Standard output is mixed up when submitting Jython jobs in non forked mode.

[mbenguig]

Standard output is mixed up if Python scripts are evaluated (Jython) in parallel within one single JVM (non forked mode). In non forked mode, reproduce it by submitting twice a single-Jython-task-based workflow with the following script:

import time
for x in range(1, 5):
print x
time.sleep(1)

The job1 output:

1
2
3

The job2 output:

1
4
2
3
4

Fabien Viale Hier à 20:56
I have some bad news, this seems to be an identified and unresolved issue in jython.

https://bugs.jython.org/issue2513

Describes exactly the same issue, with no progress.
(Need to wait for an update of jython)

[Fahim302]

Hi

[sudishtakumar]

In non-forked parallel execution using Jython within a single JVM, standard output (sys.stdout) is shared among all scripts, because Jython uses a single JVM-level System.out object. As a result, outputs from multiple concurrently running scripts can interleave or get mixed up.

This happens because Jython does not provide thread-local isolation of standard output — all threads share the same stdout by default.

Efficient solution:
A reliable and efficient way to prevent output mixing is to redirect sys.stdout and sys.stderr to a thread-local (or task-local) buffer at the beginning of each task or script. Each thread gets its own output stream, ensuring isolation.

// code
import sys
import threading
import io

Create a thread-local stdout

class ThreadLocalOutput(io.StringIO):
def write(self, msg):
sys.stdout.write(f"[{threading.current_thread().name}] {msg}")

Redirect stdout to thread-local output

sys.stdout = ThreadLocalOutput()

Normal script

import time
for x in range(1, 5):
print(x)
time.sleep(1)

[sudishtakumar]

Alternatively, if modifying stdout redirection is not feasible, another solution is to synchronize print operations by using a global print lock, which serializes access to standard output, thus avoiding interleaving:

//code here
import threading
print_lock = threading.Lock()

def safe_print(*args, **kwargs):
with print_lock:
print(*args, **kwargs)

Usage

for x in range(1, 5):
safe_print(x)

[sudishtakumar]

**Key Points to Remember
Root cause: sys.stdout is shared among threads in Jython.

Best efficient solution: Redirect sys.stdout per thread (thread-local output).

Alternate solution: Synchronize print calls with a global lock.

Why it's unresolved: Jython lacks built-in thread-local stdout management.**