Python scope is too complicated

[Max]

Yeah, I know. It's the price we pay for forsaking variable declarations.
But for java programmers like me, Py's scoping is too complicated.
Please explain what constitutes a block/namespace, and how to refer to
variables outside of it.

 

[Christos TZOTZIOY Georgiou]

Yeah, I know. It's the price we pay for forsaking variable declarations.
But for java programmers like me, Py's scoping is too complicated.
Please explain what constitutes a block/namespace, and how to refer to
variables outside of it.

I presume you read "4.1 Naming and binding" from Python reference manual.

"What constitutes a block" is answered in paragraph 2, for example. Please come
back with confusing / missing information in that chapter giving us a chance to
improve the documentation.

Cheers!

 

[Dan Bishop]

Some may disagree, but for me the easiest way to understand python's
scopes is this:

"""In Python, there are only two scopes. The global and the local.
The global scope is a dictionary while the local, in the case of a
function is extremely fast. There are no other scopes.

This isn't true anymore, now that generator comprehensions have been
added to the language.
17

 

[Ivan Van Laningham]

Hi All--

This isn't true anymore, now that generator comprehensions have been
added to the language.

17

The equivalent in list comprehensions which currently allows the x to
leak out into its containing scope is going away soon. Will that be
another scope? Or are generator and list comprehensions only one scope?

Metta,
Ivan
----------------------------------------------
Ivan Van Laningham
God N Locomotive Works
http://www.andi-holmes.com/
http://www.foretec.com/python/workshops/1998-11/proceedings.html
Army Signal Corps: Cu Chi, Class of '70
Author: Teach Yourself Python in 24 Hours

 

[jfj]

Yeah, I know. It's the price we pay for forsaking variable declarations.
But for java programmers like me, Py's scoping is too complicated.
Please explain what constitutes a block/namespace, and how to refer to
variables outside of it.

Some may disagree, but for me the easiest way to understand python's
scopes is this:

"""In Python, there are only two scopes. The global and the local.
The global scope is a dictionary while the local, in the case of a
function is extremely fast. There are no other scopes. There are
no scopes in the nested statements inside code blocks and there are
no class scopes. As a special case, nested function definitions
appear to be something like a nested scope, but in reallity this is
detected at compile-time and a strange feature called 'cell variables'
is used"""

In order to write to a global variable from a function, we have
to use:
global var

which notifies the compiler that assignment to 'var' does not make
a new local variable, but it modifies the global one (OT: wouldn't
"global.var = 3" be nicer?). On the other hand, if we just want to
read a global variable we don't have to say "global var" because
the compiler sees that there is no assignment to 'var' in the function
code and therefore intelligently concludes that it's about a global
one.

Generally, python sees everything as
exec "code" in global_dictionary, local_dictionary

In this case it uses the opcode LOAD_NAME which looks first in locals
and then in globals (and actually, then in __builtins__)

For functions it uses either LOAD_FAST for locals or LOAD_GLOBAL for
globals.

HTH

jfj

 

[Terry Reedy]

def foo(x):
y= (i for i in x)
return y

From the disassembly it seems that the generator is a code object

What is type(foo([1,2,3])) ?

but 'x' is not a cell variable. WTF?

As I understand it, the object 'x' binds to is immediately used to create
the generator object. The local name is just a dummy that is not part of
the result. I believe that the above is equivalent to

def foo(x):
def _(z):
for i in z: yield i
return _(x)

although I suspect that the implementation builds the generator more
directly.

Terry J. Reedy

 

[jfj]

17

Your example with generator expressions is interesting.
Even more interesting is:

def foo(x):
y= (i for i in x)
return y

From the disassembly it seems that the generator is a code object but
'x' is not a cell variable. WTF? How do I disassemble the generator?
Must look into this....

[ list comprehensions are different because they expand to inlined
bytecode ]

jf

 

[Greg Ewing]

def foo(x):
y= (i for i in x)
return y

From the disassembly it seems that the generator is a code object but
'x' is not a cell variable. WTF?

That's because x is not assigned to anywhere in the
body of foo. The bytecode compiler optimizes away the
creation of a cell in this case, just passing the
value of x as an implicit parameter to the generator.

How do I disassemble the generator?

You'd have to get hold of the code object for it
and disassemble that. There should be a reference to
it in one of the co_consts slots, I think.