commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Phil Steitz <>
Subject Re: [math]Re: [releases] lacking releases for .... Was: [sandbox] report
Date Mon, 15 Mar 2004 11:46:58 GMT

--- Christopher Schuck <> wrote:
> Phil, I am new to the Commons, but would like to help out the math 
> project if I could. In brief I am a mathematician turned developer 
> (principally Java), and would like to do something mathematical again. 
> What is your highest priority now? I saw the TODO list, but I imagine 
> there are some things more urgent than others. One item I didn't 
> understand was the refactoring of the solve method for the RealMatrix 
> class (I didn't follow what was meant by making it more pluggable), but 
> if that isn't what you need the most, I'd rather help out where you need 
> the help.
> Chris


The TODO list and "Straw Man Release Plan" here:
provide a decent summary of what we need to do to get a 1.0 release out the
door.  Right now, we are pretty much focussed on cleaning up what we have
so that we can make an intitial release and then get on to some new stuff. 
Depending on your interest / background / tolerance for things like
documentation, testing, etc., I would suggest that you start on one or more
of the following. All have essentially equal priority, since all must be
done prior to release.

1. Code and javadoc review -- there are lots of errors and broken links in
the javadoc and we need to get through all of this.  Pick any package and
start reviewing / submitting patches / fixing broken links.  We need to get
everything up to the standard specified in the "Documentation" section
If you look at, for example, the random or distribution packages, the
javadoc is not in bad shape; but other packages need significant work.
Obviously, a key part of this task is verifying that the code and tests
match the javadoc.

2. User Guide contributions.  If you know basic stat, filling in the
missing sections on Bivariate Regression and/or Statistical Tests in
stat.xml would be great.  The Utilities section also needs to be written.
Let me know if you are going to work on anything in the User Guide so that
we don't step on each other.

3. Test path and boundary value coverage improvements. If you have access
to R or S, for example, or have a numerics background, it would be very
helpful to do more validation testing using extreme values, badly
conditioned problems, etc.  Here again, let us know what you are working on
if you decide to take some of this on.

4. Do the research, evaluate the patch here:
and either submit a patch or recommend closing the spline implementation
issue as WONTFIX.  Test cases and good references describing the proposed
algorithms will be necessary to close this.

5. RealMatrix Solve issue. What is going on here is that the solve() method
of the RealMatrix interface enables you to solve a linear system whose
coefficient matrix is the RealMatrix instance.  The current implementation
uses LU decomposition, which is a good general method, but when you get
down to it the best solution method depends on the system, so it would be
good to make the solution method pluggable.  That is the problem statement.
 If you dig into the code, you will see that the LU decomposition is cached
and used for multiple things (e.g. determinant computation, singularity
testing) and there are some efficiencies that result from this.  I would
not want to lose these efficiencies by fully separating the decomposition.

6. Critical assessment of the APIs from a usability and extensibility
standpoint. I would really like to hold onto the 1.0 target, which means
that the APIs need to be stable and easily extensible without breaking
backward compatability.

If you have any problems getting set up, submitting patches, figuring out
maven, etc., don't hesitate to ask for help. Thanks in advance.


Do you Yahoo!?
Yahoo! Mail - More reliable, more storage, less spam

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message