[ARROW-9603] [C++][Parquet] Write Arrow relies on unspecified behavior for nested types

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.0.0
• Component/s: C++
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/25665
• Assignee: Micah Kornfield
• Reporter: Micah Kornfield
• Created: 30/Jul/20 22:29
• Updated: 11/Jan/23 08:07
• Resolved: 24/Sep/20 09:44

Description

parquet/column_writer.cc WriteArrow implementations at certain points checks null counts/required data and passes through the null bitmap for encoding.  This only works for nested data types if the if the null slot on a parent implies a null slot on the leaf.  This relationship is not required by the specifications.

 

Most paths for creating arrays follow this pattern so it would be esoteric to hit this bug, but we should still fix it.

 

All branches that rely on reading nullness should generate a new null bitmap based on definition levels if the column is nested, and decisions should be based off of that.

Issue Links

• links to: GitHub Pull Request #8219