[ORC-1160] [C++] RowReaderImpl::seekToRow() can't seek within a selected row group

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.7.0, 1.7.1, 1.7.2, 1.7.3, 1.7.4
• Fix Version/s: 1.7.5
• Component/s: C++
• Assignee: Quanlong Huang
• Reporter: Quanlong Huang
• Created: 27/Apr/22 07:29
• Updated: 16/Jun/22 19:15
• Resolved: 28/Apr/22 20:12

Description

RowReaderImpl::seekToRow() depends on startNextStripe() to seek to the start of the next selected row group when PPD is enabled. However, it doesn't seek inside the selected row group. In the case when the target rowNumber is not the first row of the row group, the seeked position is incorrect.

Related codes:
https://github.com/apache/orc/blob/v1.7.4/c%2B%2B/src/Reader.cc#L351
https://github.com/apache/orc/blob/v1.7.4/c%2B%2B/src/Reader.cc#L1045

We should either do the remaining seek in seekToRow() or startNextStripe().

 

Issue Links

• blocks: ORC-1159 [C++] Crash when the last stripe is skipped
• links to: GitHub Pull Request #1100
• links to: GitHub Pull Request #1102