wgtmac commented on code in PR #36774:
URL: https://github.com/apache/arrow/pull/36774#discussion_r1268882255
##########
cpp/src/parquet/file_reader.cc:
##########
@@ -366,13 +366,19 @@ class SerializedFile : public ParquetFileReader::Contents
{
std::make_shared<::arrow::io::internal::ReadRangeCache>(source_, ctx,
options);
std::vector<::arrow::io::ReadRange> ranges;
prebuffered_column_chunks_.clear();
+ int num_cols = file_metadata_->num_columns();
+ // a bitmap for buffered columns.
+ std::shared_ptr<Buffer> buffer_columns;
+ if (!row_groups.empty()) {
Review Comment:
Using `buffer=nullptr` as a placeholder for marking all columns are
pre-buffered seems ok. Then we should not add a row group to the map if the
none of columns in the row group are pre-buffered.
Besides, if only a single column is not pre-buffered in the row group, the
overhead is still high to set bit for other columns.
--
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]
