[Python-Dev] range objects in 3.x

[Guido van Rossum]

On Tue, Sep 27, 2011 at 1:05 PM, Alexander Belopolsky
<alexander.belopolsky at gmail.com> wrote:
> On Tue, Sep 27, 2011 at 2:53 PM, Guido van Rossum <guido at python.org> wrote:
>> On Tue, Sep 27, 2011 at 11:48 AM, Alexander Belopolsky
>> <alexander.belopolsky at gmail.com> wrote:
>>> On Tue, Sep 27, 2011 at 2:36 PM, Guido van Rossum <guido at python.org> wrote:
>>> ..
>>>> But that's exactly what we don't *want* you to do! Because (unless you
>>>> are a numerical wizard) you probably aren't doing the error analysis
>>>> needed to avoid the "unexpected extra point" problem due to floating
>>>> point inaccuracies. For your own good, we want you to state the count
>>>> and let us deliver the number of points you want.
>
> I don't disagree that the ability to provide count= option is useful.
> I am just saying that there are also cases where float step is known
> exactly and count (or stop) can be deduced from stop (or count)
> without any floating point issues.  Iteration over integers that
> happen to be represented by floats is one use case, but using integer
> range may be a better option in this case.  In US it is still popular
> to measure things in power of two fractions.  Simulating a carpenter's
> yard does not suffer from rounding when done in floats.  Counting by
> .5 and .25 has its uses too.  Maybe frange() should just signal the FP
> inexact exception if we expect users to need hand holding to such a
> degree.

But why offer an API that is an attractive nuisance? I don't think
that it is a burden to the user to have to specify "from 0 to 2 inches
in 8 steps" instead of "from 0 to 2 inches in 1/4 inch steps". (And
what if they tried to say "from 0 to 3 1/4 inches in 1/2 inch steps"
?)

-- 
--Guido van Rossum (python.org/~guido)