[PARQUET-2067] [C++] null_count and num_nulls incorrect for repeated columns

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: cpp-6.0.0
• Component/s: parquet-cpp
• Labels: pull-request-available
• Assignee: William Butler
• Reporter: Micah Kornfield
• Created: 16/Jul/21 06:58
• Updated: 07/Oct/21 07:45
• Resolved: 06/Oct/21 08:21

Description

Currently only nulls at the leaf are accounted for in the null count statstics.  For nested lists this is incorrect because null lists have zero elements and don't show up in the leaf.

 

Example from mailing list discussion

 

[[0, 1], None, [2, None, 3]]

 

should have a null count of 2 (it currently reports as 1).

Issue Links

• links to: GitHub Pull Request #11281