Endpoints for PFT details & plotting of run outputs #2665
Conversation
| if (nrow(qry_res) == 0) { | ||
| PEcAn.DB::db.close(dbcon) |
There was a problem hiding this comment.
You might want to just use on.exit to specify the db.close right when you open it to make sure that close happens even if any other sort of error occurs
There was a problem hiding this comment.
Yeah sure...that's a good idea...will try n incorporate this for all endpoints
There was a problem hiding this comment.
I think the problem is that the application does not exit, but stays open. We just need to make sure we close teh connections.
| #* @get /<run_id>/graph/<year>/<y_var> | ||
| #* @serializer contentType list(type='image/png') | ||
|
|
||
| plotResults <- function(run_id, year, y_var, x_var="time", width=800, height=600, res){ |
There was a problem hiding this comment.
Not that there's anything 'wrong' with this, but I'd say that grabbing/downloading the model output itself is going to be FAR more important to most users than downloading an image file. Probably grabbing the whole netCDF would be the best place to start, but I could see a lot of folk wanting to subset variables, locations, data ranges, and ensemble members. The latter support it built into THREDDS, but (1) not all of our PEcAn servers support THREDDS (but maybe the right thing to do is to fix that) and (2) regardless we should have some helper functions in the R package to help people download output rather than having to learn to work with THREDDS directly (which is nontrivial)
I think there's also a REALLY high need, even if it's just internal to the project, to be able to correctly grab INPUTS and POSTERIORS via the API. The former will be pretty important to a lot of enduser analyses as well (e.g. plot predicted fluxes vs meteorological inputs)
There was a problem hiding this comment.
This function is needed in the current web application. The current web application can not create the graphs we had in the VM version. This will allow me/us to call this endpoint to get an image back.
There was a problem hiding this comment.
As for the ability to download the outputs/inputs/posteriors I agree, we just have to find the right functions, but I think we have a lot of this framework in place.
There was a problem hiding this comment.
Ahh, yes that makes perfect sense.
As we discussed during last week's planning meeting, we should think about whether it makes sense to rework the whole PHP interface to use the API rather than it's current approach of direct execution of R code.
There was a problem hiding this comment.
shhhhh, but that is my plan :) and maybe even use R and leaflet or other things.
/api/models/{model_id}api/runs/{run_id}if the outputs folder exists on the hostReview Time Estimate
Types of changes
Checklist: