Instagram: 40% Py3 to 99% Py3 in 10 months (Posting On Python-List Prohibited)
Steven D'Aprano
steve at pearwood.info
Tue Jun 20 05:47:53 EDT 2017
On Tue, 20 Jun 2017 00:11:25 -0700, Lawrence D’Oliveiro wrote:
> On Tuesday, June 20, 2017 at 6:16:05 PM UTC+12, wxjm... at gmail.com wrote:
>> - Py3 on Windows just does not work.
>
> Whose fault is that?
Pay no attention to wxjmfauth and his complaints that Python 3 does not
work. He's either trolling for a reaction, or has an obsessive idée fixe
that Python 3.3+ Unicode handling is "broken" because it uses an
implementation which, way back in the earliest 3.3 alpha releases when he
first tested it, was slightly slower that 3.2 for some terrible
benchmarks.
Python 3.3 strings are optimized for memory saving: the string will use
one, two or four bytes, according to the maximum needed for that
particular string. This usually results in good memory savings and faster
code, but if you perform an artificial benchmark of creating and
destroying millions of strings as fast as you can, the overhead is a
little greater than for 3.2 and it is a bit slower.
In other words, if all you do is create millions of strings, destroy
them, then create them again, and no more processing or work, then Python
3.3's string handling is slower than 3.2.
Obviously this is a trade-off that is worth it for practical speed
increases in real code due to the smaller memory usage of many Python
strings. But jmfauth doesn't see it that way, and so he has invented in
his own head a story that Python is "broken" because it doesn't use UTF-16
everywhere, that there's some sort of conspiracy among the Python
developers to hit Europeans with higher memory use than Americans, and he
complains about perfectly normal Unicode decoding or encoding exceptions
as if they were segfaults.
It's quite sad really, if he's trolling he is the most dedicated troll
I've ever seen, continuing his efforts despite hardly any reaction from
anyone. But I think he is just a crackpot.
--
Steve
More information about the Python-list
mailing list