HuaHuaY commented on code in PR #431:
URL: https://github.com/apache/iceberg-cpp/pull/431#discussion_r2642831178
##########
src/iceberg/manifest/manifest_entry.h:
##########
@@ -178,94 +178,114 @@ struct ICEBERG_EXPORT DataFile {
/// present
std::optional<int64_t> content_size_in_bytes;
+ inline static constexpr int32_t kContentFieldId = 134;
Review Comment:
We can remove `inline` on `static constexpr`.
##########
src/iceberg/manifest/manifest_reader.cc:
##########
@@ -19,61 +19,867 @@
#include "iceberg/manifest/manifest_reader.h"
+#include <algorithm>
+#include <memory>
+#include <unordered_set>
+
+#include <nanoarrow/nanoarrow.h>
+
+#include "iceberg/arrow/nanoarrow_status_internal.h"
+#include "iceberg/arrow_c_data_guard_internal.h"
+#include "iceberg/expression/expression.h"
+#include "iceberg/expression/projections.h"
+#include "iceberg/file_format.h"
#include "iceberg/manifest/manifest_entry.h"
#include "iceberg/manifest/manifest_list.h"
#include "iceberg/manifest/manifest_reader_internal.h"
+#include "iceberg/metadata_columns.h"
+#include "iceberg/partition_spec.h"
#include "iceberg/schema.h"
-#include "iceberg/schema_internal.h"
+#include "iceberg/schema_field.h"
#include "iceberg/type.h"
+#include "iceberg/util/checked_cast.h"
#include "iceberg/util/macros.h"
namespace iceberg {
+namespace {
+
+#define PARSE_PRIMITIVE_FIELD(item, array_view, type)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetIntUnsafe(array_view, row_idx);
\
+ item = static_cast<type>(value);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_STRING_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
\
+ item = std::string(value.data, value.size_bytes);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_BINARY_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(view_of_column, row_idx)) {
\
+ item = ArrowArrayViewGetInt8Vector(array_view, row_idx);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_INTEGER_VECTOR_FIELD(item, count, array_view, type)
\
+ for (int64_t manifest_idx = 0; manifest_idx < count; manifest_idx++) {
\
+ auto offset = ArrowArrayViewListChildOffset(array_view, manifest_idx);
\
+ auto next_offset = ArrowArrayViewListChildOffset(array_view, manifest_idx
+ 1); \
+ for (int64_t offset_idx = offset; offset_idx < next_offset; offset_idx++)
{ \
+ item.emplace_back(static_cast<type>(
\
+ ArrowArrayViewGetIntUnsafe(array_view->children[0], offset_idx)));
\
+ }
\
+ }
+
+#define PARSE_MAP_FIELD(item, count, array_view, key_type, value_type,
assignment) \
+ do {
\
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_MAP) {
\
+ return InvalidManifest("Field:{} should be a map.", field_name);
\
+ }
\
+ auto view_of_map = array_view->children[0];
\
+ ASSERT_VIEW_TYPE_AND_CHILDREN(view_of_map,
ArrowType::NANOARROW_TYPE_STRUCT, 2); \
+ auto view_of_map_key = view_of_map->children[0];
\
+ ASSERT_VIEW_TYPE(view_of_map_key, key_type);
\
+ auto view_of_map_value = view_of_map->children[1];
\
+ ASSERT_VIEW_TYPE(view_of_map_value, value_type);
\
+ for (int64_t row_idx = 0; row_idx < count; row_idx++) {
\
+ auto offset = array_view->buffer_views[1].data.as_int32[row_idx];
\
+ auto next_offset = array_view->buffer_views[1].data.as_int32[row_idx +
1]; \
+ for (int32_t offset_idx = offset; offset_idx < next_offset;
offset_idx++) { \
+ auto key = ArrowArrayViewGetIntUnsafe(view_of_map_key, offset_idx);
\
+ item[key] = assignment;
\
+ }
\
+ }
\
+ } while (0)
+
+#define PARSE_INT_LONG_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_INT64, \
+ ArrowArrayViewGetIntUnsafe(view_of_map_value, offset_idx));
+
+#define PARSE_INT_BINARY_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_BINARY, \
+ ArrowArrayViewGetInt8Vector(view_of_map_value, offset_idx));
+
+#define ASSERT_VIEW_TYPE(view, type)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
+
+#define ASSERT_VIEW_TYPE_AND_CHILDREN(view, type, n_child)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
\
+ if (view->n_children != n_child) {
\
+ return InvalidManifest("Sub Field for:{} should have key&value:{}
columns.", \
+ field_name, n_child);
\
+ }
+
+std::vector<uint8_t> ArrowArrayViewGetInt8Vector(const ArrowArrayView* view,
+ int32_t offset_idx) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view, offset_idx);
+ return {buffer.data.as_char, buffer.data.as_char + buffer.size_bytes};
+}
+
+Status ParsePartitionFieldSummaryList(ArrowArrayView* view_of_column,
+ std::vector<ManifestFile>&
manifest_files) {
+ auto manifest_count = view_of_column->length;
+ // view_of_column is list<struct<PartitionFieldSummary>>
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_LIST) {
+ return InvalidManifestList("partitions field should be a list.");
+ }
+ auto view_of_list_iterm = view_of_column->children[0];
+ // view_of_list_iterm is struct<PartitionFieldSummary>
+ if (view_of_list_iterm->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifestList("partitions list field should be a list.");
+ }
+ if (view_of_list_iterm->n_children != 4) {
+ return InvalidManifestList("PartitionFieldSummary should have 4 fields.");
+ }
+ if (view_of_list_iterm->children[0]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_null should have be bool type
column.");
+ }
+ auto contains_null = view_of_list_iterm->children[0];
+ if (view_of_list_iterm->children[1]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_nan should have be bool type
column.");
+ }
+ auto contains_nan = view_of_list_iterm->children[1];
+ if (view_of_list_iterm->children[2]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("lower_bound should have be binary type
column.");
+ }
+ auto lower_bound_list = view_of_list_iterm->children[2];
+ if (view_of_list_iterm->children[3]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("upper_bound should have be binary type
column.");
+ }
+ auto upper_bound_list = view_of_list_iterm->children[3];
+ for (int64_t manifest_idx = 0; manifest_idx < manifest_count;
manifest_idx++) {
+ auto offset = ArrowArrayViewListChildOffset(view_of_column, manifest_idx);
+ auto next_offset = ArrowArrayViewListChildOffset(view_of_column,
manifest_idx + 1);
+ // partitions from offset to next_offset belongs to manifest_idx
+ auto& manifest_file = manifest_files[manifest_idx];
+ for (int64_t partition_idx = offset; partition_idx < next_offset;
partition_idx++) {
+ PartitionFieldSummary partition_field_summary;
+ if (!ArrowArrayViewIsNull(contains_null, partition_idx)) {
+ partition_field_summary.contains_null =
+ ArrowArrayViewGetIntUnsafe(contains_null, partition_idx);
+ } else {
+ return InvalidManifestList("contains_null is null at row {}",
partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(contains_nan, partition_idx)) {
+ partition_field_summary.contains_nan =
+ ArrowArrayViewGetIntUnsafe(contains_nan, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(lower_bound_list, partition_idx)) {
+ partition_field_summary.lower_bound =
+ ArrowArrayViewGetInt8Vector(lower_bound_list, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(upper_bound_list, partition_idx)) {
+ partition_field_summary.upper_bound =
+ ArrowArrayViewGetInt8Vector(upper_bound_list, partition_idx);
+ }
+
+ manifest_file.partitions.emplace_back(partition_field_summary);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestFile>> ParseManifestList(ArrowSchema* schema,
+ ArrowArray* array_in,
+ const Schema&
iceberg_schema) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestFile> manifest_files;
+ manifest_files.resize(array_in->length);
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+ switch (field_id) {
+ case ManifestFile::kManifestPathFieldId:
+ PARSE_STRING_FIELD(manifest_files[row_idx].manifest_path,
view_of_column);
+ break;
+ case ManifestFile::kManifestLengthFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].manifest_length,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSpecIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].partition_spec_id,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].content, view_of_column,
+ ManifestContent);
+ break;
+ case ManifestFile::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kMinSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].min_sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kExistingFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_files_count,
+ view_of_column, int32_t);
+ break;
+ case ManifestFile::kDeletedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kAddedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kExistingRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kDeletedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSummaryFieldId:
+ ICEBERG_RETURN_UNEXPECTED(
+ ParsePartitionFieldSummaryList(view_of_column, manifest_files));
