get each pair from a string.

[Ian Kelly]

On Tue, Oct 23, 2012 at 12:47 AM,  <wxjmfauth at gmail.com> wrote:
> The latest Python version is systematically slower
> than the previous ones as soon as one uses non "ascii
> strings".

No, it isn't.  You've previously demonstrated a *microbenchmark* where
3.3 is slower than 3.2.  This is a far cry from demonstrating that 3.3
is "systematically slower", and that larger claim is simply false.

In any case, have you been following the progress of issue 16061?
There is a patch for the str.replace regression that you identified,
which results in better performance across the board than 3.2.  So
although 3.3 is slower than 3.2 on this one microbenchmark, 3.4 will
not be.