reduce test time, and convert all UniformScalings to Diagonals

src/Observations.jl

@@ -126,13 +126,25 @@ function Observation(obs_dict::Dict)
             snew = samples_tmp  # [[1,2,3]]
         end
     end
+
     if !isa(covariances, AbstractVector) # [2 1;1 2] -> [[2 1;1 2]]
-        cnew = [covariances]
+        ctmp = [covariances]
     else
-        T = promote_type((typeof(c) for c in covariances)...)
-        cnew = [convert(T, c) for c in covariances] # to re-infer eltype
+        ctmp = covariances
     end
-
+    # additionally provide a dimension for UniformScalings for covariances
+    ctmp2 = []
+    for (id,c) in enumerate(ctmp)
+        if isa(c,UniformScaling)
+            push!(ctmp2, Diagonal(c.λ*ones(length(snew[id])))) # get dim from samples
+        else
+            push!(ctmp2, c)
+        end
+    end 
+    # then promote
+    T = promote_type((typeof(c) for c in ctmp2)...)
+    cnew = [convert(T, c) for c in ctmp2] # to re-infer eltype
+
     if !("inv_covariances" ∈ collect(keys(obs_dict)))
         inv_covariances = []
         for c in cnew # ensures its a vector
@@ -142,12 +154,22 @@ function Observation(obs_dict::Dict)
         inv_covariances = obs_dict["inv_covariances"]
     end
     if !isa(inv_covariances, AbstractVector) # [2 1;1 2] -> [[2 1;1 2]]
-        icnew = [inv_covariances]
+        ictmp = [inv_covariances]
     else
-        T = promote_type((typeof(c) for c in inv_covariances)...)
-        icnew = [convert(T, c) for c in inv_covariances] # to re-infer eltype
+        ictmp = inv_covariances
     end
-
+    # additionally provide a dimension for UniformScalings
+    ictmp2=[]
+    for (id,c) in enumerate(ictmp)
+        if isa(c,UniformScaling)
+            push!(ictmp2, Diagonal(c.λ*ones(length(snew[id])))) # get dim from samples
+        else
+            push!(ictmp2, c)
+        end
+    end 
+    T = promote_type((typeof(c) for c in ictmp2)...)
+    icnew = [convert(T, c) for c in ictmp2] # to re-infer eltype
+
     if !isa(names, AbstractVector) # "name" -> ["name"]
         nnew = [names]
     else
@@ -249,13 +271,7 @@ function get_obs_noise_cov(o::Observation; build = true)
         covs = get_covs(o)
         cov_full = zeros(maximum(indices[end]), maximum(indices[end]))
         for (idx, c) in zip(indices, covs)
-            if isa(c, UniformScaling)
-                for idxx in idx
-                    cov_full[idxx, idxx] = c.λ
-                end
-            else
-                cov_full[idx, idx] .= c
-            end
+            cov_full[idx, idx] .= c
         end
 
         return cov_full
@@ -276,13 +292,7 @@ function get_obs_noise_cov_inv(o::Observation; build = true)
         inv_covs = get_inv_covs(o)
         inv_cov_full = zeros(maximum(indices[end]), maximum(indices[end]))
         for (idx, c) in zip(indices, inv_covs)
-            if isa(c, UniformScaling)
-                for idxx in idx
-                    inv_cov_full[idxx, idxx] = c.λ
-                end
-            else
-                inv_cov_full[idx, idx] .= c
-            end
+            inv_cov_full[idx, idx] .= c            
         end
 
         return inv_cov_full

test/EnsembleKalmanProcess/runtests.jl

@@ -932,15 +932,16 @@ end
             failure_handler_method = SampleSuccGauss(),
         )
         T = 0.0
-        for i in 1:N_iter
+        N_iter_new = 5
+        for i in 1:N_iter_new
             params_i = get_ϕ_final(prior, ekiobj)
             g_ens = G_test(params_i)
             dt = @elapsed EKP.update_ensemble!(ekiobj, g_ens)
             T += dt
         end
         # Skip timing of first due to precompilation
         if i >= 2
-            @info "ETKI with $n_obs_test observations took $T seconds. (average update taking: $(T/Float64(N_iter)))"
+            @info "$N_iter_new iterations of ETKI with $n_obs_test observations took $T seconds. (avg update: $(T/Float64(N_iter)))"
         end
     end
 end