On 4/12/11 5:13 AM, Jason Baldridge wrote:
Thanks everyone for your thoughts. I think the first step is to refactor the
package sticking with Java and then we'll see about moving to a Scala/Java
mix after that (but only for the opennlp machine learning package, currently
opennlp-maxent).
I was actually sort of appalled looking through the code yesterday and
seeing so many global variables used all over the place, making it hard to
know exactly what every method had access to. I think this was sort of an
artifact of how I used Trove functions a loooong time ago to enable quick
iteration over the data structures (which required some objects to be
global). That is obviously gone now, but the global variables didn't go
away... hope I'll find time to improve things over the next 5-6 months.
I know there are lots of global variables, but I think that is easy to
refactor
and to test afterward.
Jörn
