We need a few graph data sets to develop more robust and interesting unit tests
on top of. It is possible for a data set to be generated by a program, as long
as the structure is defined precisely enough that tests can operate on it. It
is also possible to have them in some data files and load them in a fresh
database before a bunch of tests are run.
Perhaps there are two data sets that will be useful: one for a classic graph
(binary edges, node and edge sets being disjoint) and another for an actual
hypergraph. The classic graph would be useful for testing standard traversals
and algorithms that mostly make sense in that setting. And the hypergraph will
test more involved AL generators in traversals, hyper-traversals, and more
complex pattern matching queries (when we have the pattern matching developed).
This is by no means a trivial task. The data sets need to exhibit enough
diversity to cover interesting structures and possibly catch limit cases in the
tests. Yet one should be able to describe them so as to write tests against
them.
Original issue reported on code.google.com by borislav...@gmail.com on 24 Jun 2012 at 5:03
Original issue reported on code.google.com by
borislav...@gmail.comon 24 Jun 2012 at 5:03