+ break;
+ case ManifestFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_files[row_idx].key_metadata,
view_of_column);
+ break;
+ case ManifestFile::kFirstRowIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].first_row_id,
view_of_column,
+ int64_t);
+ break;
+ default:
+ return InvalidManifestList("Unsupported field: {} in manifest file.",
field_name);
+ }
+ }
+ return manifest_files;
+}
+
+Status ParseLiteral(ArrowArrayView* view_of_partition, int64_t row_idx,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_partition->storage_type == ArrowType::NANOARROW_TYPE_BOOL) {
+ auto value = ArrowArrayViewGetUIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Boolean(value
!= 0));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_INT32) {
+ auto value = ArrowArrayViewGetIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Int(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_INT64) {
+ auto value = ArrowArrayViewGetIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Long(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_FLOAT) {
+ auto value = ArrowArrayViewGetDoubleUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Float(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_DOUBLE) {
+ auto value = ArrowArrayViewGetDoubleUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Double(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_STRING) {
+ auto value = ArrowArrayViewGetStringUnsafe(view_of_partition, row_idx);
+ manifest_entries[row_idx].data_file->partition.AddValue(
+ Literal::String(std::string(value.data, value.size_bytes)));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_BINARY) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view_of_partition, row_idx);
+ manifest_entries[row_idx].data_file->partition.AddValue(
+ Literal::Binary(std::vector<uint8_t>(buffer.data.as_char,
+ buffer.data.as_char +
buffer.size_bytes)));
+ } else {
+ return InvalidManifest("Unsupported field type: {} in data file
partition.",
+
static_cast<int32_t>(view_of_partition->storage_type));
+ }
+ return {};
+}
+
+Status ParseDataFile(const std::shared_ptr<StructType>& data_file_schema,
+ ArrowArrayView* view_of_column, std::optional<int64_t>&
first_row_id,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifest("DataFile field should be a struct.");
+ }
+ if (view_of_column->n_children != data_file_schema->fields().size()) {
+ return InvalidManifest("DataFile schema size:{} not match with ArrayArray
columns:{}",
+ data_file_schema->fields().size(),
view_of_column->n_children);
+ }
+ for (int64_t col_idx = 0; col_idx < view_of_column->n_children; ++col_idx) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field,
data_file_schema->GetFieldByIndex(col_idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
col_idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto array_view = view_of_column->children[col_idx];
+ auto manifest_entry_count = array_view->length;
+
+ switch (field_id) {
+ case DataFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content,
array_view,
+ DataFile::Content);
+ break;
+ case DataFile::kFilePathFieldId:
+ PARSE_STRING_FIELD(manifest_entries[row_idx].data_file->file_path,
array_view);
+ break;
+ case DataFile::kFileFormatFieldId:
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
+ std::string_view path_str(value.data, value.size_bytes);
+
ICEBERG_ASSIGN_OR_RAISE(manifest_entries[row_idx].data_file->file_format,
+ FileFormatTypeFromString(path_str));
+ }
+ }
+ break;
+ case DataFile::kPartitionFieldId: {
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifest("Field:{} should be a struct.", field_name);
+ }
+ if (array_view->n_children > 0) {
+ for (int64_t partition_idx = 0; partition_idx <
array_view->n_children;
+ partition_idx++) {
+ auto view_of_partition = array_view->children[partition_idx];
+ for (int64_t row_idx = 0; row_idx < view_of_partition->length;
row_idx++) {
+ if (ArrowArrayViewIsNull(view_of_partition, row_idx)) {
+ break;
+ }
+ ICEBERG_RETURN_UNEXPECTED(
+ ParseLiteral(view_of_partition, row_idx, manifest_entries));
+ }
+ }
+ }
+ } break;
+ case DataFile::kRecordCountFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->record_count,
+ array_view, int64_t);
+ break;
+ case DataFile::kFileSizeFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->file_size_in_bytes,
+ array_view, int64_t);
+ break;
+ case DataFile::kColumnSizesFieldId:
+ // key&value should have the same offset
+ // HACK(xiao.dong) workaround for arrow bug:
+ // ArrowArrayViewListChildOffset can not get the correct offset for map
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->column_sizes,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kNullValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->null_value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kNanValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->nan_value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kLowerBoundsFieldId:
+
PARSE_INT_BINARY_MAP_FIELD(manifest_entries[row_idx].data_file->lower_bounds,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kUpperBoundsFieldId:
+
PARSE_INT_BINARY_MAP_FIELD(manifest_entries[row_idx].data_file->upper_bounds,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_entries[row_idx].data_file->key_metadata,
array_view);
+ break;
+ case DataFile::kSplitOffsetsFieldId:
+ PARSE_INTEGER_VECTOR_FIELD(
+ manifest_entries[manifest_idx].data_file->split_offsets,
manifest_entry_count,
+ array_view, int64_t);
+ break;
+ case DataFile::kEqualityIdsFieldId:
+
PARSE_INTEGER_VECTOR_FIELD(manifest_entries[manifest_idx].data_file->equality_ids,
+ manifest_entry_count, array_view, int32_t);
+ break;
+ case DataFile::kSortOrderIdFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->sort_order_id,
+ array_view, int32_t);
+ break;
+ case DataFile::kFirstRowIdFieldId: {
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->first_row_id,
+ array_view, int64_t);
+ if (first_row_id.has_value()) {
+ std::ranges::for_each(manifest_entries,
[&first_row_id](ManifestEntry& entry) {
+ if (entry.status != ManifestStatus::kDeleted &&
+ !entry.data_file->first_row_id.has_value()) {
+ entry.data_file->first_row_id = first_row_id.value();
+ first_row_id = first_row_id.value() +
entry.data_file->record_count;
+ }
+ });
+ } else {
+ // data file's first_row_id is null when the manifest's first_row_id
is null
+ std::ranges::for_each(manifest_entries, [](ManifestEntry& entry) {
+ entry.data_file->first_row_id = std::nullopt;
+ });
+ }
+ break;
+ }
+ case DataFile::kReferencedDataFileFieldId:
+
PARSE_STRING_FIELD(manifest_entries[row_idx].data_file->referenced_data_file,
+ array_view);
+ break;
+ case DataFile::kContentOffsetFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content_offset,
+ array_view, int64_t);
+ break;
+ case DataFile::kContentSizeFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content_size_in_bytes,
+ array_view, int64_t);
+ break;
+ default:
+ return InvalidManifest("Unsupported field: {} in data file.",
field_name);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestEntry>> ParseManifestEntry(
+ ArrowSchema* schema, ArrowArray* array_in, const Schema& iceberg_schema,
+ std::optional<int64_t>& first_row_id) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifest("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifest("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestEntry> manifest_entries;
+ manifest_entries.resize(array_in->length);
+ for (int64_t i = 0; i < array_in->length; i++) {
+ manifest_entries[i].data_file = std::make_shared<DataFile>();
+ }
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for manifest entry
schema", idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+
+ switch (field_id) {
+ case ManifestEntry::kStatusFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].status, view_of_column,
+ ManifestStatus);
+ break;
+ case ManifestEntry::kSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestEntry::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestEntry::kFileSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].file_sequence_number,
+ view_of_column, int64_t);
+ break;
+ case ManifestEntry::kDataFileFieldId: {
+ auto data_file_schema =
+ internal::checked_pointer_cast<StructType>(field->get().type());
+ ICEBERG_RETURN_UNEXPECTED(ParseDataFile(data_file_schema,
view_of_column,
+ first_row_id,
manifest_entries));
+ break;
+ }
+ default:
+ return InvalidManifest("Unsupported field: {} in manifest entry.",
field_name);
+ }
+ }
+ return manifest_entries;
+}
+
+const std::unordered_set<std::string> kStatsColumns = {
Review Comment:
I think `std::vector<std::string>` is enough.
