refactor: move type visitor classes to type_util (#347)

Author: HeartLinked

src/iceberg/CMakeLists.txt

@@ -77,6 +77,7 @@ set(ICEBERG_SOURCES
     util/temporal_util.cc
     util/timepoint.cc
     util/truncate_util.cc
+    util/type_util.cc
     util/uuid.cc)
 
 set(ICEBERG_STATIC_BUILD_INTERFACE_LIBS)

src/iceberg/meson.build

@@ -99,6 +99,7 @@ iceberg_sources = files(
     'util/temporal_util.cc',
     'util/timepoint.cc',
     'util/truncate_util.cc',
+    'util/type_util.cc',
     'util/uuid.cc',
 )
 

src/iceberg/partition_spec.cc

@@ -108,7 +108,7 @@ bool PartitionSpec::Equals(const PartitionSpec& other) const {
 }
 
 Status PartitionSpec::Validate(const Schema& schema, bool allow_missing_fields) const {
-  std::unordered_map<int32_t, int32_t> parents = indexParents(schema);
+  std::unordered_map<int32_t, int32_t> parents = IndexParents(schema);
   for (const auto& partition_field : fields_) {
     ICEBERG_ASSIGN_OR_RAISE(auto source_field,
                             schema.FindFieldById(partition_field.source_id()));

src/iceberg/schema.cc

@@ -27,93 +27,12 @@
 #include "iceberg/schema_internal.h"
 #include "iceberg/type.h"
 #include "iceberg/util/formatter.h"  // IWYU pragma: keep
-#include "iceberg/util/formatter_internal.h"
 #include "iceberg/util/macros.h"
+#include "iceberg/util/type_util.h"
 #include "iceberg/util/visit_type.h"
 
 namespace iceberg {
 
-class IdToFieldVisitor {
- public:
-  explicit IdToFieldVisitor(
-      std::unordered_map<int32_t, std::reference_wrapper<const SchemaField>>&
-          id_to_field);
-  Status Visit(const PrimitiveType& type);
-  Status Visit(const NestedType& type);
-
- private:
-  std::unordered_map<int32_t, std::reference_wrapper<const SchemaField>>& id_to_field_;
-};
-
-class NameToIdVisitor {
- public:
-  explicit NameToIdVisitor(
-      std::unordered_map<std::string, int32_t, StringHash, std::equal_to<>>& name_to_id,
-      bool case_sensitive = true,
-      std::function<std::string(std::string_view)> quoting_func = {});
-  Status Visit(const ListType& type, const std::string& path,
-               const std::string& short_path);
-  Status Visit(const MapType& type, const std::string& path,
-               const std::string& short_path);
-  Status Visit(const StructType& type, const std::string& path,
-               const std::string& short_path);
-  Status Visit(const PrimitiveType& type, const std::string& path,
-               const std::string& short_path);
-  void Finish();
-
- private:
-  std::string BuildPath(std::string_view prefix, std::string_view field_name,
-                        bool case_sensitive);
-
- private:
-  bool case_sensitive_;
-  std::unordered_map<std::string, int32_t, StringHash, std::equal_to<>>& name_to_id_;
-  std::unordered_map<std::string, int32_t, StringHash, std::equal_to<>> short_name_to_id_;
-  std::function<std::string(std::string_view)> quoting_func_;
-};
-
-class PositionPathVisitor {
- public:
-  Status Visit(const PrimitiveType& type) {
-    if (current_field_id_ == kUnassignedFieldId) {
-      return InvalidSchema("Current field id is not assigned, type: {}", type.ToString());
-    }
-
-    if (auto ret = position_path_.try_emplace(current_field_id_, current_path_);
-        !ret.second) {
-      return InvalidSchema("Duplicate field id found: {}, prev path: {}, curr path: {}",
-                           current_field_id_, ret.first->second, current_path_);
-    }
-
-    return {};
-  }
-
-  Status Visit(const StructType& type) {
-    for (size_t i = 0; i < type.fields().size(); ++i) {
-      const auto& field = type.fields()[i];
-      current_field_id_ = field.field_id();
-      current_path_.push_back(i);
-      ICEBERG_RETURN_UNEXPECTED(VisitTypeInline(*field.type(), this));
-      current_path_.pop_back();
-    }
-    return {};
-  }
-
-  // Non-struct types are not supported yet, but it is not an error.
-  Status Visit(const ListType& type) { return {}; }
-  Status Visit(const MapType& type) { return {}; }
-
-  std::unordered_map<int32_t, std::vector<size_t>> Finish() {
-    return std::move(position_path_);
-  }
-
- private:
-  constexpr static int32_t kUnassignedFieldId = -1;
-  int32_t current_field_id_ = kUnassignedFieldId;
-  std::vector<size_t> current_path_;
-  std::unordered_map<int32_t, std::vector<size_t>> position_path_;
-};
-
 Schema::Schema(std::vector<SchemaField> fields, std::optional<int32_t> schema_id)
     : StructType(std::move(fields)), schema_id_(schema_id) {}
 
