A little comment from me, because I just need to use this.

I'm wondering why people in opensource projects so much like to complicate simple things and like to break the rules of OOP, making a lot obstacles for users, like package visibility, almost everything private, and making final classes or methods. This just prevents to quick use the existing code, override class and make anything you want with this the existing code.

It looks that this is because we should only use existing code, but not to change it any way, right?

I want to use PolishAnalyzer, which is already implemented, and add a simple improvement to this class, but I cannot do it now. I need to copy all existing class to my package and write there the extension.

Great "improvement" for me in this ticket.

Many thanks.

Esmond: hi, what you are doing here is exactly the reason why we made it final.

By subclassing StandardAnalyzer in this way, the indexer is no longer able to reuse tokenstreams, making analysis very slow and inefficient.

The easiest way to get your PorterStemAnalyzer is to just use EnglishAnalyzer, which does just this.

Otherwise if you really want to do it yourself, do it like this:

Analyzer analyzer = new ReusableAnalyzerBase() {
  protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
    Tokenizer tokenizer = new StandardTokenizer(...);
    TokenStream filteredStream = new StandardFilter(tokenizer, ...);
    filteredStream = new LowerCaseFilterFilter(filteredStream, ...);
    filteredStream = new StopFilterFilter(filteredStream, ...);
    filteredStream = new PorterStemFilter(filteredStream, ...);
    return new TokenStreamComponents(tokenizer, filteredStream);
  }
};

Please see LUCENE-3055 for more examples and a more thorough explanation.

The good news is if you implement your analyzer like this, you will see performance improvements!

Steve

Thanks. Maybe you could have a look at this. How do you suggest I recode it?
I wrote this 7 years ago and cannot now remember anything about it. Quite
possibly the entire thing is now obsolete, but I've been carting it around
since before Lucene was even at Apache. All I've ever done is adjust the
version number.

==========================================================
public class PorterStemAnalyzer extends StandardAnalyzer
{
/**

• Construct a new instance of PorterStemAnalyzer.
  */
  public PorterStemAnalyzer() { super(Version.LUCENE_30); }

@Override
public final TokenStream tokenStream(String fieldName, Reader
reader)

}
============================================================

EJP

Hi Esmond,

Take a look at the source code for StandardAnalyzer. Fewer than 50 lines of code there, if you take out the comments. Copy/paste suddenly seems doable. Lucene's Analyzers are best thought of as examples.

Steve

Did somebody implement this for 3.1.0? StandardAnalyzer became final between 3.0.3 and 3.1.0. This is not acceptable. Binary compatibility must be preserved and to be frank I do not give a good goddam how ugly the code inside looks compared to this requirement.

committed in revision 888799

thanks robert for review

patch looks good to me!

Updated to latest trunk

Simon, if you want I will take care of contrib/collation tomorrow. This might make things simpler because I plan on moving it around if no one objects to the patch: LUCENE-2124 has details.

+1

(just my nitpick of s/TokenStreamsm/TokenStreams/ typo in CHANGES...)

Thanks!

will commit this tomorrow if nobody objects

I plan to commit this until 12/09/09 if nobody objects

There are not many analyzers in core anyway (is it just StandardAnalyzer that is not final?)

Three of them:

• StandardAnalyzer
• KeywordAnalyzer
• PerFieldAnalyzerWrapper

My motivation for those was so we could get rid of the deprecated setOverridesTokenStreamMethod method.

+1 this make me mad each time I look at those analyzers

Hi Simon, this sounds good to me if we clean up contrib first. There are not many analyzers in core anyway (is it just StandardAnalyzer that is not final?)

My motivation for those was so we could get rid of the deprecated setOverridesTokenStreamMethod method.

This patch marks all analyzers in contrib as final and removes the backwards compat tests checking if subclasses implement reusableTokenStream.

Simon what do you suggest? Instead of breaking in 3.1

I suggest to move the core analyzer into a separate issue and link those. That way we can make progress here as the bw policy is not that strict or people do not care that much than they do for core analyzers. I doubt that many people have subclassed StandardAnalyzer and if they do they might do something wrong though. Lets have two issues so we can drive the discussion independently from contrib.
My personal feeling is that we should break it in 3.1 lets see what the other devs object.

Simon what do you suggest? Instead of breaking in 3.1, should we add a warning 'this will become final in X.Y, please do not subclass it, because it is not a good idea' ???

Uwe did some similar thing with making all the TokenStreams final before but I don't remember what the approach was (maybe just a break)

In my opinion this applies to StandardAnalyzer also (or other non-final analyzers in core)

I would +1 this! Yet is is still a BW-Compat break and I doubt that we can do this in 3.1.

In my opinion this applies to StandardAnalyzer also (or other non-final analyzers in core)
Otherwise we can never remove the deprecated setOverridesTokenStreamMethod method.