-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of John Maddock Sent: 16 May 2007 18:36 To: boost@lists.boost.org Subject: Re: [boost] [math toolkit] Review results
Paul A Bristow wrote:
If we want it to fail at compile time, IMO it MUST be an option becaue there are occasions when it is vital to be able treat all distributions alike. A typical case is our C# .NET (cue hisses and boos) 'applet' to calculate distribution parameters - for any distribution chosen from a dropdown. (This useful program will be posted somewhere soon).
Two other options: we could remove the definitions of accessor functions that are undefined for that distribution, and then add our own private "catch all" versions when we need them. Or we could add the "catch all" versions to a separate namespace, and users could bring them into scope when needed.
These seem more complicated to me, unless I am misunderstanding, again. There are only quite few *undefined* accessors (mean, etc).
Do we have customised (as in meaningful and distribution specific) error messages for the current functions though? If so we should try and keep them.
The compiler outputs: I:\Boost-sandbox\math_toolkit\boost/math/distributions/cauchy.hpp(218) : error C2027: use of undefined type 'boost::STATIC_ASSERTION_FAILURE<x>' with [ x=false ] .\test_cauchy_mean.cpp(47) : see reference to function template instantiation 'RealType boost::math::mean<double>(const boost::math::cauchy_distribution<RealType> &)' being compiled with [ RealType=double ] but if you click on cauchy.hpp line 218 you see the pretty self-explanatory code below: template <class RealType> inline RealType mean(const cauchy_distribution<RealType>& ) { // There is no mean: #ifdef BOOST_MATH_COMPILE_FAIL_IF_UNDEFINED # ifndef BOOST_MSVC # error Mean of the Cauchy distribution is undefined! # else BOOST_STATIC_ASSERT(sizeof(RealType) == 0); # endif #endif return tools::domain_error<RealType>( BOOST_CURRENT_FUNCTION, "The Cauchy distribution does not have a mean: " "the only possible return value is %1%.", std::numeric_limits<RealType>::quiet_NaN()); }
So to provide an option, a macro seems sensible, perhaps BOOST_MATH_COMPILE_FAIL_IF_UNDEFINED?
The consensus seems to be for making compiler errors the default, so maybe BOOST_MATH_DEFINE_ALL_ACCESSORS to make them available regardless of whether they are mathematically defined for the distribution.
OK - I assume this is just the inverse of BOOST_MATH_COMPILE_FAIL_IF_UNDEFINED Are we *sure* this is the consensus? IMO providing NaN may be more useful?
BOOST_STATIC_ASSERT(false); has been one way of providing a compile time trap and fails with a rather long message, although clicking on the 1st line goes straight to the source code above, so it soon becomes pretty clear.
It would have to be BOOST_STATIC_ASSERT(sizeof(T) == 0) as some compilers realise that BOOST_STATIC_ASSERT(false) will always fail and issue a diagnostic irrespective of whether the template is instantiated.
OK - but I hope compilers don't get even smarter.
For MSVC at least one can use the #error pragma thus
#ifdef BOOST_MATH_COMPILE_FAIL_IF_UNDEFINED # ifdef BOOST_MSVC # error Mean of the Cauchy distribution is undefined! # else BOOST_STATIC_ASSERT(false); # endif #endif
and this produces the rather more explicit message:
I:\Boost-sandbox\math_toolkit\boost/math/distributions/cauchy.hpp(216)
: fatal error C1189: #error : Mean of the Cauchy distribution is undefined!
Surely that just gives an error whenever you include the header?
Seems to work for MSVC 8.0.
Or we could simply not provide the definitions of these functions at all: but then the error messages might be rather cryptic?
A definite step backwards, IMO. Paul --- Paul A Bristow Prizet Farmhouse, Kendal, Cumbria UK LA8 8AB +44 1539561830 & SMS, Mobile +44 7714 330204 & SMS pbristow@hetp.u-net.com