The issue came up while trying to get some Sympy code running on CLPython.
class C: exec "a = 3" print locals()
1. Is it guaranteed that class C gets an attribute "a", i.e. that the locals printed include {'a': 3}? 2. It it (also) guaranteed if it were in a function scope?
The complete syntax of exec is: exec CODE in Y, Z where Y, Z are optional.
The documentation of "exec" says "if the optional parts are omitted,the code is executed in the current scope." There are at least two different interpretations:
a. The code is executed in the current class scope, so the assignment must have an effect on the class scope.
b. The scope defaults to the local scope, by which is meant the mapping returned by locals(), and of locals() the documentation says that changes made to it may not influence the interpreter. (The documentation of exec suggests using globals() and locals() as arguments to exec, which seems hint at this interpretation.)
The relevant documentation: exec: http://docs.python.org/reference/simple_stmts.html#grammar-token-exec_stmt locals: http://docs.python.org/library/functions.html#locals
- Willem
Willem Broekema wrote:
The issue came up while trying to get some Sympy code running on CLPython.
class C: exec "a = 3" print locals()
- Is it guaranteed that class C gets an attribute "a", i.e. that the
locals printed include {'a': 3}? 2. It it (also) guaranteed if it were in a function scope?
The complete syntax of exec is: exec CODE in Y, Z where Y, Z are optional.
The documentation of "exec" says "if the optional parts are omitted,the code is executed in the current scope." There are at least two different interpretations:
a. The code is executed in the current class scope, so the assignment must have an effect on the class scope.
b. The scope defaults to the local scope, by which is meant the mapping returned by locals(), and of locals() the documentation says that changes made to it may not influence the interpreter. (The documentation of exec suggests using globals() and locals() as arguments to exec, which seems hint at this interpretation.)
The relevant documentation: exec: http://docs.python.org/reference/simple_stmts.html#grammar-token-exec_stmt locals: http://docs.python.org/library/functions.html#locals
The 3.0 doc for exec() has this warning: "Warning The default locals act as described for function locals() below: modifications to the default locals dictionary should not be attempted. Pass an explicit locals dictionary if you need to see effects of the code on locals after function exec() returns."
This implies interpretation b.
However, is spite of the warning, class locals is a dict and locals() is that dict, so a is available for further use in class code.
So the answer to question 1 for current CPython is yes.
Whether that is guaranteed for all implementations and versions is another story.
Functions are much trickier. The local namespace is not a dict, and modifying the locals() dict does not modify the namespace. The answer to 2. is No, not even now.
def f():
exec('a=3') print(locals()) return a
f()
{'a': 3} Traceback (most recent call last): File "<pyshell#37>", line 1, in <module> f() File "<pyshell#36>", line 4, in f return a NameError: global name 'a' is not defined
But why then is 'a' printed in the second call to locals (the implied one in exec being the first)? It appears that a function or code object can have only only one repeatedly used shadow dict. The 3.0 (and 2.5) doc says "locals() Update and return a dictionary representing the current local symbol table." Note "update"; I had missed that before. To see this...
def g():
a = locals() b = locals() return id(a), id(b), a,b
g()
(20622048, 20622048, {'a': {...}}, {'a': {...}})
Inserting "print(a['a'])" between the locals calls raises KeyError.
Terry Jan Reedy
Hi Terry,
On Wed, Oct 8, 2008 at 9:17 PM, Terry Reedy tjreedy@udel.edu wrote:
Willem Broekema wrote:
The issue came up while trying to get some Sympy code running on CLPython.
class C: exec "a = 3" print locals()
- Is it guaranteed that class C gets an attribute "a", i.e. that the
locals printed include {'a': 3}? 2. It it (also) guaranteed if it were in a function scope?
The complete syntax of exec is: exec CODE in Y, Z where Y, Z are optional.
The documentation of "exec" says "if the optional parts are omitted,the code is executed in the current scope." There are at least two different interpretations:
a. The code is executed in the current class scope, so the assignment must have an effect on the class scope.
b. The scope defaults to the local scope, by which is meant the mapping returned by locals(), and of locals() the documentation says that changes made to it may not influence the interpreter. (The documentation of exec suggests using globals() and locals() as arguments to exec, which seems hint at this interpretation.)
The relevant documentation: exec: http://docs.python.org/reference/simple_stmts.html#grammar-token-exec_stmt locals: http://docs.python.org/library/functions.html#locals
The 3.0 doc for exec() has this warning: "Warning The default locals act as described for function locals() below: modifications to the default locals dictionary should not be attempted. Pass an explicit locals dictionary if you need to see effects of the code on locals after function exec() returns."
This implies interpretation b.
However, is spite of the warning, class locals is a dict and locals() is that dict, so a is available for further use in class code.
So the answer to question 1 for current CPython is yes.
Whether that is guaranteed for all implementations and versions is another story.
Functions are much trickier. The local namespace is not a dict, and modifying the locals() dict does not modify the namespace. The answer to 2. is No, not even now.
def f():
exec('a=3') print(locals()) return a
f()
{'a': 3} Traceback (most recent call last): File "<pyshell#37>", line 1, in <module> f() File "<pyshell#36>", line 4, in f return a NameError: global name 'a' is not defined
But why then is 'a' printed in the second call to locals (the implied one in exec being the first)? It appears that a function or code object can have only only one repeatedly used shadow dict. The 3.0 (and 2.5) doc says "locals() Update and return a dictionary representing the current local symbol table." Note "update"; I had missed that before. To see this...
def g():
a = locals() b = locals() return id(a), id(b), a,b
g()
(20622048, 20622048, {'a': {...}}, {'a': {...}})
Inserting "print(a['a'])" between the locals calls raises KeyError.
