"Eric Friedman" wrote:
<DIV>[snip]</DIV>
<DIV>&gt; <BR>&gt; * Generic visitation by any functionable object.<BR>&gt; <BR>&gt; &nbsp; This is *very* powerful; nearly all of variant's internal features<BR>&gt; &nbsp; are implemented in terms of generic (templated) visitation.<BR>&gt; <BR>(1) A short comment regarding proper credits: The technique of using the type index for selecting the correct 'variant-to-visitor bridge' function</DIV>
<DIV>was first introduced in my post: <A href="http://lists.boost.org/MailArchives/boost/msg30415.php">http://lists.boost.org/MailArchives/boost/msg30415.php</A>.&nbsp;At the same discussion, Doug&nbsp;Gregor made the suggestion, that&nbsp;operator() will be used as the visitor's 'visit_me()' function.&nbsp;&nbsp;The initial inspiration came from the vtable idiom described&nbsp;at Andrei<BR>Alexandrescu's paper: "An implementation of discriminated unions in C++".</DIV>
<DIV>&nbsp;</DIV>
<DIV>&nbsp;</DIV>
<DIV>(2) I'd like to repeat my view: given the fact that the two varaint implementations are getting more and more identical, the two implementations should be integrated into single library.&nbsp;I still believe that the most effective approach, would be for us to co-develop this (unified) variant library.</DIV>
<DIV>&nbsp;</DIV>
<DIV>-Itay</DIV>
<DIV>&nbsp;</DIV><p><br><hr size=1><b>Do You Yahoo!?</b><br>
New! <a href="http://pa.yahoo.com/*http://rd.yahoo.com/sbcyahoo/consumer/evt=640/*http://sbc.yahoo.com" target="_blank">SBC Yahoo! Dial</a> - 1st Month Free &amp; unlimited access