feat: implement manifest group

src/iceberg/CMakeLists.txt

@@ -42,6 +42,7 @@ set(ICEBERG_SOURCES
     json_internal.cc
     manifest/manifest_adapter.cc
     manifest/manifest_entry.cc
+    manifest/manifest_group.cc
     manifest/manifest_list.cc
     manifest/manifest_reader.cc
     manifest/manifest_writer.cc

src/iceberg/constants.h

@@ -19,12 +19,24 @@
 
 #pragma once
 
+/// \file iceberg/constants.h
+/// This file defines constants used commonly and shared across multiple
+/// source files.  It is mostly useful to add constants that are used as
+/// default values in the class definitions in the header files without
+/// including other headers just for the constant definitions.
+
 #include <cstdint>
 #include <string_view>
 
 namespace iceberg {
 
 constexpr std::string_view kParquetFieldIdKey = "PARQUET:field_id";
 constexpr int64_t kInvalidSnapshotId = -1;
+/// \brief Stand-in for the current sequence number that will be assigned when the commit
+/// is successful. This is replaced when writing a manifest list by the ManifestFile
+/// wrapper
+constexpr int64_t kUnassignedSequenceNumber = -1;
+
+// TODO(gangwu): move other commonly used constants here.
 
 }  // namespace iceberg

src/iceberg/delete_file_index.cc

@@ -453,34 +453,32 @@ Result<std::shared_ptr<DataFile>> DeleteFileIndex::FindDV(
 }
 
 Result<DeleteFileIndex::Builder> DeleteFileIndex::BuilderFor(
-    std::shared_ptr<FileIO> io, std::vector<ManifestFile> delete_manifests) {
+    std::shared_ptr<FileIO> io, std::shared_ptr<Schema> schema,
+    std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id,
+    std::vector<ManifestFile> delete_manifests) {
   ICEBERG_PRECHECK(io != nullptr, "FileIO cannot be null");
-  return Builder(std::move(io), std::move(delete_manifests));
+  ICEBERG_PRECHECK(schema != nullptr, "Schema cannot be null");
+  ICEBERG_PRECHECK(!specs_by_id.empty(), "Partition specs cannot be empty");
+  return Builder(std::move(io), std::move(schema), std::move(specs_by_id),
+                 std::move(delete_manifests));
 }
 
 // Builder implementation
 
-DeleteFileIndex::Builder::Builder(std::shared_ptr<FileIO> io,
-                                  std::vector<ManifestFile> delete_manifests)
-    : io_(std::move(io)), delete_manifests_(std::move(delete_manifests)) {}
+DeleteFileIndex::Builder::Builder(
+    std::shared_ptr<FileIO> io, std::shared_ptr<Schema> schema,
+    std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id,
+    std::vector<ManifestFile> delete_manifests)
+    : io_(std::move(io)),
+      schema_(std::move(schema)),
+      specs_by_id_(std::move(specs_by_id)),
+      delete_manifests_(std::move(delete_manifests)) {}
 
 DeleteFileIndex::Builder::~Builder() = default;
 DeleteFileIndex::Builder::Builder(Builder&&) noexcept = default;
 DeleteFileIndex::Builder& DeleteFileIndex::Builder::operator=(Builder&&) noexcept =
     default;
 
-DeleteFileIndex::Builder& DeleteFileIndex::Builder::SpecsById(
-    std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id) {
-  specs_by_id_ = std::move(specs_by_id);
-  return *this;
-}
-
-DeleteFileIndex::Builder& DeleteFileIndex::Builder::WithSchema(
-    std::shared_ptr<Schema> schema) {
-  schema_ = std::move(schema);
-  return *this;
-}
-
 DeleteFileIndex::Builder& DeleteFileIndex::Builder::AfterSequenceNumber(int64_t seq) {
   min_sequence_number_ = seq;
   return *this;
@@ -721,10 +719,6 @@ Status DeleteFileIndex::Builder::AddEqualityDelete(
 
 Result<std::unique_ptr<DeleteFileIndex>> DeleteFileIndex::Builder::Build() {
   ICEBERG_RETURN_UNEXPECTED(CheckErrors());
-  ICEBERG_PRECHECK(io_ != nullptr, "FileIO is required to load delete files");
-  ICEBERG_PRECHECK(schema_ != nullptr, "Schema is required to load delete files");
-  ICEBERG_PRECHECK(!specs_by_id_.empty(),
-                   "Partition specs are required to load delete files");
 
   std::vector<ManifestEntry> entries;
   if (!delete_manifests_.empty()) {

src/iceberg/delete_file_index.h

@@ -268,10 +268,14 @@ class ICEBERG_EXPORT DeleteFileIndex {
   /// \brief Create a builder for constructing a DeleteFileIndex from manifest files.
   ///
   /// \param io The FileIO to use for reading manifests
+  /// \param schema Current table schema
+  /// \param specs_by_id Partition specs by their IDs
   /// \param delete_manifests The delete manifests to index
   /// \return A Builder instance
-  static Result<Builder> BuilderFor(std::shared_ptr<FileIO> io,
-                                    std::vector<ManifestFile> delete_manifests);
+  static Result<Builder> BuilderFor(
+      std::shared_ptr<FileIO> io, std::shared_ptr<Schema> schema,
+      std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id,
+      std::vector<ManifestFile> delete_manifests);
 
  private:
   friend class Builder;
@@ -318,7 +322,9 @@ class ICEBERG_EXPORT DeleteFileIndex {
 class ICEBERG_EXPORT DeleteFileIndex::Builder : public ErrorCollector {
  public:
   /// \brief Construct a builder from manifest files.
-  Builder(std::shared_ptr<FileIO> io, std::vector<ManifestFile> delete_manifests);
+  Builder(std::shared_ptr<FileIO> io, std::shared_ptr<Schema> schema,
+          std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id,
+          std::vector<ManifestFile> delete_manifests);
 
   ~Builder() override;
 
@@ -327,15 +333,6 @@ class ICEBERG_EXPORT DeleteFileIndex::Builder : public ErrorCollector {
   Builder(const Builder&) = delete;
   Builder& operator=(const Builder&) = delete;
 
-  /// \brief Set the partition specs by ID.
-  Builder& SpecsById(
-      std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id);
-
-  /// \brief Set the table schema.
-  ///
-  /// Required for filtering and expression evaluation.
-  Builder& WithSchema(std::shared_ptr<Schema> schema);
-
   /// \brief Set the minimum sequence number for delete files.
   ///
   /// Only delete files with sequence number > min_sequence_number will be included.
@@ -384,10 +381,10 @@ class ICEBERG_EXPORT DeleteFileIndex::Builder : public ErrorCollector {
       ManifestEntry&& entry);
 
   std::shared_ptr<FileIO> io_;
+  std::shared_ptr<Schema> schema_;
+  std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id_;
   std::vector<ManifestFile> delete_manifests_;
   int64_t min_sequence_number_ = 0;
-  std::unordered_map<int32_t, std::shared_ptr<PartitionSpec>> specs_by_id_;
-  std::shared_ptr<Schema> schema_;
   std::shared_ptr<Expression> data_filter_;
   std::shared_ptr<Expression> partition_filter_;
   std::shared_ptr<PartitionSet> partition_set_;