@@ -210,216 +129,6 @@ Result<std::unique_ptr<StructLikeAccessor>> Schema::GetAccessorById(
   return NotFound("Cannot get accessor for field id: {}", field_id);
 }
 
-IdToFieldVisitor::IdToFieldVisitor(
-    std::unordered_map<int32_t, std::reference_wrapper<const SchemaField>>& id_to_field)
-    : id_to_field_(id_to_field) {}
-
-Status IdToFieldVisitor::Visit(const PrimitiveType& type) { return {}; }
-
-Status IdToFieldVisitor::Visit(const NestedType& type) {
-  const auto& nested = internal::checked_cast<const NestedType&>(type);
-  const auto& fields = nested.fields();
-  for (const auto& field : fields) {
-    auto it = id_to_field_.try_emplace(field.field_id(), std::cref(field));
-    if (!it.second) {
-      return InvalidSchema("Duplicate field id found: {}", field.field_id());
-    }
-    ICEBERG_RETURN_UNEXPECTED(VisitTypeInline(*field.type(), this));
-  }
-  return {};
-}
-
-NameToIdVisitor::NameToIdVisitor(
-    std::unordered_map<std::string, int32_t, StringHash, std::equal_to<>>& name_to_id,
-    bool case_sensitive, std::function<std::string(std::string_view)> quoting_func)
-    : case_sensitive_(case_sensitive),
-      name_to_id_(name_to_id),
-      quoting_func_(std::move(quoting_func)) {}
-
-Status NameToIdVisitor::Visit(const ListType& type, const std::string& path,
-                              const std::string& short_path) {
-  const auto& field = type.fields()[0];
-  std::string new_path = BuildPath(path, field.name(), case_sensitive_);
-  std::string new_short_path;
-  if (field.type()->type_id() == TypeId::kStruct) {
-    new_short_path = short_path;
-  } else {
-    new_short_path = BuildPath(short_path, field.name(), case_sensitive_);
-  }
-  auto it = name_to_id_.try_emplace(new_path, field.field_id());
-  if (!it.second) {
-    return InvalidSchema("Duplicate path found: {}, prev id: {}, curr id: {}",
-                         it.first->first, it.first->second, field.field_id());
-  }
-  short_name_to_id_.try_emplace(new_short_path, field.field_id());
-  ICEBERG_RETURN_UNEXPECTED(
-      VisitTypeInline(*field.type(), this, new_path, new_short_path));
-  return {};
-}
-
-Status NameToIdVisitor::Visit(const MapType& type, const std::string& path,
-                              const std::string& short_path) {
-  std::string new_path, new_short_path;
-  const auto& fields = type.fields();
-  for (const auto& field : fields) {
-    new_path = BuildPath(path, field.name(), case_sensitive_);
-    if (field.name() == MapType::kValueName &&
-        field.type()->type_id() == TypeId::kStruct) {
-      new_short_path = short_path;
-    } else {
-      new_short_path = BuildPath(short_path, field.name(), case_sensitive_);
-    }
-    auto it = name_to_id_.try_emplace(new_path, field.field_id());
-    if (!it.second) {
-      return InvalidSchema("Duplicate path found: {}, prev id: {}, curr id: {}",
-                           it.first->first, it.first->second, field.field_id());
-    }
-    short_name_to_id_.try_emplace(new_short_path, field.field_id());
-    ICEBERG_RETURN_UNEXPECTED(
-        VisitTypeInline(*field.type(), this, new_path, new_short_path));
-  }
-  return {};
-}
-
-Status NameToIdVisitor::Visit(const StructType& type, const std::string& path,
-                              const std::string& short_path) {
