enable and test general array or number type observables

You can measure complex matrices now!

src/accumulator.jl

@@ -4,7 +4,7 @@ using Statistics
 
 const binning_output_chunk_size = 1000
 
-mutable struct Accumulator{T<:AbstractFloat,N,M}
+mutable struct Accumulator{T<:Number,N,M}
     const bin_length::Int64
     bins::ElasticArray{T,N,M,Vector{T}}
 

src/results.jl

@@ -1,13 +1,13 @@
 import JSON
 
 """Result of a Carlo Monte Carlo calculation containing the mean, statistical error and autocorrelation time."""
-mutable struct ResultObservable{T<:AbstractFloat,N,M}
+mutable struct ResultObservable{T<:Number,R<:Real,N,M}
     internal_bin_length::Int64
     rebin_length::Int64
 
     mean::Array{T,N}
-    error::Array{T,N}
-    autocorrelation_time::Array{T,N}
+    error::Array{R,N}
+    autocorrelation_time::Array{R,N}
 
     rebin_means::Array{T,M}
 end

test/test_merge.jl

@@ -18,7 +18,7 @@ function create_mock_data(
     idx = 1
     for run = 1:runs
         nsamples = samples_per_run + extra_samples * (run == 1)
-        samples = zeros(0)
+        samples = []
         h5open(filenames[run], "w") do file
             meas = Carlo.Measurements(internal_binsize)
             for i = 1:nsamples
@@ -65,7 +65,7 @@ end
                 end
 
                 filenames2, _ = create_mock_data(
-                    idx -> [idx, 1.0];
+                    idx -> [idx+1.0im 1.0; 1.0im 0];
                     runs = runs,
                     obsname = :vec_test,
                     internal_binsize = internal_binsize,
@@ -101,7 +101,7 @@ end
                         vec_obs = results2[:vec_test]
                         @test iszero(vec_obs.error[2])
                         @test vec_obs.error[1] ≈ count_obs.error[1]
-                        @test vec_obs.mean ≈ [count_obs.mean[1], 1.0]
+                        @test vec_obs.mean ≈ [count_obs.mean[1]+1.0im 1.0; 1.0im 0]
                     end
                 end
             end