Spark: parameterized queries fail with UNBOUND_SQL_PARAMETER when Iceberg SQL extensions are enabled

[j1wonpark]

Apache Iceberg version

1.11.0 (latest release)

Query engine

Spark

Please describe the bug 🐞

Summary

On Spark 4.1, when the Iceberg SQL extensions are enabled, parameterized queries fail with UNBOUND_SQL_PARAMETER. Any parameterized statement submitted via PySpark's spark.sql(sqlText, args=...) (both classic PySpark and Spark Connect) leaves its positional/named parameter markers unbound. The Iceberg extensions parser doesn't forward the parameter context to Spark; details and fix are in the linked PR.

How to reproduce

Start Spark 4.1 with the Iceberg SQL extensions enabled, then run a parameterized query from PySpark:

from pyspark.sql import SparkSession

spark = (
    SparkSession.builder
    .config(
        "spark.sql.extensions",
        "org.apache.iceberg.spark.extensions.IcebergSparkSessionExtensions",
    )
    .getOrCreate()
)

# No catalog or table is needed — enabling the extension alone triggers the bug,
# since the parser wrapper intercepts all SQL.

# positional marker
spark.sql("SELECT ? AS id", args=[42]).show()

# named marker
spark.sql("SELECT :id AS id", args={"id": 42}).show()

Actual result

Both calls raise AnalysisException. The unresolved plan still contains posparameter(7) / namedparameter(id), confirming the parameter context was dropped during parsing (id#N is a session-dependent attribute id):

# positional — spark.sql("SELECT ? AS id", args=[42])
[UNBOUND_SQL_PARAMETER] Found the unbound parameter: _7. Please, fix `args` and provide a mapping of the parameter to either a SQL literal or collection constructor functions such as `map()`, `array()`, `struct()`. SQLSTATE: 42P02; line 1 pos 7;
'Project [posparameter(7) AS id#N]
+- OneRowRelation

# named — spark.sql("SELECT :id AS id", args={"id": 42})
[UNBOUND_SQL_PARAMETER] Found the unbound parameter: id. Please, fix `args` and provide a mapping of the parameter to either a SQL literal or collection constructor functions such as `map()`, `array()`, `struct()`. SQLSTATE: 42P02; line 1 pos 7;
'Project [namedparameter(id) AS id#N]
+- OneRowRelation

The identical queries bind correctly when the Iceberg extensions are not installed, which pins the wrapper parser as the regression point.

Expected result

Parameter markers are bound and the query returns the supplied arguments (42), exactly as without the Iceberg extensions.

Willingness to contribute

• I can contribute a fix for this bug independently
• I would be willing to contribute a fix for this bug with guidance from the Iceberg community
• I cannot contribute a fix for this bug at this time