[Baypiggies] Question about breaking out of a loop

Hy Carrinski hcarrinski at gmail.com
Wed Jun 8 02:03:26 CEST 2011

Jeremy's function g() does produce outputs equivalent to the sample
code without using a conditional. Thank you also for including
examples. As a generator, it actually could serve as a three parameter
wrapper for itertools.combinations().

I hope that this present email can better specify the question. I very
much appreciate the good thoughts thus far.

1. The only information we have about the generator is that its
outputs have a defined order. I find this abstraction useful because
it may help any answers to this question to be more generally
applicable, and also contains the complexity of the code for the
generator itself by not considering multiple starting or ending

> There is generator that yields tuples in a defined order into a loop
> that performs a calculation.

2. I wrote the gist in order to answer the primary question:

    Is it an antipattern to change a datatype to cause an exception?

3. I also began to think that itertools.takewhile() might be a better
option, but does not seem to be in this case. First, it introduces a
need for something larger than an integer to compare, for the case
where the threshold is None. Second, while it looked nice, it also
caused a significant slowdown (~50%)


On Tue, Jun 7, 2011 at 1:57 PM, Jeremy Fishman
<jeremy.r.fishman at gmail.com> wrote:
> Yes thank you Hy for pointing out f() function will not break early.  I am
> not sure why I changed the implementation as I had intended f() to be a copy
> of your for-loop from loop_fcn_v2() to demonstrate equivalency.
> I believe g() generates exactly the values expected and no more.
>   - Jeremy
> On Tue, Jun 7, 2011 at 1:42 PM, Hy Carrinski <hcarrinski at gmail.com> wrote:
>> I agree that this optimization is not critical. Premature optimization
>> is certainly something that I try to avoid, and is an easy path to
>> follow.
>> My actual code has gone through a few rounds of refactoring and some
>> optimization. This is question involves a specific area which my
>> profiling has shown can introduce around a 10% decrease in runtime
>> based only on eliminating this conditional. These rounds have involved
>> effort to maintain and increase clarity. The computationally intensive
>> functions actually make use of caching so they do not consume much
>> time.
>> I did not include many of these details in the original posting.
>> By introducing the if statement, the runtime of the profiled actual
>> code increases by around 20%. While that increase is not too
>> important, the structure with the if statement does not seem right to
>> me because it introduces the overhead whether or not the threshold
>> option is exercised. Is there a more Pythonic way to introduce the
>> break without the commensurate increase in overhead? By the way, I do
>> think that this loop is the most appropriate place to introduce a
>> threshold.
>> w.r.t. Jeremy's recent interesting suggestions. I think that filtering
>> on the generator may not actually result in StopIteration without
>> computing the values. My actual generator makes a sequence of only the
>> values that I care about. After writing it, I found a posting by Tim
>> Peters from a few years ago at
>> http://code.activestate.com/recipes/218332/ which uses an algorithm
>> similar to mine (for the generator). But, please note that link is
>> pretty far astray from the present questions.
>> Thank you,
>> Hy
>> On Tue, Jun 7, 2011 at 1:06 PM, Dirk Bergstrom <krid at otisbean.com> wrote:
>> > On 06/07/2011 12:40 PM, Hy Carrinski wrote:
>> >>
>> >> There is generator that yields tuples in a defined order into a loop
>> >> that performs a calculation. I would like to provide an option to stop
>> >> the calculation when the threshold is reached.
>> >> The function in version 3 is pretty close to my current solution, but
>> >> the functions combinations(), f() and g() are standing in for more
>> >> computationally intensive functions.
>> >
>> > This seems like a perfect example of premature optimization.  You've got
>> > a
>> > loop with two computationally intensive operations per cycle, and you're
>> > worried about optimizing away a single if-equals check per cycle.  Will
>> > that
>> > single if statement really make so much difference once you put the real
>> > (and presumably much more time consuming) functions in place?
>> >
>> > --
>> >       --------------------------------------
>> >      Dirk Bergstrom           krid at otisbean.com
>> >             http://otisbean.com/
>> >

More information about the Baypiggies mailing list