Mauro Talevi recently proposed a new package for general linear
regression (https://issues.apache.org/jira/browse/MATH-203). This patch
needs Java 5, mainly for annotations.
Mauro suggested to take the opportunity of the next [math] major version
to switch to Java 5. A major version seems appropriate for such a
change, but do we want to do it now ?
My personal opinion is that sticking to Java 1.3 is really obsolete and
difficult. When I upgraded my Linux box recently, I had to search old
backups to reinstall a JDK manually. Dropping this could simplify some
codes (exceptions for example) and fix some errors (there is a known
issue with unit tests since Java 1.3 does not compute trigonometric
functions as it should).
If we decide to change minimal Java version, I would choose to target
1.5. It is widely adopted and deployed now and has many features which
would be useful for a mathematical library:
- new Math functions (log10, cbrt, ulp, signum, cosh, sinh, tanh,
hypot, expm1, log1p)
- autoboxing
- MathContext, RoundingMode
In addition, there are the many features that are interesting for any
type of Java development (enums, generics, annotations).
Java 6 brings even more Math functions (copysign, getExponent,
nextAfter, nextUp, scalb), some of which we needed to add ourselves in
MathUtils. However, I'm not sure it is as widely deployed than Java5.
Perhaps Java 7 would bring even more functions (asinh, acosh and atanh
are still missing ...)
What do you think ?
Luc
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
