Move indicator parameters to yaml

[karpfen]

Currently, most indicator parameters (e.g. min_size in src/gfw.R) are hard coded in the individual scripts in src. I propose to extend the config.yml file to include these parameters. I'm thinking of something that looks like this in the end:

default:
  input: ./data/WDPA_WDOECM_Jun2024_Public.parquet
  ncores: 12
  data_path: ./data
  output_path: ./output
  progress: true
  by_region: false
  timeout: 600
  overwrite: false
  wdpa_ver: "latest"
  params:
    resources:
      get_key_biodiversity_areas:
        path: "./kbas.gpkg"
      get_gfw_treecover:
        version: "GFC-2023-v1.11"
    indicators:
      calc_carbon:
        stats: c("min", "mean", "max")
    calc_treecover_area:
      years: 2000:2010
      min_size: 20
      min_cover: 50

Does that make sense to you, @goergen95?

[karpfen]

@goergen95 I noticed one potential issue when I played around with this approach yesterday. Ideally, everything that's set under params should be validated against the package imho.

This means in the case above:

• Check if resource get_key_biodiversity_areas exists
• Check if resource get_key_biodiversity_areas has a parameter named path
• Check if resource get_gfw_treecover exists
• Check if resource get_gfw_treecover has a parameter named version
• .....

However, available_resources() does not give us the actual function name (e.g. get_gfw_treecover), but instead the "alias" gfw_treecover. We need the alias to check against available_resources()$name, but for the actual call we then need the real function name; so we would need to get that mapping somewhere. I don't think that's possible right now, no?

[goergen95]

We will require validation, but I see this happening in two stages:

• validation of the yaml content: this would only check if parameters in the yaml are specified correctly without any further checks against the software. We would need to develop a schema and some tooling for validation (see e.g. here).
• during runtime, we would need to construct mapme.biodiversity function calls, so from your example we would construct get_key_biodiversity__areas(path = "./kbas.gpkg"). If errors occur here we should expose those to the user

In summary, I do not see the need to go over available_resources() for this to work.

[goergen95]

Something like this should get the job done:

f <- getFromNamespace("get_key_biodiversity_areas", "mapme.biodiversity")
f(path = "./kbas.gpkg")

[goergen95]

Taking first steps here. Validating the configuration via JSON Schema looks fine. Next steps are to figure out how to represent resources/indicators as objects in JSON Schema.

Update: That seems to work now, too. Next up is some R logic to construct proper function calls from the validated yaml and start the processing.

[goergen95]

I think the schema branch is ready for testing now. :crossed_fingers: