Re: [boost] [Random] Question on random_test.cpp

Subject: Re: [boost] [Random] Question on random_test.cpp
From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2009-03-27 14:47:43

This is known from lexical cast loopback test - see

http://connect.microsoft.com/VisualStudio/feedback/ViewFeedback.aspx?Feedbac
kID=98770 (requires sign-in)

For my money, this is a 'buglet' in the istream library - but MS say it is a
feature.

There is no language guarantee, but my tests showed that it did work only if

you use enough decimal digits (17) for 64 bit doubles. << setprecision(17) 

AND for MSVC you need to use << scientific format.

I am told that GCC library is OK.

Paul

---
Paul A. Bristow
Prizet Farmhouse
Kendal, UK   LA8 8AB
[phone], mobile [phone]
pbristow_at_[hidden]
> -----Original Message-----
> From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]]
On
> Behalf Of n1driver_at_[hidden]
> Sent: 27 March 2009 17:59
> To: boost_at_[hidden]
> Subject: [boost] [Boost][Random] Question on random_test.cpp
> 
> The following is an excerpt from the random_test.cpp file (part of the
> instantiate_urng function):
> 
>  urng2 = urng;
> 
>   {
>     // narrow stream first
>     std::ostringstream file;
>     file << urng;
>     // move forward
>     urng();
>     // restore old state
>     std::istringstream input(file.str());
>     input >> urng;
>     // std::cout << file.str() << std::endl;
> #if !defined(BOOST_MSVC) || BOOST_MSVC > 1300 // MSVC brokenness
>     // advance some more so that state is exercised
>     for(int i = 0; i < 10000; ++i) {
>       urng();
>       urng2();
>     }
>     BOOST_CHECK(urng == urng2);
> #endif // BOOST_MSVC
>   }
> 
> My concern is that when a random number generator with an underlying
double type
> (like fibonacci607) is stored to the stream and then pulled back out, then
> precision may be lost, and the underlying data is not guaranteed to be the
same.
> If a double is written out to a stream in decimal form and then read back
in the
> value may be changed.  Am I missing some sort of language guarantee that
this is
> supposed to work?  I've tried this on AIX and the test failed at this
point for
> fibonacci607 because some (but not all) of the underlying doubles had
changed
> after the object was stored and retrieved.
> 
> Thanks!
> _______________________________________________
> Unsubscribe & other changes:
http://lists.boost.org/mailman/listinfo.cgi/boost
> 
> 
> 
> 
> 
> E-mail message checked by Spyware Doctor (6.0.0.386)
> Database version: 5.12060
> http://www.pctools.com/uk/spyware-doctor-antivirus/
E-mail message checked by Spyware Doctor (6.0.0.386)
Database version: 5.12060
http://www.pctools.com/uk/spyware-doctor-antivirus/