[ARROW-13982] [C++] Async scanner stalls if a fragment generates no batches

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 5.0.0
• Fix Version/s: 6.0.0
• Component/s: C++
• Labels: pull-request-available query-engine
• Environment: ubuntu 18.04 LTS
• External issue URL: https://github.com/apache/arrow/issues/29583

Language

C++

• Assignee: David Li
• Reporter: Huxley Hu
• Created: 13/Sep/21 11:48
• Updated: 11/Jan/23 08:36
• Resolved: 15/Sep/21 17:09

Description

Reading parquet files using dataset scanner may stall due to a never-finished future. 

To reproduce this case, one needs two parquet files and sets the filter expression to something that could filter one file completely.  After that, calling `AsyncScanner::ToRecordBatchReader` and read data continually. 

I also have dug this bug a little. It's caused by the `MakeEmptyGenerator<std::shared_ptr<RecordBatch>>` when filtered row groups is empty, which's ignored by `FragmentToBatches` and causes SequencingGenerator to stall.

A quick fix is to return a record batch with 0 rows instead of returning a nullptr there.

Issue Links

• links to: GitHub Pull Request #11145