Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #67226 > unrolled thread

Re: extend methods of decimal module

Back to article view | Back to comp.lang.python

Contents

  Re: extend methods of decimal module Wolfgang Maier <wolfgang.maier@biologie.uni-freiburg.de> - 2014-02-28 14:41 +0000
    Re: extend methods of decimal module Wolfgang <xpysol@gmail.com> - 2014-02-28 06:52 -0800
    Re: extend methods of decimal module "Mark H. Harris" <harrismh777@gmail.com> - 2014-02-28 10:17 -0800

#67226 — Re: extend methods of decimal module

Mark H. Harris <harrismh777 <at> gmail.com> writes:
> 
> If you get a chance, take a look at the  dmath.py  code on:
> 
>    https://code.google.com/p/pythondecimallibrary/ 
> 

Hi Mark,
here is an enhancement for your epx function.
Your current version comes with the disadvantage of potentially storing
extremely large values in n and d because of the multiplications in the
while loop:

    q = D(x)
    n = q
    c = D(1)
    d = D(1)
    ex = 1 + q
    prev_ex = D(0)
    while (ex != prev_ex):
        prev_ex = ex
        c += 1
        d *= c    # can become a huge number
        n *= q    # this as well
        ex += n/d

in general, large numbers are handled well by the Decimal class, but there
is a certain burden on these calculations and with VERY large numbers you
can also get a decimal.Overflow:

>>> exp(200000)
Traceback (most recent call last):
  File "<pyshell#46>", line 1, in <module>
    epx(190000)
  File "C:\Python34\dmath_rev.py", line 27, in epx
    n *= q
decimal.Overflow: [<class 'decimal.Overflow'>]

My re-write of the part above is:

    q = D(x)
    c = 1
    new_element = q
    ex = 1 + new_element
    prev_ex = D(0)
    while (ex != prev_ex):
        prev_ex = ex
        c += 1
        # every new element in the series is a known fraction of the
        # previous element,
        # so there is no need to store large numbers
        new_element *= q/c
        ex += new_element

in my hands, this simple change increases performance (exact timing left to
you) and boost the range of possible calculations:

>>> epx2(1000000)
Decimal('3.033215396802087545086402141E+434294')

Cheers,
Wolfgang

[toc] | [next] | [standalone]

#67228

Uhh, the curse of not copy-pasting everything:

> >>> exp(200000)

should, of course, read
>>> epx(190000)

> Traceback (most recent call last):
>   File "<pyshell#46>", line 1, in <module>
>     epx(190000)
>   File "C:\Python34\dmath_rev.py", line 27, in epx
>     n *= q
> decimal.Overflow: [<class 'decimal.Overflow'>]
> 

[toc] | [prev] | [next] | [standalone]

#67238

On Friday, February 28, 2014 8:41:49 AM UTC-6, Wolfgang Maier wrote:

> Hi Mark,
> 
> here is an enhancement for your epx function.
> 

> Wolfgang

hi Wolfgang,  thanks much!   As a matter of point in fact, I ran into this little snag and didn't understand it, because I was thinking that outside of memory there should not be any overflow for decimal in python /   which means, I suppose, that the C accelerated decimal does have some built-in limitations.  So, yes, I got some overflows too...

Also, thanks for your comment about my D(str()) function...   from Oscar to you, and everyone else... well that has to go.  

Well, now I gotta get busy and clean this thing up...

Thanks again.
marcus

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web