operator.attrgetter is slower than a lambda

[florentx]

BPO	17194
Nosy	@florentx, @serhiy-storchaka

^{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 2013-02-12.22:30:54.343>
created_at = <Date 2013-02-12.16:43:15.354>
labels = ['invalid', 'performance']
title = 'operator.attrgetter is slower than a lambda'
updated_at = <Date 2013-02-12.22:30:54.341>
user = 'https://github.com/florentx'

bugs.python.org fields:

activity = <Date 2013-02-12.22:30:54.341>
actor = 'flox'
assignee = 'none'
closed = True
closed_date = <Date 2013-02-12.22:30:54.343>
closer = 'flox'
components = []
creation = <Date 2013-02-12.16:43:15.354>
creator = 'flox'
dependencies = []
files = []
hgrepos = []
issue_num = 17194
keywords = []
message_count = 3.0
messages = ['181967', '181978', '181979']
nosy_count = 2.0
nosy_names = ['flox', 'serhiy.storchaka']
pr_nums = []
priority = 'low'
resolution = 'not a bug'
stage = None
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue17194'
versions = []

[florentx]

When two implementations give the same result, I use to run micro benchmarks to give me an hint.

I just noticed that attrgetter is slower than a lambda here:

$ python3.3 -m timeit -s 'from operator import attrgetter; n1 = attrgetter("__name__"); n2 = lambda s: s.__name__' 'rv = n1(int)'
1000000 loops, best of 3: 0.275 usec per loop
$ python3.3 -m timeit -s 'from operator import attrgetter; n1 = attrgetter("__name__"); n2 = lambda s: s.__name__' 'rv = n2(int)'
1000000 loops, best of 3: 0.347 usec per loop

(verified with 2.6, 2.7 and 3.3. But for 2.5 attrgetter is faster)

The function operator.itemgetter does not have same issue.
$ python3.3 -m timeit -s 'from operator import itemgetter; n1 = itemgetter("foot"); n2 = lambda s: s["foot"]; d = {"foot": 42}' 'rv = n1(d)'
10000000 loops, best of 3: 0.122 usec per loop
$ python3.3 -m timeit -s 'from operator import itemgetter; n1 = itemgetter("foot"); n2 = lambda s: s["foot"]; d = {"foot": 42}' 'rv = n2(d)'
10000000 loops, best of 3: 0.176 usec per loop

[serhiy-storchaka]

Are you sure? 0.275 < 0.347

[florentx]

You're right. I've misinterpreted the figures.

Only 2.6 and 2.7 are affected --> closing the issue.