Misc small code cleanups

apis/python/src/tiledb/vector_search/index.py

@@ -542,10 +542,15 @@ def consolidate_updates(self, retrain_index: bool = False, **kwargs):
             tiledb.consolidate(self.updates_array_uri, config=conf)
 
         # We don't copy the centroids if self.partitions=0 because this means our index was previously empty.
-        should_pass_copy_centroids_uri = (
-            self.index_type == "IVF_FLAT" and not retrain_index and self.partitions > 0
-        )
-        if should_pass_copy_centroids_uri:
+        copy_centroids_uri = None
+        if (
+            (self.index_type == "IVF_FLAT" or self.index_type == "IVF_PQ")
+            and not retrain_index
+            and self.partitions > 0
+        ):
+            if self.index_type == "IVF_FLAT":
+                # TODO(paris): Update so that IVF_PQ can also copy the centroids. We also need to pass the PQ-centroids.
+                copy_centroids_uri = self.centroids_uri
             # Make sure the user didn't pass an incorrect number of partitions.
             if "partitions" in kwargs and self.partitions != kwargs["partitions"]:
                 raise ValueError(
@@ -565,9 +570,7 @@ def consolidate_updates(self, retrain_index: bool = False, **kwargs):
             index_timestamp=max_timestamp,
             distance_metric=self.distance_metric,
             storage_version=self.storage_version,
-            copy_centroids_uri=self.centroids_uri
-            if should_pass_copy_centroids_uri
-            else None,
+            copy_centroids_uri=copy_centroids_uri,
             config=self.config,
             **kwargs,
         )

apis/python/src/tiledb/vector_search/ingestion.py

@@ -1738,9 +1738,7 @@ def ingest_vectors_udf(
             if part_end > end:
                 part_end = end
 
-            str(part) + "-" + str(part_end)
             part_id = int(part / batch)
-            part_id * (partitions + 1)
 
             logger.debug("Input vectors start_pos: %d, end_pos: %d", part, part_end)
             updated_ids = read_updated_ids(

apis/python/test/test_index.py

@@ -495,28 +495,28 @@ def test_index_with_incorrect_num_of_query_columns_complex(tmp_path):
     # number of columns in the indexed data.
     size = 1000
     indexes = ["FLAT", "IVF_FLAT", "VAMANA", "IVF_PQ"]
-    num_columns_in_vector = [1, 2, 3, 4, 5, 10]
+    dimensions_in_ingestion = [1, 2, 3, 4, 5, 10]
     for index_type in indexes:
-        for num_columns in num_columns_in_vector:
-            index_uri = os.path.join(tmp_path, f"array_{index_type}_{num_columns}")
-            dataset_dir = os.path.join(tmp_path, f"dataset_{index_type}_{num_columns}")
+        for dimensions in dimensions_in_ingestion:
+            index_uri = os.path.join(tmp_path, f"array_{index_type}_{dimensions}")
+            dataset_dir = os.path.join(tmp_path, f"dataset_{index_type}_{dimensions}")
             create_random_dataset_f32_only_data(
-                nb=size, d=num_columns, centers=1, path=dataset_dir
+                nb=size, d=dimensions, centers=1, path=dataset_dir
             )
             index = ingest(
                 index_type=index_type,
                 index_uri=index_uri,
                 source_uri=os.path.join(dataset_dir, "data.f32bin"),
-                num_subspaces=num_columns,
+                num_subspaces=dimensions,
                 partitions=1,
             )
 
-            # We have created a dataset with num_columns in each vector. Let's try creating queries
+            # We have created a dataset with dimensions in each vector. Let's try creating queries
             # with different numbers of columns and confirming incorrect ones will throw.
-            for num_columns_for_query in range(1, num_columns + 2):
-                query_shape = (1, num_columns_for_query)
+            for dimensions_in_query in range(1, dimensions + 2):
+                query_shape = (1, dimensions_in_query)
                 query = np.random.rand(*query_shape).astype(np.float32)
-                if query.shape[1] == num_columns:
+                if query.shape[1] == dimensions:
                     index.query(query, k=1, nprobe=1)
                 else:
                     with pytest.raises(TypeError):

apis/python/test/test_type_erased_module.py

@@ -234,6 +234,19 @@ def test_numpy_to_feature_vector_array_with_ids():
     assert b.ids_type_string() == "uint64"
 
