Repository navigation
duckdb Python Extension Analysis Report #408
Description
What happens?
I ran cext-review-toolkit on duckdb-python and got a report with some issues:
The primary risks are: (1) Python API calls without GIL in the parallel numpy scan path (data corruption/crash), (2)
PyTuple_Packreference leak inmap.cpp(~11 KB/call), (3) uncheckedPyNumber_RshiftNULL in 128-bit integer conversion (segfault), (4)PyErr_PrintExclobbering exceptions in map and filesystem, and (5) UDF exception handling that optionally swallowsMemoryError/KeyboardInterrupt.
The full report is available at https://gist.github.com/devdanzin/5c7600153c8e76afee3ef1a2d516ad70
To Reproduce
Some reproducers are available in the Reproducers Appencix.
OS:
Linux x86_64
DuckDB Package Version:
1.5.1
Python Version:
3.14.3+ debug build
Full Name:
Daniel Silva Diniz
Affiliation:
None
What is the latest build you tested with? If possible, we recommend testing with the latest nightly build.
I have tested with a stable release
Did you include all relevant data sets for reproducing the issue?
Not applicable - the reproduction does not require a data set
Did you include all code required to reproduce the issue?
- Yes, I have
Did you include all relevant configuration to reproduce the issue?
- Yes, I have
Activity
Fix python error propagation (duckdb#679)
evertlammerts commentedon Oct 19, 2026
Thanks. On the 2.0 branch the int conversion no longer calls a subclass's
__rshift__,return_nullforwardsMemoryErrorandKeyboardInterrupt,the map.cpp tuple leak is gone and the pandas string scan holds the GIL. I verified the first two against 1.5.6 and the current dev build. One finding is still open:rel.mapprints the traceback and raises a genericInvalidInputExceptioninstead of chaining your exception. I will track that one separately.