Rhys Ulerich-2 wrote:
<br/>&gt; A total guess... &nbsp;You have two 4D multi_arrays used in an algorithm.
<br/>&gt; Algorithm checks that their fastest dimension has the same size. &nbsp;It
<br/>&gt; does. &nbsp;The algorithm is called recursively on a 3D subproblem. &nbsp;The
<br/>&gt; algorithm checks that the two 3D sub-multi_arrays have a fastest
<br/>&gt; dimension with the same size. &nbsp;They don't. &nbsp;Now what should the
<br/>&gt; generic algorithm do? &nbsp;If it resizes the 3D subproblem bad things
<br/>&gt; happen to the original 4D multi_array.
<br/><br/>I completely understand that many generic algorithms require that sizes 
<br/>match. And we agree on that they should not resize any of the 
<br/>(sub-)multi_array. But I'm afraid this is missing the point. My feeling 
<br/>is that generic algorithm should check that the sizes match if it 
<br/>requires so, NOT the assignment operator. What the assignment operator 
<br/>should require seems largely independent to me.
<br/><br/>Etienne
<br/>
	
<br/><hr align="left" width="300" />
View this message in context: <a href="http://boost.2283326.n4.nabble.com/multi-array-operator-requires-resize-why-tp2563752p4630885.html">Re: [multi_array] operator= requires resize - why?</a><br/>
Sent from the <a href="http://boost.2283326.n4.nabble.com/Boost-Users-f2553780.html">Boost - Users mailing list archive</a> at Nabble.com.<br/>