[ARROW-8128] [C#] NestedType children serialized on wrong length

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.17.0
• Component/s: C#
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/17088
• Assignee: Takashi Hashida
• Reporter: Takashi Hashida
• Created: 16/Mar/20 05:01
• Updated: 11/Jan/23 07:58
• Resolved: 16/Mar/20 07:01

Description

Each node of NestedType children is serialized on a previous node Length and NullCount.

This causes wrong data access at ListArray.GetValueOffset and so on.

 

https://github.com/apache/arrow/blob/master/csharp/src/Apache.Arrow/Ipc/ArrowReaderImplementation.cs#L219

 

Flatbuf.FieldNode childFieldNode = recordBatchEnumerator.CurrentNode;
recordBatchEnumerator.MoveNextNode();

At this lines, MoveNextNode should be executed before assigning CurrentNode.

this can be reproduced by changing TestData.ArrayCreator.Visit(ListType type) like below and execute ArrowFileReaderTests.

public void Visit(ListType type)
 {
 var builder = new ListArray.Builder(type.ValueField).Reserve(Length);

//Todo : Support various types
 var valueBuilder = (Int64Array.Builder)builder.ValueBuilder.Reserve(Length);

for (var i = 0; i < Length; i++)
 {
 builder.Append();
 valueBuilder.Append(i);
 }

//Add a value to check if Values length can exceed ListArray length
 valueBuilder.Append(0);

Array = builder.Build();
 }

Issue Links

• links to: GitHub Pull Request #6628