##########
src/iceberg/manifest/manifest_reader.cc:
##########
@@ -19,61 +19,867 @@
#include "iceberg/manifest/manifest_reader.h"
+#include <algorithm>
+#include <memory>
+#include <unordered_set>
+
+#include <nanoarrow/nanoarrow.h>
+
+#include "iceberg/arrow/nanoarrow_status_internal.h"
+#include "iceberg/arrow_c_data_guard_internal.h"
+#include "iceberg/expression/expression.h"
+#include "iceberg/expression/projections.h"
+#include "iceberg/file_format.h"
#include "iceberg/manifest/manifest_entry.h"
#include "iceberg/manifest/manifest_list.h"
#include "iceberg/manifest/manifest_reader_internal.h"
+#include "iceberg/metadata_columns.h"
+#include "iceberg/partition_spec.h"
#include "iceberg/schema.h"
-#include "iceberg/schema_internal.h"
+#include "iceberg/schema_field.h"
#include "iceberg/type.h"
+#include "iceberg/util/checked_cast.h"
#include "iceberg/util/macros.h"
namespace iceberg {
+namespace {
+
+#define PARSE_PRIMITIVE_FIELD(item, array_view, type)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetIntUnsafe(array_view, row_idx);
\
+ item = static_cast<type>(value);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_STRING_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
\
+ item = std::string(value.data, value.size_bytes);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_BINARY_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(view_of_column, row_idx)) {
\
+ item = ArrowArrayViewGetInt8Vector(array_view, row_idx);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_INTEGER_VECTOR_FIELD(item, count, array_view, type)
\
+ for (int64_t manifest_idx = 0; manifest_idx < count; manifest_idx++) {
\
+ auto offset = ArrowArrayViewListChildOffset(array_view, manifest_idx);
\
+ auto next_offset = ArrowArrayViewListChildOffset(array_view, manifest_idx
+ 1); \
+ for (int64_t offset_idx = offset; offset_idx < next_offset; offset_idx++)
{ \
+ item.emplace_back(static_cast<type>(
\
+ ArrowArrayViewGetIntUnsafe(array_view->children[0], offset_idx)));
\
+ }
\
+ }
+
+#define PARSE_MAP_FIELD(item, count, array_view, key_type, value_type,
assignment) \
+ do {
\
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_MAP) {
\
+ return InvalidManifest("Field:{} should be a map.", field_name);
\
+ }
\
+ auto view_of_map = array_view->children[0];
\
+ ASSERT_VIEW_TYPE_AND_CHILDREN(view_of_map,
ArrowType::NANOARROW_TYPE_STRUCT, 2); \
+ auto view_of_map_key = view_of_map->children[0];
\
+ ASSERT_VIEW_TYPE(view_of_map_key, key_type);
\
+ auto view_of_map_value = view_of_map->children[1];
\
+ ASSERT_VIEW_TYPE(view_of_map_value, value_type);
\
+ for (int64_t row_idx = 0; row_idx < count; row_idx++) {
\
+ auto offset = array_view->buffer_views[1].data.as_int32[row_idx];
\
+ auto next_offset = array_view->buffer_views[1].data.as_int32[row_idx +
1]; \
+ for (int32_t offset_idx = offset; offset_idx < next_offset;
offset_idx++) { \
+ auto key = ArrowArrayViewGetIntUnsafe(view_of_map_key, offset_idx);
\
+ item[key] = assignment;
\
+ }
\
+ }
\
+ } while (0)
+
+#define PARSE_INT_LONG_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_INT64, \
+ ArrowArrayViewGetIntUnsafe(view_of_map_value, offset_idx));
+
+#define PARSE_INT_BINARY_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_BINARY, \
+ ArrowArrayViewGetInt8Vector(view_of_map_value, offset_idx));
+
+#define ASSERT_VIEW_TYPE(view, type)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
+
+#define ASSERT_VIEW_TYPE_AND_CHILDREN(view, type, n_child)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
\
+ if (view->n_children != n_child) {
\
+ return InvalidManifest("Sub Field for:{} should have key&value:{}
columns.", \
+ field_name, n_child);
\
+ }
+
+std::vector<uint8_t> ArrowArrayViewGetInt8Vector(const ArrowArrayView* view,
+ int32_t offset_idx) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view, offset_idx);
+ return {buffer.data.as_char, buffer.data.as_char + buffer.size_bytes};
+}
+
+Status ParsePartitionFieldSummaryList(ArrowArrayView* view_of_column,
+ std::vector<ManifestFile>&
manifest_files) {
+ auto manifest_count = view_of_column->length;
+ // view_of_column is list<struct<PartitionFieldSummary>>
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_LIST) {
+ return InvalidManifestList("partitions field should be a list.");
+ }
+ auto view_of_list_iterm = view_of_column->children[0];
+ // view_of_list_iterm is struct<PartitionFieldSummary>
+ if (view_of_list_iterm->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifestList("partitions list field should be a list.");
+ }
+ if (view_of_list_iterm->n_children != 4) {
+ return InvalidManifestList("PartitionFieldSummary should have 4 fields.");
+ }
+ if (view_of_list_iterm->children[0]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_null should have be bool type
column.");
+ }
+ auto contains_null = view_of_list_iterm->children[0];
+ if (view_of_list_iterm->children[1]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_nan should have be bool type
column.");
+ }
+ auto contains_nan = view_of_list_iterm->children[1];
+ if (view_of_list_iterm->children[2]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("lower_bound should have be binary type
column.");
+ }
+ auto lower_bound_list = view_of_list_iterm->children[2];
+ if (view_of_list_iterm->children[3]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("upper_bound should have be binary type
column.");
+ }
+ auto upper_bound_list = view_of_list_iterm->children[3];
+ for (int64_t manifest_idx = 0; manifest_idx < manifest_count;
manifest_idx++) {
+ auto offset = ArrowArrayViewListChildOffset(view_of_column, manifest_idx);
+ auto next_offset = ArrowArrayViewListChildOffset(view_of_column,
manifest_idx + 1);
+ // partitions from offset to next_offset belongs to manifest_idx
+ auto& manifest_file = manifest_files[manifest_idx];
+ for (int64_t partition_idx = offset; partition_idx < next_offset;
partition_idx++) {
+ PartitionFieldSummary partition_field_summary;
+ if (!ArrowArrayViewIsNull(contains_null, partition_idx)) {
+ partition_field_summary.contains_null =
+ ArrowArrayViewGetIntUnsafe(contains_null, partition_idx);
+ } else {
+ return InvalidManifestList("contains_null is null at row {}",
partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(contains_nan, partition_idx)) {
+ partition_field_summary.contains_nan =
+ ArrowArrayViewGetIntUnsafe(contains_nan, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(lower_bound_list, partition_idx)) {
+ partition_field_summary.lower_bound =
+ ArrowArrayViewGetInt8Vector(lower_bound_list, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(upper_bound_list, partition_idx)) {
+ partition_field_summary.upper_bound =
+ ArrowArrayViewGetInt8Vector(upper_bound_list, partition_idx);
+ }
+
+ manifest_file.partitions.emplace_back(partition_field_summary);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestFile>> ParseManifestList(ArrowSchema* schema,
+ ArrowArray* array_in,
+ const Schema&
iceberg_schema) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestFile> manifest_files;
+ manifest_files.resize(array_in->length);
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+ switch (field_id) {
+ case ManifestFile::kManifestPathFieldId:
+ PARSE_STRING_FIELD(manifest_files[row_idx].manifest_path,
view_of_column);
+ break;
+ case ManifestFile::kManifestLengthFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].manifest_length,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSpecIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].partition_spec_id,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].content, view_of_column,
+ ManifestContent);
+ break;
+ case ManifestFile::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kMinSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].min_sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kExistingFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_files_count,
+ view_of_column, int32_t);
+ break;
+ case ManifestFile::kDeletedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kAddedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kExistingRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kDeletedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSummaryFieldId:
+ ICEBERG_RETURN_UNEXPECTED(
+ ParsePartitionFieldSummaryList(view_of_column, manifest_files));
+ break;
+ case ManifestFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_files[row_idx].key_metadata,
view_of_column);
+ break;
+ case ManifestFile::kFirstRowIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].first_row_id,
view_of_column,
+ int64_t);
+ break;
+ default:
+ return InvalidManifestList("Unsupported field: {} in manifest file.",
field_name);
+ }
+ }
+ return manifest_files;
+}
+
+Status ParseLiteral(ArrowArrayView* view_of_partition, int64_t row_idx,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_partition->storage_type == ArrowType::NANOARROW_TYPE_BOOL) {
+ auto value = ArrowArrayViewGetUIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Boolean(value
!= 0));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_INT32) {
+ auto value = ArrowArrayViewGetIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Int(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_INT64) {
+ auto value = ArrowArrayViewGetIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Long(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_FLOAT) {
+ auto value = ArrowArrayViewGetDoubleUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Float(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_DOUBLE) {
+ auto value = ArrowArrayViewGetDoubleUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Double(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_STRING) {
+ auto value = ArrowArrayViewGetStringUnsafe(view_of_partition, row_idx);
+ manifest_entries[row_idx].data_file->partition.AddValue(
