Uploaded image for project: 'Lucene - Core'
  1. Lucene - Core
  2. LUCENE-1693

AttributeSource/TokenStream API improvements


    • Improvement
    • Status: Closed
    • Minor
    • Resolution: Fixed
    • None
    • 2.9
    • modules/analysis
    • None
    • New, Patch Available


      This patch makes the following improvements to AttributeSource and

      • introduces interfaces for all Attributes. The corresponding
        implementations have the postfix 'Impl', e.g. TermAttribute and
        TermAttributeImpl. AttributeSource now has a factory for creating
        the Attribute instances; the default implementation looks for
        implementing classes with the postfix 'Impl'. Token now implements
        all 6 TokenAttribute interfaces.
      • new method added to AttributeSource:
        addAttributeImpl(AttributeImpl). Using reflection it walks up in the
        class hierarchy of the passed in object and finds all interfaces
        that the class or superclasses implement and that extend the
        Attribute interface. It then adds the interface->instance mappings
        to the attribute map for each of the found interfaces.
      • removes the set/getUseNewAPI() methods (including the standard
        ones). Instead it is now enough to only implement the new API,
        if one old TokenStream implements still the old API (next()/next(Token)),
        it is wrapped automatically. The delegation path is determined via
        reflection (the patch determines, which of the three methods was
      • Token is no longer deprecated, instead it implements all 6 standard
        token interfaces (see above). The wrapper for next() and next(Token)
        uses this, to automatically map all attribute interfaces to one
        TokenWrapper instance (implementing all 6 interfaces), that contains
        a Token instance. next() and next(Token) exchange the inner Token
        instance as needed. For the new incrementToken(), only one
        TokenWrapper instance is visible, delegating to the currect reusable
        Token. This API also preserves custom Token subclasses, that maybe
        created by very special token streams (see example in Backwards-Test).
      • AttributeImpl now has a default implementation of toString that uses
        reflection to print out the values of the attributes in a default
        formatting. This makes it a bit easier to implement AttributeImpl,
        because toString() was declared abstract before.
      • Cloning is now done much more efficiently in
        captureState. The method figures out which unique AttributeImpl
        instances are contained as values in the attributes map, because
        those are the ones that need to be cloned. It creates a single
        linked list that supports deep cloning (in the inner class
        AttributeSource.State). AttributeSource keeps track of when this
        state changes, i.e. whenever new attributes are added to the
        AttributeSource. Only in that case will captureState recompute the
        state, otherwise it will simply clone the precomputed state and
        return the clone. restoreState(AttributeSource.State) walks the
        linked list and uses the copyTo() method of AttributeImpl to copy
        all values over into the attribute that the source stream
        (e.g. SinkTokenizer) uses.
      • Tee- and SinkTokenizer were deprecated, because they use
        Token instances for caching. This is not compatible to the new API
        using AttributeSource.State objects. You can still use the old
        deprecated ones, but new features provided by new Attribute types
        may get lost in the chain. A replacement is a new TeeSinkTokenFilter,
        which has a factory to create new Sink instances, that have compatible
        attributes. Sink instances created by one Tee can also be added to
        another Tee, as long as the attribute implementations are compatible
        (it is not possible to add a sink from a tee using one Token instance
        to a tee using the six separate attribute impls). In this case UOE is thrown.

      The cloning performance can be greatly improved if not multiple
      AttributeImpl instances are used in one TokenStream. A user can
      e.g. simply add a Token instance to the stream instead of the individual
      attributes. Or the user could implement a subclass of AttributeImpl that
      implements exactly the Attribute interfaces needed. I think this
      should be considered an expert API (addAttributeImpl), as this manual
      optimization is only needed if cloning performance is crucial. I ran
      some quick performance tests using Tee/Sink tokenizers (which do
      cloning) and the performance was roughly 20% faster with the new
      API. I'll run some more performance tests and post more numbers then.

      Note also that when we add serialization to the Attributes, e.g. for
      supporting storing serialized TokenStreams in the index, then the
      serialization should benefit even significantly more from the new API
      than cloning.

      This issue contains one backwards-compatibility break:
      TokenStreams/Filters/Tokenizers should normally be final
      (see LUCENE-1753 for the explaination). Some of these core classes are
      not final and so one could override the next() or next(Token) methods.
      In this case, the backwards-wrapper would automatically use
      incrementToken(), because it is implemented, so the overridden
      method is never called. To prevent users from errors not visible
      during compilation or testing (the streams just behave wrong),
      this patch makes all implementation methods final
      (next(), next(Token), incrementToken()), whenever the class
      itsself is not final. This is a BW break, but users will clearly see,
      that they have done something unsupoorted and should better
      create a custom TokenFilter with their additional implementation
      (instead of extending a core implementation).

      For further changing contrib token streams the following procedere should be used:

      • rewrite and replace next(Token)/next() implementations by new API
      • if the class is final, no next(Token)/next() methods needed (must be removed!!!)
      • if the class is non-final add the following methods to the class:
              /** @deprecated Will be removed in Lucene 3.0. This method is final, as it should
               * not be overridden. Delegates to the backwards compatibility layer. */
              public final Token next(final Token reusableToken) throws java.io.IOException {
                return super.next(reusableToken);
              /** @deprecated Will be removed in Lucene 3.0. This method is final, as it should
               * not be overridden. Delegates to the backwards compatibility layer. */
              public final Token next() throws java.io.IOException {
                return super.next();

        Also the incrementToken() method must be final in this case
        (and the new method end() of LUCENE-1448)


        1. LUCENE-1693-TokenizerAttrFactory.patch
          1 kB
          Uwe Schindler
        2. lucene-1693.patch
          224 kB
          Michael Busch
        3. LUCENE-1693.patch
          219 kB
          Uwe Schindler
        4. LUCENE-1693.patch
          211 kB
          Uwe Schindler
        5. PerfTest3.java
          4 kB
          Uwe Schindler
        6. LUCENE-1693.patch
          199 kB
          Uwe Schindler
        7. lucene-1693.patch
          194 kB
          Michael Busch
        8. LUCENE-1693.patch
          196 kB
          Uwe Schindler
        9. lucene-1693.patch
          174 kB
          Michael Busch
        10. LUCENE-1693.patch
          172 kB
          Uwe Schindler
        11. LUCENE-1693.patch
          172 kB
          Uwe Schindler
        12. TestAPIBackwardsCompatibility.java
          17 kB
          Michael Busch
        13. LUCENE-1693.patch
          172 kB
          Uwe Schindler
        14. LUCENE-1693.patch
          142 kB
          Uwe Schindler
        15. LUCENE-1693.patch
          140 kB
          Uwe Schindler
        16. TestCompatibility.java
          8 kB
          Michael Busch
        17. LUCENE-1693.patch
          125 kB
          Uwe Schindler
        18. TestCompatibility.java
          8 kB
          Michael Busch
        19. LUCENE-1693.patch
          126 kB
          Uwe Schindler
        20. LUCENE-1693.patch
          109 kB
          Uwe Schindler
        21. TestCompatibility.java
          3 kB
          Michael Busch
        22. LUCENE-1693.patch
          109 kB
          Uwe Schindler
        23. LUCENE-1693.patch
          106 kB
          Uwe Schindler
        24. TestCompatibility.java
          3 kB
          Michael Busch
        25. LUCENE-1693.patch
          96 kB
          Uwe Schindler
        26. lucene-1693.patch
          111 kB
          Michael Busch

        Issue Links



              michaelbusch Michael Busch
              michaelbusch Michael Busch
              0 Vote for this issue
              1 Start watching this issue