This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

classification
Title: odd exec() behavior or documentation
Type: behavior Stage:
Components: Documentation Versions: Python 3.1
process
Status: closed Resolution: works for me
Dependencies: Superseder:
Assigned To: georg.brandl Nosy List: georg.brandl, ncoghlan, pmawhorter
Priority: normal Keywords:

Created on 2009-11-08 06:39 by pmawhorter, last changed 2022-04-11 14:56 by admin. This issue is now closed.

Files
File name Uploaded Description Edit
test.py pmawhorter, 2009-11-08 06:39 Demonstration script
Messages (3)
msg95036 - (view) Author: Peter Mawhorter (pmawhorter) Date: 2009-11-08 06:39
The behavior of the built-in exec() function differs in Python3.1 from
the behavior in Python 2.6 when only a single argument is passed.
Additionally, the documentation for the function does not suggest the
behavior that it has. In Python2.6, an exec statement could both access
and change the values of local variables. In Python3.1, it can read
values from the local scope, but can't change them. Run the attached
script to see examples of this behavior. Note especially that in the
first function, the exec'd code changes the value of a variable called
'value'. However, the change isn't visible after the call to exec(),
because the value changed was a local variable and presumably, the
update is dropped after the exec() finishes. To me, this conflicts with
the statement in the documentation that "In all cases, if the optional
parts are omitted, the code is executed in the current scope." At the
very least, a sentence or two should be added to the documentation
explaining that exec() can't affect the value of local variables. In
addition, it's strange that, for example, in a function declaring
"global value", an exec() statement that modifies 'value' doesn't affect
the value of the global variable. In this sense, the code being exec'd
doesn't seem to be running in a perfect copy of the function's scope.

If it were up to me, I'd allow code being executed by an exec()
statement to modify the value of local variables, because that keeps
with the paradigm of the exec'd code being simply dynamic code that
behaves as if it were written in-line. It also means that programmers
who need to change things with an exec statement aren't limited to
pushing their changes through global variables, which can lead to
conflicts with name sharing (for example, what if my exec() statement is
in a recursive function?). Of course, implementing this may be
difficult... so it may be better just to add to the documentation to
explain this behavior of exec().
msg95214 - (view) Author: Nick Coghlan (ncoghlan) * (Python committer) Date: 2009-11-13 22:38
This is a largely inevitable consequence of the conversion from a
statement to a function - the interpreter eval loop for the exec
statement used to do some fancy footwork to make locals() modifications
work, but with exec becoming just another builtin function that special
casing is gone.

Accordingly, in 3.x, exec() is dependent on the implementation defined
behaviour of locals(), just like all other code. (The upside to this is
that it means the compiler always knows the identity of all local
variables).

I have no idea how well this change has been documented though (to be
honest, I'd forgotten that exec had been converted to a function for 3.x)
msg95353 - (view) Author: Georg Brandl (georg.brandl) * (Python committer) Date: 2009-11-16 17:19
The docs already contain the following note:

   .. note::

      The default *locals* act as described for function :func:`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 :func:`exec` returns.
History
Date User Action Args
2022-04-11 14:56:54adminsetgithub: 51535
2009-11-16 17:19:37georg.brandlsetstatus: open -> closed
resolution: works for me
messages: + msg95353
2009-11-13 22:38:20ncoghlansetnosy: + ncoghlan
messages: + msg95214
components: - Interpreter Core
2009-11-08 06:39:04pmawhortercreate