Yicong Huang created SPARK-55507:
------------------------------------
Summary: is_geometry and is_geography crash with TypeError when
struct field has no metadata
Key: SPARK-55507
URL: https://issues.apache.org/jira/browse/SPARK-55507
Project: Spark
Issue Type: Bug
Components: PySpark
Affects Versions: 4.2.0
Reporter: Yicong Huang
{{is_geometry()}} and {{is_geography()}} in
{{python/pyspark/sql/pandas/types.py}} crash when a PyArrow struct contains a
field named {{wkb}} without metadata.
PyArrow fields have {{metadata=None}} by default (not an empty dict). The
current code does:
{code:python}
field.name == "wkb"
and b"geometry" in field.metadata # TypeError if metadata is None
and field.metadata[b"geometry"] == b"true"
{code}
This raises {{TypeError: argument of type 'NoneType' is not iterable}}.
These functions are called from {{from_arrow_type()}} for every struct type,
which is used in {{toPandas()}}, {{pandas_udf}}, and Spark Connect schema
parsing. Any user with a struct containing a field named {{wkb}} without
geometry metadata will hit this crash.
The fix is to add a {{field.metadata is not None}} check before the {{in}}
operator:
{code:python}
field.name == "wkb"
and field.metadata is not None
and b"geometry" in field.metadata
and field.metadata[b"geometry"] == b"true"
{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
