New submission from Garrett Cooper:
I was a bit surprised when I ran into this issue when porting some nose tests
from Windows to Linux:
#!/usr/bin/env python
with open('/etc/services') as fd:
lines = fd.readlines()
lines.append('')
SERVICES = [line.split()[0] for line in lines
if (line and not line.startswith('#'))]
$ python list_comprehension.py
Traceback (most recent call last):
File list_comprehension.py, line 5, in module
if (line and not line.startswith('#'))]
IndexError: list index out of range
$ python3.2 list_comprehension.py
Traceback (most recent call last):
File list_comprehension.py, line 4, in module
SERVICES = [line.split()[0] for line in lines
File list_comprehension.py, line 5, in listcomp
if (line and not line.startswith('#'))]
IndexError: list index out of range
$ python -V
Python 2.7.5
$ python3.2 -V
Python 3.2.5
This is occurring of course because the .split() is being done on an empty line.
The docs don't note (at least in the list comprehension section [*]) that
if-statements are evaluated after the value is generated for the current index
in the loop. This seems very backwards because generating a value could in fact
be very expensive, whereas determining whether or not a precondition has been
met should be less expensive.
What could/should be done is one of two things: 1. evaluation order should be
clearly spelled out in the docs, or 2. the list comprehension handling code
should be changed to support evaluating the conditional statements before
calculating a result. Otherwise discouraging use of [map+]filter (at least
pylint does that) seems incredibly unwise as I can get the functionality I want
in a single line as opposed to an unrolled loop.
[*] http://docs.python.org/2/tutorial/datastructures.html#list-comprehensions
--
messages: 203409
nosy: yaneurabeya
priority: normal
severity: normal
status: open
title: List comprehension conditional evaluation order seems backwards
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue19657
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
