2to3 Slight Patch

[nedds]

BPO	3182
Files	pytree.py: new version of pytree.py

^{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 2008-08-21.22:27:30.913>
created_at = <Date 2008-06-23.17:25:43.387>
labels = ['expert-2to3', 'performance']
title = '2to3 Slight Patch'
updated_at = <Date 2008-08-21.22:27:30.912>
user = 'https://bugs.python.org/nedds'

bugs.python.org fields:

activity = <Date 2008-08-21.22:27:30.912>
actor = 'benjamin.peterson'
assignee = 'collinwinter'
closed = True
closed_date = <Date 2008-08-21.22:27:30.913>
closer = 'benjamin.peterson'
components = ['2to3 (2.x to 3.x conversion tool)']
creation = <Date 2008-06-23.17:25:43.387>
creator = 'nedds'
dependencies = []
files = ['10713']
hgrepos = []
issue_num = 3182
keywords = []
message_count = 5.0
messages = ['68641', '68692', '68694', '69632', '69633']
nosy_count = 2.0
nosy_names = ['collinwinter', 'nedds']
pr_nums = []
priority = 'normal'
resolution = 'rejected'
stage = None
status = 'closed'
superseder = None
type = 'performance'
url = 'https://bugs.python.org/issue3182'
versions = ['Python 2.6']

[nedds]

This is a small patch to the 2to3 tool, replacing some calls to
isinstance (x,y) with type(x) is y in the file pytree.py. Although there
is only a slight performance increase for each time this change is made,
the recursive nature of pattern matching means that isinstance was being
calling hundreds of thousands of times per file, so overall these minor
changes result in a 3-4% speed increase in my limited testing. It
currently fails only one test due to line 432 in test_pytree, because I
had to rename the type parameter to new_type so I could call type() on
something. But with the line in the test file changed, it passes all tests.

[collinwinter]

I'm not wild about making this change for such a minimal improvement
(which I would guess falls within the margin of error), since it will
limit extensibility and testability. I think I'd be fine with it if the
benefit were >10%, but 3% doesn't sound significant enough.

[nedds]

I don't think the improvement falls in margin of error, but maybe I'm
wrong. The problem was that even when just run on a single file,
isinstance was being called upwards of 100,000 times. I guess it doesn't
merit inclusion on its own though, but with other things I'm working on,
I think it makes sense.

[collinwinter]

So, revisiting this...

On the face of it, I'm not convinced that the isinstance(x, Leaf) ->
type(x) is Leaf changes are correct: certain fixers have in the past
utilized their own subclasses of Node, and I can foresee this being done
again in the future. Reverting those changes seems to remove the speedup
you observed.

[nedds]

Fair enough. I guess that even though there's a little bit of a
performance improvement from this, it does hurt extensibility, so its
probably not a worthwhile change after all.