Set default values for *_CPLHIST_DIR

[ekluzek]

I suggest setting the default_value for the DATA Components _CPLHIST_DIR to what would be the users normal archive directory for their cases using the component CPLHIST_CASE settings.

So for example for DATM_CPLHIST_DIR:

$CIME_OUTPUT_ROOT/archive/$DATM_CPLHIST_CASE/cpl/hist

This makes it easier for a user to run a case and then run CPLHIST mode from it. And it makes it easier for them to figure out where to point to for someone else's case as well.

[billsacks]

@ekluzek I imagine you have given this a lot more thought than I have, but my gut reaction is to disagree with this. I think defaults are important in situations where most people will want to use the default and don't want to think about it - or don't have enough information to think about it. But in this case, I feel like it's best to force a user to think about this and set it explicitly: In many cases, I suspect users will not want to point to the default you suggest, and if you're running a cplhist case, I feel you should be forced to think about and explicitly set where you want to get the forcing data (i.e., the cplhist_dir). I'm concerned that, with your suggestion, many users will run into confusing errors where the default directory doesn't exist. Or - less commonly but worse - the directory does happen to exist but that's not actually the right place to get the final data.

[ekluzek]

@ekluzek I imagine you have given this a lot more thought than I have, but my gut reaction is to disagree with this. I think defaults are important in situations where most people will want to use the default and don't want to think about it - or don't have enough information to think about it. But in this case, I feel like it's best to force a user to think about this and set it explicitly: In many cases, I suspect users will not want to point to the default you suggest, and if you're running a cplhist case, I feel you should be forced to think about and explicitly set where you want to get the forcing data (i.e., the cplhist_dir). I'm concerned that, with your suggestion, many users will run into confusing errors where the default directory doesn't exist. Or - less commonly but worse - the directory does happen to exist but that's not actually the right place to get the final data.

These are good points that I'd like to take in. And @billsacks and I just went through some of this. I proposed a compromise that you support. First I looked into the code for your question about the directories -- and the code doesn't handle this gracefully. But, I think it should. It should check for the directories existence and die if it doesn't exist. And that should be done in the code no matter what. I'd like to propose a PR where I just do this for DATM, and we'll see if it's OK there. And later we could either back it out, or other components could then think about adding something similar if it's successful. The other thing we talked about was asking a few CTSM folks about this. And we thought @slevis-lmwg @olyson and @wwieder are the ones to ask there. And we need to hear from Ocean/Sea-ice folks so that's probably: @klindsay28 @mnlevy1981 @alperaltuntas and possibly @dabail10 for CICE. So I'll work on that.

[slevis-lmwg]

I may need some follow-up to understand the details, but I support a compromise that makes the handling more graceful and user-friendly, while at the same time improving the chances that the user will end up pointing to the correct files.