+ Literal::String(std::string(value.data, value.size_bytes)));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_BINARY) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view_of_partition, row_idx);
+ manifest_entries[row_idx].data_file->partition.AddValue(
+ Literal::Binary(std::vector<uint8_t>(buffer.data.as_char,
+ buffer.data.as_char +
buffer.size_bytes)));
+ } else {
+ return InvalidManifest("Unsupported field type: {} in data file
partition.",
+
static_cast<int32_t>(view_of_partition->storage_type));
+ }
+ return {};
+}
+
+Status ParseDataFile(const std::shared_ptr<StructType>& data_file_schema,
+ ArrowArrayView* view_of_column, std::optional<int64_t>&
first_row_id,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifest("DataFile field should be a struct.");
+ }
+ if (view_of_column->n_children != data_file_schema->fields().size()) {
+ return InvalidManifest("DataFile schema size:{} not match with ArrayArray
columns:{}",
+ data_file_schema->fields().size(),
view_of_column->n_children);
+ }
+ for (int64_t col_idx = 0; col_idx < view_of_column->n_children; ++col_idx) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field,
data_file_schema->GetFieldByIndex(col_idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
col_idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto array_view = view_of_column->children[col_idx];
+ auto manifest_entry_count = array_view->length;
+
+ switch (field_id) {
+ case DataFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content,
array_view,
+ DataFile::Content);
+ break;
+ case DataFile::kFilePathFieldId:
+ PARSE_STRING_FIELD(manifest_entries[row_idx].data_file->file_path,
array_view);
+ break;
+ case DataFile::kFileFormatFieldId:
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
+ std::string_view path_str(value.data, value.size_bytes);
+
ICEBERG_ASSIGN_OR_RAISE(manifest_entries[row_idx].data_file->file_format,
+ FileFormatTypeFromString(path_str));
+ }
+ }
+ break;
+ case DataFile::kPartitionFieldId: {
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifest("Field:{} should be a struct.", field_name);
+ }
+ if (array_view->n_children > 0) {
+ for (int64_t partition_idx = 0; partition_idx <
array_view->n_children;
+ partition_idx++) {
+ auto view_of_partition = array_view->children[partition_idx];
+ for (int64_t row_idx = 0; row_idx < view_of_partition->length;
row_idx++) {
+ if (ArrowArrayViewIsNull(view_of_partition, row_idx)) {
+ break;
+ }
+ ICEBERG_RETURN_UNEXPECTED(
+ ParseLiteral(view_of_partition, row_idx, manifest_entries));
+ }
+ }
+ }
+ } break;
+ case DataFile::kRecordCountFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->record_count,
+ array_view, int64_t);
+ break;
+ case DataFile::kFileSizeFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->file_size_in_bytes,
+ array_view, int64_t);
+ break;
+ case DataFile::kColumnSizesFieldId:
+ // key&value should have the same offset
+ // HACK(xiao.dong) workaround for arrow bug:
+ // ArrowArrayViewListChildOffset can not get the correct offset for map
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->column_sizes,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kNullValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->null_value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kNanValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->nan_value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kLowerBoundsFieldId:
+
PARSE_INT_BINARY_MAP_FIELD(manifest_entries[row_idx].data_file->lower_bounds,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kUpperBoundsFieldId:
+
PARSE_INT_BINARY_MAP_FIELD(manifest_entries[row_idx].data_file->upper_bounds,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_entries[row_idx].data_file->key_metadata,
array_view);
+ break;
+ case DataFile::kSplitOffsetsFieldId:
+ PARSE_INTEGER_VECTOR_FIELD(
+ manifest_entries[manifest_idx].data_file->split_offsets,
manifest_entry_count,
+ array_view, int64_t);
+ break;
+ case DataFile::kEqualityIdsFieldId:
+
PARSE_INTEGER_VECTOR_FIELD(manifest_entries[manifest_idx].data_file->equality_ids,
+ manifest_entry_count, array_view, int32_t);
+ break;
+ case DataFile::kSortOrderIdFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->sort_order_id,
+ array_view, int32_t);
+ break;
+ case DataFile::kFirstRowIdFieldId: {
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->first_row_id,
+ array_view, int64_t);
+ if (first_row_id.has_value()) {
+ std::ranges::for_each(manifest_entries,
[&first_row_id](ManifestEntry& entry) {
+ if (entry.status != ManifestStatus::kDeleted &&
+ !entry.data_file->first_row_id.has_value()) {
+ entry.data_file->first_row_id = first_row_id.value();
+ first_row_id = first_row_id.value() +
entry.data_file->record_count;
+ }
+ });
+ } else {
+ // data file's first_row_id is null when the manifest's first_row_id
is null
+ std::ranges::for_each(manifest_entries, [](ManifestEntry& entry) {
+ entry.data_file->first_row_id = std::nullopt;
+ });
+ }
+ break;
+ }
+ case DataFile::kReferencedDataFileFieldId:
+
PARSE_STRING_FIELD(manifest_entries[row_idx].data_file->referenced_data_file,
+ array_view);
+ break;
+ case DataFile::kContentOffsetFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content_offset,
+ array_view, int64_t);
+ break;
+ case DataFile::kContentSizeFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content_size_in_bytes,
+ array_view, int64_t);
+ break;
+ default:
+ return InvalidManifest("Unsupported field: {} in data file.",
field_name);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestEntry>> ParseManifestEntry(
+ ArrowSchema* schema, ArrowArray* array_in, const Schema& iceberg_schema,
+ std::optional<int64_t>& first_row_id) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifest("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifest("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestEntry> manifest_entries;
+ manifest_entries.resize(array_in->length);
+ for (int64_t i = 0; i < array_in->length; i++) {
+ manifest_entries[i].data_file = std::make_shared<DataFile>();
+ }
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for manifest entry
schema", idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+
+ switch (field_id) {
+ case ManifestEntry::kStatusFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].status, view_of_column,
+ ManifestStatus);
+ break;
+ case ManifestEntry::kSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestEntry::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestEntry::kFileSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].file_sequence_number,
+ view_of_column, int64_t);
+ break;
+ case ManifestEntry::kDataFileFieldId: {
+ auto data_file_schema =
+ internal::checked_pointer_cast<StructType>(field->get().type());
+ ICEBERG_RETURN_UNEXPECTED(ParseDataFile(data_file_schema,
view_of_column,
+ first_row_id,
manifest_entries));
+ break;
+ }
+ default:
+ return InvalidManifest("Unsupported field: {} in manifest entry.",
field_name);
+ }
+ }
+ return manifest_entries;
+}
+
+const std::unordered_set<std::string> kStatsColumns = {
+ "value_counts", "null_value_counts", "nan_value_counts",
+ "lower_bounds", "upper_bounds", "record_count"};
+
+bool RequireStatsProjection(const std::shared_ptr<Expression>& row_filter,
+ const std::vector<std::string>& columns) {
+ if (!row_filter || row_filter->op() == Expression::Operation::kTrue) {
+ return false;
+ }
+ if (columns.empty()) {
+ return false;
+ }
+ const std::unordered_set<std::string> selected(columns.cbegin(),
columns.cend());
+ if (std::ranges::all_of(
+ ManifestReader::kAllColumns,
+ [&selected](const std::string& col) { return selected.contains(col);
})) {
+ return false;
+ }
+ if (std::ranges::all_of(kStatsColumns, [&selected](const std::string& col) {
+ return selected.contains(col);
+ })) {
+ return false;
+ }
+ return true;
+}
+
+Result<std::shared_ptr<Schema>> ProjectSchema(std::shared_ptr<Schema> schema,
+ const std::vector<std::string>&
columns,
+ bool case_sensitive) {
+ if (!columns.empty()) {
+ return schema->Select(columns, case_sensitive);
+ }
+ return schema;
+}
+
+} // namespace
+
+std::vector<std::string> ManifestReader::WithStatsColumns(
+ const std::vector<std::string>& columns) {
+ if (std::ranges::all_of(ManifestReader::kAllColumns,
+ [&columns](const std::string& col) {
+ return std::ranges::find(columns, col) !=
columns.cend();
+ })) {
Review Comment:
Use `std::ranges::contains` instead.
##########
src/iceberg/manifest/manifest_list.cc:
##########
@@ -19,22 +19,24 @@
#include "iceberg/manifest/manifest_list.h"
-#include "iceberg/schema.h"
+#include <memory>
+
+#include "iceberg/type.h"
namespace iceberg {
-const StructType& PartitionFieldSummary::Type() {
- static const StructType kInstance{{
+const std::shared_ptr<StructType>& PartitionFieldSummary::Type() {
Review Comment:
Why do we need to change the return type from `const T&` to `const
std::shared_ptr<T>&`?
##########
src/iceberg/manifest/manifest_reader.cc:
##########
@@ -19,61 +19,870 @@
#include "iceberg/manifest/manifest_reader.h"
+#include <algorithm>
+#include <memory>
+#include <unordered_set>
+
+#include <nanoarrow/nanoarrow.h>
+
+#include "iceberg/arrow/nanoarrow_status_internal.h"
+#include "iceberg/arrow_c_data_guard_internal.h"
+#include "iceberg/expression/expression.h"
+#include "iceberg/expression/projections.h"
+#include "iceberg/file_format.h"
#include "iceberg/manifest/manifest_entry.h"
#include "iceberg/manifest/manifest_list.h"
#include "iceberg/manifest/manifest_reader_internal.h"
+#include "iceberg/metadata_columns.h"
+#include "iceberg/partition_spec.h"
#include "iceberg/schema.h"
-#include "iceberg/schema_internal.h"
+#include "iceberg/schema_field.h"
#include "iceberg/type.h"
+#include "iceberg/util/checked_cast.h"
#include "iceberg/util/macros.h"
namespace iceberg {
+namespace {
+
+#define PARSE_PRIMITIVE_FIELD(item, array_view, type)
\
Review Comment:
Can we change these macros to template functions?