-  const auto& fields = type.fields();
-  std::string new_path, new_short_path;
-  for (const auto& field : fields) {
-    new_path = BuildPath(path, field.name(), case_sensitive_);
-    new_short_path = BuildPath(short_path, field.name(), case_sensitive_);
-    auto it = name_to_id_.try_emplace(new_path, field.field_id());
-    if (!it.second) {
-      return InvalidSchema("Duplicate path found: {}, prev id: {}, curr id: {}",
-                           it.first->first, it.first->second, field.field_id());
-    }
-    short_name_to_id_.try_emplace(new_short_path, field.field_id());
-    ICEBERG_RETURN_UNEXPECTED(
-        VisitTypeInline(*field.type(), this, new_path, new_short_path));
-  }
-  return {};
-}
-
-Status NameToIdVisitor::Visit(const PrimitiveType& type, const std::string& path,
-                              const std::string& short_path) {
-  return {};
-}
-
-std::string NameToIdVisitor::BuildPath(std::string_view prefix,
-                                       std::string_view field_name, bool case_sensitive) {
-  std::string quoted_name;
-  if (!quoting_func_) {
-    quoted_name = std::string(field_name);
-  } else {
-    quoted_name = quoting_func_(field_name);
-  }
-  if (case_sensitive) {
-    return prefix.empty() ? quoted_name : std::string(prefix) + "." + quoted_name;
-  }
-  return prefix.empty() ? StringUtils::ToLower(quoted_name)
-                        : std::string(prefix) + "." + StringUtils::ToLower(quoted_name);
-}
-
-void NameToIdVisitor::Finish() {
-  for (auto&& it : short_name_to_id_) {
-    name_to_id_.try_emplace(it.first, it.second);
-  }
-}
-
-/// \brief Visitor for pruning columns based on selected field IDs.
-///
-/// This visitor traverses a schema and creates a projected version containing only
-/// the specified fields. When `select_full_types` is true, a field with all its
-/// sub-fields are selected if its field-id has been selected; otherwise, only leaf
-/// fields of selected field-ids are selected.
-///
-/// \note It returns an error when projection is not successful.
-class PruneColumnVisitor {
- public:
-  PruneColumnVisitor(const std::unordered_set<int32_t>& selected_ids,
-                     bool select_full_types)
-      : selected_ids_(selected_ids), select_full_types_(select_full_types) {}
-
-  Result<std::shared_ptr<Type>> Visit(const std::shared_ptr<Type>& type) const {
-    switch (type->type_id()) {
-      case TypeId::kStruct:
-        return Visit(internal::checked_pointer_cast<StructType>(type));
-      case TypeId::kList:
-        return Visit(internal::checked_pointer_cast<ListType>(type));
-      case TypeId::kMap:
-        return Visit(internal::checked_pointer_cast<MapType>(type));
-      default:
-        return nullptr;
-    }
-  }
-
-  Result<std::shared_ptr<Type>> Visit(const SchemaField& field) const {
-    if (selected_ids_.contains(field.field_id())) {
-      return (select_full_types_ || field.type()->is_primitive()) ? field.type()
-                                                                  : Visit(field.type());
-    }
-    return Visit(field.type());
-  }
-
-  static SchemaField MakeField(const SchemaField& field, std::shared_ptr<Type> type) {
-    return {field.field_id(), std::string(field.name()), std::move(type),
