[r-t] Big searches

[Don Morrison]

Mark Davies wrote:
> I do remember I had problems with my fragment search loop, though. I wonder,
> is this the sort of thing you have developed for your Rutland search Don?

Nope. I've never done anything like that.

The ten seconds of breadth first search I referred to is much stupider, though 
in some circumstances surprisingly helpful. As part of setting up a problem I 
search backwards from the end node breadth first, merging nodes so that the end 
of the search will be just one, really big, node. In practice this turns out to 
be a win generally only where I'm trying to meet some threshold of some 
property, typically wanting a solution to have at least N instances of some set 
of musical rows. I've never really worked through the theory of where it helps 
and where it doesn't, and just have empirical heuristics for when it is or is 
not likely to help.




-- 
Don Morrison <dfm at ringing.org>
"Enum is actually a generic class defined as Enum<T extends Enum<T>>.
This circular definition is probably the most confounding generic type
definition you are likely to encounter. We're assured by the type
theorists that this is quite valid and significant, and that we should
simply not think about it too much, for which we are grateful."
            -- Ken Arnold and David Holmes, _The Java Programming Language_