Function GetOriginSchema

cpp/src/parquet/arrow/schema.cc:939–982

Get the original Arrow schema, as serialized in the Parquet metadata

// Get the original Arrow schema, as serialized in the Parquet metadata
Status GetOriginSchema(const std::shared_ptr<const KeyValueMetadata>& metadata,
                       std::shared_ptr<const KeyValueMetadata>* clean_metadata,
                       std::shared_ptr<::arrow::Schema>* out) {
  if (metadata == nullptr) {
    *out = nullptr;
    *clean_metadata = nullptr;
    return Status::OK();
  }

  static const std::string kArrowSchemaKey = "ARROW:schema";
  int schema_index = metadata->FindKey(kArrowSchemaKey);
  if (schema_index == -1) {
    *out = nullptr;
    *clean_metadata = metadata;
    return Status::OK();
  }

  // The original Arrow schema was serialized using the store_schema option.
  // We deserialize it here and use it to inform read options such as
  // dictionary-encoded fields.
  ARROW_ASSIGN_OR_RAISE(auto decoded,
                        ::arrow::util::base64_decode(metadata->value(schema_index)));
  auto schema_buf = std::make_shared<Buffer>(decoded);

  ::arrow::ipc::DictionaryMemo dict_memo;
  ::arrow::io::BufferReader input(schema_buf);

  ARROW_ASSIGN_OR_RAISE(*out, ::arrow::ipc::ReadSchema(&input, &dict_memo));

  if (metadata->size() > 1) {
    // Copy the metadata without the schema key
    auto new_metadata = ::arrow::key_value_metadata({}, {});
    new_metadata->reserve(metadata->size() - 1);
    for (int64_t i = 0; i < metadata->size(); ++i) {
      if (i == schema_index) continue;
      new_metadata->Append(metadata->key(i), metadata->value(i));
    }
    *clean_metadata = new_metadata;
  } else {
    // No other keys, let metadata be null
    *clean_metadata = nullptr;
  }
  return Status::OK();
}

// Restore original Arrow field information that was serialized as Parquet metadata
// but that is not necessarily present in the field reconstituted from Parquet data