UDF Memory Leak

[paultiq]

What happens?

Successive calls to UDFs leak memory, even across connections.

Running the below reproducer will lead to an OOM.

To Reproduce

import duckdb
import gc
import resource

def get_rss_mb():
    return resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / 1024

def udf_process_text(x):
    return f'processed_data_{x}' * 1024

def mre_udf_leak():
    con = duckdb.connect()
    initial = get_rss_mb()
    print(f"RSS: {initial:.0f} MB")

    i=0
    while i:=i+1:
        with duckdb.connect() as con: 
            con.create_function("udf_process_text", udf_process_text, ["BIGINT"], "VARCHAR")

            rows = 1000
            con.execute(f"SELECT udf_process_text(range) FROM range({rows})")

            rss = get_rss_mb()

            if i%10 == 0: 
                print(f"{i:>5} {rss:>10.0f} MB")
                gc.collect()  # - here just to demonstrate that it doesn't matter. 

if __name__ == "__main__":
    mre_udf_leak()

Output:

RSS: 52 MB
10 341 MB
20 520 MB
...
4140 62503 MB
4150 62503 MB
Killed

OS:

Linux/Ubuntu

DuckDB Package Version:

1.4.3

Python Version:

3.13.9

Full Name:

Paul T

Affiliation:

Iqmo

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?

Yes

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