Revamping of PMF

[mrshirts]

Here's my proposed revamping of PMF functionality. I'd love to get comments so we can merge this in for pymbar4. I anticipate there may be some requested changes, so I'd love to get started on them.

[Lnaden]

Be more specific what I should look at when you say "check the test coverage".

I mean we should make sure the tests are updated in a way to maximally hit lines of code, especially since this is a new module. The existing tests I think just run through the PMF computation, but not necessarily all the functions and code paths.

We can do this in another PR to avoid making this one too large, especially since I want to change from Nose -> Pytest

[mrshirts]

We can do this in another PR to avoid making this one too large, especially since I want to change from Nose -> Pytest

OK, added issues to switch testing frameworks (#330, @Lnaden ), and after to add more PMF tests (#331 @mrshirts )

[Lnaden]

I'm of the opinion lets get this in, then we can work on creating tests and whatnot after I handle #330. I have some time today so I'd like to get the test framework switched over, and start on #332 as well.

Since this is all on the pymbar4 branch, I feel safe making rapid changes now, improve the CI/test/coverage framework, then going back and ensuring we're covering all our bases.

[mrshirts]

@jchodera Do you want to take an overall look at how this is structured? There's a couple of things that need to be done with the usability of the MCMC sampling of parameters, but it would be good to know if there are any showstoppers about the general usability and way it is presented. Check the examples to see how PMF's are now generated for most users.

[mrshirts]

mbar_computePMF test obviously failing, since the function doesn't exist.

Should be completely fixed in the near future by adding a test_pmf.py file (test_bar.py and test_exp.py need to be done as well!), but for now I have removed the broken function call.

[Lnaden]

Great! If this is good to go in your book, go ahead and merge!

[mrshirts]

Note: address issue with a deprecated KDE functionality being used.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (pymbar4@adadee3). Click here to learn what that means.
The diff coverage is 41.93%.

@@            Coverage Diff             @@
##             pymbar4     #327   +/-   ##
==========================================
  Coverage           ?   42.92%           
==========================================
  Files              ?       14           
  Lines              ?     2947           
  Branches           ?        0           
==========================================
  Hits               ?     1265           
  Misses             ?     1682           
  Partials           ?        0

Flag	Coverage Δ
#unittests	42.92% <41.93%> (?)

Impacted Files	Coverage Δ
pymbar/testsystems/gaussian_work.py	12.5% <ø> (ø)
pymbar/exp.py	26% <ø> (ø)
pymbar/old_mbar.py	20.91% <0%> (ø)
pymbar/bar.py	9.43% <0%> (ø)
pymbar/confidenceintervals.py	3.43% <0%> (ø)
pymbar/mbar_solvers.py	86.66% <100%> (ø)
pymbar/utils.py	66.66% <20%> (ø)
pymbar/pmf.py	45.13% <57.14%> (ø)
pymbar/mbar.py	73.37% <61.9%> (ø)
pymbar/utils_for_testing.py	89.13% <83.33%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update adadee3...070ec1e. Read the comment docs.

[mrshirts]

OK, failed because the conda-recipe (I think?) doesn't have sklearn. So it will need to be added to run kernel density calculations.

[mrshirts]

Let's go with merging this now with basic tests now implemented?