Set process metadata?

[Matthias Mohr]

In JS I can do something like:

let builder = await connection.buildProcess();

// Set the metadata for the process
builder.description = "# Moisture Stress Index\nCalculate MSI for an area specifying co...";
builder.id = "MSI";
builder.categories = ["foo", "bar"];
builder.summary = "MSI calculation";

let loadcollection1 = builder.load_collection(...);

print(builder.toJSON());

Which allows me to print the full metadata with all the additional metadata such as id, summary, categories. I'm not sure how I can set the metadata in R though. Is it possible? Or is this a missing feature?

I need this for code generation for example.

[Florian Lahn]

I try to state an example how it is currently implemented by creating the NDVI on GEE:

  gee_host_url = "https://earthengine.openeo.org"
  gee = connect(host = gee_host_url)
  p = processes()
  data1 = p$load_collection(id = "COPERNICUS/S2",
                            spatial_extent = list(west=-2.7634,south=43.0408,east=-1.121,north=43.8385),
                            temporal_extent = c("2018-04-30","2018-06-26"),
                            bands = c("B4","B8"))
 
  ndvi = p$reduce_dimension(data = data1, dimension = "bands",reducer = function(x, context) {
    B4 = x["B4"]
    B8 = x["B8"]
    return(p$normalized_difference(x = B4,y = B8))
  })
  reducer = p$reduce_dimension(data = ndvi,dimension = "t", reducer = function(x, context) {
    min(x)
  })
  apply_linear_transform = p$apply(data = reducer, process = function(x,context) {
    p$linear_scale_range(x = x, inputMin = -1, inputMax = 1,outputMin = 0,outputMax = 255)
  })
  final = p$save_result(data = apply_linear_transform,format = "PNG")
  udp = as(final,"Process")

In this case udp is the user defined process and it is a R6 object (the object oriented approach of R). As such I have implemented the functions setSummary and setDescription on that object (getter respectively). You can set the meta data manually with:

udp$setSummary("summary")
udp$setDescription("description")

However, in R you normally don't need the description here, so I have put those arguments in create_user_process() and update_user_process(), where you can state those information.

I think the field categories was not defined in API 1.0.0, so this will be implemented, once I start to catch up to the changes for 1.1.0.

[Matthias Mohr]

Thanks, I'll try to make this work in code gen.

Categories were available in 1.0.0 already, see https://api.openeo.org/1.0.0/index.html#operation/list-processes

So basically, I'd need setId, setSummary, setDescription, setCategories, setParameters, setReturns, setDeprecated, setExperimental, setExceptions, setExamples and setLinks (although setParameters and setReturns may be offered in a different way).

[Florian Lahn]

Yes, parameters and returns are derived automatically. Parameters are simply variables in the graph (create_variable) and "return" is derived from the final node. For the others setId, setSummary and setDescription are realized. Unfortunately the others are currently not supported, yet.

[Matthias Mohr]

The use case here is to transform a well-defined process (e.g. with pre-defined parameter descriptions and return values) from JSON to R code automatically, so the auto-derivation doesn't really make sense as it would override what the user already wrote. Anyway, I'll do it step-by-step then and start with what is available and add the other stuff later once we make progress with this issue.