Using CMEPS outside of CESM framework is getting way more complex

[uturuncoglu]

I think the current design of CMEPS is too complex. To compile CMEPS outside of the CESM framework you need to do lots of tricks/hacks and also get bunch of code from different places such as https://github.com/ESCOMP/CESM_share even if you are not using them. Even for experienced users it is not easy and getting more complex every day. Sometimes, you also need to provide CDEPS shared code and/or different compile flags also involves to generate template codes or use some capability which is already exit in CMEPS (like CESM aoflux calculation, why it is designed only used by CESM?). I think we need to think carefully to reduce this complexity if we really want to be a community software (or specifically mediator). My impression is that this approach is not sustainable and have makes the debugging and testing more and more difficult. Maybe in the past I raised same concerns but my impression is that CMEPS needs to be self contained software package that could be easily port to any application. This is my opinion but I am using CMEPS mostly outside of CESM framework and this is my observation. Anyway, I am happy to discuss more if you would like.

[mvertens]

@uturuncoglu - thank you for bringing this up. I totally agree. I have advocated in the past moving the cesm layer to an external repository and the same for the computation of atm/ocn fluxes. @billsacks - it would be useful to have your input as well and try to set up a time to discuss further.

[uturuncoglu]

@mvertens Thanks. Yes, it would be nice to discuss more about it. As an example, I have just synced with CMEPS head of develop and now I need to add lots of extra shared code related with water isotopes etc. even if I am not using it. It also requires lots of shared staff including template files like https://github.com/ESCOMP/CESM_share/blob/main/src/shr_assert_mod.F90.in that needs to be processed before compiling. So, I am not sure anyone except me tried to compile CMEPS outside of CESM but this is really challenging and you need to do it to understand its complexity. Do we really need all those staff to compile just one component? Since those are not changing too frequently, maybe it would be defined as an external dependency library (like FMS for MOM6) which can be installed standalone and make it available through the module files. So, you could use it easily. But again, the ideal way to have minimal dependency and have self contained component for easy to use.

[billsacks]

I just talked with @uturuncoglu . I very much support the goals here. We should talk soon to come up with a vision for improving this.

As for the specific issue of water isotopes / tracers, I hope that if you use the version here - https://github.com/ESCOMP/CMEPS/blob/main/ufs/wtracers_mod.F90 - then that will solve that issue for you. That file is about to expand significantly as soon as I merge #638 , but hopefully will still support the standalone build you're trying to do - at least, that's the intent of the design, and I believe that's what's being tested in the GitHub action that builds CMEPS as part of each PR.