feat: add position_deletes metadata table

Author: wirybeaver

src/iceberg/CMakeLists.txt

@@ -55,6 +55,7 @@ set(ICEBERG_SOURCES
     geospatial.cc
     inspect/history_table.cc
     inspect/metadata_table.cc
+    inspect/position_deletes_table.cc
     inspect/snapshots_table.cc
     inheritable_metadata.cc
     json_serde.cc

src/iceberg/arrow_c_data_util.cc

@@ -220,6 +220,12 @@ Status AppendValue(const ArrowSchema& input_schema, const ArrowArray& input_arra
 
 }  // namespace
 
+Status AppendArrayValue(const ArrowSchema& input_schema, const ArrowArray& input_array,
+                        const ArrowArrayView& input_view, int64_t row_index,
+                        ArrowArray* output_array) {
+  return AppendValue(input_schema, input_array, input_view, row_index, output_array);
+}
+
 ProjectionContext::ProjectionContext(ProjectionContext&& other) noexcept
     : input_schema_(std::exchange(other.input_schema_, nullptr)),
       output_schema_(std::exchange(other.output_schema_, nullptr)),

src/iceberg/arrow_c_data_util_internal.h

@@ -35,6 +35,8 @@
 #include "iceberg/result.h"
 #include "iceberg/type_fwd.h"
 
+struct ArrowArrayView;
+
 namespace iceberg {
 
 /// \brief Cached state for ProjectBatch over one input/output schema pair.
@@ -242,4 +244,10 @@ ICEBERG_EXPORT Result<ArrowArray> ProjectBatch(ArrowArray* input_batch,
                                                std::span<const int32_t> row_indices,
                                                ProjectionContext& projection);
 
+/// \brief Append one value from an Arrow array into a compatible nanoarrow builder.
+ICEBERG_EXPORT Status AppendArrayValue(const ArrowSchema& input_schema,
+                                       const ArrowArray& input_array,
+                                       const ArrowArrayView& input_view,
+                                       int64_t row_index, ArrowArray* output_array);
+
 }  // namespace iceberg

src/iceberg/deletes/position_delete_index.cc

@@ -136,6 +136,10 @@ int64_t PositionDeleteIndex::Cardinality() const {
   return static_cast<int64_t>(bitmap_.Cardinality());
 }
 
+void PositionDeleteIndex::ForEach(const std::function<void(int64_t)>& fn) const {
+  bitmap_.ForEach(fn);
+}
+
 void PositionDeleteIndex::Merge(const PositionDeleteIndex& other) {
   bitmap_.Or(other.bitmap_);
   delete_files_.insert(delete_files_.end(), other.delete_files_.begin(),

src/iceberg/deletes/position_delete_index.h

@@ -23,6 +23,7 @@
 /// Index of deleted row positions for a data file.
 
 #include <cstdint>
+#include <functional>
 #include <memory>
 #include <span>
 #include <vector>
@@ -66,6 +67,9 @@ class ICEBERG_EXPORT PositionDeleteIndex {
   /// \brief Get the number of deleted positions.
   int64_t Cardinality() const;
 
+  /// \brief Iterate over deleted positions in ascending order.
+  void ForEach(const std::function<void(int64_t)>& fn) const;
+
   /// \brief Merge another index into this one.
   /// \param other The index to merge (union operation)
   void Merge(const PositionDeleteIndex& other);

src/iceberg/inspect/meson.build

@@ -16,6 +16,11 @@
 # under the License.
 
 install_headers(
-    ['history_table.h', 'metadata_table.h', 'snapshots_table.h'],
+    [
+        'history_table.h',
+        'metadata_table.h',
+        'position_deletes_table.h',
+        'snapshots_table.h',
+    ],
     subdir: 'iceberg/inspect',
 )

src/iceberg/inspect/metadata_table.cc

@@ -23,6 +23,7 @@
 #include <utility>
 
 #include "iceberg/inspect/history_table.h"
+#include "iceberg/inspect/position_deletes_table.h"
 #include "iceberg/inspect/snapshots_table.h"
 
 namespace iceberg {
@@ -35,6 +36,10 @@ MetadataTable::MetadataTable(std::shared_ptr<Table> source_table,
 
 MetadataTable::~MetadataTable() = default;
 
+Result<ArrowArray> MetadataTable::Scan(const Schema& /*projected_schema*/) {
+  return NotSupported("Scan is not supported for this metadata table type");
+}
+
 Result<std::unique_ptr<MetadataTable>> MetadataTable::Make(std::shared_ptr<Table> table,
                                                            Kind kind) {
   if (table == nullptr) [[unlikely]] {
@@ -46,6 +51,8 @@ Result<std::unique_ptr<MetadataTable>> MetadataTable::Make(std::shared_ptr<Table
       return SnapshotsTable::Make(table);
     case Kind::kHistory:
       return HistoryTable::Make(table);
+    case Kind::kPositionDeletes:
+      return PositionDeletesTable::Make(table);
   }
 
   return NotSupported("Unsupported metadata table type");

src/iceberg/inspect/metadata_table.h

@@ -24,6 +24,7 @@
 
 #include <memory>
 
+#include "iceberg/arrow_c_data.h"
 #include "iceberg/iceberg_export.h"
 #include "iceberg/result.h"
 #include "iceberg/table_identifier.h"
@@ -37,6 +38,7 @@ class ICEBERG_EXPORT MetadataTable {
   enum class Kind {
     kSnapshots,
     kHistory,
+    kPositionDeletes,
   };
 
   static Result<std::unique_ptr<MetadataTable>> Make(std::shared_ptr<Table> table,
@@ -46,6 +48,14 @@ class ICEBERG_EXPORT MetadataTable {
 
   virtual Kind kind() const noexcept = 0;
 
+  /// \brief Scan all rows using the metadata table's full schema.
+  Result<ArrowArray> Scan() { return Scan(*schema_); }
+
+  /// \brief Scan all rows projected to the requested top-level fields.
+  ///
+  /// The default implementation returns NotSupported.
+  virtual Result<ArrowArray> Scan(const Schema& projected_schema);
+
   const TableIdentifier& name() const { return identifier_; }
 
   const std::shared_ptr<Schema>& schema() const { return schema_; }