Refactor fallbacks (#27)

Author: juliohm

src/models.jl

@@ -33,7 +33,13 @@ Predict one or multiple variables `vars` at geometry `gₒ` with
 given geostatistical `model`.
 """
 predict(model::FittedGeoStatsModel, var::AbstractString, gₒ) = predict(model, Symbol(var), gₒ)
-predict(model::FittedGeoStatsModel, vars, gₒ) = [predict(model, var, gₒ) for var in vars]
+function predict(model::FittedGeoStatsModel, vars, gₒ)
+  if length(vars) > 1
+    throw(ArgumentError("cannot use univariate model to predict multiple variables"))
+  else
+    [predict(model, first(vars), gₒ)]
+  end
+end
 
 """
     predictprob(model, vars, gₒ)
@@ -42,7 +48,13 @@ Predict distribution of one or multiple variables `vars` at
 geometry `gₒ` with given geostatistical `model`.
 """
 predictprob(model::FittedGeoStatsModel, var::AbstractString, gₒ) = predictprob(model, Symbol(var), gₒ)
-predictprob(model::FittedGeoStatsModel, vars, gₒ) = [predictprob(model, var, gₒ) for var in vars]
+function predictprob(model::FittedGeoStatsModel, vars, gₒ)
+  if length(vars) > 1
+    throw(ArgumentError("cannot use univariate model to predict multiple variables"))
+  else
+    [predictprob(model, first(vars), gₒ)]
+  end
+end
 
 """
     status(fitted)

test/idw.jl

@@ -23,7 +23,7 @@
     @test pred isa Composition
   end
 
-  @testset "Single/Multiple" begin
+  @testset "Fallbacks" begin
     d = georef((; z=[1.0, 0.0, 1.0]), [(0.0, 0.0), (1.0, 0.0), (1.0, 1.0)])
     idw = GeoStatsModels.fit(IDW(), d)
     pred1 = GeoStatsModels.predict(idw, :z, Point(0.0, 0.0))

test/krig.jl

@@ -395,7 +395,7 @@
     @test all(≥(0), var.(dkdist))
   end
 
-  @testset "Single/Multiple" begin
+  @testset "Fallbacks" begin
     d = georef((; z=[1.0, 0.0, 1.0]), [(0.0, 0.0), (1.0, 0.0), (1.0, 1.0)])
     γ = GaussianVariogram()
     ok = GeoStatsModels.fit(OK(γ), d)

test/lwr.jl

@@ -23,7 +23,7 @@
     @test pred isa Composition
   end
 
-  @testset "Single/Multiple" begin
+  @testset "Fallbacks" begin
     d = georef((; z=[1.0, 0.0, 1.0]), [(0.0, 0.0), (1.0, 0.0), (1.0, 1.0)])
     lwr = GeoStatsModels.fit(LWR(), d)
     pred1 = GeoStatsModels.predict(lwr, :z, Point(0.0, 0.0))

test/misc.jl

@@ -3,10 +3,9 @@
 
   # fitpredict with IDW
   pset = PointSet(rand(rng, Point, 3))
-  gtb = georef((a=[1, 2, 3], b=[4, 5, 6]), pset)
+  gtb = georef((z=[1, 2, 3],), pset)
   pred = GeoStatsModels.fitpredict(IDW(), gtb, pset, neighbors=false)
-  @test pred.a == gtb.a
-  @test pred.b == gtb.b
+  @test pred.z == gtb.z
   @test pred.geometry == gtb.geometry
 
   # also works with views

test/nn.jl

@@ -34,7 +34,7 @@
     @test pred isa Composition
   end
 
-  @testset "Single/Multiple" begin
+  @testset "Fallbacks" begin
     d = georef((; z=[1.0, 0.0, 1.0]), [(0.0, 0.0), (1.0, 0.0), (1.0, 1.0)])
     nn = GeoStatsModels.fit(NN(), d)
     pred1 = GeoStatsModels.predict(nn, :z, Point(0.0, 0.0))

test/poly.jl

@@ -36,12 +36,11 @@
 
     # correct schema
     rng = StableRNG(42)
-    d = georef((a=rand(rng, 10), b=rand(rng, 10)), rand(rng, Point, 10))
+    d = georef((z=rand(rng, 10),), rand(rng, Point, 10))
     d̄ = fitpredict(Polynomial(), d)
     t̄ = values(d̄)
-    @test propertynames(t̄) == (:a, :b)
-    @test eltype(t̄.a) == Float64
-    @test eltype(t̄.b) == Float64
+    @test propertynames(t̄) == (:z,)
+    @test eltype(t̄.z) == Float64
 
     # latlon coordinates
     d = georef((; z=[1, 2, 3]), Point.([LatLon(0, 0), LatLon(0, 1), LatLon(1, 0)]))
@@ -68,7 +67,7 @@
     @test pred isa Composition
   end
 
-  @testset "Single/Multiple" begin
+  @testset "Fallbacks" begin
     d = georef((; z=[1.0, 0.0, 1.0]), [(0.0, 0.0), (1.0, 0.0), (1.0, 1.0)])
     poly = GeoStatsModels.fit(Polynomial(), d)
     pred1 = GeoStatsModels.predict(poly, :z, Point(0.0, 0.0))