It looks to me as if, when I obtain a DefaultSimilarity object and
attempt to call its tf method (to weight term frequency), it
mishandles the case where it's passed a Python float. (My application
does this because it's doing the same thing as a Java
SimilarityDelegator, and it just delegates the tf method without
alteration.) A workaround is to pass a Python integer.
I've attached tf.py, which attempts to demonstrate the problem, and
lucene.cpp.patch (against r290), which seems to fix the problem for me.
Thanks,
Aaron Lav ([EMAIL PROTECTED] / http://www.pobox.com/~asl2)
import sys
import PyLucene
store = PyLucene.RAMDirectory()
analyzer = PyLucene.StandardAnalyzer()
writer = PyLucene.IndexWriter(store, analyzer, True)
similarity = writer.getSimilarity()
print similarity, similarity.tf(2.0), similarity.tf(2)
assert similarity.tf(2.0) == similarity.tf(2)
Index: lucene.cpp
===================================================================
--- lucene.cpp (revision 290)
+++ lucene.cpp (working copy)
@@ -9430,7 +9430,7 @@
}
if (!parseArg(arg, "d", &d))
{
- OBJ_CALL(f = self->object->tf((jfloat) f));
+ OBJ_CALL(f = self->object->tf((jfloat) d));
return PyFloat_FromDouble((double) f);
}
_______________________________________________
pylucene-dev mailing list
[email protected]
http://lists.osafoundation.org/mailman/listinfo/pylucene-dev
