Add pickle test and revert theano var names (hopefully temporarily)

[canyon289]

This commit does a couple of things per comments in #4112

1. Adding tests to cover this issue
2. "Comment out" reverts changes in adding meaningful str representations to PyMC3 objects #4076

Proposed steps, pending successful CI, is that we merge this to fix broken SMC parallel capability and ArviZ pipelines, and then try reimplementing this feature in a way that doesn't break pickling of pm.Deterministic

[codecov]

Codecov Report

Merging #4117 into master will decrease coverage by 0.28%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #4117      +/-   ##
==========================================
- Coverage   88.75%   88.46%   -0.29%     
==========================================
  Files          89       89              
  Lines       14040    14028      -12     
==========================================
- Hits        12461    12410      -51     
- Misses       1579     1618      +39     

Impacted Files	Coverage Δ
pymc3/distributions/distribution.py	92.80% <ø> (-1.63%)	⬇️
pymc3/model.py	88.07% <ø> (-1.26%)	⬇️
pymc3/util.py	91.86% <100.00%> (-2.53%)	⬇️
pymc3/tests/models.py	79.13% <0.00%> (-8.64%)	⬇️
pymc3/step_methods/hmc/base_hmc.py	95.00% <0.00%> (-2.50%)	⬇️
pymc3/distributions/continuous.py	92.35% <0.00%> (-0.58%)	⬇️
pymc3/distributions/multivariate.py	80.78% <0.00%> (-0.15%)	⬇️

[canyon289]

This is ready for review, I guess our code coverage tolerance is set pretty high hence the CI failure

[AlexAndorra]

Pinging @Spaak, as I did in the associated issue

[canyon289]

Sounds good, going to ask if we can merge in 24 hours if no strong objection is raised, as this right now put ArviZ at risk

[Spaak]

Thanks for the ping also here, and for the tests. The issue appears to be with the "type hacking" I added to Deterministic in order to force str override to work with non-class-based types, and thankfully not with the new string machinery in general. I'll file a PR for a more targeted fix in a few hours.

[MarcoGorelli]

@Spaak do you have a test which fails if you just do

    var.__str__ = functools.partial(_repr_deterministic_rv, var, formatting='plain')

? For me, TestStrAndLatexRepr from pymc3/tests/test_distributions.py passes like this:

$ pytest pymc3/tests/test_distributions.py -k TestStrAndLatexRepr -q
...                                                                                                                                                                                                [100%]
============================================================================================ warnings summary ============================================================================================
pymc3/tests/test_distributions.py::TestStrAndLatexRepr::test__repr_latex_
  /home/marco/miniconda/envs/pymc3-dev/lib/python3.8/site-packages/theano/tensor/var.py:468: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3, and in 3.9 it will stop working
    isinstance(args_el, collections.Iterable)):

pymc3/tests/test_distributions.py: 48 warnings
  /home/marco/pymc3-dev/pymc3/util.py:149: DeprecationWarning: np.asscalar(a) is deprecated since NumPy v1.16, use a.item() instead
    return asscalar(value)

-- Docs: https://docs.pytest.org/en/stable/warnings.html
3 passed, 144 deselected, 49 warnings in 2.52s

[Spaak]

@Spaak do you have a test which fails if you just do

    var.__str__ = functools.partial(_repr_deterministic_rv, var, formatting='plain')

?

@MarcoGorelli Yes, I do now, see #4120 . Patching __str__ like that on the fly was the first thing I tried, but unfortunately Python does not always use an instance's obj.__str__() when str(obj) is called, but instead does some other type of lookup, which is why I had to add the type-patching. See also https://stackoverflow.com/a/5918210/1692028

[canyon289]

Thanks @Spaak!!