Are range iterators thread safe?

[Steven D'Aprano]

Using Python 3, are range_iterator objects thread-safe?

I have tried this, and it seems to be safe:
.... print("result =", next(x))
....

threads = [Thread(target=doit, args=(x,)) for i in range(4)]
for t in threads:

.... t.start()
....
result = 0
result = 1
result = 2
result = 3

 

[Stefan Behnel]

Steven D'Aprano, 20.10.2011 10:04:

Using Python 3, are range_iterator objects thread-safe?

I have tried this, and it seems to be safe:
... print("result =", next(x))
...

threads = [Thread(target=doit, args=(x,)) for i in range(4)]
for t in threads:

... t.start()
...
result = 0
result = 1
result = 2
result = 3

The GIL ensures it's thread safe.

Stefan

 

[Stefan Behnel]

Ben Finney, 20.10.2011 13:23:

The GIL applies only to CPython.

and PyPy.

What is the answer for other Python
implementations which don't have a GIL?

That would basically be Jython and IronPython.

Note that none of the three alternative implementations currently supports
Python language version 3. So, the current answer for all existing Python 3
implementations is: the GIL ensures that it's thread safe.

Stefan

 

[Ian Kelly]

Steven D'Aprano, 20.10.2011 10:04:

Using Python 3, are range_iterator objects thread-safe?

I have tried this, and it seems to be safe:

from threading import Thread
x = iter(range(4))
def doit(x):

...     print("result =", next(x))
...

threads = [Thread(target=doit, args=(x,)) for i in range(4)]
for t in threads:

...     t.start()
...
result = 0
result = 1
result = 2
result = 3

The GIL ensures it's thread safe.

Because range_iterator objects are implemented in C and so calling the
__next__ method is only a single bytecode instruction, correct? If
they were implemented in Python the GIL would make no such assurance.