Huanian-Vector-Li wrote: > Hi Steve and team, > > I have checked CStyle.py and used profile module to get the time > bottleneck, which seems it is re.search()/re.match(). > Pls see attached file (getbottleneck.log) to get the detail. > > Also, I investigated the ctyle.pl and the implement of regular > expression(RE for short) in perl. > RE is implemented via C in perl while it is implemented in re.py in python. > So, it seems reasonable that the performance of CStyle.py is lower than > ctyle.pl. > At least I think the performance of perl's RE is higher than python RE > module. > > Last, I have two questions here, > 1. What the acceptable performance of CStyle.py in comparison with > cstyle.pl?
Ideally, we would like for CStyle.py to be at least on par with cstyle.pl > 2. Any other possible bottleneck of CStyle.py? > Please add your suggestion/comments, thanks! When I ported CStyle.py, I basically did a one-to-one port of the old Perl logic (in order to achieve correctness in a minimal amount of time) - so I believe we could achieve some performance gains by properly refactoring it a bit as more of a proper Python program. Can we eliminate the use of RE's in some places in favour of more standard string.* methods? cheers, steve -- stephen lau // stevel at sun.com | 650.786.0845 | http://whacked.net opensolaris // solaris kernel development
