[Robbybp]

@mrmundt Yep!