[ARROW-18229] [C++][Python] RecordBatchReader can be created with a 'dict' schema which then crashes on use

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 10.0.0
• Fix Version/s: 11.0.0
• Component/s: C++, Python
• Labels: pull-request-available triaged
• External issue URL: https://github.com/apache/arrow/issues/33414
• Assignee: Joris Van den Bossche
• Reporter: David Li
• Created: 02/Nov/22 15:11
• Updated: 11/Jan/23 11:58
• Resolved: 09/Nov/22 16:27

Description

Presumably we should disallow this or convert it to a schema?

https://github.com/duckdb/duckdb/issues/5143

>>> import pyarrow as pa
>>> pa.__version__
'10.0.0'
>>> reader = pa.RecordBatchReader.from_batches({"a": pa.int8()}, [])
>>> reader.schema
fish: Job 1, 'python3' terminated by signal SIGSEGV (Address boundary error)

(gdb) bt
#0  0x00007ffff4247580 in arrow::Schema::num_fields() const ()
   from /home/lidavidm/miniconda3/lib/python3.9/site-packages/pyarrow/libarrow.so.1000
#1  0x00007ffff42b93f7 in arrow::(anonymous namespace)::SchemaPrinter::Print()
    ()
   from /home/lidavidm/miniconda3/lib/python3.9/site-packages/pyarrow/libarrow.so.1000
#2  0x00007ffff42b98a7 in arrow::PrettyPrint(arrow::Schema const&, arrow::PrettyPrintOptions const&, std::string*) ()
   from /home/lidavidm/miniconda3/lib/python3.9/site-packages/pyarrow/libarrow.so.1000
#3  0x00007ffff64f814b in __pyx_pw_7pyarrow_3lib_6Schema_52to_string(_object*, _object*, _object*) ()

Issue Links

• links to: GitHub Pull Request #14583