I am pleased to announce the release of TextIndexNG 1.05alpha1. TextIndexNG is a pluggable index for the ZCatalog that enhances the fulltext indexing capabilities of Zope by providing the following features: * support for document converters (HTML, PDF, WinWord, PowerPoint, Postscript). Custom converters can be easily added * stemmer support for 12 languages * optional support for right truncation * similarity search (soundex, metaphone support) (for english) * NEAR search * phrase search * pluggable query parsers (two parsers included) * stop words support * new test tab for interactive testing * faster than Zopes old TextIndex * full unicode support (new) * normalization support (new) * new similarity algorithm: double metaphone (new) Requirements: * Zope 2.5 or Zope CVS trunk checkout Documentation: * http://www.zope.org/Members/ajung/TextIndexNG/wiki Download: * http://www.zope.org/Members/ajung/TextIndexNG/ or * http://sourceforge.net/project/showfiles.php?group_id=50052 Changes: * added full wildcard support for CLLexicon and StandardLexicon * rewrote Stemmer module (now fully unicode compliant) * unittests code cleanup * query evaluation refactored * Parser API changed to return a parse tree instead of a Python expression * new parse tree evaluator added * PyQueryParser: now accepts a minus sign as prefix of a word to indicate NOT. Searching for "foo -bar" will be recognized as "foo AND NOT bar". In addition the syntax for "ANDNOT" has been changed to "AND NOT". * stopword handling through registry * added double metaphone algorithm for similarity search * Splitter handling changed: The new TXNGSplitter has been added. It supports both strings and unicode strings and supercedes the functionalities of all other existing splitters for Zope. TXNGSplitter is the only splitter that will be used by TextIndexNG. The "index numbers" options has been removed both from the splitter and the ZMI. In addition the splitter now accepts an optional set of characters that are recognized to be valid inside words. This allows you to index common words like "C++" or "python-22.lib" when you specify "+.-" as valid word characters. * Python C extensions compile now under Windows (Binary distriution will be available for Windows) * normalizer support added * full unicode support * the add form for TextIndexNG now uses the registries to obtain informations about registered componenents instead of hardcoded values. * lots of code cleanup * bug fixes... Note: I will not be reachable during July because of a longer vaction. Please report any problems or bugs to the tracker on Sourceforge project page. --------------------------------------------------------------------- - Andreas Jung http://www.andreas-jung.com - - EMail: andreas at andreas-jung.com - - "Life is too short to (re)write parsers" - ---------------------------------------------------------------------