##########
src/iceberg/manifest/manifest_reader.cc:
##########
@@ -19,61 +19,867 @@
#include "iceberg/manifest/manifest_reader.h"
+#include <algorithm>
+#include <memory>
+#include <unordered_set>
+
+#include <nanoarrow/nanoarrow.h>
+
+#include "iceberg/arrow/nanoarrow_status_internal.h"
+#include "iceberg/arrow_c_data_guard_internal.h"
+#include "iceberg/expression/expression.h"
+#include "iceberg/expression/projections.h"
+#include "iceberg/file_format.h"
#include "iceberg/manifest/manifest_entry.h"
#include "iceberg/manifest/manifest_list.h"
#include "iceberg/manifest/manifest_reader_internal.h"
+#include "iceberg/metadata_columns.h"
+#include "iceberg/partition_spec.h"
#include "iceberg/schema.h"
-#include "iceberg/schema_internal.h"
+#include "iceberg/schema_field.h"
#include "iceberg/type.h"
+#include "iceberg/util/checked_cast.h"
#include "iceberg/util/macros.h"
namespace iceberg {
+namespace {
+
+#define PARSE_PRIMITIVE_FIELD(item, array_view, type)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetIntUnsafe(array_view, row_idx);
\
+ item = static_cast<type>(value);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_STRING_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
\
+ item = std::string(value.data, value.size_bytes);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_BINARY_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(view_of_column, row_idx)) {
\
+ item = ArrowArrayViewGetInt8Vector(array_view, row_idx);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_INTEGER_VECTOR_FIELD(item, count, array_view, type)
\
+ for (int64_t manifest_idx = 0; manifest_idx < count; manifest_idx++) {
\
+ auto offset = ArrowArrayViewListChildOffset(array_view, manifest_idx);
\
+ auto next_offset = ArrowArrayViewListChildOffset(array_view, manifest_idx
+ 1); \
+ for (int64_t offset_idx = offset; offset_idx < next_offset; offset_idx++)
{ \
+ item.emplace_back(static_cast<type>(
\
+ ArrowArrayViewGetIntUnsafe(array_view->children[0], offset_idx)));
\
+ }
\
+ }
+
+#define PARSE_MAP_FIELD(item, count, array_view, key_type, value_type,
assignment) \
+ do {
\
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_MAP) {
\
+ return InvalidManifest("Field:{} should be a map.", field_name);
\
+ }
\
+ auto view_of_map = array_view->children[0];
\
+ ASSERT_VIEW_TYPE_AND_CHILDREN(view_of_map,
ArrowType::NANOARROW_TYPE_STRUCT, 2); \
+ auto view_of_map_key = view_of_map->children[0];
\
+ ASSERT_VIEW_TYPE(view_of_map_key, key_type);
\
+ auto view_of_map_value = view_of_map->children[1];
\
+ ASSERT_VIEW_TYPE(view_of_map_value, value_type);
\
+ for (int64_t row_idx = 0; row_idx < count; row_idx++) {
\
+ auto offset = array_view->buffer_views[1].data.as_int32[row_idx];
\
+ auto next_offset = array_view->buffer_views[1].data.as_int32[row_idx +
1]; \
+ for (int32_t offset_idx = offset; offset_idx < next_offset;
offset_idx++) { \
+ auto key = ArrowArrayViewGetIntUnsafe(view_of_map_key, offset_idx);
\
+ item[key] = assignment;
\
+ }
\
+ }
\
+ } while (0)
+
+#define PARSE_INT_LONG_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_INT64, \
+ ArrowArrayViewGetIntUnsafe(view_of_map_value, offset_idx));
+
+#define PARSE_INT_BINARY_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_BINARY, \
+ ArrowArrayViewGetInt8Vector(view_of_map_value, offset_idx));
+
+#define ASSERT_VIEW_TYPE(view, type)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
+
+#define ASSERT_VIEW_TYPE_AND_CHILDREN(view, type, n_child)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
\
+ if (view->n_children != n_child) {
\
+ return InvalidManifest("Sub Field for:{} should have key&value:{}
columns.", \
+ field_name, n_child);
\
+ }
+
+std::vector<uint8_t> ArrowArrayViewGetInt8Vector(const ArrowArrayView* view,
+ int32_t offset_idx) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view, offset_idx);
+ return {buffer.data.as_char, buffer.data.as_char + buffer.size_bytes};
+}
+
+Status ParsePartitionFieldSummaryList(ArrowArrayView* view_of_column,
+ std::vector<ManifestFile>&
manifest_files) {
+ auto manifest_count = view_of_column->length;
+ // view_of_column is list<struct<PartitionFieldSummary>>
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_LIST) {
+ return InvalidManifestList("partitions field should be a list.");
+ }
+ auto view_of_list_iterm = view_of_column->children[0];
+ // view_of_list_iterm is struct<PartitionFieldSummary>
+ if (view_of_list_iterm->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifestList("partitions list field should be a list.");
+ }
+ if (view_of_list_iterm->n_children != 4) {
+ return InvalidManifestList("PartitionFieldSummary should have 4 fields.");
+ }
+ if (view_of_list_iterm->children[0]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_null should have be bool type
column.");
+ }
+ auto contains_null = view_of_list_iterm->children[0];
+ if (view_of_list_iterm->children[1]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_nan should have be bool type
column.");
+ }
+ auto contains_nan = view_of_list_iterm->children[1];
+ if (view_of_list_iterm->children[2]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("lower_bound should have be binary type
column.");
+ }
+ auto lower_bound_list = view_of_list_iterm->children[2];
+ if (view_of_list_iterm->children[3]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("upper_bound should have be binary type
column.");
+ }
+ auto upper_bound_list = view_of_list_iterm->children[3];
+ for (int64_t manifest_idx = 0; manifest_idx < manifest_count;
manifest_idx++) {
+ auto offset = ArrowArrayViewListChildOffset(view_of_column, manifest_idx);
+ auto next_offset = ArrowArrayViewListChildOffset(view_of_column,
manifest_idx + 1);
+ // partitions from offset to next_offset belongs to manifest_idx
+ auto& manifest_file = manifest_files[manifest_idx];
+ for (int64_t partition_idx = offset; partition_idx < next_offset;
partition_idx++) {
+ PartitionFieldSummary partition_field_summary;
+ if (!ArrowArrayViewIsNull(contains_null, partition_idx)) {
+ partition_field_summary.contains_null =
+ ArrowArrayViewGetIntUnsafe(contains_null, partition_idx);
+ } else {
+ return InvalidManifestList("contains_null is null at row {}",
partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(contains_nan, partition_idx)) {
+ partition_field_summary.contains_nan =
+ ArrowArrayViewGetIntUnsafe(contains_nan, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(lower_bound_list, partition_idx)) {
+ partition_field_summary.lower_bound =
+ ArrowArrayViewGetInt8Vector(lower_bound_list, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(upper_bound_list, partition_idx)) {
+ partition_field_summary.upper_bound =
+ ArrowArrayViewGetInt8Vector(upper_bound_list, partition_idx);
+ }
+
+ manifest_file.partitions.emplace_back(partition_field_summary);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestFile>> ParseManifestList(ArrowSchema* schema,
+ ArrowArray* array_in,
+ const Schema&
iceberg_schema) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestFile> manifest_files;
+ manifest_files.resize(array_in->length);
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+ switch (field_id) {
+ case ManifestFile::kManifestPathFieldId:
+ PARSE_STRING_FIELD(manifest_files[row_idx].manifest_path,
view_of_column);
+ break;
+ case ManifestFile::kManifestLengthFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].manifest_length,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSpecIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].partition_spec_id,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].content, view_of_column,
+ ManifestContent);
+ break;
+ case ManifestFile::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kMinSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].min_sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kExistingFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_files_count,
+ view_of_column, int32_t);
+ break;
+ case ManifestFile::kDeletedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kAddedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kExistingRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kDeletedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSummaryFieldId:
+ ICEBERG_RETURN_UNEXPECTED(
+ ParsePartitionFieldSummaryList(view_of_column, manifest_files));
+ break;
+ case ManifestFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_files[row_idx].key_metadata,
view_of_column);
+ break;
+ case ManifestFile::kFirstRowIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].first_row_id,
view_of_column,
+ int64_t);
+ break;
+ default:
+ return InvalidManifestList("Unsupported field: {} in manifest file.",
field_name);
+ }
+ }
+ return manifest_files;
+}
+
+Status ParseLiteral(ArrowArrayView* view_of_partition, int64_t row_idx,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_partition->storage_type == ArrowType::NANOARROW_TYPE_BOOL) {
Review Comment:
Use `switch` instead of multiple `if`?
##########
src/iceberg/manifest/manifest_reader.cc:
##########
@@ -19,61 +19,867 @@
#include "iceberg/manifest/manifest_reader.h"
+#include <algorithm>
+#include <memory>
+#include <unordered_set>
+
+#include <nanoarrow/nanoarrow.h>
+
+#include "iceberg/arrow/nanoarrow_status_internal.h"
+#include "iceberg/arrow_c_data_guard_internal.h"
+#include "iceberg/expression/expression.h"
+#include "iceberg/expression/projections.h"
+#include "iceberg/file_format.h"
#include "iceberg/manifest/manifest_entry.h"
#include "iceberg/manifest/manifest_list.h"
#include "iceberg/manifest/manifest_reader_internal.h"
+#include "iceberg/metadata_columns.h"
+#include "iceberg/partition_spec.h"
#include "iceberg/schema.h"
-#include "iceberg/schema_internal.h"
+#include "iceberg/schema_field.h"
#include "iceberg/type.h"
+#include "iceberg/util/checked_cast.h"
#include "iceberg/util/macros.h"
namespace iceberg {
+namespace {
+
+#define PARSE_PRIMITIVE_FIELD(item, array_view, type)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetIntUnsafe(array_view, row_idx);
\
+ item = static_cast<type>(value);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_STRING_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
\
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
\
+ item = std::string(value.data, value.size_bytes);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_BINARY_FIELD(item, array_view)
\
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
\
+ if (!ArrowArrayViewIsNull(view_of_column, row_idx)) {
\
+ item = ArrowArrayViewGetInt8Vector(array_view, row_idx);
\
+ } else if (required) {
\
+ return InvalidManifestList("Field {} is required but null at row {}",
field_name, \
+ row_idx);
\
+ }
\
+ }
+
+#define PARSE_INTEGER_VECTOR_FIELD(item, count, array_view, type)
\
+ for (int64_t manifest_idx = 0; manifest_idx < count; manifest_idx++) {
\
+ auto offset = ArrowArrayViewListChildOffset(array_view, manifest_idx);
\
+ auto next_offset = ArrowArrayViewListChildOffset(array_view, manifest_idx
