atomic section in code

[Ahmad Humayun]

Hi there,

I need to create an atomic section in Python code i.e. there is no
context switch to any other thread during the running of that piece of
code. Would would do the trick?


thanks,

[Fredrik Lundh]

Re: atomic section in code

Ahmad Humayun wrote:

I need to create an atomic section in Python code i.e. there is no
context switch to any other thread during the running of that piece of
code. Would would do the trick?

use a lock, and make sure that anyone that needs access to the shared
state you're manipulating in that section uses the same lock.

lock = threading.Lock( ) # or RLock() etc [1]

with lock:
section

this only works if everyone honors the lock, of course; there's no way
in Python to lock out non-cooperating external threads.

</F>

1) see http://effbot.org/zone/thread-synchronization.htm

[Ahmad Humayun]

Re: atomic section in code

On Sep 5, 1:59 pm, Fredrik Lundh <fred...@python ware.comwrote:

Ahmad Humayun wrote:

I need to create an atomic section in Python code i.e. there is no
context switch to any other thread during the running of that piece of
code. Would would do the trick?

>
use a lock, and make sure that anyone that needs access to the shared
state you're manipulating in that section uses the same lock.
>
     lock = threading.Lock( ) # or RLock() etc [1]
>
     with lock:
         section
>
this only works if everyone honors the lock, of course; there's no way
in Python to lock out non-cooperating external threads.
>
</F>
>
1) seehttp://effbot.org/zone/thread-synchronization .htm

Thats true, but this will ensure mutual exclusion; not atomicity

thanks again,

[Diez B. Roggisch]

Re: atomic section in code

Ahmad Humayun schrieb:

On Sep 5, 1:59 pm, Fredrik Lundh <fred...@python ware.comwrote:

>Ahmad Humayun wrote:

>>I need to create an atomic section in Python code i.e. there is no
>>context switch to any other thread during the running of that piece of
>>code. Would would do the trick?

>use a lock, and make sure that anyone that needs access to the shared
>state you're manipulating in that section uses the same lock.
>>
> lock = threading.Lock( ) # or RLock() etc [1]
>>
> with lock:
> section
>>
>this only works if everyone honors the lock, of course; there's no way
>in Python to lock out non-cooperating external threads.
>>
></F>
>>
>1) seehttp://effbot.org/zone/thread-synchronization .htm

>
Thats true, but this will ensure mutual exclusion; not atomicity

AFAIC that kind of mutual exclusion is what atomicity is about. What
else do you expect to happen?

Diez

[Michele Simionato]

Re: atomic section in code

On Sep 5, 10:59 am, Fredrik Lundh <fred...@python ware.comwrote:

>
1) seehttp://effbot.org/zone/thread-synchronization .htm

The page you link here is WAYS better than the standard documentation
of the threading module.
Generally speaking, the effbot zone contains a lot of improvements
over the standard docs
which are lacking in various areas.
I have always wondered why they are kept separated. Wouldn't be nice
to have the standard
docs integrated with the effbot docs? Are there plans in this sense
and if not, why not?

Michele Simionato

[Fredrik Lundh]

Documentation (was Re: atomic section in code)

Michele Simionato wrote:

The page you link here is WAYS better than the standard documentation
of the threading module.
Generally speaking, the effbot zone contains a lot of improvements
over the standard docs which are lacking in various areas.
I have always wondered why they are kept separated. Wouldn't be nice
to have the standard
docs integrated with the effbot docs? Are there plans in this sense
and if not, why not?

There are tons of great supplementary material out there, on blogs,
personal documentation collections (such as effbot.org), cookbook sites,
etc. I don't think all that material absolutely must be posted to
python.org, but Python users would definitely benefit from improved
cross-linking.

And this is, of course, something I've lobbied for many times, e.g.

http://mail.python.org/pipermail/python-list/2005-May/322224.html

http://mail.python.org/pipermail/python-list/2005-December/355625.html

http://effbot.org/zone/idea-seealso.htm

but I haven't yet figured out how to get something to happen [1]. All
ideas are welcome.

</F>

1) Well, I guess the new Sphinx tool might have gotten some inspiration
by my work in this domain:

http://mail.python.org/pipermail/python-dev/2006-January/059978.html

and Andrew Kuchling did some preliminary work for the old Latex work flow:

http://svn.python.org/view/sandbox/trunk/seealso/

[Fredrik Lundh]

Re: atomic section in code

Diez B. Roggisch wrote:

AFAIC that kind of mutual exclusion is what atomicity is about. What
else do you expect to happen?

sounds like he wants/needs non-cooperative, mandatory locking.

</F>