 
+def test_numpy_to_feature_vector_single_item():
+    a = np.array([1], dtype=np.float32)
+    assert a.ndim == 1
+    assert a.shape == (1,)
+    b = vspy.FeatureVector(a)
+    assert b.dimensions() == 1
+    assert b.feature_type_string() == "float32"
+    c = np.array(b)
+    assert c.ndim == 1
+    assert c.shape == (1,)
+    assert (a == c).all()
+
+
 def test_TemporalPolicy():
     temporal_policy = vspy.TemporalPolicy()
     assert temporal_policy.timestamp_start() == 0

src/include/index/ivf_pq_group.h

@@ -163,34 +163,38 @@ class ivf_pq_group : public base_index_group<index_type> {
   /*****************************************************************************
    * Inverted index information: centroids, index, pq_parts, ids
    ****************************************************************************/
-  [[nodiscard]] auto cluster_centroids_uri() const {
-    return this->array_key_to_uri("cluster_centroids_array_name");
-  }
   [[nodiscard]] auto flat_ivf_centroids_uri() const {
     return this->array_key_to_uri("flat_ivf_centroids_array_name");
   }
-  [[nodiscard]] auto pq_ivf_indices_uri() const {
-    return this->array_key_to_uri("pq_ivf_indices_array_name");
-  }
-  [[nodiscard]] auto pq_ivf_ids_uri() const {
-    return this->array_key_to_uri("pq_ivf_ids_array_name");
-  }
-  [[nodiscard]] auto pq_ivf_vectors_uri() const {
-    return this->array_key_to_uri("pq_ivf_vectors_array_name");
+  [[nodiscard]] auto flat_ivf_centroids_array_name() const {
+    return this->array_key_to_array_name("flat_ivf_centroids_array_name");
   }
 
+  [[nodiscard]] auto cluster_centroids_uri() const {
+    return this->array_key_to_uri("cluster_centroids_array_name");
+  }
   [[nodiscard]] auto cluster_centroids_array_name() const {
     return this->array_key_to_array_name("cluster_centroids_array_name");
   }
-  [[nodiscard]] auto flat_ivf_centroids_array_name() const {
-    return this->array_key_to_array_name("flat_ivf_centroids_array_name");
+
+  [[nodiscard]] auto pq_ivf_indices_uri() const {
+    return this->array_key_to_uri("pq_ivf_indices_array_name");
   }
   [[nodiscard]] auto pq_ivf_indices_array_name() const {
     return this->array_key_to_array_name("pq_ivf_indices_array_name");
   }
+
+  [[nodiscard]] auto pq_ivf_ids_uri() const {
+    return this->array_key_to_uri("pq_ivf_ids_array_name");
+  }
   [[nodiscard]] auto pq_ivf_ids_array_name() const {
     return this->array_key_to_array_name("pq_ivf_ids_array_name");
   }
+
+  [[nodiscard]] auto pq_ivf_vectors_uri() const {
+    return this->array_key_to_uri("pq_ivf_vectors_array_name");
+  }
+
   [[nodiscard]] auto pq_ivf_vectors_array_name() const {
     return this->array_key_to_array_name("pq_ivf_vectors_array_name");
   }
@@ -205,6 +209,9 @@ class ivf_pq_group : public base_index_group<index_type> {
     metadata_.num_subspaces_ = num_subspaces;
   }
 
+  uint32_t get_sub_dimensions() const {
+    return metadata_.sub_dimensions_;
+  }
   void set_sub_dimensions(uint32_t sub_dimensions) {
     metadata_.sub_dimensions_ = sub_dimensions;
   }

