apacheGH-43427: Parquet Deprecate ColumnChunk::file_offset field

mapleFU · Jul 26, 2024 · ddd5f16 · ddd5f16
1 parent f54b120
commit ddd5f16
Showing 1 changed file with 4 additions and 3 deletions.
diff --git a/cpp/src/parquet/metadata.cc b/cpp/src/parquet/metadata.cc
@@ -1536,10 +1536,11 @@ class ColumnChunkMetaDataBuilder::ColumnChunkMetaDataBuilderImpl {
               const std::shared_ptr<Encryptor>& encryptor) {
     if (dictionary_page_offset > 0) {
       column_chunk_->meta_data.__set_dictionary_page_offset(dictionary_page_offset);
-      column_chunk_->__set_file_offset(dictionary_page_offset + compressed_size);
-    } else {
-      column_chunk_->__set_file_offset(data_page_offset + compressed_size);
     }
+    // https://github.com/apache/parquet-format/pull/440
+    // The `file_offset` field is deprecated and should be set to 0 for writer
+    // if the column chunk has not been written outsidethe footer.
+    column_chunk_->__set_file_offset(0);
     column_chunk_->__isset.meta_data = true;
     column_chunk_->meta_data.__set_num_values(num_values);
     if (index_page_offset >= 0) {