futurize potentially changes meaning of map(...)

[AndersBlomdell]

Before futurizing, the following:

#!/usr/bin/python
a='right'
print a 
v=['a', 'b', 'wrong']
m=map(lambda a: a+a, v)
print a

gives the following output:

right
right

while the futurized code:

#!/usr/bin/python
from __future__ import print_function
a='right'
print(a)
v=['a', 'b', 'wrong']
m=[a+a for a in v]
print(a)

gives the following output:

right
wrong

See http://python-history.blogspot.se/2010/06/from-list-comprehensions-to-generator.html for further explanation.

[edschofield]

Thanks for filing this issue, Anders! Interesting ...

Would you suggest that futurize work around this problem by changing all map(f, seq) calls into list((f(item) for item in seq)) instead of [f(item) for item in seq]? This would work, but it seems a little drastic ... The other option would be a note in the docs warning about a potential change in the semantics due to Py2's leakage of list-comprehension temporary variables ...

[AndersBlomdell]

I personally would go the drastic way (in fact I have started modifying all my python2 code accordingly), I do not read documentation close enough to catch this :-(, especially not if not aware of the problem beforehand.