Merge #135230

135230: vecindex: support deleting vectors from C-SPANN index r=drewkimball a=andy-kimball

Add Delete method to the vector index, which attempts to remove a vector from the index, given its value and primary key. Delete may not be able to locate the vector in the index, leaving a "dangling vector" reference which other methods need to take care not to ignore.

Epic: CRDB-42943

Release note: None

Co-authored-by: Andrew Kimball <[email protected]>

pkg/sql/vecindex/testdata/delete.ddt

@@ -1,5 +1,141 @@
 # ----------
-# Test deleting vectors from primary index, but not from secondary index.
+# Construct new index with one vector in the root.
+# ----------
+new-index min-partition-size=1 max-partition-size=3 beam-size=2
+vec1: (1, 2)
+----
+• 1 (0, 0)
+│
+└───• vec1 (1, 2)
+
+# Delete remaining vector in the root.
+delete
+vec1
+----
+• 1 (0, 0)
+
+# ----------
+# Construct new index with only duplicate vectors.
+# ----------
+new-index min-partition-size=1 max-partition-size=3 beam-size=2
+vec1: (1, 2)
+vec2: (1, 2)
+vec3: (1, 2)
+vec4: (1, 2)
+vec5: (1, 2)
+vec6: (1, 2)
+----
+• 1 (1, 2)
+│
+├───• 2 (1, 2)
+│   │
+│   ├───• vec1 (1, 2)
+│   └───• vec2 (1, 2)
+│
+└───• 3 (1, 2)
+    │
+    ├───• vec3 (1, 2)
+    ├───• vec4 (1, 2)
+    ├───• vec5 (1, 2)
+    └───• vec6 (1, 2)
+
+# Ensure the correct duplicates are deleted (i.e. with matching keys).
+delete
+vec1
+vec5
+----
+• 1 (1, 2)
+│
+├───• 2 (1, 2)
+│   │
+│   └───• vec2 (1, 2)
+│
+└───• 3 (1, 2)
+    │
+    ├───• vec3 (1, 2)
+    ├───• vec4 (1, 2)
+    └───• vec6 (1, 2)
+
+# ----------
+# Construct new index with multiple levels.
+# ----------
+new-index min-partition-size=1 max-partition-size=3 beam-size=1
+vec1: (1, 2)
+vec2: (7, 4)
+vec3: (4, 3)
+vec4: (-4, 5)
+vec5: (1, 11)
+vec6: (1, -6)
+vec7: (0, 4)
+vec8: (-2, 8)
+vec9: (2, 8)
+----
+• 1 (1.5, 1.875)
+│
+├───• 2 (1, -2)
+│   │
+│   ├───• vec1 (1, 2)
+│   └───• vec6 (1, -6)
+│
+├───• 4 (1.75, 4)
+│   │
+│   ├───• vec3 (4, 3)
+│   ├───• vec4 (-4, 5)
+│   ├───• vec7 (0, 4)
+│   └───• vec2 (7, 4)
+│
+└───• 5 (0.3333, 9)
+    │
+    ├───• vec5 (1, 11)
+    ├───• vec8 (-2, 8)
+    └───• vec9 (2, 8)
+
+# Test case where initial search fails to find vector to delete and it must be
+# retried.
+delete
+vec1: (0, 8)
+----
+• 1 (1.5, 1.875)
+│
+├───• 2 (1, -2)
+│   │
+│   └───• vec6 (1, -6)
+│
+├───• 4 (1.75, 4)
+│   │
+│   ├───• vec3 (4, 3)
+│   ├───• vec4 (-4, 5)
+│   ├───• vec7 (0, 4)
+│   └───• vec2 (7, 4)
+│
+└───• 5 (0.3333, 9)
+    │
+    ├───• vec5 (1, 11)
+    ├───• vec8 (-2, 8)
+    └───• vec9 (2, 8)
+
+# Delete multiple vectors.
+delete
+vec4
+vec5
+vec6
+----
+• 1 (1.5, 1.875)
+│
+├───• 2 (1, -2)
+├───• 4 (1.75, 4)
+│   │
+│   ├───• vec3 (4, 3)
+│   ├───• vec2 (7, 4)
+│   └───• vec7 (0, 4)
+│
+└───• 5 (0.3333, 9)
+    │
+    ├───• vec9 (2, 8)
+    └───• vec8 (-2, 8)
+
+# ----------
+# Construct new index with multiple levels.
 # ----------
 new-index min-partition-size=1 max-partition-size=3 beam-size=2
 vec1: (1, 2)

