nodecounts sequences for backtracking algorithms

[magictour at free.fr]

nodecounts sequences for backtracking algorithms


for most sequences Axxxxxx in the oeis (and other not included
sequences) and for any small n we can write some computer programs
to compute Axxxxxx(n).
This will typically be a backtracking or recursive algorithm
which will also work for other n.
Such a program gives naturally a (finite) sequence of nodecounts
node(1..k) where the counter node(i) is increased whenever
the recurrence-subroutine is called at nested level i,
that is : with i return-addresses to itself on the stack.
For backtracking versions, the counts of node(i) can usually
also be invoked easily.

Comparing these node-sequences is often an easy method
to detect bugs in a program.
It's also a measure for the complexity and "pruning-characteristics"
of a program.

So it would be useful to have a database of such node-sequences.

You could also make normal,1d  infinite sequences by adding all the
nodecounts, so e.g. a simple straightforward program to generate
all permutations would give A007526 as (sum of nodecounts)(n).
The same in 2d for latin squares gives 1,8,93,5680,2314165,...
which is not in the oeis.

So it would be useful to have at least a database of such
node-sum-sequences.  Even for the most common boguous
implementations.

Has anyone ever heard about such a project ?



-Guenter (sterten.(at).aol.com)