improves support for new database abstractions

Project.toml

@@ -1,7 +1,7 @@
 name = "KNearestCenters"
 uuid = "4dca28ae-43b8-11eb-1f5e-d55054101997"
 authors = ["Eric S. Tellez"]
-version = "0.6.2"
+version = "0.7.0"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
@@ -17,7 +17,7 @@ StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 
 [compat]
 CategoricalArrays = "0.8, 0.9, 0.10"
-KCenters = "0.5.0"
+KCenters = "0.6"
 MLDataUtils = "0.5"
 Parameters = "0.12"
 SearchModels = "0.3"

src/knc.jl

@@ -54,7 +54,7 @@ end
 
 Creates a Knc classifier using the given configuration and data.
 """
-function Knc(config::KncConfig, X, y::CategoricalArray; verbose=true)
+function Knc(config::KncConfig, X::AbstractDatabase, y::CategoricalArray; verbose=true)
     # computes a set of #labels centers using labels for clustering
     D = kcenters(config.kernel.dist, X, y, config.centerselection)
     @assert length(levels(y)) == length(D.centers)

src/kncproto.jl

@@ -126,9 +126,8 @@ end
 
 Creates a KncProto classifier using the given configuration and data.
 """
-function KncProto(config::KncProtoConfig, X, y::CategoricalArray; verbose=true)
+function KncProto(config::KncProtoConfig, X::AbstractDatabase, y::CategoricalArray; verbose=true)
     config.ncenters == 0 && error("invalid ncenter $ncenters; ncenters <= -2 or 2 <= ncenters; please use plain Knc otherwise")
-    X = convert(AbstractDatabase, X)
     if config.ncenters > 0
         # computes a set of ncenters for all dataset
         verbose && println(stderr, "KncProto> clustering data without knowing labels", config)
@@ -144,7 +143,7 @@ function KncProto(config::KncProtoConfig, X, y::CategoricalArray; verbose=true)
         ncenters = abs(config.ncenters)
         verbose && println(stderr, "KncProto> clustering data with label division", config)
         nclasses = length(levels(y))
-        centers = VectorDatabase(eltype(X))
+        centers = eltype(X)[]
         dmax = Float32[]
         class_map = Int32[]
         nclasses = length(levels(y))
@@ -170,19 +169,19 @@ function KncProto(config::KncProtoConfig, X, y::CategoricalArray; verbose=true)
             end
         end
 
-        KncProto(config, centers, dmax, class_map, convert(Int32, nclasses), KnnResult(1))
+        KncProto(config, VectorDatabase(centers), dmax, class_map, convert(Int32, nclasses), KnnResult(1))
     end
 end
 
 function KncProto(
         config::KncProtoConfig,
         input_clusters::ClusteringData,
-        train_X,
+        train_X::AbstractDatabase,
         train_y::CategoricalArray;
         verbose=false
     )
     train_X = convert(AbstractDatabase, train_X)
-    centers = VectorDatabase(eltype(train_X)) # clusters
+    centers = eltype(train_X)[] # clusters
     classes = Int32[] # class mapping between clusters and classes
     dmax = Float32[]
     ncenters = length(input_clusters.centers)
@@ -243,7 +242,7 @@ function KncProto(
     end
 
     verbose && println(stderr, "finished with $(length(centers)) centers; started with $(length(input_clusters.centers))")
-    KncProto(config, centers, dmax, classes, convert(Int32, nclasses), KnnResult(1))
+    KncProto(config, VectorDatabase(centers), dmax, classes, convert(Int32, nclasses), KnnResult(1))
 end
 
 """

test/loaddata.jl

@@ -1,6 +1,6 @@
 # This file is a part of KNearestCenters.jl
 
-using Test
+using Test, SimilaritySearch
 
 function loadiris()
     url = "https://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data"
@@ -19,7 +19,7 @@ function loadiris()
         push!(y, arr[end])
     end
 
-    X, y
+    VectorDatabase(X), y
 end