<html>
John,<br>
<br>
Our application would not need scaling. It needs to extract the last N
feasible (extreme) points. I think we can store the first N feasible
points in a FIFO queue with capacity = N and update the queue with each
new feasible point along the way to optimality.<br>
<br>
Thanks,<br>
<br>
Yiming<br>
<br>
<br>
At 01:12 PM 7/12/2003 -0400, John J Forrest wrote:<br>
<br>
<blockquote type=cite class=cite cite><font size=2>Yiming,</font> <br>
<br>
<font size=2>I am not sure what would be the simplest way to do what you
want, but I think the best way will be if I create a new class
ClpSimplexMessageHandler inheriting from CoinMessageHandler. This
would have a pointer to simplex information plus a void pointer for
user. Then when certain messages occur you could copy solution or
basic information across. This would have more general uses than
your particular need. There are some issues like what happens
if scaling is on.</font> <br>
<br>
<font size=2>If you give me more details then I can use that as part of
the unit test, and I can probably address scaling issue.</font> <br>
<br>
<font size=2>On your other question - Presolve obviously gets rid of
redundant rows but it may do too much. Coding in
ClpSimplex::tightenPrimalBounds identifies many redundancies (but does
nothing about them as it is not going to take them out). It would
be simple to write something to create a new model without those
rows.</font> <br>
<br>
<font size=2>John Forrest</font> </blockquote><br>
</html>