Change Input File Format from `.py` to `.yaml`

JacksonBurns commented 8 months ago

Motivation or Problem

Right now to run RMG, users write a functional Python script which is then actually executed (using exec) by RMG to load the requested reactors, species etc. into memory. The problems with this are:

this is formatting sensitive and tedious
exec is a security risk
exec is difficult to debug
exec requires passing the global and local namespace to the executed file, which is slow

Desired Solution

I mentioned in #2633 that the Green group has been planning to switch the input file format to YAML, and with the opening of #2633 it seems like a good time to announce this formally.

In line with cantera, RMS, and so many other modern tools, users will specify their simulation setup in a yaml format like:

 - reactor:
    - type: simple
    - temperature: 200 K
 - species:
    - smiles: C
    - smiles: CC
 - constraints:
    - max_size: 4

which will then be loaded by RMG.

Potential Alternatives

The easy thing to do would be to just not do this, but for the outlined reasons this is a bad idea.

etc.

Following the merger of #2553 we can now begin to move to Python 3.11+ (see https://github.com/ReactionMechanismGenerator/RMG-Py/issues/2445 and https://github.com/orgs/ReactionMechanismGenerator/projects/11/views/1?layout=board) which will make doing this easier, so we will implement this new loader based on that work.
Should we provide a script to convert old .py files to the new .yaml format? I suspect this would be more difficult than it seems, and it is not a significant burden to ask people to rewrite their input files in the new simpler format (IMO, welcome thoughts here).

jonwzheng commented 8 months ago

This is a great initiative and I'm on board with this. A couple of initial thoughts:

ARC already uses the .yml input format and overlaps strongly / utilizes RMG & Arkane in its codebase. It might be worth chatting with people who've worked on ARC (@alongd) because some of the work might already be "done" in ARC.
No strong feelings yet on if we should supply a conversion script. My initial thought is to allow both .py and .yml for now, with a DepreciationWarning if .py is used, and slowly transition to .yml in a future release. I agree a conversion script might be a lot of work, and input files tend not to be super complex, so I would also lean towards not providing it, as long as we provide some examples of "before" and "after" input files.

github-actions[bot] commented 5 months ago

This issue is being automatically marked as stale because it has not received any interaction in the last 90 days. Please leave a comment if this is still a relevant issue, otherwise it will automatically be closed in 30 days.

github-actions[bot] commented 2 months ago

This issue is being automatically marked as stale because it has not received any interaction in the last 90 days. Please leave a comment if this is still a relevant issue, otherwise it will automatically be closed in 30 days.

JacksonBurns commented 2 months ago

Should probably reformat this as a RIP

ReactionMechanismGenerator / RMG-Py