Invalid doctest in PyROS - unclear what behavior should be

[mrmundt]

Summary

During tracking down a totally different problem, I noticed that there is a doctest from uncertainty_sets that is invalid code and doesn't run. We don't fail on warnings, so we hadn't noticed before.

Steps to reproduce the issue

I noticed this in the doctests when I was looking at something else entirely

D:\a\pyomo\pyomo\doc\OnlineDocs\..\..\pyomo\contrib\pyros\uncertainty_sets.py.rst: 
WARNING: ignoring invalid doctest code: 
'>>> conf_ellipsoid = EllipsoidalSet(\n...     center=np.zeros(4),\n...     shape_matrix=np.diag(range(1, 5)),\n...     scale=None,\n...     gaussian_conf_lvl=0.95,\n... )\n>>> conf_ellipsoid.center\narray([0, 0, 0, 0])\n>>> conf_ellipsoid.shape_matrix\narray([[1, 0, 0, 0]],\n       [0, 2, 0, 0]],\n       [0, 0, 3, 0]],\n       [0, 0, 0. 4]])\n>>> conf_ellipsoid.scale\n...9.4877...\n>>> conf_ellipsoid.gaussian_conf_lvl\n0.95'
...

In looking at that code, there is, in fact, a problem. Here is the code now:

    A 4D 95% confidence ellipsoid:

    >>> conf_ellipsoid = EllipsoidalSet(
    ...     center=np.zeros(4),
    ...     shape_matrix=np.diag(range(1, 5)),
    ...     scale=None,
    ...     gaussian_conf_lvl=0.95,
    ... )
    >>> conf_ellipsoid.center
    array([0, 0, 0, 0])
    >>> conf_ellipsoid.shape_matrix
    array([[1, 0, 0, 0]],
           [0, 2, 0, 0]],
           [0, 0, 3, 0]],
           [0, 0, 0. 4]])
    >>> conf_ellipsoid.scale
    ...9.4877...
    >>> conf_ellipsoid.gaussian_conf_lvl
    0.95

I think it should actually be:

    A 4D 95% confidence ellipsoid:

    >>> conf_ellipsoid = EllipsoidalSet(
    ...     center=np.zeros(4),
    ...     shape_matrix=np.diag(range(1, 5)),
    ...     scale=None,
    ...     gaussian_conf_lvl=0.95,
    ... )
    >>> conf_ellipsoid.center
    array([0., 0., 0., 0.])
    >>> conf_ellipsoid.shape_matrix
    array([[1, 0, 0, 0],
           [0, 2, 0, 0],
           [0, 0, 3, 0],
           [0, 0, 0, 4]])
    >>> conf_ellipsoid.scale
    9.4877...
    >>> conf_ellipsoid.gaussian_conf_lvl
    0.95

The thing that is unclear to me, however - this test fails because conf_ellipsoid.scale outputs np.float64(9.4877...). Is it intended that XX.scale can return an np.float object, or is it intended to be cast to a raw Python float? I didn't want to make an assumption, so I hand this off to the PyROS folks to decide what the right behavior is.

Error Message

Document: api/pyomo.contrib.pyros.uncertainty_sets.EllipsoidalSet
-----------------------------------------------------------------
**********************************************************************
File "..\..\pyomo\contrib\pyros\uncertainty_sets.py", line ?, in default
Failed example:
    conf_ellipsoid.scale
Expected:
    9.4877...
Got:
    np.float64(9.487729036781154)
**********************************************************************
1 item had failures:
   1 of  15 in default
15 tests in 1 item.
14 passed and 1 failed.
***Test Failed*** 1 failure.

Information on your system

Pyomo version: 6.9.2.dev0
Python version: Multiple
Operating system: Windows (NOTE: Only seems to happen on Windows)
How Pyomo was installed (PyPI, conda, source): source
Solver (if applicable):

[mrmundt]

@shermanjasonaf @natalieisenberg @jas-yao

[shermanjasonaf]

@mrmundt Thanks for pointing this out. I will discuss with the PyROS Dev Team on 2025-03-28.