Is global really global.

[Steve Holden]

Steve Juranich wrote:
> 
> So far, I really love Python.  However, the most confusing part of it for me
> is definitely the whole namespace business.
> 
Well, you only have three to worry about at any one time ...

> I'm writing a suite of modules that are supposed to work in concert.  In the
> "main" module, I read in a file that specifies all of the run-time paramters
> that the program is supposed to use.  Here is where I define the "options"
> structure:
> 
>     global options
>     options = read_options(parfile)
> 
> Where the read_options function works a little magic on the file and returns
> the table (one of my own classes) that contains all of the information that
> the user specified in a parameter file (or "parfile").  This all happens in
> a module called "Peaks.py".
> 
So options is global _to module main_.

> Later on, I try and test some values in "options" via:
> 
>         if options.no_coda:
>             # Eventually, here's where I'd remove the coda, but for
>             # now, just...
>             raise NotImplementedError, \
>                   "I can't deal with the no_coda option yet."
> 
> Please ignore the fact that I don't have the option implemented yet. <0.5 wink>

Sure.  Some of my best code is the unimplemented stuff.

> The test above happens in a module called "Locator.py".  But the above test
> when I try to run it, gives me:
> 
> condor (peaks)$ Peaks.py parfile.peaks first_list First
> Traceback (most recent call last):
>   File "./Peaks.py", line 75, in ?
>     main_loop()
>   File "./Peaks.py", line 66, in main_loop
>     peaks = Locator.f0_peak_locator(frames)
>   File "./Locator.py", line 23, in f0_peak_locator
>     if options.no_coda:
> NameError: options
> 
That would seem logical, captain.  Presumably Locator.py imports the
Peaks module?

> Why is this happenning even though I specified "options" as a global?  This
> causes me to suspect that "global" means something different in Python than
> it does in C (which isn't a bad idea).
> 
Correct.  Global means "not local".  It is typically used to indicate in the
code of a function that a reference should be taken to refer to a variable
in the namespace of the surrounding module rather than the namespace of the
function.

By default Python assumes that names in a function are local if they are
assigned to.  Here's a demonstration:

Python 1.5.2 (#0, Apr 13 1999, 10:51:12) [MSC 32 bit (Intel)] on win32
Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam
IDLE 0.5 -- press F1 for help
>>> def f(a):
	print a
	print x
	x = 3
	print x

	
>>> def g(a):
	print a
	print x

	
>>> x = 39
>>> g(12)
12
39
>>> f(12)
12
Traceback (innermost last):
  File "<pyshell#12>", line 1, in ?
    f(12)
  File "<pyshell#5>", line 3, in f
    print x
NameError: x
>>> 

Note that the reference to x in the second print statement of the f function
is assumed to be local (because of the assigment which follows), while the
reference in teh g function is assumed global, because there is never an
assignment to it.

> Am I misunderstanding the meaning of "global", or is something else afoot?
> How do I work around this?  I know using globals isn't good anyway, but I'm
> to the point that I _really_ need to show my professor that I've been doing
> more than sitting on my thumb for the past couple of weeks (read: I need to
> get something done _quickly_).
> 

You seem to have understood the essentials pretty well.

The easiest way would be refer to options as a qualifier to the module name
in which it's defined.  i.e.:

         if Peaks.options.no_coda:
             # Eventually, here's where I'd remove the coda, but for
             # now, just...
             raise NotImplementedError, \
                   "I can't deal with the no_coda option yet."

There are other ways, all equivalent but avoiding a little verbosity and a
few dictionary lookups.  For example, after you've called the read_options
function in your Peaks module, you could just assign the value of the
result to a local variabel called options:

	options = Peaks.options

After that, the rest of your code should work as it is.


> Thanks for the help.
> 
A pleasure.

> trying-to-adjust-to-life-in-grad-school-ly y'rs,
> 
> ----------------------------------------------------------------------
> Stephen W. Juranich                         sjuranic at ee.washington.edu
> Electrical Engineering         http://students.washington.edu/sjuranic
> University of Washington             http://rcs.ee.washington.edu/ssli

trying-to-forget-supervising-grad-students-ly y'rs - steve
-- 
Helping people meet their information needs with training and technology.
703 967 0887      sholden at bellatlantic.net      http://www.holdenweb.com/