Why not use valarray<valarray<T> > ?

[Jim West]

The GNU implementation of the valarray header (g++ version 3.3.2)
includes the comment:

// This complication is so to make valarray<valarray<T> > work
// even though it is not required by the standard. Nobody should
// be saying valarray<valarray<T> > anyway. See the specs.

Unfortunately, "See the specs" does not give me a whole lot of insight
to why this is a Bad Idea(tm). Can someone please enlighten me?

[Buster]

"Jim West" wrote

The GNU implementation of the valarray header (g++ version 3.3.2)
includes the comment:

// This complication is so to make valarray<valarray<T> > work
// even though it is not required by the standard. Nobody should
// be saying valarray<valarray<T> > anyway. See the specs.

Unfortunately, "See the specs" does not give me a whole lot of insight
to why this is a Bad Idea(tm). Can someone please enlighten me?

You use slices to make multidimensional arrays with valarray. See the
standard or a good textbook.

Regards,
Buster.

[Jim West]

In article <bs**********@news6.svr.pol.co.uk>, Buster wrote:

"Jim West" wrote

The GNU implementation of the valarray header (g++ version 3.3.2)
includes the comment:

// This complication is so to make valarray<valarray<T> > work
// even though it is not required by the standard. Nobody should
// be saying valarray<valarray<T> > anyway. See the specs.

Unfortunately, "See the specs" does not give me a whole lot of insight
to why this is a Bad Idea(tm). Can someone please enlighten me?

You use slices to make multidimensional arrays with valarray. See the
standard or a good textbook.

Thanks for the response. I do indeed use slices (or Blitz) when using
multidimensional arrays. The nature of the comment made me think that
there might be some non-obvious reason why valarray<valarray<T> > would
give incorrect results, rather than just be inefficient.