[ARROW-14156] [C++] StructArray::Flatten is incorrect in some cases

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 5.0.0
• Fix Version/s: 6.0.0
• Component/s: C++
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/29744
• Assignee: David Li
• Reporter: Micah Kornfield
• Created: 28/Sep/21 16:56
• Updated: 11/Jan/23 08:38
• Resolved: 07/Oct/21 07:49

Description

When trying to flatten a struct that has children that were sliced we see incorrect results.

 

Bar.java

import pyarrow as pa

a = py.array([1,2,3])

sliceds = a.slice(1)

composed_struct = pa.StructArray.from_buffers(pa.struct([pa.field("a", sliceds.type)]), len(sliceds), [pa.array([True, False]).buffers()[1]], children=[sliceds])

>>> composed_struct
<pyarrow.lib.StructArray object at 0x7f5502a74880>
-- is_valid:
  [
    true,
    false
  ]

-- child 0 type: int64
  [
    2,
    3
  ]

>>> composed_struct.flatten()
[<pyarrow.lib.Int64Array object at 0x7f5502a74460>
[
  null,
  null
]]

 
I believe the problems is here the copy does not account for child array offset.

Issue Links

• links to: GitHub Pull Request #11344