[ARROW-18405] [Ruby] Raw table converter rebuilds chunked arrays

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 10.0.0
• Fix Version/s: 11.0.0
• Component/s: Ruby
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/33568
• Assignee: Kouhei Sutou
• Reporter: Sten Larsson
• Created: 25/Nov/22 15:08
• Updated: 11/Jan/23 12:00
• Resolved: 29/Nov/22 06:54

Description

Consider the following Ruby script:

require 'arrow'
data = Arrow::ChunkedArray.new([Arrow::Int64Array.new([1])])
table = Arrow::Table.new('column' => data)
puts table['column'].data_type

This prints "int64" with red-arrow 9.0.0 and "uint8" in 10.0.0.

From my understanding it is due to this commit: https://github.com/apache/arrow/commit/913d9c0a9a1a4398ed5f56d713d586770b4f702c#diff-f7f19bbc3945ea30ba06d851705f2d58f7666507bb101c4e151014ca398bd635R42

The old version would not call ArrayBuilder.build on a ChunkedArray, but the new version does. This is a problem for us, because we need the column to stay int64.

A workaround is to specify a schema and list of arrays instead to bypass the raw table converter:

table = Arrow::Table.new([{name: 'column', type: 'int64'}], [data])

Issue Links

• links to: GitHub Pull Request #14738