+ 1); \
+ for (int64_t offset_idx = offset; offset_idx < next_offset; offset_idx++)
{ \
+ item.emplace_back(static_cast<type>(
\
+ ArrowArrayViewGetIntUnsafe(array_view->children[0], offset_idx)));
\
+ }
\
+ }
+
+#define PARSE_MAP_FIELD(item, count, array_view, key_type, value_type,
assignment) \
+ do {
\
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_MAP) {
\
+ return InvalidManifest("Field:{} should be a map.", field_name);
\
+ }
\
+ auto view_of_map = array_view->children[0];
\
+ ASSERT_VIEW_TYPE_AND_CHILDREN(view_of_map,
ArrowType::NANOARROW_TYPE_STRUCT, 2); \
+ auto view_of_map_key = view_of_map->children[0];
\
+ ASSERT_VIEW_TYPE(view_of_map_key, key_type);
\
+ auto view_of_map_value = view_of_map->children[1];
\
+ ASSERT_VIEW_TYPE(view_of_map_value, value_type);
\
+ for (int64_t row_idx = 0; row_idx < count; row_idx++) {
\
+ auto offset = array_view->buffer_views[1].data.as_int32[row_idx];
\
+ auto next_offset = array_view->buffer_views[1].data.as_int32[row_idx +
1]; \
+ for (int32_t offset_idx = offset; offset_idx < next_offset;
offset_idx++) { \
+ auto key = ArrowArrayViewGetIntUnsafe(view_of_map_key, offset_idx);
\
+ item[key] = assignment;
\
+ }
\
+ }
\
+ } while (0)
+
+#define PARSE_INT_LONG_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_INT64, \
+ ArrowArrayViewGetIntUnsafe(view_of_map_value, offset_idx));
+
+#define PARSE_INT_BINARY_MAP_FIELD(item, count, array_view) \
+ PARSE_MAP_FIELD(item, count, array_view, ArrowType::NANOARROW_TYPE_INT32, \
+ ArrowType::NANOARROW_TYPE_BINARY, \
+ ArrowArrayViewGetInt8Vector(view_of_map_value, offset_idx));
+
+#define ASSERT_VIEW_TYPE(view, type)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
+
+#define ASSERT_VIEW_TYPE_AND_CHILDREN(view, type, n_child)
\
+ if (view->storage_type != type) {
\
+ return InvalidManifest("Sub Field:{} should be a {}.", field_name, #type);
\
+ }
\
+ if (view->n_children != n_child) {
\
+ return InvalidManifest("Sub Field for:{} should have key&value:{}
columns.", \
+ field_name, n_child);
\
+ }
+
+std::vector<uint8_t> ArrowArrayViewGetInt8Vector(const ArrowArrayView* view,
+ int32_t offset_idx) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view, offset_idx);
+ return {buffer.data.as_char, buffer.data.as_char + buffer.size_bytes};
+}
+
+Status ParsePartitionFieldSummaryList(ArrowArrayView* view_of_column,
+ std::vector<ManifestFile>&
manifest_files) {
+ auto manifest_count = view_of_column->length;
+ // view_of_column is list<struct<PartitionFieldSummary>>
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_LIST) {
+ return InvalidManifestList("partitions field should be a list.");
+ }
+ auto view_of_list_iterm = view_of_column->children[0];
+ // view_of_list_iterm is struct<PartitionFieldSummary>
+ if (view_of_list_iterm->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifestList("partitions list field should be a list.");
+ }
+ if (view_of_list_iterm->n_children != 4) {
+ return InvalidManifestList("PartitionFieldSummary should have 4 fields.");
+ }
+ if (view_of_list_iterm->children[0]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_null should have be bool type
column.");
+ }
+ auto contains_null = view_of_list_iterm->children[0];
+ if (view_of_list_iterm->children[1]->storage_type !=
ArrowType::NANOARROW_TYPE_BOOL) {
+ return InvalidManifestList("contains_nan should have be bool type
column.");
+ }
+ auto contains_nan = view_of_list_iterm->children[1];
+ if (view_of_list_iterm->children[2]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("lower_bound should have be binary type
column.");
+ }
+ auto lower_bound_list = view_of_list_iterm->children[2];
+ if (view_of_list_iterm->children[3]->storage_type !=
ArrowType::NANOARROW_TYPE_BINARY) {
+ return InvalidManifestList("upper_bound should have be binary type
column.");
+ }
+ auto upper_bound_list = view_of_list_iterm->children[3];
+ for (int64_t manifest_idx = 0; manifest_idx < manifest_count;
manifest_idx++) {
+ auto offset = ArrowArrayViewListChildOffset(view_of_column, manifest_idx);
+ auto next_offset = ArrowArrayViewListChildOffset(view_of_column,
manifest_idx + 1);
+ // partitions from offset to next_offset belongs to manifest_idx
+ auto& manifest_file = manifest_files[manifest_idx];
+ for (int64_t partition_idx = offset; partition_idx < next_offset;
partition_idx++) {
+ PartitionFieldSummary partition_field_summary;
+ if (!ArrowArrayViewIsNull(contains_null, partition_idx)) {
+ partition_field_summary.contains_null =
+ ArrowArrayViewGetIntUnsafe(contains_null, partition_idx);
+ } else {
+ return InvalidManifestList("contains_null is null at row {}",
partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(contains_nan, partition_idx)) {
+ partition_field_summary.contains_nan =
+ ArrowArrayViewGetIntUnsafe(contains_nan, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(lower_bound_list, partition_idx)) {
+ partition_field_summary.lower_bound =
+ ArrowArrayViewGetInt8Vector(lower_bound_list, partition_idx);
+ }
+ if (!ArrowArrayViewIsNull(upper_bound_list, partition_idx)) {
+ partition_field_summary.upper_bound =
+ ArrowArrayViewGetInt8Vector(upper_bound_list, partition_idx);
+ }
+
+ manifest_file.partitions.emplace_back(partition_field_summary);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestFile>> ParseManifestList(ArrowSchema* schema,
+ ArrowArray* array_in,
+ const Schema&
iceberg_schema) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifestList("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestFile> manifest_files;
+ manifest_files.resize(array_in->length);
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+ switch (field_id) {
+ case ManifestFile::kManifestPathFieldId:
+ PARSE_STRING_FIELD(manifest_files[row_idx].manifest_path,
view_of_column);
+ break;
+ case ManifestFile::kManifestLengthFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].manifest_length,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSpecIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].partition_spec_id,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].content, view_of_column,
+ ManifestContent);
+ break;
+ case ManifestFile::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kMinSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].min_sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kAddedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kExistingFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_files_count,
+ view_of_column, int32_t);
+ break;
+ case ManifestFile::kDeletedFilesCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_files_count,
view_of_column,
+ int32_t);
+ break;
+ case ManifestFile::kAddedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].added_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kExistingRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].existing_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kDeletedRowsCountFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].deleted_rows_count,
view_of_column,
+ int64_t);
+ break;
+ case ManifestFile::kPartitionSummaryFieldId:
+ ICEBERG_RETURN_UNEXPECTED(
+ ParsePartitionFieldSummaryList(view_of_column, manifest_files));
+ break;
+ case ManifestFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_files[row_idx].key_metadata,
view_of_column);
+ break;
+ case ManifestFile::kFirstRowIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_files[row_idx].first_row_id,
view_of_column,
+ int64_t);
+ break;
+ default:
+ return InvalidManifestList("Unsupported field: {} in manifest file.",
field_name);
+ }
+ }
+ return manifest_files;
+}
+
+Status ParseLiteral(ArrowArrayView* view_of_partition, int64_t row_idx,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_partition->storage_type == ArrowType::NANOARROW_TYPE_BOOL) {
+ auto value = ArrowArrayViewGetUIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Boolean(value
!= 0));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_INT32) {
+ auto value = ArrowArrayViewGetIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Int(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_INT64) {
+ auto value = ArrowArrayViewGetIntUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Long(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_FLOAT) {
+ auto value = ArrowArrayViewGetDoubleUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Float(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_DOUBLE) {
+ auto value = ArrowArrayViewGetDoubleUnsafe(view_of_partition, row_idx);
+
manifest_entries[row_idx].data_file->partition.AddValue(Literal::Double(value));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_STRING) {
+ auto value = ArrowArrayViewGetStringUnsafe(view_of_partition, row_idx);
+ manifest_entries[row_idx].data_file->partition.AddValue(
+ Literal::String(std::string(value.data, value.size_bytes)));
+ } else if (view_of_partition->storage_type ==
ArrowType::NANOARROW_TYPE_BINARY) {
+ auto buffer = ArrowArrayViewGetBytesUnsafe(view_of_partition, row_idx);
+ manifest_entries[row_idx].data_file->partition.AddValue(
+ Literal::Binary(std::vector<uint8_t>(buffer.data.as_char,
+ buffer.data.as_char +
buffer.size_bytes)));
+ } else {
+ return InvalidManifest("Unsupported field type: {} in data file
partition.",
+
static_cast<int32_t>(view_of_partition->storage_type));
+ }
+ return {};
+}
+
+Status ParseDataFile(const std::shared_ptr<StructType>& data_file_schema,
+ ArrowArrayView* view_of_column, std::optional<int64_t>&
first_row_id,
+ std::vector<ManifestEntry>& manifest_entries) {
+ if (view_of_column->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifest("DataFile field should be a struct.");
+ }
+ if (view_of_column->n_children != data_file_schema->fields().size()) {
+ return InvalidManifest("DataFile schema size:{} not match with ArrayArray
columns:{}",
+ data_file_schema->fields().size(),
view_of_column->n_children);
+ }
+ for (int64_t col_idx = 0; col_idx < view_of_column->n_children; ++col_idx) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field,
data_file_schema->GetFieldByIndex(col_idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for data file schema",
col_idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto array_view = view_of_column->children[col_idx];
+ auto manifest_entry_count = array_view->length;
+
+ switch (field_id) {
+ case DataFile::kContentFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content,
array_view,
+ DataFile::Content);
+ break;
+ case DataFile::kFilePathFieldId:
+ PARSE_STRING_FIELD(manifest_entries[row_idx].data_file->file_path,
array_view);
+ break;
+ case DataFile::kFileFormatFieldId:
+ for (int64_t row_idx = 0; row_idx < array_view->length; row_idx++) {
+ if (!ArrowArrayViewIsNull(array_view, row_idx)) {
+ auto value = ArrowArrayViewGetStringUnsafe(array_view, row_idx);
+ std::string_view path_str(value.data, value.size_bytes);
+
ICEBERG_ASSIGN_OR_RAISE(manifest_entries[row_idx].data_file->file_format,
+ FileFormatTypeFromString(path_str));
