gh-1168 [skip-ci] WIP unit test sempath builder

usecases/sempath/builder.go

@@ -22,6 +22,7 @@ import (
 	"strings"
 	"time"
 
+	"github.com/davecgh/go-spew/spew"
 	"github.com/etiennedi/go-tsne/tsne"
 	"github.com/pkg/errors"
 	"github.com/semi-technologies/weaviate/entities/models"
@@ -118,6 +119,8 @@ func (f *PathBuilder) calculatePathPerObject(obj search.Result, allObjects []sea
 		}
 	}
 
+	spew.Dump(projectedNeighbors)
+
 	path := f.buildPath(projectedNeighbors, projectedSearchVector, projectedTargetVector)
 	return f.addDistancesToPath(path, neighbors, params.SearchVector, obj.Vector)
 }
@@ -427,5 +430,7 @@ func cosineDist(a, b []float32) (float32, error) {
 		return 0, err
 	}
 
+	fmt.Printf("input 1: %f\ninput 2: %f\n: sim: %f\n\n", a, b, sim)
+
 	return 1 - sim, nil
 }

usecases/sempath/builder_test.go

@@ -0,0 +1,98 @@
+package sempath
+
+import (
+	"context"
+	"testing"
+
+	"github.com/davecgh/go-spew/spew"
+	"github.com/semi-technologies/weaviate/entities/models"
+	"github.com/semi-technologies/weaviate/entities/schema/kind"
+	"github.com/semi-technologies/weaviate/entities/search"
+	"github.com/stretchr/testify/require"
+)
+
+func TestSemanticPathBuilder(t *testing.T) {
+	c11y := &fakeC11y{}
+	b := New(c11y)
+
+	b.fixedSeed = 1000 // control randomness in unit test
+
+	input := []search.Result{
+		search.Result{
+			ID:        "7fe919ed-2ef6-4087-856c-a307046bf895",
+			Kind:      kind.Thing,
+			ClassName: "Foo",
+			Vector:    []float32{1, 0.1},
+		},
+	}
+	searchVector := []float32{0.3, 0.3}
+
+	c11y.neighbors = []*models.NearestNeighbors{
+		&models.NearestNeighbors{
+			Neighbors: []*models.NearestNeighbor{
+				&models.NearestNeighbor{
+					Concept: "good1",
+					Vector:  []float32{0.5, 0.1},
+				},
+				&models.NearestNeighbor{
+					Concept: "good2",
+					Vector:  []float32{0.7, 0.2},
+				},
+				&models.NearestNeighbor{
+					Concept: "good3",
+					Vector:  []float32{0.9, 0.1},
+				},
+				&models.NearestNeighbor{
+					Concept: "good4",
+					Vector:  []float32{0.55, 0.1},
+				},
+				&models.NearestNeighbor{
+					Concept: "good5",
+					Vector:  []float32{0.77, 0.2},
+				},
+				&models.NearestNeighbor{
+					Concept: "good6",
+					Vector:  []float32{0.99, 0.1},
+				},
+				&models.NearestNeighbor{
+					Concept: "bad1",
+					Vector:  []float32{-0.1, -3},
+				},
+				&models.NearestNeighbor{
+					Concept: "bad2",
+					Vector:  []float32{-0.15, -2.75},
+				},
+				&models.NearestNeighbor{
+					Concept: "bad3",
+					Vector:  []float32{-0.22, -2.35},
+				},
+				&models.NearestNeighbor{
+					Concept: "bad4",
+					Vector:  []float32{0.1, -3.3},
+				},
+				&models.NearestNeighbor{
+					Concept: "bad5",
+					Vector:  []float32{0.15, -2.5},
+				},
+				&models.NearestNeighbor{
+					Concept: "bad6",
+					Vector:  []float32{-0.4, -2.25},
+				},
+			},
+		},
+	}
+
+	res, err := b.CalculatePath(input, &Params{SearchVector: searchVector})
+	require.Nil(t, err)
+
+	spew.Dump(res)
+	t.Fail()
+}
+
+type fakeC11y struct {
+	neighbors []*models.NearestNeighbors
+}
+
+func (f *fakeC11y) MultiNearestWordsByVector(ctx context.Context, vectors [][]float32, k, n int) ([]*models.NearestNeighbors, error) {
+	return f.neighbors, nil
+}