An error in bindings of closures

[FrederickRoss]

BPO	14891
Nosy	@amauryfa

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2012-05-23.18:52:29.570>
created_at = <Date 2012-05-23.16:19:36.268>
labels = ['type-bug', 'invalid']
title = 'An error in bindings of closures'
updated_at = <Date 2012-05-23.18:52:29.569>
user = 'https://bugs.python.org/FrederickRoss'

bugs.python.org fields:

activity = <Date 2012-05-23.18:52:29.569>
actor = 'amaury.forgeotdarc'
assignee = 'none'
closed = True
closed_date = <Date 2012-05-23.18:52:29.570>
closer = 'amaury.forgeotdarc'
components = ['None']
creation = <Date 2012-05-23.16:19:36.268>
creator = 'Frederick.Ross'
dependencies = []
files = []
hgrepos = []
issue_num = 14891
keywords = []
message_count = 4.0
messages = ['161432', '161437', '161442', '161449']
nosy_count = 2.0
nosy_names = ['amaury.forgeotdarc', 'Frederick.Ross']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue14891'
versions = ['Python 2.7']

[FrederickRoss]

The following code throws an UnboundLocal error:

def f(x):
    def g():
        x = x + "a"
        return x
    return g()
f("b")

[amauryfa]

This is expected behavior: http://docs.python.org/faq/programming.html#why-am-i-getting-an-unboundlocalerror-when-the-variable-has-a-value

[FrederickRoss]

Assignment in Python creates a new binding. Whether the new binding shadows or replaces an old binding should be irrelevant. This behavior is inconsistent with that. Please fix expectations, and then Python interpreter.

[amauryfa]

What fails here is the evaluation of "x", not the assignment!

You are right concerning the assignment, the outer definition has no effect at all. The very presence of "x = " in the function code turns x into a local variable for the whole function; so 'x + "a"' fails because the local variable has not value yet.

Python, unlike Lisp, defines scopes lexically, not dynamically. There is no "previous" binding, but "local" or "outer" scopes.