the underlying issue for most of the failures is a problem where just indexing one document in a pythondirectory, doesn't actually get written to the index, it seems like one needs about 10 documents, to force the directory to be written to.
That was because of a bug in the python code (formerly fdir) which didn't properly implement Directory closing (all open streams need to be closed upon Directory.close).
another issue is that simulating multiple incremental indexs, ie opening the python directory, creating a writer, analyzer, indexing a doc, closing the writer and directory, in a loop will cause an aborted process.
With my latest fixes this test seems to be working just fine.
i'm also seeing some regression in a few of the other tests, not using pythondirectory, ( test_Analyzers.py is one), the exact api call varies, but the error message is the same. SystemError: NULL result without error in PyObject_Call
Indeed, that was because of a half-finished fix of mine. I had added a strict unicode type check in PythonIO.cpp by didn't actually set the error in case of a type mismatch. This also caused me to fix the StringReader helper class which was erroneously returning an 'str' instead of a 'unicode' instance when returning an empty string.
I integrated your new unit tests, Kapil. Thanks ! Andi.. _______________________________________________ pylucene-dev mailing list [email protected] http://lists.osafoundation.org/mailman/listinfo/pylucene-dev