pkg/sql/vecindex/testdata/insert.ddt

@@ -1,3 +1,6 @@
+# ----------
+# Construct empty index.
+# ----------
 new-index min-partition-size=1 max-partition-size=4 beam-size=2
 ----
 • 1 (0, 0)

pkg/sql/vecindex/vecstore/in_memory_store.go

@@ -352,6 +352,15 @@ func (s *InMemoryStore) DeleteVector(txn Txn, key PrimaryKey) {
 	delete(s.mu.vectors, string(key))
 }
 
+// GetVector returns a single vector from the store, by its primary key. This
+// is used for testing.
+func (s *InMemoryStore) GetVector(key PrimaryKey) vector.T {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	return s.mu.vectors[string(key)]
+}
+
 // GetAllVectors returns all vectors that have been added to the store as key
 // and vector pairs. This is used for testing.
 func (s *InMemoryStore) GetAllVectors() []VectorWithKey {

pkg/sql/vecindex/vecstore/search_set.go

@@ -6,6 +6,7 @@
 package vecstore
 
 import (
+	"bytes"
 	"sort"
 
 	"github.com/cockroachdb/cockroach/pkg/util/container/heap"
@@ -168,6 +169,10 @@ type SearchSet struct {
 	// among the best results.
 	MaxExtraResults int
 
+	// MatchKey, if non-nil, filters out all search candidates that do not have
+	// a matching primary key.
+	MatchKey PrimaryKey
+
 	// Stats tracks useful information about the search, such as how many vectors
 	// and partitions were scanned.
 	Stats SearchStats
@@ -180,6 +185,11 @@ type SearchSet struct {
 // Add includes a new candidate in the search set. If set limits have been
 // reached, then the candidate with the farthest distance will be discarded.
 func (ss *SearchSet) Add(candidate *SearchResult) {
+	if ss.MatchKey != nil && !bytes.Equal(ss.MatchKey, candidate.ChildKey.PrimaryKey) {
+		// Filter out candidates without a matching primary key.
+		return
+	}
+
 	// Fast path where no pruning is necessary.
 	if len(ss.results) < ss.MaxResults {
 		heap.Push[*SearchResult](&ss.results, candidate)

pkg/sql/vecindex/vecstore/search_set_test.go

@@ -132,4 +132,9 @@ func TestSearchSet(t *testing.T) {
 	otherSet.MaxExtraResults = 1
 	otherSet.AddAll(SearchResults{result1, result2, result3, result4, result5, result6, result7})
 	require.Equal(t, SearchResults{result3, result1, result4, result7}, otherSet.PopResults())
+
+	// Ignore results without a matching primary key.
+	otherSet = SearchSet{MaxResults: 2, MatchKey: []byte{60}}
+	otherSet.AddAll(SearchResults{result1, result2, result3, result4, result5, result6, result7})
+	require.Equal(t, SearchResults{result6}, otherSet.PopResults())
 }

pkg/sql/vecindex/vector_index.go

@@ -209,6 +209,63 @@ func (vi *VectorIndex) Insert(
 	return vi.insertHelper(&parentSearchCtx, childKey, true /* allowRetry */)
 }
 