+ }
+ }
+ break;
+ case DataFile::kPartitionFieldId: {
+ if (array_view->storage_type != ArrowType::NANOARROW_TYPE_STRUCT) {
+ return InvalidManifest("Field:{} should be a struct.", field_name);
+ }
+ if (array_view->n_children > 0) {
+ for (int64_t partition_idx = 0; partition_idx <
array_view->n_children;
+ partition_idx++) {
+ auto view_of_partition = array_view->children[partition_idx];
+ for (int64_t row_idx = 0; row_idx < view_of_partition->length;
row_idx++) {
+ if (ArrowArrayViewIsNull(view_of_partition, row_idx)) {
+ break;
+ }
+ ICEBERG_RETURN_UNEXPECTED(
+ ParseLiteral(view_of_partition, row_idx, manifest_entries));
+ }
+ }
+ }
+ } break;
+ case DataFile::kRecordCountFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->record_count,
+ array_view, int64_t);
+ break;
+ case DataFile::kFileSizeFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->file_size_in_bytes,
+ array_view, int64_t);
+ break;
+ case DataFile::kColumnSizesFieldId:
+ // key&value should have the same offset
+ // HACK(xiao.dong) workaround for arrow bug:
+ // ArrowArrayViewListChildOffset can not get the correct offset for map
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->column_sizes,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kNullValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->null_value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kNanValueCountsFieldId:
+
PARSE_INT_LONG_MAP_FIELD(manifest_entries[row_idx].data_file->nan_value_counts,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kLowerBoundsFieldId:
+
PARSE_INT_BINARY_MAP_FIELD(manifest_entries[row_idx].data_file->lower_bounds,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kUpperBoundsFieldId:
+
PARSE_INT_BINARY_MAP_FIELD(manifest_entries[row_idx].data_file->upper_bounds,
+ manifest_entry_count, array_view);
+ break;
+ case DataFile::kKeyMetadataFieldId:
+ PARSE_BINARY_FIELD(manifest_entries[row_idx].data_file->key_metadata,
array_view);
+ break;
+ case DataFile::kSplitOffsetsFieldId:
+ PARSE_INTEGER_VECTOR_FIELD(
+ manifest_entries[manifest_idx].data_file->split_offsets,
manifest_entry_count,
+ array_view, int64_t);
+ break;
+ case DataFile::kEqualityIdsFieldId:
+
PARSE_INTEGER_VECTOR_FIELD(manifest_entries[manifest_idx].data_file->equality_ids,
+ manifest_entry_count, array_view, int32_t);
+ break;
+ case DataFile::kSortOrderIdFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->sort_order_id,
+ array_view, int32_t);
+ break;
+ case DataFile::kFirstRowIdFieldId: {
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->first_row_id,
+ array_view, int64_t);
+ if (first_row_id.has_value()) {
+ std::ranges::for_each(manifest_entries,
[&first_row_id](ManifestEntry& entry) {
+ if (entry.status != ManifestStatus::kDeleted &&
+ !entry.data_file->first_row_id.has_value()) {
+ entry.data_file->first_row_id = first_row_id.value();
+ first_row_id = first_row_id.value() +
entry.data_file->record_count;
+ }
+ });
+ } else {
+ // data file's first_row_id is null when the manifest's first_row_id
is null
+ std::ranges::for_each(manifest_entries, [](ManifestEntry& entry) {
+ entry.data_file->first_row_id = std::nullopt;
+ });
+ }
+ break;
+ }
+ case DataFile::kReferencedDataFileFieldId:
+
PARSE_STRING_FIELD(manifest_entries[row_idx].data_file->referenced_data_file,
+ array_view);
+ break;
+ case DataFile::kContentOffsetFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content_offset,
+ array_view, int64_t);
+ break;
+ case DataFile::kContentSizeFieldId:
+
PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].data_file->content_size_in_bytes,
+ array_view, int64_t);
+ break;
+ default:
+ return InvalidManifest("Unsupported field: {} in data file.",
field_name);
+ }
+ }
+ return {};
+}
+
+Result<std::vector<ManifestEntry>> ParseManifestEntry(
+ ArrowSchema* schema, ArrowArray* array_in, const Schema& iceberg_schema,
+ std::optional<int64_t>& first_row_id) {
+ if (schema->n_children != array_in->n_children) {
+ return InvalidManifest("Columns size not match between schema:{} and
array:{}",
+ schema->n_children, array_in->n_children);
+ }
+ if (iceberg_schema.fields().size() != array_in->n_children) {
+ return InvalidManifest("Columns size not match between schema:{} and
array:{}",
+ iceberg_schema.fields().size(),
array_in->n_children);
+ }
+
+ ArrowError error;
+ ArrowArrayView array_view;
+ auto status = ArrowArrayViewInitFromSchema(&array_view, schema, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ internal::ArrowArrayViewGuard view_guard(&array_view);
+ status = ArrowArrayViewSetArray(&array_view, array_in, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+ status = ArrowArrayViewValidate(&array_view,
NANOARROW_VALIDATION_LEVEL_FULL, &error);
+ ICEBERG_NANOARROW_RETURN_UNEXPECTED_WITH_ERROR(status, error);
+
+ std::vector<ManifestEntry> manifest_entries;
+ manifest_entries.resize(array_in->length);
+ for (int64_t i = 0; i < array_in->length; i++) {
+ manifest_entries[i].data_file = std::make_shared<DataFile>();
+ }
+
+ for (int64_t idx = 0; idx < array_in->n_children; idx++) {
+ ICEBERG_ASSIGN_OR_RAISE(auto field, iceberg_schema.GetFieldByIndex(idx));
+ ICEBERG_CHECK(field.has_value(), "Invalid index {} for manifest entry
schema", idx);
+ auto field_name = field->get().name();
+ auto field_id = field->get().field_id();
+ auto required = !field->get().optional();
+ auto view_of_column = array_view.children[idx];
+
+ switch (field_id) {
+ case ManifestEntry::kStatusFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].status, view_of_column,
+ ManifestStatus);
+ break;
+ case ManifestEntry::kSnapshotIdFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].snapshot_id,
view_of_column,
+ int64_t);
+ break;
+ case ManifestEntry::kSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].sequence_number,
view_of_column,
+ int64_t);
+ break;
+ case ManifestEntry::kFileSequenceNumberFieldId:
+ PARSE_PRIMITIVE_FIELD(manifest_entries[row_idx].file_sequence_number,
+ view_of_column, int64_t);
+ break;
+ case ManifestEntry::kDataFileFieldId: {
+ auto data_file_schema =
+ internal::checked_pointer_cast<StructType>(field->get().type());
+ ICEBERG_RETURN_UNEXPECTED(ParseDataFile(data_file_schema,
view_of_column,
+ first_row_id,
manifest_entries));
+ break;
+ }
+ default:
+ return InvalidManifest("Unsupported field: {} in manifest entry.",
field_name);
+ }
+ }
+ return manifest_entries;
+}
+
+const std::unordered_set<std::string> kStatsColumns = {
+ "value_counts", "null_value_counts", "nan_value_counts",
+ "lower_bounds", "upper_bounds", "record_count"};
+
+bool RequireStatsProjection(const std::shared_ptr<Expression>& row_filter,
+ const std::vector<std::string>& columns) {
+ if (!row_filter || row_filter->op() == Expression::Operation::kTrue) {
+ return false;
+ }
+ if (columns.empty()) {
+ return false;
+ }
+ const std::unordered_set<std::string> selected(columns.cbegin(),
columns.cend());
+ if (std::ranges::all_of(
+ ManifestReader::kAllColumns,
+ [&selected](const std::string& col) { return selected.contains(col);
})) {
+ return false;
+ }
+ if (std::ranges::all_of(kStatsColumns, [&selected](const std::string& col) {
+ return selected.contains(col);
+ })) {
+ return false;
+ }
+ return true;
+}
+
+Result<std::shared_ptr<Schema>> ProjectSchema(std::shared_ptr<Schema> schema,
+ const std::vector<std::string>&
columns,
+ bool case_sensitive) {
+ if (!columns.empty()) {
+ return schema->Select(columns, case_sensitive);
+ }
+ return schema;
+}
+
+} // namespace
+
+std::vector<std::string> ManifestReader::WithStatsColumns(
+ const std::vector<std::string>& columns) {
+ if (std::ranges::all_of(ManifestReader::kAllColumns,
+ [&columns](const std::string& col) {
+ return std::ranges::find(columns, col) !=
columns.cend();
+ })) {
+ return columns;
+ } else {
+ std::vector<std::string> updated_columns{columns};
+ updated_columns.insert(updated_columns.end(), kStatsColumns.begin(),
+ kStatsColumns.end());
+ return updated_columns;
+ }
+}
+
+// ManifestReaderImpl constructor
+ManifestReaderImpl::ManifestReaderImpl(
+ std::string manifest_path, std::optional<int64_t> manifest_length,
+ std::shared_ptr<FileIO> file_io, std::shared_ptr<Schema> schema,
+ std::shared_ptr<PartitionSpec> spec,
+ std::unique_ptr<InheritableMetadata> inheritable_metadata,
+ std::optional<int64_t> first_row_id)
+ : manifest_path_(std::move(manifest_path)),
+ manifest_length_(manifest_length),
+ file_io_(std::move(file_io)),
+ schema_(std::move(schema)),
+ spec_(std::move(spec)),
+ inheritable_metadata_(std::move(inheritable_metadata)),
+ first_row_id_(first_row_id) {}
+
+ManifestReader& ManifestReaderImpl::Select(const std::vector<std::string>&
columns) {
+ columns_ = columns;
+ return *this;
+}
+
+ManifestReader&
ManifestReaderImpl::FilterPartitions(std::shared_ptr<Expression> expr) {
+ part_filter_ = expr ? std::move(expr) : True::Instance();
+ return *this;
+}
+
+ManifestReader& ManifestReaderImpl::FilterPartitions(
+ std::shared_ptr<PartitionSet> partition_set) {
+ partition_set_ = std::move(partition_set);
+ return *this;
+}
+
+ManifestReader& ManifestReaderImpl::FilterRows(std::shared_ptr<Expression>
expr) {
+ row_filter_ = expr ? std::move(expr) : True::Instance();
+ return *this;
+}
+
+ManifestReader& ManifestReaderImpl::CaseSensitive(bool case_sensitive) {
+ case_sensitive_ = case_sensitive;
+ return *this;
+}
+
+bool ManifestReaderImpl::HasPartitionFilter() const {
+ return part_filter_ && part_filter_->op() != Expression::Operation::kTrue;
+}
+
+bool ManifestReaderImpl::HasRowFilter() const {
+ return row_filter_ && row_filter_->op() != Expression::Operation::kTrue;
+}
+
+Result<Evaluator*> ManifestReaderImpl::GetEvaluator() {
+ if (!evaluator_) {
+ auto projection_evaluator = Projections::Inclusive(*spec_, *schema_,
case_sensitive_);
+ ICEBERG_ASSIGN_OR_RAISE(auto projected,
projection_evaluator->Project(row_filter_));
+ ICEBERG_ASSIGN_OR_RAISE(auto final_part_filter,
+ And::Make(std::move(projected), part_filter_));
+
+ ICEBERG_ASSIGN_OR_RAISE(auto partition_type,
spec_->PartitionType(*schema_));
+ auto partition_schema = partition_type->ToSchema();
+ ICEBERG_ASSIGN_OR_RAISE(
+ evaluator_, Evaluator::Make(*partition_schema,
std::move(final_part_filter),
+ case_sensitive_));
+ }
+ return evaluator_.get();
+}
+
+Result<InclusiveMetricsEvaluator*> ManifestReaderImpl::GetMetricsEvaluator() {
+ if (!metrics_evaluator_) {
+ ICEBERG_ASSIGN_OR_RAISE(
+ metrics_evaluator_,
+ InclusiveMetricsEvaluator::Make(row_filter_, *schema_,
case_sensitive_));
+ }
+ return metrics_evaluator_.get();
+}
+
+bool ManifestReaderImpl::InPartitionSet(const DataFile& file) const {
+ if (!partition_set_) {
+ return true;
+ }
+ return partition_set_->contains(file.partition_spec_id, file.partition);
+}
+
+Status ManifestReaderImpl::OpenReader(std::shared_ptr<Schema> projection) {
+ ICEBERG_PRECHECK(projection != nullptr, "Projection schema cannot be null");
+
+ // Ensure required fields are present in the projected data file schema.
