Prefer uint64_t instead of size_t

Author: Nikos Papailiou

apis/python/src/tiledb/vector_search/module.cc

@@ -279,7 +279,7 @@ static void declare_ivf_index_tdb(py::module& m, const std::string& suffix) {
         }, py::keep_alive<1,2>());
 }
 
-template <class T=float, class U=size_t>
+template <class T=float, class U=uint64_t>
 static void declareFixedMinPairHeap(py::module& mod) {
   using PyFixedMinPairHeap = py::class_<fixed_min_pair_heap<T, U>>;
   PyFixedMinPairHeap cls(mod, "FixedMinPairHeap", py::buffer_protocol());
@@ -356,7 +356,7 @@ void declareStdVector(py::module& m, const std::string& suffix) {
     });
 }
 
-template <typename T, typename indices_type = size_t>
+template <typename T, typename indices_type = uint64_t>
 void declarePartitionIvfIndex(py::module& m, const std::string& suffix) {
   m.def(("partition_ivf_index_" + suffix).c_str(),
         [](ColMajorMatrix<float>& centroids,
@@ -496,7 +496,7 @@ PYBIND11_MODULE(_tiledbvspy, m) {
         [](ColMajorMatrix<float>& data,
            ColMajorMatrix<float>& query_vectors,
            int k,
-           size_t nthreads) -> std::tuple<ColMajorMatrix<float>, ColMajorMatrix<size_t>> {
+           size_t nthreads) -> std::tuple<ColMajorMatrix<float>, ColMajorMatrix<uint64_t>> {
           auto r = detail::flat::vq_query_heap(data, query_vectors, k, nthreads);
           return r;
         });
@@ -505,13 +505,13 @@ PYBIND11_MODULE(_tiledbvspy, m) {
         [](tdbColMajorMatrix<uint8_t>& data,
            ColMajorMatrix<float>& query_vectors,
            int k,
-           size_t nthreads) -> std::tuple<ColMajorMatrix<float>, ColMajorMatrix<size_t>> {
+           size_t nthreads) -> std::tuple<ColMajorMatrix<float>, ColMajorMatrix<uint64_t>> {
           auto r = detail::flat::vq_query_heap(data, query_vectors, k, nthreads);
           return r;
         });
 
   m.def("validate_top_k_u64",
-      [](const ColMajorMatrix<size_t>& top_k,
+      [](const ColMajorMatrix<uint64_t>& top_k,
          const ColMajorMatrix<int32_t>& ground_truth) -> bool {
         return validate_top_k(top_k, ground_truth);
       });

src/include/detail/flat/qv.h

@@ -128,7 +128,7 @@ auto qv_query_heap(
 template <class DB, class Q>
 auto qv_query_heap(const DB& db, const Q& q, int k_nn, unsigned nthreads) {
   return qv_query_heap(
-      without_ids{}, db, q, std::vector<size_t>{}, k_nn, nthreads);
+      without_ids{}, db, q, std::vector<uint64_t>{}, k_nn, nthreads);
 }
 
 template <class DB, class Q, class ID>
@@ -209,7 +209,7 @@ auto qv_query_heap_tiled(
 template <class DB, class Q>
 auto qv_query_heap_tiled(DB& db, const Q& q, int k_nn, unsigned nthreads) {
   return qv_query_heap_tiled(
-      without_ids{}, db, q, std::vector<size_t>{}, k_nn, nthreads);
+      without_ids{}, db, q, std::vector<uint64_t>{}, k_nn, nthreads);
 }
 
 template <class DB, class Q, class ID>
@@ -226,11 +226,11 @@ auto qv_query_heap_tiled(
     [[maybe_unused]] const ID& ids,
     int k_nn,
     unsigned nthreads) {
-  
+
   // using feature_type = typename std::remove_reference_t<decltype(db)>::value_type;
   using id_type = typename std::remove_reference_t<decltype(ids)>::value_type;
   using score_type = float;
-  
+
   if constexpr (is_loadable_v<decltype(db)>) {
     db.load();
   }

src/include/detail/flat/vq.h

@@ -63,7 +63,7 @@ auto vq_query_heap(
 template <class DB, class Q>
 auto vq_query_heap(DB& db, const Q& q, int k_nn, unsigned nthreads) {
   return vq_query_heap(
-      without_ids{}, db, q, std::vector<size_t>{}, k_nn, nthreads);
+      without_ids{}, db, q, std::vector<uint64_t>{}, k_nn, nthreads);
 }
 
 template <class DB, class Q, class ID>
@@ -83,11 +83,11 @@ auto vq_query_heap(
     unsigned nthreads) {
   // @todo Need to get the total number of queries, not just the first block
   // @todo Use Matrix here rather than vector of vectors
-  
+
   // using feature_type = typename std::remove_reference_t<decltype(db)>::value_type;
   using id_type = typename std::remove_reference_t<decltype(ids)>::value_type;
   using score_type = float;
-  
+
   std::vector<std::vector<fixed_min_pair_heap<score_type, id_type>>> scores(
       nthreads,
       std::vector<fixed_min_pair_heap<score_type, id_type>>(
@@ -173,7 +173,7 @@ auto vq_query_heap_tiled(
 template <class DB, class Q>
 auto vq_query_heap_tiled(DB& db, const Q& q, int k_nn, unsigned nthreads) {
   return vq_query_heap_tiled(
-      without_ids{}, db, q, std::vector<size_t>{}, k_nn, nthreads);
+      without_ids{}, db, q, std::vector<uint64_t>{}, k_nn, nthreads);
 }
 
 template <class DB, class Q, class ID>
@@ -192,11 +192,11 @@ auto vq_query_heap_tiled(
     unsigned nthreads) {
   // @todo Need to get the total number of queries, not just the first block
   // @todo Use Matrix here rather than vector of vectors
-  
+
   // using feature_type = typename std::remove_reference_t<decltype(db)>::value_type;
   using id_type = typename std::remove_reference_t<decltype(ids)>::value_type;
   using score_type = float;
-  
+
   std::vector<std::vector<fixed_min_pair_heap<score_type, id_type>>> scores(
       nthreads,
       std::vector<fixed_min_pair_heap<score_type, id_type>>(
@@ -255,7 +255,7 @@ auto vq_query_heap_2(
 template <class DB, class Q>
 auto vq_query_heap_2(DB& db, const Q& q, int k_nn, unsigned nthreads) {
   return vq_query_heap_2(
-      without_ids{}, db, q, std::vector<size_t>{}, k_nn, nthreads);
+      without_ids{}, db, q, std::vector<uint64_t>{}, k_nn, nthreads);
 }
 
 template <class DB, class Q, class ID>