[New-bugs-announce] [issue7826] support caching for 2to3
report at bugs.python.org
Mon Feb 1 01:59:12 CET 2010
New submission from Brian Harring <ferringb at gmail.com>:
Bit like unittest, right now it's rather hard to extend 2to3 for caching support (and other outputs) w/out duplicating the main function.
Attached is a patch that allows the refactoring tool class to be passed in- at the very least, this patch is enough to remove some of the nastier monkey patching tricks I had to level to inline caching support into 2to3.
Actual caching patch will follow shortly; roughly what it does is track the md5 of the original source and use that as a lookup to the transformed version.
That caching support isn't useful to the majority of users, but for developers w/ buildslaves it's a very useful reduction in runtime- for example, for 6 buildslaves I run for pkgcore (a py2k source that we translate upon install to py3k if the target is py3k), if the cache is primed this is a reduction of 20s to 1.8s. Two minutes cpu time across the slaves brought down to ~11s.
As said, I understand it's a corner case, so getting the caching into 2to3 directly may not be possible.
This initial patch however makes it *way* easier to do inline the caching into 2to3 without having to copy/paste main (or do some heinous monkeypatching).
Patch is against svn trunk; doesn't apply cleanly to 3.1/2.6, but that's just a minor modification to make it so.
components: 2to3 (2.x to 3.0 conversion tool)
title: support caching for 2to3
type: feature request
versions: Python 3.2
Added file: http://bugs.python.org/file16077/allow_alternate_output_tools.patch
Python tracker <report at bugs.python.org>
More information about the New-bugs-announce