Version 0.7.1 release planning

[Paul Romano]

Hi all,

We've had a pretty productive two months since the last release, so I'd like to get another release (0.7.1) out the door soon. Things that have already gone into develop since the last release:

- Complex cell support

- "quadric" surface type

- Secondary particle bank

- Binary output refactor, solely HDF5 now

- Multi-group cross section generation module (openmc.mgxs)

- Collision estimators

- Delayed fission neutron production score / delayed group filter

- Inverse velocity score

- Performance improvement for binary search

Are there any features/fixes that you'd really like to get into the code for this release? Keep in mind that our next release (0.8) will probably take a bit longer than this one and may break backwards compatibility, so if there are small things you want to get in, now is the time to do it.

Thanks,

Paul

[Adam Nelson]

Man, this is going to be a good release.

Should we make a decision on whether or not to do the isotropic scattering, since it will be relatively easy to incorporate once we make the decision?

[Sterling Harper]

Agreed, this will be an awesome release.  The last one was good too.  We're really killing this software development thing :)

I think we should merge in the isotropic scattering.  We're getting a lot of features that mesh with deterministic codes so I think isotropic scattering is a natural feature to include.

[Adam Nelson]

I guess I'm in no position to argue with that.

[Paul Romano]

I'm personally not crazy about features that make the physics wrong, but I understand the motivation and since we have at least three developers who feel this is a worthwhile thing to be in develop, I'll go ahead and finish the review for 0.7.1. Thanks all!

[Will Boyd]

I'm also interested in the iso-in-lab capability since it can be very useful to highlight inherent biases in MGXS for deterministic codes. I'll try to jump in and package up that PR over the next few days.

I want/need to implement finer-grained control over domains for openmc.mgxs.Library, and nuclides for openmc.mgxs.MGXS. I will likely do this next week, but can't make any promises, so don't let it hold up v0.7.1.

Another minor feature I was thinking of implementing is an automated, randomized, and reproducible color spec scheme for the Python API's plots. Right now if a user does not manually specify a color spec to a plot (and who does that on a routine basis), the colors used in the plot will change with each run. It would be nice to randomly generate a color scheme for a plot based on an analysis of the geometry object (cell IDs, material IDs, ...). In addition, such a scheme should be perspective / scale invariant. Not a requirement for v0.7.1, but something I'd like to see in the code at some point.

[Will Boyd]

All of the items that I mentioned in my last post have now been implemented in the develop branch. This weekend I want to take a few hours to craft a more methodical suite of IPython Notebooks to showcase the openmc.mgxs module, in particular, the Library class. It would be great to hold off on the v0.7.1 release until these notebooks are included in the primary documentation, since a number of folks are starting to use openmc.mgxs (some of whom aren't Git / Sphinx pros and who need the online documentation to reflect the capabilities that I claim are in openmc.mgxs). Paul, did you have any other items on your wish list for the v0.7.1 release? 

[Paul Romano]

I'm fine holding off for some more notebooks. The only other item I have marked is pull request #500 which I still need a reviewer for. If you're comfortable handling that Will, that's great. A lot of the changes are related to Fortran syntax though, so it would be good to have one of the more experienced Fortran developers (Adam, Sterling, Bryan, Jon) take a quick look as well.

[Will Boyd]

Ok great, I'll try to package up a suite of notebooks on MGXS generation before Thanksgiving. Before I plunge headfirst into this, I wanted to gauge everyone's thoughts on the following outline for the notebook suite - disclaimer: I had a little fun coming up with not-so-witty rhymes :-)

• Part I - The Hungry for MGXS Games
• Introduce basic equations for MGXS (general rxns, transport, scatter matrix, and chi)
• Use fine-grained API to model MGXS for an infinite homogeneous medium
• Show how to use PyNe to plot micro MGXS atop CEXS for comparison
• Part II - Catching Fire with openmc.mgxs
• Use fine-grained API to model MGXS for a BEAVRS pin cell
• Show API for energy condensation
• Use tally arithmetic to combine / compare derived MGXS tallies (i.e., abs xs + scatter xs = total xs)
• Validate fine and coarse group MGXS in OpenMOC
• Part III - MockNJOYay
• Show how to use Library class to generate MGXS for a fuel assembly
• Validate fine and coarse group MGXS in OpenMOC
• Plot OpenMC (structured mesh) and OpenMOC (unstructured mesh) scalar fluxes, fission rates side-by-side

Let me know what you guys think of this outline and if there are any ideas for content that you think should be added here. Also, if you think that the demonstrations with OpenMOC - esp. Part III which will show the seamless injection of an OpenMC MGXS library directly into OpenMOC - are too preferential to OpenMOC let me know.

Also, I added Sam Shaner to this email thread since he is becoming more involved in OpenMC development. I'm a n00b and can't seem to figure out how to add him to the Google Group, so perhaps you can do that Paul? Sorry, I'm incompetent at technology sometimes... 

[Adam Nelson]

Yes, these look really good to me.  Including the names...

[Paul Romano]

Sounds good to me too. Looking forward to the PR!

[Will Boyd]

So I've not yet finished the Notebook suite. I think it's safe to say that I will definitely have a PR within a week though. I think it's worth the wait - plus, it is probably a good idea to wait to include Sam's PR to improve tally arithmetic in the v0.7.1 release as well.

[Paul Romano]

No problem -- I'm ok waiting one more week for the notebooks and the improved tally arithmetic.

[Paul Romano]

Thanks all for the latest pull requests / reviews. Everything we said we're going to merge for version 0.7.1 has been merged, so I've gone ahead and created a branch called release-0.7.1 for testing. If you have time in the next week, please checkout this branch and confirm that you don't run into any issues with it.

Thanks,

Paul

[Will Boyd]

I assume this new branch is the same as develop? At least it looks that way. If so, everything is working well for me!

[Paul Romano]

Yes, same as develop other than a few doc changes and incrementing the version number. I also just made a one-line fix on the branch to avoid a DeprecationWarning from clean_xml.py.

[Sam Shaner]

I preparing a small PR that will be ready later today that fixes a bug in tally arithmetic and adds a tally arithmetic test. Should I PR it into develop or release-0.7.1? 

--

Regards,

Sam Shaner

Graduate Student

MIT Nuclear Science and Engineering

samuel...@gmail.com | (805)748-7684 | 24-607

[Paul Romano]

You can PR it to release-0.7.1 directly.