multi-variate distribution

[Daniel Krügler]

2012/9/3 Ulrich Drepper <drepper@gmail.com>:
> The distribution functions included so far are only the basic ones.
> For simulations what is needed are multiple variables and they usually
> are correlated.  This is what multi-variate distributions are about.
> The most commonly used is the multi-variate normal distribution.  The
> attached patch implements such a distribution.

[..]

> Any opinions?  Otherwise I'll submit the patch for inclusion.

I don't think that

template<typename _InputIterator1, typename _InputIterator2>
+	  param_type(_InputIterator1 __meanbegin, _InputIterator1 __meanend,
+		     _InputIterator2 __varcovbegin, _InputIterator2 __varcovend)

should name the parameters as _InputIterator1 and _InputIterator2, resp,
because the code does multi-pass traversal. Suggested change would
be

template<typename _ForwardIterator1, typename _ForwardIterator2>
+	  param_type(_ForwardIterator1 __meanbegin, _ForwardIterator1 __meanend,
+		     _ForwardIterator2 __varcovbegin, _ForwardIterator2 __varcovend)

This also means that e.g.

+      template<typename _InputIterator1, typename _InputIterator2>
+	normal_mv_distribution(_InputIterator1 __meanbegin,
+			       _InputIterator1 __meanend,
+			       _InputIterator2 __varcovbegin,
+			       _InputIterator2 __varcovend)

should change its parameter names.

When looking into

 template<std::size_t _Dimen, typename _RealType>
+    template<typename _InputIterator1, typename _InputIterator2>
+      void normal_mv_distribution<_Dimen, _RealType>::param_type::
+      _M_init_full(_InputIterator1 __meanbegin, _InputIterator1 __meanend,
+		   _InputIterator2 __varcovbegin, _InputIterator2 __varcovend)

it seems that _InputIterator1 is properly chosen, but _InputIterator2 not: The
code imposes random access traversal. Is this really necessary? IMO,
the code could simply use advance on __varcovbegin.

- Daniel