func TestFilterSmallClusters(t *testing.T) {
g := FilterSmallClusters([]*clustering.Cluster{}, 5)
if g == nil {
t.Errorf("result should not be nil")
}
if len(g) != 0 {
t.Errorf("zero group input should result in zero group output")
}
testSet := []*clustering.Cluster{
clustering.NewCluster(&event{Location: geo.NewPoint(1, 1)}),
clustering.NewCluster(&event{Location: geo.NewPoint(1, 1)}, &event{Location: geo.NewPoint(2, 2)}),
}
g = FilterSmallClusters(testSet, 5)
if l := len(g); l != 0 {
t.Errorf("should filter out small groups, but got %d", l)
}
if l := len(testSet); l != 2 {
t.Errorf("should not change test set, but got length %d", l)
}
g = FilterSmallClusters(testSet, 2)
if l := len(g); l != 1 {
t.Errorf("should filter out small groups, but got %d", l)
}
if l := len(testSet); l != 2 {
t.Errorf("should not change test set, but got length %d", l)
}
}
// RemoveOutlierPointersByQuadkey will bucket all pointers by quad key (defined by the level)
// and remove the buckets with less than threshold pointers. The buckets become the resulting point_clustering.Clusters.
func RemoveOutlierPointersByQuadkey(pointers []geo.Pointer, level, threshold int) []*clustering.Cluster {
buckets := make(map[int64][]geo.Pointer)
for _, p := range pointers {
key := p.Point().Quadkey(level)
buckets[key] = append(buckets[key], p)
}
clusters := make([]*clustering.Cluster, 0, len(buckets))
for _, b := range buckets {
if len(b) >= threshold {
clusters = append(clusters, clustering.NewCluster(b...))
}
}
return clusters
}
// RematchPointersToClusters will take a set of pointers and map them to the closest cluster.
// Basically creates a new cluster from that one point and does the ClusterDistance between them.
// Will return a new list.
func RematchPointersToClusters(
clusters []*clustering.Cluster,
pointers []geo.Pointer,
distancer clustering.ClusterDistancer,
threshold float64,
) []*clustering.Cluster {
if len(clusters) == 0 {
return []*clustering.Cluster{}
}
newClusters := make([]*clustering.Cluster, 0, len(clusters))
// clear the current members
for _, c := range clusters {
newClusters = append(newClusters, clustering.NewClusterWithCentroid(c.Centroid))
}
// remap all the groupers to these new groups
for _, pointer := range pointers {
minDist := math.MaxFloat64
index := 0
pointerCluster := clustering.NewCluster(pointer)
// find the closest group
for i, c := range newClusters {
if d := distancer.ClusterDistance(c, pointerCluster); d < minDist {
minDist = d
index = i
}
}
if minDist < threshold {
// leaves the center as found by the previous clustering
newClusters[index].Pointers = append(newClusters[index].Pointers, pointer)
}
}
return newClusters
}