Path: csiph.com!usenet.pasdenom.info!weretis.net!feeder4.news.weretis.net!ecngs!feeder2.ecngs.de!newsfeed.freenet.ag!news2.euro.net!newsgate.cistron.nl!newsgate.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
Date: Sat, 21 Jul 2012 12:32:53 +0300
From: Jan Riechers <janpeterr@freenet.de>
User-Agent: Mozilla/5.0 (Windows NT 6.1; rv:14.0) Gecko/20120713 Thunderbird/14.0
MIME-Version: 1.0
To: python-list@python.org
Subject: Re: Basic question about speed/coding style/memory
References: <mailman.2364.1342856185.4697.python-list@python.org> <500a711f$0$29978$c3e8da3$5496439d@news.astraweb.com>
In-Reply-To: <500a711f$0$29978$c3e8da3$5496439d@news.astraweb.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: list
Reply-To: janpeterr@freenet.de
Newsgroups: comp.lang.python
Message-ID: <mailman.2373.1342863358.4697.python-list@python.org>
Lines: 66
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:25735

On 21.07.2012 12:06, Steven D'Aprano wrote:
>
> But in general, you're worrying too much about trivia. One way or the
> other, any speed difference will be trivial. Write whatever style reads
> and writes most naturally, and only worry about what's faster where it
> actually counts.
>

>
> Notice that I try to make each function do the same amount of work, so
> that we're seeing only the difference between "else" vs "no else".
>
> Now let's test the speed difference with Python 2.7. Because this is
> timing small code snippets, we should use the timeit module to time the
> code:
>
> from timeit import Timer
> setup = "from __main__ import with_else, without_else"
> t1 = Timer("for i in (0, 1): result = with_else(i)", setup)
> t2 = Timer("for i in (0, 1): result = without_else(i)", setup)
>
> Each snippet calls the function twice, once to take the if branch, then
> to take the else branch.
>
> Now we time how long it takes to run each code snippet 1000000 times. We
> do that six times each, and print the best (lowest) speed:
>
> py> min(t1.repeat(repeat=6))
> 0.9761919975280762
> py> min(t2.repeat(repeat=6))
> 0.9494419097900391
>
> So there is approximately 0.03 second difference per TWO MILLION
> if...else blocks, or about 15 nanoseconds each. This is highly unlikely
> to be the bottleneck in your code. Assuming the difference is real, and
> not just measurement error, the difference is insignificant.
>
> So, don't worry about which is faster. Write whichever is more natural,
> easier to read and write.
>
>

Hello Steven,

very nice example and thank you very much for also for the Timeit test!
Actually it confirms my assumption in some way:

[SNIP myself]
So if there is some overhead in some fashion in case we don't offer the
else, assuming the interpreter has to exit the evaluation of the
"if"-statement clause and return to a "normal parsing code"-state
outside the if statement itself.
[SNAP]

Without having looked at Andrew's bytecode excecution hint, using the 
dis module, to see how the interpreter handles the task on lower level.

But fare enough for me :)

But I agree, the return in my example is misleading and it would be 
illegal outside of a function call. I just added it to make clear that 
the fellow code below the return should not be executed in comparison to 
the 2nd example.

Thank you very much
Jan