I'm the one who forked ECJ to .NET a while ago, and I'm only responding to
this "feedback" request for one reason:
When I was converting the Java code to C# I found that it was all essential
to create a very large number of unit tests (500+), to sanity-check what I
Can you imagine the immense benefit to ECJ (not to mention MASON) if time
spent checking on float to double conversion was instead directed to
building a full suite of unit tests instead?
I'm currently very heavily involved in a new open-source project for
advanced compression of time series data ("Random Walk Compression").
Otherwise, I would myself contribute time to building a test suite for ECJ.
I think this new-fangled idea of rigorous testing might really have some
long-term merit. ;-)
As a professional "quant" software developer, I can't imagine building and
maintaining a sophisticated body of code without it.
For those contributing to the project who are non-professional coders, a
quick Google search on "Test First" might provide some motivation and
Academics can think of this as "hypothesis before proof". The scientific
method would be completely lost without it.
Ben Stabile, MCSD
From: ECJ Evolutionary Computation Toolkit
[mailto:[log in to unmask]] On Behalf Of Sean Luke
Sent: Monday, March 24, 2014 10:48 AM
To: [log in to unmask]
Subject: Any feedback on the float->double conversion in ECJ?
The offers of help in testing or double-checking my updates to ECJ have died
down and I fear are gone. Anyone have any feedback or test results with
regard to the big float->double updates I made recently out on SVN?