Thanks very much for the thorough answer. The reason for Willem's question is this code that we currently have in sympy (see [1] for the whole thread):
class Basic(AssumeMeths): ... for k in AssumeMeths._assume_defined: exec "is_%s = property(make__get_assumption('Basic', '%s'))" % (k,k)
Which works in CPython but fails in CLPython. From your answer it seems to me that this code is not nice and we should not use it and should rather use something like:
class Basic(AssumeMeths): ...
for k in AssumeMeths._assume_defined: setattr(Basic, 'is_%s' % k, property(make__get_assumption('Basic', '%s' % k)))
which should work on all platforms. What do you think?
Ondrej
Well, I don't recall what CLPython is, but I believe it is broken and that code should work -- there are (or used to be) examples of using exec to populate classes in the standard library so while it may look dodgy it really is exected to work...
On Wed, Oct 8, 2008 at 4:40 PM, Ondrej Certik ondrej@certik.cz wrote:
Hi Terry,
On Wed, Oct 8, 2008 at 9:17 PM, Terry Reedy tjreedy@udel.edu wrote:
Willem Broekema wrote:
The issue came up while trying to get some Sympy code running on CLPython.
class C: exec "a = 3" print locals()
- Is it guaranteed that class C gets an attribute "a", i.e. that the
locals printed include {'a': 3}? 2. It it (also) guaranteed if it were in a function scope?
The complete syntax of exec is: exec CODE in Y, Z where Y, Z are optional.
The documentation of "exec" says "if the optional parts are omitted,the code is executed in the current scope." There are at least two different interpretations:
a. The code is executed in the current class scope, so the assignment must have an effect on the class scope.
b. The scope defaults to the local scope, by which is meant the mapping returned by locals(), and of locals() the documentation says that changes made to it may not influence the interpreter. (The documentation of exec suggests using globals() and locals() as arguments to exec, which seems hint at this interpretation.)
The relevant documentation: exec: http://docs.python.org/reference/simple_stmts.html#grammar-token-exec_stmt locals: http://docs.python.org/library/functions.html#locals
The 3.0 doc for exec() has this warning: "Warning The default locals act as described for function locals() below: modifications to the default locals dictionary should not be attempted. Pass an explicit locals dictionary if you need to see effects of the code on locals after function exec() returns."
This implies interpretation b.
However, is spite of the warning, class locals is a dict and locals() is that dict, so a is available for further use in class code.
So the answer to question 1 for current CPython is yes.
Whether that is guaranteed for all implementations and versions is another story.
Functions are much trickier. The local namespace is not a dict, and modifying the locals() dict does not modify the namespace. The answer to 2. is No, not even now.
def f():
exec('a=3') print(locals()) return a
f()
{'a': 3} Traceback (most recent call last): File "<pyshell#37>", line 1, in <module> f() File "<pyshell#36>", line 4, in f return a NameError: global name 'a' is not defined
But why then is 'a' printed in the second call to locals (the implied one in exec being the first)? It appears that a function or code object can have only only one repeatedly used shadow dict. The 3.0 (and 2.5) doc says "locals() Update and return a dictionary representing the current local symbol table." Note "update"; I had missed that before. To see this...
def g():
a = locals() b = locals() return id(a), id(b), a,b
g()
(20622048, 20622048, {'a': {...}}, {'a': {...}})
Inserting "print(a['a'])" between the locals calls raises KeyError.
Thanks very much for the thorough answer. The reason for Willem's question is this code that we currently have in sympy (see [1] for the whole thread):
class Basic(AssumeMeths): ... for k in AssumeMeths._assume_defined: exec "is_%s = property(make__get_assumption('Basic', '%s'))" % (k,k)
Which works in CPython but fails in CLPython. From your answer it seems to me that this code is not nice and we should not use it and should rather use something like:
class Basic(AssumeMeths): ...
for k in AssumeMeths._assume_defined: setattr(Basic, 'is_%s' % k, property(make__get_assumption('Basic', '%s' % k)))
which should work on all platforms. What do you think?
Ondrej
[1] http://code.google.com/p/sympy/issues/detail?id=1134 _______________________________________________ Python-Dev mailing list Python-Dev@python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/guido%40python.org
Guido van Rossum wrote:
Well, I don't recall what CLPython is, but I believe it is broken and that code should work -- there are (or used to be) examples of using exec to populate classes in the standard library so while it may look dodgy it really is exected to work...
I think this behaviour (modifying locals() at class scope) was actually implicitly blessed in PEP 3115, even though that PEP doesn't explicitly state locals() in a class body is required to grant access to the namespace returned by __prepare__().
Perhaps the time is right for the locals() documentation to be more explicit regarding when modifying its contents will affect the current namespace? - yes at module scope (effectively the same as globals()) - yes at class scope - maybe (but typically not) at function scope
Cheers, Nick.
Ondrej Certik wrote:
Which works in CPython but fails in CLPython. From your answer it seems to me that this code is not nice and we should not use it and should rather use something like:
class Basic(AssumeMeths): ...
for k in AssumeMeths._assume_defined: setattr(Basic, 'is_%s' % k, property(make__get_assumption('Basic', '%s' % k)))
which should work on all platforms. What do you think?
That is what setattr is for. Many consider exec a last resort. I think any further discussion should move to the general python list or c.l.p since this is not a develop-core-python issue.