Repository navigation
Don't needlessly change refcounts of dummy objects for sets #62997
Description
| BPO | 18797 |
|---|---|
| Nosy | @tim-one, @rhettinger, @pitrou, @markshannon |
| Files |
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 2013-08-22.16:01:38.849>
created_at = <Date 2013-08-21.07:55:45.530>
labels = ['interpreter-core', 'performance']
title = "Don't needlessly change refcounts of dummy objects for sets"
updated_at = <Date 2013-08-22.16:01:38.848>
user = 'https://github.com/rhettinger'bugs.python.org fields:
activity = <Date 2013-08-22.16:01:38.848>
actor = 'rhettinger'
assignee = 'none'
closed = True
closed_date = <Date 2013-08-22.16:01:38.849>
closer = 'rhettinger'
components = ['Interpreter Core']
creation = <Date 2013-08-21.07:55:45.530>
creator = 'rhettinger'
dependencies = []
files = ['31394', '31402']
hgrepos = []
issue_num = 18797
keywords = ['patch']
message_count = 4.0
messages = ['195751', '195753', '195810', '195904']
nosy_count = 5.0
nosy_names = ['tim.peters', 'rhettinger', 'pitrou', 'Mark.Shannon', 'python-dev']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue18797'
versions = ['Python 3.4']
Activity
rhettinger commentedon Aug 21, 2013
AFAICT, there is no reason for sets to incref and decref dummy objects. The dummy object address is used as placeholders in the hash table but it is never accessed by set the logic. As long the one reference is held at the time the dummy object is created, nothing further is served by the increfs and decrefs.
I can take them out entirely or use ifdefs to keep them for debug builds.
Does anyone know of any issues?
pitrou commentedon Aug 21, 2013
I can't think of any counter-indication but I think we shouldn't distinguish between debug and non-debug mode. That way the debug hooks can check that the refcounting optimization is right.
rhettinger commentedon Aug 21, 2013
That makes good sense.
Attaching a new patch without the #ifdefs.
python-dev commentedon Aug 22, 2013
New changeset ac6dab1819c4 by Raymond Hettinger in branch 'default':
bpo-18797: Remove unneeded refcount adjustments for dummy objects.
http://hg.python.org/cpython/rev/ac6dab1819c4