add tests

Author: Datseris

README.md

@@ -16,12 +16,11 @@ which interplay with the functions:
 To create new clustering algorithms simply create a new
 subtype of `ClusteringAlgorithm` that extends `cluster`
 so that it returns a new subtype of `ClusteringResult`.
-The result must extend `cluster_number, cluster_labels`
+This result must extend `cluster_number, cluster_labels`
 and optionally `cluster_probs`.
 
-Note that data input type must always be `AbstractVector` of vectors
-(anything that can have distance defined).
-Two helper functions `each_data_point, input_data_size` can help
-making this harmonious with matrix inputs.
+For developers: see two helper functions `each_data_point, input_data_size`
+so that you can support matrix input while abiding the declared api
+of iterable of vectors as input.
 
 For more, see the docstring of `cluster`.

src/ClusteringAPI.jl

@@ -66,8 +66,14 @@ Return the cluster probabilities of the data points used in [`cluster`](@ref).
 They are length-`n` vectors containing the "probabilities" or "score" of each point
 belonging to one of the created clusters (used with fuzzy clustering algorithms).
 """
-function cluster_labels(cr::ClusteringResults)
-    return cr.labels # typically there
+function cluster_probs(cr::ClusteringResults)
+    labels = cluster_labels(cr)
+    n = cluster_number(cr)
+    probs = [zeros(Real, n) for _ in 1:length(labels)]
+    for (i, label) in enumerate(labels)
+        probs[i][label] = 1
+    end
+    return probs
 end
 
 # two helper functions for agnostic input data type

test/runtests.jl

@@ -1,5 +1,25 @@
 using Test
 using ClusteringAPI
 
-@test true
+struct TestClustering <: ClusteringAlgorithm
+end
+struct TestResults <: ClusteringResults
+    labels::Vector{Int}
+    n::Int
+end
 
+function ClusteringAPI.cluster(::TestClustering, data)
+    return TestResults(fill(1, length(data)), 2)
+end
+
+cr = cluster(TestClustering(), randn(100))
+@test cluster_number(cr) == 1
+@test cluster_labels(cr) == fill(1, 100)
+@test cluster_probs(cr) == fill([1.0], 100)
+
+@test ClusteringAPI.input_data_size([rand(3) for _ in 1:30]) == (3, 30)
+@test ClusteringAPI.input_data_size(rand(3,30)) == (3, 30)
+
+v = [ones(3) for _ in 1:30]
+@test ClusteringAPI.each_data_point(v) == v
+@test ClusteringAPI.each_data_point(ones(3,30)) == v