Le mercredi 26 septembre 2012 11:55:16 UTC+2, Chris Angelico a écrit : > On Wed, Sep 26, 2012 at 7:31 PM, <wxjmfa...@gmail.com> wrote: > > > you are correct. But the price you pay for this is extremely > > > high. Now, practically all characters are affected, espacially > > > those *in* the Basic *** Multilingual*** Plane, these characters > > > used by non "American" user (No offense here, I just use this > > > word for ascii/latin-1). > > > > > > I'm ready to be considered as an idiot, but I'm not blind. > > > As soon as I tested these characters, Py3.3 performs really > > > badly. It seems to me it is legitimate to consider, there > > > is a serious problem here. > > > > We've been over this thread. The only reason you're counting 3.3 as > > worse is because you're comparing against a narrow build of Python > > 3.2. Narrow builds are **BUGGY** and this needed to be resolved. > > > > When you compare against a wide build, semantics of 3.2 and 3.3 are > > identical, and then - and ONLY then - can you sanely compare > > performance. And 3.3 stacks up much better. > > > > ChrisA
No, I'm comparing Py33 with Py32 narrow build [*]. And I am not a Python newbie. Others in a previous discussion have pointed "bad" numbers and even TR wrote something like "I'm baffled (?) by these numbers". I took a look at the test suites, unfortunatelly they are mainly testing "special cases", something like one of the 3 internal representations, eg "latin-1". I can also add to this, that it is not only one of the internal representation which may be suspect (it is probably different now, Py32/Py33) but also the "switch" between these representations which is causing troubles. [*] I have not the knowledge to compile a wide build and I do not wish to spend my time in something that will be most probably a nightmare for me. I'm reacting like a "normal" Python user. jmf -- http://mail.python.org/mailman/listinfo/python-list