+// Delete attempts to remove a vector from the index, given its value and
+// primary key. This is called within the scope of a transaction so that the
+// index does not appear to change during the delete.
+//
+// NOTE: Delete may not be able to locate the vector in the index, meaning a
+// "dangling vector" reference will be left in the tree. Vector index methods
+// handle this rare case by checking for duplicates when returning search
+// results.
+func (vi *VectorIndex) Delete(
+	ctx context.Context, txn vecstore.Txn, vector vector.T, key vecstore.PrimaryKey,
+) error {
+	// Search for the vector in the index.
+	searchCtx := searchContext{
+		Txn:      txn,
+		Original: vector,
+		Level:    vecstore.LeafLevel,
+		Options: SearchOptions{
+			SkipRerank: vi.options.DisableErrorBounds,
+		},
+	}
+	searchCtx.Ctx = internal.WithWorkspace(ctx, &searchCtx.Workspace)
+
+	// Randomize the vector if required by the quantizer.
+	tempRandomized := searchCtx.Workspace.AllocVector(vi.quantizer.GetRandomDims())
+	defer searchCtx.Workspace.FreeVector(tempRandomized)
+	vi.quantizer.RandomizeVector(ctx, vector, tempRandomized, false /* invert */)
+	searchCtx.Randomized = tempRandomized
+
+	searchSet := vecstore.SearchSet{MaxResults: 1, MatchKey: key}
+
+	// Search with the base beam size. If that fails to find the vector, try again
+	// with a larger beam size, in order to minimize the chance of dangling
+	// vector references in the index.
+	baseBeamSize := max(vi.options.BaseBeamSize, 1)
+	for {
+		searchCtx.Options.BaseBeamSize = baseBeamSize
+
+		err := vi.searchHelper(&searchCtx, &searchSet, true /* allowRetry */)
+		if err != nil {
+			return err
+		}
+		results := searchSet.PopResults()
+		if len(results) == 0 {
+			// Retry search with significantly higher beam size.
+			if baseBeamSize == vi.options.BaseBeamSize {
+				baseBeamSize *= 8
+				continue
+			}
+			return nil
+		}
+
+		// Remove the vector from its partition in the store.
+		_, err = vi.removeFromPartition(ctx, txn, results[0].ParentPartitionKey, results[0].ChildKey)
+		return err
+	}
+}
+
 // Search finds vectors in the index that are closest to the given query vector
 // and returns them in the search set. Set searchSet.MaxResults to limit the
 // number of results. This is called within the scope of a transaction so that

pkg/sql/vecindex/vector_index_test.go

@@ -269,31 +269,52 @@ func (s *testState) Delete(d *datadriven.TestData) string {
 		}
 	}
 
-	txn := beginTransaction(s.Ctx, s.T, s.InMemStore)
-	defer commitTransaction(s.Ctx, s.T, s.InMemStore, txn)
+	for i, line := range strings.Split(d.Input, "\n") {
+		line = strings.TrimSpace(line)
+		if len(line) == 0 {
+			continue
+		}
 
-	// Get root in order to acquire partition lock.
-	_, err := s.InMemStore.GetPartition(s.Ctx, txn, vecstore.RootKey)
-	require.NoError(s.T, err)
+		// If vector to delete has a colon, then its value is specified as well
+		// as its name. This is useful for forcing a certain value to delete.
+		var key vecstore.PrimaryKey
+		var vec vector.T
+		parts := strings.Split(line, ":")
+		if len(parts) == 1 {
+			// Get the value from the store.
+			key = vecstore.PrimaryKey(line)
+			vec = s.InMemStore.GetVector(key)
+		} else {
+			require.Len(s.T, parts, 2)
+			// Parse the value after the colon.
+			key = vecstore.PrimaryKey(parts[0])
+			vec = s.parseVector(parts[1])
+		}
 
-	if notFound {
-		for _, line := range strings.Split(d.Input, "\n") {
-			line = strings.TrimSpace(line)
-			if len(line) == 0 {
-				continue
-			}
+		// Delete within the scope of a transaction.
+		txn := beginTransaction(s.Ctx, s.T, s.InMemStore)
 
-			// Simulate case where the vector is deleted in the primary index, but
-			// it cannot be found in the secondary index.
-			s.InMemStore.DeleteVector(txn, []byte(line))
+		// If notFound=true, then simulate case where the vector is deleted in
+		// the primary index, but it cannot be found in the secondary index.
+		if !notFound {
+			err := s.Index.Delete(s.Ctx, txn, vec, key)
+			require.NoError(s.T, err)
+		}
+		s.InMemStore.DeleteVector(txn, key)
+
+		commitTransaction(s.Ctx, s.T, s.InMemStore, txn)
+
+		if (i+1)%s.Options.MaxPartitionSize == 0 {
+			// Periodically, run synchronous fixups so that test results are
+			// deterministic.
+			require.NoError(s.T, s.Index.fixups.runAll(s.Ctx))
 		}
 	}
 
-	// TODO(andyk): Add code to delete vector from index.
+	// Handle any remaining fixups.
+	require.NoError(s.T, s.Index.fixups.runAll(s.Ctx))
 
-	tree, err := s.Index.Format(s.Ctx, txn, FormatOptions{PrimaryKeyStrings: true})
-	require.NoError(s.T, err)
-	return tree
+	return s.FormatTree(d)
 }
 
 func (s *testState) Recall(d *datadriven.TestData) string {