dallan-keylogic

Thank you for profiling it @Robbybp . The main motivation was trying to read the old code, finding it unnecessarily complex, seeing a triple `for` loop in Python, and deciding...

On a different note, there's the failing test. I am also having problems with `test_display_near_parallel_variables`, which I know is changed in #1380 . The original test expects only `(v1, v2)`,...

This is what I get with my latest version. Indeed, creating `norms` was the bottleneck, probably because it wasn't preallocated. ``` Identifier ncalls cumtime percall % ------------------------------------------------------- new 1 58.482...

> > Frankly, I'm shocked that the (v1, v3) pair was missed by the old method (and that seems to have become undone with minor changes). > > This is...

I did some (probably excessive) optimization, and now we have: ``` Identifier ncalls cumtime percall % ------------------------------------------------------- new 1 0.211 0.211 3.6 -------------------------------------------------- check-dotprods 1 0.145 0.145 68.5 get-jacobian 1...

The test failures in Python 3.8 are because some element of Numpy or Scipy decided to change the order it reports elements. The set of parallel vectors is the same,...

@Robbybp , any additional comments about this new method before I bug @andrewlee94 about it?

> I have a diagnostics example I'm working on that relies on `check_parallel_jacobian`. When I run with the current implementation, I get 11 pairs of parallel constraints, but when I...

At the moment, this PR reveals many of the unit models to be poorly scaled because checking for numerical issues is carried out on the model before a Pyomo scaling...

@Robbybp @bpaul4 could one of you review this PR now that it's ready to be merged?