nicer experiment config

src/ekp_interface.jl

@@ -10,17 +10,20 @@ import ClimaComms
 export ExperimentConfig
 
 struct ExperimentConfig
-    id
-    n_iterations
-    ensemble_size
-    observations
-    noise
-    prior
-    output_dir
+    id::AbstractString
+    n_iterations::Integer
+    ensemble_size::Integer
+    observations::Any
+    noise::Any
+    prior::Any
+    output_dir::Any
 end
 
 """
-    ExperimentConfig(experiment_id; 
+    ExperimentConfig(experiment_id)
+
+    ExperimentConfig(
+        experiment_id,
         n_iterations,
         ensemble_size,
         observations,
@@ -29,38 +32,38 @@ end
         output_dir,
     )
 
-ExperimentConfig constructor. If an individual keyword argument is not given, 
-default is obtained from the YAML at `get_ekp_yaml(experiment_id)`.
+ExperimentConfig stores the configuration for a specific experiment.
+If only the `experiment_id` is passed, the config will be loaded from
+`experiments/experiment_id/ekp_config.yml`
 """
-function ExperimentConfig(experiment_id; kwargs...)
-    # config_dict is a Dict read from a YAML file
-    config_dict = get_ekp_yaml(experiment_id)
+function ExperimentConfig(experiment_id)
+    config_yaml = joinpath("experiments", experiment_id, "ekp_config.yml")
+    config_dict = isfile(config_yaml) ? Dict() : YAML.load_file(config_yaml)
+
+    required_fields = ["n_iterations",
+    "ensemble_size",
+    "prior_path",
+    "observations",
+    "noise"]
+    @assert issubset(required_fields, keys(config_dict))
 
     default_output =
         haskey(ENV, "CI") ? experiment_id : joinpath("output", experiment_id)
-    config_dict["output_dir"] = get(config_dict, "output_dir", default_output)
+    output_dir = get(config_dict, "output_dir", default_output)
 
-    for key in ["observations", "noise"]
-        if haskey(config_dict, key)
-            config_dict[key] = JLD2.load_object(config_dict[key])
-        end
-    end
-
-    if haskey(config_dict, "prior_path") 
-        config_dict["prior"] = get_prior(config_dict["prior_path"])
-    end
+    observations = JLD2.load_object(config_dict["observations"])
+    noise = JLD2.load_object(config_dict["noise"])
 
-    config_kwargs = Dict(Symbol(key) => config_dict[key] for key in keys(config_dict))
-    merged_kwargs = merge(config_kwargs, kwargs)
+    prior = get_prior(config_dict["prior_path"])
 
     return ExperimentConfig(
         experiment_id,
-        merged_kwargs[:n_iterations],
-        merged_kwargs[:ensemble_size],
-        merged_kwargs[:observations],
-        merged_kwargs[:noise],
-        merged_kwargs[:prior],
-        merged_kwargs[:output_dir],
+        config_dict["n_iterations"],
+        config_dict["ensemble_size"],
+        observations,
+        noise,
+        prior,
+        output_dir,
     )
 end
 
@@ -90,15 +93,6 @@ function get_prior(param_dict::AbstractDict; names = nothing)
     return prior
 end
 
-"""
-    get_ekp_yaml(experiment_id)
-
-Load the EKP configuration for a given `experiment_id`. If no file is found, return an empty Dict()
-"""
-function get_ekp_yaml(experiment_id)
-    config_yaml = joinpath("experiments", experiment_id, "ekp_config.yml")
-    return isfile(config_yaml) ? Dict() : YAML.load_file(config_yaml)
-end
 """
     save_G_ensemble(experiment_id, iteration, G_ensemble)