Feature: Implement constrained k-mer generation and clique search

This commit introduces core functionality for oligonucleotide family tagging, constrained k-mer generation, and clique finding. It also includes updates to the Go language version and refactors the logic for k-mer dissimilarity metrics.

pkg/obikmer/kmer_dissimilarity_test.go

@@ -0,0 +1,113 @@
+package obikmer_test
+
+import (
+	"context"
+	"math/rand"
+	"testing"
+	"time"
+
+	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obigraph"
+	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obikmer"
+)
+
+func allKmers(k int) []uint64 {
+	kmers := make([]uint64, 0, 1<<(2*uint(k)))
+	for km := range obikmer.IterConstrainedKmers(k, obikmer.NoKmerConstraints()) {
+		kmers = append(kmers, km)
+	}
+	return kmers
+}
+
+func TestDissimilarityFilterMatchesDistances(t *testing.T) {
+	const k = 7
+	rng := rand.New(rand.NewSource(1))
+	kmers := make([]uint64, 300)
+	for i := range kmers {
+		kmers[i] = rng.Uint64() & (1<<(2*k) - 1)
+	}
+	candidates := make([]int32, 0, len(kmers))
+	for i := 1; i < len(kmers); i++ {
+		candidates = append(candidates, int32(i))
+	}
+
+	for _, metric := range []obikmer.KmerMetric{obikmer.HammingMetric, obikmer.LCSMetric} {
+		for _, dmin := range []int{1, 3, 5} {
+			filter := obikmer.NewKmerDissimilarityFilter(kmers, k, metric, dmin)
+			got := filter(0, candidates, nil)
+
+			var want []int32
+			for _, c := range candidates {
+				var d int
+				if metric == obikmer.HammingMetric {
+					d = obikmer.HammingDistance(kmers[0], kmers[c], k)
+				} else {
+					d = obikmer.KmerLCS(kmers[0], kmers[c], k).Distance()
+				}
+				if d >= dmin {
+					want = append(want, c)
+				}
+			}
+			if len(got) != len(want) {
+				t.Fatalf("%v d>=%d: got %d vertices, want %d", metric, dmin, len(got), len(want))
+			}
+			for i := range got {
+				if got[i] != want[i] {
+					t.Fatalf("%v d>=%d: mismatch at %d", metric, dmin, i)
+				}
+			}
+		}
+	}
+}
+
+// A family of 5-mers pairwise at Hamming distance >= 3, built with FindClique.
+func TestFindClique_HammingFamily(t *testing.T) {
+	const k, dmin, target = 5, 3, 32
+	kmers := allKmers(k)
+	filter := obikmer.NewKmerDissimilarityFilter(kmers, k, obikmer.HammingMetric, dmin)
+
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+	r := obigraph.FindClique(ctx, len(kmers), filter, target, obigraph.CliqueOptions{Workers: 2})
+	if !r.Reached {
+		t.Fatalf("family of %d not found (best %d, exhausted %v)", target, len(r.Clique), r.Exhausted)
+	}
+	for i, a := range r.Clique {
+		for _, b := range r.Clique[i+1:] {
+			if obikmer.HammingDistance(kmers[a], kmers[b], k) < dmin {
+				t.Fatalf("%d and %d are too close", kmers[a], kmers[b])
+			}
+		}
+	}
+}
+
+func TestFindClique_LCSFamily(t *testing.T) {
+	const k, dmin, target = 6, 3, 20
+	kmers := allKmers(k)
+	filter := obikmer.NewKmerDissimilarityFilter(kmers, k, obikmer.LCSMetric, dmin)
+
+	ctx, cancel := context.WithTimeout(context.Background(), 60*time.Second)
+	defer cancel()
+	r := obigraph.FindClique(ctx, len(kmers), filter, target, obigraph.CliqueOptions{Workers: 2})
+	if !r.Reached {
+		t.Fatalf("family of %d not found (best %d, exhausted %v)", target, len(r.Clique), r.Exhausted)
+	}
+	for i, a := range r.Clique {
+		for _, b := range r.Clique[i+1:] {
+			if obikmer.KmerLCS(kmers[a], kmers[b], k).Distance() < dmin {
+				t.Fatalf("%d and %d are too close", kmers[a], kmers[b])
+			}
+		}
+	}
+}
+
+func TestParseKmerMetric(t *testing.T) {
+	if m, err := obikmer.ParseKmerMetric("LCS"); err != nil || m != obikmer.LCSMetric {
+		t.Fatal("lcs not parsed")
+	}
+	if m, err := obikmer.ParseKmerMetric("hamming"); err != nil || m != obikmer.HammingMetric {
+		t.Fatal("hamming not parsed")
+	}
+	if _, err := obikmer.ParseKmerMetric("levenshtein"); err == nil {
+		t.Fatal("unknown metric must be refused")
+	}
+}