On 10/17/2014 10:46 AM, "Ola Fosheim Grøstad"
<[email protected]>" wrote:
On Friday, 17 October 2014 at 17:16:29 UTC, Ola Fosheim Grøstad wrote:
Accurate is slower, but not this:
sum(a)/len(a)
Forgot to point out that the original point with mentioning Python in the thread
is:
1. Compiled static languages still have a long way to go with expressiveness.
D has closed that gap quite a bit.
2. Generic operations on double will lead to wrong results. What happens if your
first value is very large? You loose the accumulation of the smaller values, in
the worst case you only get the first value. Thus you will have to sort the
values by exponent before accumulating or convert it into a different format.
Or use the much-hated much-misunderstood 80 bit reals!
In any case, just insert a .sort operation. It's still generic.
So yes, C++ iterators and D ranges are kind of cool, but cannot beat a well
engineered library and a good mapping to it on the language side. (Python is of
course not optimal in any way.)
80 bit reals can compensate for a lot of loss-of-precision issues.