src/include/index/ivf_pq_index.h

@@ -1665,6 +1665,8 @@ class ivf_pq_index {
                 << " num_vectors(rhs): " << ::num_vectors(rhs) << std::endl;
       std::cout << "dimensions(lhs): " << ::dimensions(lhs)
                 << " dimensions(rhs): " << ::dimensions(rhs) << std::endl;
+      debug_matrix(lhs, "[ivf_pq_index@compare_feature_vector_arrays] lhs");
+      debug_matrix(rhs, "[ivf_pq_index@compare_feature_vector_arrays] rhs");
       return false;
     }
     for (size_t i = 0; i < ::num_vectors(lhs); ++i) {
@@ -1737,6 +1739,9 @@ class ivf_pq_index {
       return true;
     }
     if (!partitioned_pq_vectors_ || !rhs.partitioned_pq_vectors_) {
+      std::cout << "[ivf_pq_index@compare_ivf_index] partitioned_pq_vectors_ "
+                   "|| rhs.partitioned_pq_vectors_ is nullptr"
+                << std::endl;
       return false;
     }
     return compare_feature_vectors(
@@ -1750,6 +1755,9 @@ class ivf_pq_index {
       return true;
     }
     if (!partitioned_pq_vectors_ || !rhs.partitioned_pq_vectors_) {
+      std::cout << "[ivf_pq_index@compare_ivf_ids] partitioned_pq_vectors_ || "
+                   "rhs.partitioned_pq_vectors_ is nullptr"
+                << std::endl;
       return false;
     }
     return compare_feature_vectors(

src/include/test/unit_api_ivf_pq_index.cc

@@ -509,11 +509,11 @@ TEST_CASE("infer dimension", "[api_ivf_pq_index]") {
 
 TEST_CASE("write and read", "[api_ivf_pq_index]") {
   auto ctx = tiledb::Context{};
-  std::string api_ivf_pq_index_uri =
+  std::string index_uri =
       (std::filesystem::temp_directory_path() / "api_ivf_pq_index").string();
   tiledb::VFS vfs(ctx);
-  if (vfs.is_dir(api_ivf_pq_index_uri)) {
-    vfs.remove_dir(api_ivf_pq_index_uri);
+  if (vfs.is_dir(index_uri)) {
+    vfs.remove_dir(index_uri);
   }
 
   auto a = IndexIVFPQ(std::make_optional<IndexOptions>(
@@ -524,9 +524,9 @@ TEST_CASE("write and read", "[api_ivf_pq_index]") {
   auto training_set = FeatureVectorArray(ctx, siftsmall_inputs_uri);
   a.train(training_set);
   a.add(training_set);
-  a.write_index(ctx, api_ivf_pq_index_uri);
+  a.write_index(ctx, index_uri);
 
-  auto b = IndexIVFPQ(ctx, api_ivf_pq_index_uri);
+  auto b = IndexIVFPQ(ctx, index_uri);
 
   CHECK(dimensions(a) == dimensions(b));
   CHECK(a.feature_type() == b.feature_type());
@@ -561,10 +561,10 @@ TEST_CASE("read index and query", "[api_ivf_pq_index]") {
 
   size_t k_nn = 10;
 
-  std::string api_ivf_pq_index_uri =
+  std::string index_uri =
       (std::filesystem::temp_directory_path() / "api_ivf_pq_index").string();
-  if (vfs.is_dir(api_ivf_pq_index_uri)) {
-    vfs.remove_dir(api_ivf_pq_index_uri);
+  if (vfs.is_dir(index_uri)) {
+    vfs.remove_dir(index_uri);
   }
 
   auto a = IndexIVFPQ(std::make_optional<IndexOptions>(
@@ -576,19 +576,19 @@ TEST_CASE("read index and query", "[api_ivf_pq_index]") {
   auto training_set = FeatureVectorArray(ctx, siftsmall_inputs_uri);
   a.train(training_set);
   a.add(training_set);
-  a.write_index(ctx, api_ivf_pq_index_uri);
+  a.write_index(ctx, index_uri);
 
   auto query_set = FeatureVectorArray(ctx, siftsmall_query_uri);
   auto groundtruth_set = FeatureVectorArray(ctx, siftsmall_groundtruth_uri);
 
   std::unique_ptr<IndexIVFPQ> b;
   SECTION("infinite") {
-    b = std::make_unique<IndexIVFPQ>(ctx, api_ivf_pq_index_uri);
+    b = std::make_unique<IndexIVFPQ>(ctx, index_uri);
   }
   SECTION("finite") {
     size_t upper_bound = GENERATE(500, 1000);
     b = std::make_unique<IndexIVFPQ>(
-        ctx, api_ivf_pq_index_uri, IndexLoadStrategy::PQ_OOC, upper_bound);
+        ctx, index_uri, IndexLoadStrategy::PQ_OOC, upper_bound);
     CHECK(b->upper_bound() == upper_bound);
   }