-            field.optional(), std::string(field.doc())};
-  }
-
-  Result<std::shared_ptr<Type>> Visit(const std::shared_ptr<StructType>& type) const {
-    bool same_types = true;
-    std::vector<SchemaField> selected_fields;
-    for (const auto& field : type->fields()) {
-      ICEBERG_ASSIGN_OR_RAISE(auto child_type, Visit(field));
-      if (child_type) {
-        same_types = same_types && (child_type == field.type());
-        selected_fields.emplace_back(MakeField(field, std::move(child_type)));
-      }
-    }
-
-    if (selected_fields.empty()) {
-      return nullptr;
-    } else if (same_types && selected_fields.size() == type->fields().size()) {
-      return type;
-    }
-    return std::make_shared<StructType>(std::move(selected_fields));
-  }
-
-  Result<std::shared_ptr<Type>> Visit(const std::shared_ptr<ListType>& type) const {
-    const auto& elem_field = type->fields()[0];
-    ICEBERG_ASSIGN_OR_RAISE(auto elem_type, Visit(elem_field));
-    if (elem_type == nullptr) {
-      return nullptr;
-    } else if (elem_type == elem_field.type()) {
-      return type;
-    }
-    return std::make_shared<ListType>(MakeField(elem_field, std::move(elem_type)));
-  }
-
-  Result<std::shared_ptr<Type>> Visit(const std::shared_ptr<MapType>& type) const {
-    const auto& key_field = type->fields()[0];
-    const auto& value_field = type->fields()[1];
-    ICEBERG_ASSIGN_OR_RAISE(auto key_type, Visit(key_field));
-    ICEBERG_ASSIGN_OR_RAISE(auto value_type, Visit(value_field));
-
-    if (key_type == nullptr && value_type == nullptr) {
-      return nullptr;
-    } else if (value_type == value_field.type() &&
-               (key_type == key_field.type() || key_type == nullptr)) {
-      return type;
-    } else if (value_type == nullptr) {
-      return InvalidArgument("Cannot project Map without value field");
-    }
-    return std::make_shared<MapType>(
-        (key_type == nullptr ? key_field : MakeField(key_field, std::move(key_type))),
-        MakeField(value_field, std::move(value_type)));
-  }
-
- private:
-  const std::unordered_set<int32_t>& selected_ids_;
-  const bool select_full_types_;
-};
-
 Result<std::unique_ptr<Schema>> Schema::Select(std::span<const std::string> names,
                                                bool case_sensitive) const {
   const std::string kAllColumns = "*";

src/iceberg/test/type_test.cc

@@ -659,7 +659,7 @@ TEST(TypeTest, IndexParents) {
       points,
   });
 
-  std::unordered_map<int32_t, int32_t> parent_index = iceberg::indexParents(root_struct);
+  std::unordered_map<int32_t, int32_t> parent_index = iceberg::IndexParents(root_struct);
 
   // Verify top-level fields have no parent
   ASSERT_EQ(parent_index.find(1), parent_index.end());

src/iceberg/util/meson.build

@@ -17,19 +17,24 @@
 
 install_headers(
     [
+        'bucket_util.h',
         'checked_cast.h',
         'config.h',
+        'conversions.h',
         'decimal.h',
         'endian.h',
         'formattable.h',
         'formatter.h',
         'int128.h',
+        'lazy.h',
         'macros.h',
         'partition_value_util.h',
         'string_util.h',
+        'temporal_util.h',
         'timepoint.h',
         'truncate_util.h',
         'type_util.h',
+        'uuid.h',
         'visitor_generate.h',
         'visit_type.h',
     ],