[Snowball-discuss] Name Convention and abstract Method "stem()"

[Richard Boulton]

On Sat, 2005-08-20 at 20:15 +0200, Jens Krefeldt wrote:
> 2. I have an improvement suggestion for the JAVA Snowball Stemmer: If you 
> want to use a SnowballStemmer you have to prepare something like this:
> 
>             String name = ...;
>             Class stemClass = Class.forName("org.tartarus.snowball.ext." + 
> name + "Stemmer");
>             stemmer = (SnowballProgram) stemClass.newInstance();
>             stemMethod = stemClass.getMethod("stem", new Class[0]);
>             stemMethod.invoke(stemmer, null);
> 
> My Question: Why doesn't the SnowballProgram class has an abstract stem() 
> method like
> 
> public abstract class SnowballProgram {
>     ...
>     /**
>      * Every derived <CODE>SnowballProgram</CODE> has to implement this 
> method
>      * to initialize the appropiate stem algorithm.
>      */
>     public abstract boolean stem();
> 
>     ...
> }

SnowballProgram represents a general snowball program.  Although the
snowball programs presented on www.snowball.tartarus.org are all
stemming algorithms with a single function "stem" as an entry point,
there is no requirement in the snowball language for there to be a
function called stem.  It is conceivable that future algorithms might
have more than one entry point, for example.

Therefore, I do not think it's appropriate to add a stem() method to
SnowballProgram. However, the idea is reasonable: it just shouldn't be a
subclass of SnowballProgram.

Actually, what I would like to do is define a new class,
"SnowballStemmer", which is passed the name (or ISO language code) of a
stemming algorithm in its constructor, and has a stem method as you
suggest.  This will happen at some point, when I find the time.

-- 
Richard