On Tue, Jan 4, 2011 at 5:15 PM, richard emberson<
richard.ember...@gmail.com
wrote:
Dev Wicketers,
What: I have ported Wicket to Scala
A couple of months ago I took a 1.5 snapshot and ported to Scala.
This encompasses all of the source and test code. As successive 1.5
snapshots were released, I ported those differences to my Scala
version. I am current with 1.5 M3.
The Java 137,791 loc in 1.5 M3 are now 100,077 loc Scala (not
counting all the println statements I put into the Scala code
for debugging). I used cloc (http://cloc.sourceforge.net/) to
count lines of code.
I haven't used CLOC before. I've used Ohcount (
http://www.ohloh.net/p/ohcount) and like it. I'll have to give this a
try.
I have also replaced all of the Java collection classes with
Scala collection classes (though a small number of Java collection
classes remain that did not have comparable Scala implementations).
I have changed many method return types from the Java returning
some "object" or "null" to Scala returning "Some(object)" or "None"
(using the Scala Option[return-type] construct) - trying to
eliminate nulls.
Lastly, I pushed the IModel[T] typing down to the Component class
making get/set DefaultModel and get/set DefaultModelObject strong
typed. This included using Scala companion object apply methods
which eliminated having to explicitly declare type parameters in
most end-user code (I had read that one of the objections to
pushing strong typing down to the Component class in Wicket was
that there were "too many notes", end-user code was too verbose).
It can not interoperate with Java Wicket because Scala compiles to
JVM class files and so all of the classes in Java Wicket also
appear in Scala-Wicket.
I have an "internal" name for my Scala port of Wicket which
acknowledges its Wicket heritage as well as advertises its
enterprise level capabilities. For external communications,
I am currently simply call it Scala-Wicket.
Why: Scala is a better Java
I was introduced to Scala 9 months ago and quickly determined that
it was a better Java (at least IMO). For Scala to succeed it
requires more programmers to use it. Many on the Scala mailing
lists were from a functional background and seemed not to recognize
that Haskell and Lisp are not blindingly successful but, rather,
niche languages and that the heavy selling of Scala's function and
typing capabilities might turn off Java programmers.
Scala struck me in many ways as a strong-typed JavaScript, at
least, much of the code did not have to have type declarations
because the compiler could infer types in many cases. In addition,
a whole lot of the Java boil-plate code was not needed. As such,
it could be sold as simply a better Java; a more-to-the-point
object oriented language with functional programming in-the-small.
To get more Java programmers to try Scala I looked for a
significant Java application with a strong support and user
community that I could port to Scala. I ended up with Wicket.
Wicket is an enterprise level web framework (unlike existing
Scale web frameworks which place restrictions on enterprise IT
organizations, e.g., by requiring sticky sessions). It is well
documented. And, as it turned out, very, very importantly it had
a large number of unit tests (the unit tests saved my butt,
without them I would never had succeeded in getting a port that
worked).
No, Really, Why:
I like Scala and I took the time to learn it. Right now about
20% of programmers use Java while only some 0.4% use Scala.
I did not want my effort of learning Scala to be wasted so my
solution is to increase the number of Scala programmers. Where
to get them? Again, my solution is from the existing horde of
Java programmers.
Plans: Release, Evolve and Proselytize
I would like to release Scala-Wicket.
I do not know if Apache hosts anything other than Java code.
Growing a community is important.
Still Todo:
Comments: All of the existing class and inline comments are
still Java related. This would have to be a long, on-going
task to edit the comments so they reflect the code's
Scala usage.
Package path: The code still uses the "org.apache.wicket"
package path and this must be changed - unless this became
an Apache project.
Author: I have quite literally looked at and touched every line
of code but I have not yet felt comfortable with adding
myself as an author since, well, many changes were
syntactic and not semantic.
Refactor with Traits: Currently the port uses Scala traits like
Java interfaces but it should be possible to factor the
common code out into the traits. This would result in many
of the interfaces, the "I" files, such as IModel.scala,
going away.
Some general refactoring:
As an example, consider factoring out the IModel[T] from
Component. Half the time a user wants a Component with
no model, so, if there was a HasModel trait:
class Model[T](var value: T) {
def getObject: T = value
def setObject(value: T): Unit = this.value = value
}
trait HasModel[T] {
var model: Model[T]
def getDefaultModel: IModel[T] = model
def setDefaultModel(model: IModel[T]): this.type = {
....
this
}
def getDefaultModelObject: Option[T] = {
....
}
def setDefaultModelObject(obj: T): this.type = {
....
this
}
}
The Component hierarchy would have no model support.
The user could add model support when needed:
val form = new Form("hi")
with HasModel[Int] { var model = new Model(42) }
Just an Idea.
STM: There are a number of Scala STM projects and I do not know
if it is useful to add STM capabilities to Scala-Wicket.
RBAC: I've written a Scala implementation of the NIST RBAC
recommended standard and might consider adding it.
Logging: Adding a Scala-based logging framework to aid user
debugging.
Monitoring and stats: In the last couple of years many web
sites have added monitoring and statistics gathering
capabilities (e.g., who clicks what, where, how long, on
what page does the visitor exit the web site, etc.) in
order to know how the web site is being used and then
improve the web site.
Significant Memory Usage Reduction: I've an idea that would
significantly decrease the memory usage of Scala-Wicket and
I plan to do a test implementation.
Replace Java features: There are still some Java-isms that can
be replaced with Scala equivalents.
Port additional Java Wicket libraries to Scala.
Enable multiple instances of a unit tests to be run at once.
More: ????????????
I want to avoid using some of the WTF features of Scala (when a
Java programmer looks at the code and says "WTF") in order to
ease and accelerate acceptance by Java programmers; as
examples, implicits can make code hard to understand and
advanced Scala type usages, as James Gosling said, "makes one's
head spin".
Help and Advice: How should Scala-Wicket be extended and released
Scala-Wicket is a port and evolution of Wicket, not a ground-up
re-write. Given that, what would you do differently in Wicket
now that there are years of experience using it?
How best to get a hosting site, release the code and build a
community?
If you're looking for a place to host it, I'd recommend starting with
Github. Git is where the crowd is headed, and Github is the easiest
place
to get up and running with it these days.
You mentioned earlier the idea of it being an Apache project. If you
wanted
it to be an Apache project, you would start at the Incubator (
http://incubator.apache.org/). The one barrier you'll have initially