+ std::vector<SchemaField> fields(projection->fields().begin(),
+ projection->fields().end());
+ ICEBERG_ASSIGN_OR_RAISE(auto record_count_field,
+
projection->FindFieldById(DataFile::kRecordCount.field_id()));
+ ICEBERG_ASSIGN_OR_RAISE(auto first_row_id_field,
+
projection->FindFieldById(DataFile::kFirstRowId.field_id()));
+ if (!record_count_field.has_value()) {
+ fields.push_back(DataFile::kRecordCount);
+ }
+ if (!first_row_id_field.has_value()) {
+ fields.push_back(DataFile::kFirstRowId);
+ }
+ /// FIXME: this is not supported yet.
+ // fields.push_back(MetadataColumns::kRowPosition);
+ auto data_file_type = std::make_shared<StructType>(std::move(fields));
+
+ // Wrap the projected data file schema into manifest entry schema
+ file_schema_ =
+
ManifestEntry::TypeFromDataFileType(std::move(data_file_type))->ToSchema();
+
+ ReaderOptions options;
+ options.path = manifest_path_;
+ options.io = file_io_;
+ options.projection = file_schema_;
+ if (manifest_length_.has_value()) {
+ options.length = manifest_length_;
+ }
+
+ ICEBERG_ASSIGN_OR_RAISE(file_reader_,
+ ReaderFactoryRegistry::Open(FileFormatType::kAvro,
options));
+
+ return {};
+}
+
+Result<std::vector<ManifestEntry>> ManifestReaderImpl::Entries() {
+ return ReadEntries(/*only_live=*/false);
+}
+
+Result<std::vector<ManifestEntry>> ManifestReaderImpl::LiveEntries() {
+ return ReadEntries(/*only_live=*/true);
+}
+
+Result<std::vector<ManifestEntry>> ManifestReaderImpl::ReadEntries(bool
only_live) {
+ ICEBERG_ASSIGN_OR_RAISE(auto partition_type, spec_->PartitionType(*schema_));
+ auto data_file_schema =
DataFile::Type(std::move(partition_type))->ToSchema();
+
+ std::shared_ptr<Schema> projected_data_file_schema;
+ bool needs_filtering =
+ HasRowFilter() || HasPartitionFilter() || partition_set_ != nullptr;
+ if (needs_filtering) {
+ // ensure stats columns are present for metrics evaluation
+ bool requires_stats_projection = RequireStatsProjection(row_filter_,
columns_);
+ ICEBERG_ASSIGN_OR_RAISE(
+ projected_data_file_schema,
+ ProjectSchema(
+ std::move(data_file_schema),
+ (requires_stats_projection ?
ManifestReader::WithStatsColumns(columns_)
+ : columns_),
+ case_sensitive_));
+ } else {
+ ICEBERG_ASSIGN_OR_RAISE(
+ projected_data_file_schema,
+ ProjectSchema(std::move(data_file_schema), columns_, case_sensitive_));
+ }
+
+ ICEBERG_RETURN_UNEXPECTED(OpenReader(std::move(projected_data_file_schema)));
+ ICEBERG_DCHECK(file_reader_ != nullptr, "File reader should be initialized");
+
+ std::vector<ManifestEntry> manifest_entries;
+ ICEBERG_ASSIGN_OR_RAISE(auto arrow_schema, file_reader_->Schema());
+ internal::ArrowSchemaGuard schema_guard(&arrow_schema);
+
+ // Get evaluators if needed
+ Evaluator* evaluator = nullptr;
+ InclusiveMetricsEvaluator* metrics_evaluator = nullptr;
+ if (HasPartitionFilter() || HasRowFilter()) {
+ ICEBERG_ASSIGN_OR_RAISE(evaluator, GetEvaluator());
+ }
+ if (HasRowFilter()) {
+ ICEBERG_ASSIGN_OR_RAISE(metrics_evaluator, GetMetricsEvaluator());
+ }
+
+ while (true) {
+ ICEBERG_ASSIGN_OR_RAISE(auto result, file_reader_->Next());
+ if (!result.has_value()) {
+ break; // EOF
+ }
+
+ internal::ArrowArrayGuard array_guard(&result.value());
+ ICEBERG_ASSIGN_OR_RAISE(
+ auto entries,
+ ParseManifestEntry(&arrow_schema, &result.value(), *file_schema_,
first_row_id_));
+
+ for (auto& entry : entries) {
+ ICEBERG_RETURN_UNEXPECTED(inheritable_metadata_->Apply(entry));
+
+ if (only_live && !entry.IsAlive()) {
+ continue;
+ }
+
+ if (needs_filtering) {
+ ICEBERG_DCHECK(entry.data_file != nullptr, "Data file cannot be null");
+ if (evaluator) {
+ ICEBERG_ASSIGN_OR_RAISE(bool partition_match,
+
evaluator->Evaluate(entry.data_file->partition));
+ if (!partition_match) {
+ continue;
+ }
+ }
+ if (metrics_evaluator) {
+ ICEBERG_ASSIGN_OR_RAISE(bool metrics_match,
+
metrics_evaluator->Evaluate(*entry.data_file));
+ if (!metrics_match) {
+ continue;
+ }
+ }
+ if (!InPartitionSet(*entry.data_file)) {
+ continue;
+ }
+ }
+
+ manifest_entries.push_back(std::move(entry));
+ }
+ }
+
+ return manifest_entries;
+}
+
+Result<std::vector<ManifestFile>> ManifestListReaderImpl::Files() const {
+ std::vector<ManifestFile> manifest_files;
+ ICEBERG_ASSIGN_OR_RAISE(auto arrow_schema, reader_->Schema());
+ internal::ArrowSchemaGuard schema_guard(&arrow_schema);
+ while (true) {
+ ICEBERG_ASSIGN_OR_RAISE(auto result, reader_->Next());
+ if (result.has_value()) {
+ internal::ArrowArrayGuard array_guard(&result.value());
+ ICEBERG_ASSIGN_OR_RAISE(
+ auto parse_result, ParseManifestList(&arrow_schema, &result.value(),
*schema_));
+ manifest_files.insert(manifest_files.end(),
+ std::make_move_iterator(parse_result.begin()),
+ std::make_move_iterator(parse_result.end()));
+ } else {
+ // eof
+ break;
+ }
Review Comment:
```suggestion
if (!result.has_value()) {
// eof
break;
}
internal::ArrowArrayGuard array_guard(&result.value());
ICEBERG_ASSIGN_OR_RAISE(
auto parse_result, ParseManifestList(&arrow_schema, &result.value(),
*schema_));
manifest_files.insert(manifest_files.end(),
std::make_move_iterator(parse_result.begin()),
std::make_move_iterator(parse_result.end()));
```
I prefer to write this way, with one less tab.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
