2021_09_29 Thompson/Wagner check in

General updates

• JW – Thanks for firefighting parmed this week

• JW – I’m going to continue working on project planning – Effort estimate review today?

  • Something I sent the MolSSI MMIC devs earlier this week (also I sent them your way for a chat on Interchange)

We're in a somewhat serious time deficit at OpenFF where I basically evaluate how expensive indecision on a question will be, and if it's greater than $10,000 I answer immediately, if it's greater than $1000 I answer on Weds/Thurs (my "meeting days), and if it's less than that and takes more than 15 minutes I don't answer at all. This is a bad place to be and I'm trying to reorganize a bit so this doesn't continue into 2022... MolSSI seems to be doing WAY better in this regard.

So, this PR is in the "if we make the wrong choice here and break the object structure/API later, it'll cost more than $1,000 but less than $10,000" bucket. So this is a good Weds/Thurs task if it can be ready this week (otherwise, we can't break what we don't release, so if it's NOT ready this week and you're out in October, there's not much cost to delaying to November)

• MT – I’m in touch with MMIC devs as well, will have one meeting with them but not commit to doing stuff for them. I want more guidance from OpenFF on whether we’re collaborating with MolSSI/whether I have to do work for them/whether I can or should assign work to them/whether I’m responsible for the success of their project/whether they’re responsible for the success of my project.

PR reviews

• Toolkit #1096 - Approved

• Toolkit #1094 - Approved

• Toolkit #1088 - Soft approved / Requested final review from LWang

• Toolkit #949 - JW will look over, likely delegate to JMitchell.

Planning

• JW – Keeping things on track for coordinating with Mitchell on interchange docs.

  • MT – I pinged JM a few days ago on help with the users guide, I think he’s working on other stuff right now.

  • JW – Interchange user guide is a great use of his time. In general your requests to Josh M should be equivalent to mine, I’ll chat with him later today and tell him this.

• High-level ideas

  • “Interchange is a parmed replacement”

    • Input

      • what’s minimal input?

      • what’s maximal input?

    • State

      • Positions

      • Box vectors

      • Topology

        • May be chemical or atomtyped

        • What’s a “TopologyKey”?

          •  

      • Handlers

        • Parameter provenance?

          • can-be-traced-to-original-parameter (How much is atomtype-based assignment “a trivial lookup”? complications from wildcards?)

          • we-don’t-know-where-this-came-from (a lot of imports, stuff like ToolkitAM1BCC)

          • this-came-from-a-neural-net-and-we’re-storing-the-whole-darn-thing

        • Some formats store cutoffs/other globals separately from the per-particle parameters

    • Output

      • If the state/input is underinformed, what do we have to “make up”?

  • What is a “system”?

    • Relative to existing formats (amber prmtop/crd files, ParmEd Structure)

  • When can two systems be added?

    • In an inequivalent combination, what’s “safe” to make up?

  • Parameter provenance

    • Storage formats that do vs. don’t track parameter provenance

  • Molecule representations

    • Atomtyped vs. chemical vs. coarse grained

      • Can SOMETIMES jump between them, but not always

  • Units/Pydantic/other validation/enumerated types (like combining rules)

  • Safety

    • Should there be a “I just want to run the simulation” mode and a “paranoid” mode? Or something analagous to “maxwarn”?

• MT will refine the above outline, add more detail, convert into markdown, and add to PR, tagging Josh M to see if that can be a good handoff.