execution time with global variables

[sarmin kho]

Hi Pythoners,

i have been using a lot of global variables in the python script i am working on. the global variables are shared and used by all various 'definitions' :

def name ():
global all globs...
.....

my question is: 'is it safe using global variables in term of its execution time?

i cant think of any other ways not to use the globs because some variables simply have to be shared among those definitions..

many thanks..
sarmin

 

[Duncan Booth]

i have been using a lot of global variables in the python script i am
working on. the global variables are shared and used by all various
'definitions' :

def name ():
global all globs...
.....

my question is: 'is it safe using global variables in term of its
execution time?

Is the program running fast enough? If so, the answer is 'yes'. If not then
using local variables might provide some speedup, but usually changing your
algorithm is a better way to optimise.

i cant think of any other ways not to use the globs because some
variables simply have to be shared among those definitions..

The usual solution is to define a class and then all your globals simply
become attributes within the class. Try doing it that way and see if you
like it.

 

[Peter Otten]

Hi Pythoners,

i have been using a lot of global variables in the python script i am
working on. the global variables are shared and used by all various
'definitions' :

def name ():
global all globs...
.....

my question is: 'is it safe using global variables in term of its
execution time?

Here is may attempt to measure it:

<time_global.py>
def f():
x = 1
x
x
x
x
x
x
x
x
x
x


def g():
global x
x = 1
x
x
x
x
x
x
x
x
x
x
</time_global.py>

$ timeit.py -s"from time_global import f" "f()"
1000000 loops, best of 3: 0.972 usec per loop
$ timeit.py -s"from time_global import g" "g()"
1000000 loops, best of 3: 1.13 usec per loop

If the methodology isn't flawed, the effect is hardly noticable, as a real
function will have other operations that take much longer than a variable
lookup.

i cant think of any other ways not to use the globs because some variables
simply have to be shared among those definitions..

For a one-off script I see no problem, but otherwise I'd suggest you think
again and stay away from global variables as much as possible; the cost in
maintenance time may by far eclipse the extra execution time.
As a dumb "better than nothing" recipe I suggest you bundle the globals in a
Bunch-like class and pass that explicitly to your functions, or convert the
functions into methods and keep the "global" state as instance attributes.

Peter

 

[Peter Otten]

As a dumb "better than nothing" recipe I suggest you bundle the globals in
a Bunch-like class and pass that explicitly to your functions, or convert
the functions into methods and keep the "global" state as instance
attributes.

Of course this is even slower:

<time_global.py continued>
class C:
pass

def h(c):
c.x = 1
c.x
c.x
c.x
c.x
c.x
c.x
c.x
c.x
c.x
c.x
</time_global.py>

$ timeit.py -s"from time_global import h, C" -s"c=C()" "h(c)"
1000000 loops, best of 3: 1.94 usec per loop

Peter