Closed mphoward closed 2 months ago
Is this ever going to happen? HOOMD has already progressed to 3.5.0 and there are a lot of improvements, including to the build process
Yes, we are still planning to do this, but there is considerable effort needed to do it. HOOMD v3 implemented multiple API-breaking changes that we need to accommodate. I agree that the improved build process is a major benefit.
Features to remove:
bond.fene
restrain
: what to keep (if anything)?special_pair
pair.lj124
pair.lj96
pair.slj
Features to rename:
pair.ashbaugh
-> PerturbedLJ
(or something like that)Features to move to pure Python:
analyze.rdf
: or maybe remove?variant.sphere_area
Only a few years longer than planned… 🙂
A beta release of HOOMD 3.0 is now available, and we should start the process of migrating our code.
As discussed with @astatt, PR #40 will be the last feature added for HOOMD 2.9, and it will be included in our last 0.x minor release. From there, all new features should target HOOMD 3.0, and we will begin the process of supporting the new API. We will keep a v0.11 (or hoomd-29? name TBD) branch open to support bug fixes for old code for so long as it is practical.
Steps that need to be taken:
HOOMD now uses
pytest
rather thanunittest
for python-level testing. We could consider using some of their test fixtures, since they should be installed alongside the code, or we can stick with our existing unit tests. Either way, I do think we should try to consolidate tests where possible (#34).We should also hit the goal of fully hosting documentation and finishing packaging for this for the official 1.0 release. I'm sure more things will pop up, and we can add to this list as needed.