After an exceedingly pleasant dinner hosted by Laura and Willie, I came home to look at the results of today's tests.
The results are encouraging: cleaning up three of the filters to avoid a bug cleared up big hunks of the discrepancies. All of the remaining discrepancies between the two programs can be traced to a common cause.
The bad news: The program whose results I'm trying to match with my program is inconsistent about its own results. Viewed one way, it says that 1034 documents pass a given filter. But when I dump the output to XML, it says that only 830 documents pass. (My program gives the 1034 result, but the XML output is what I'm diffing against.) Argh!
In the words of Emily Dresner-Thornber, "Fetch Daddy's blue fright wig. I must be handsome when I unleash my rage." (It's a pay site; you won't be able to read it unless you're a Pyramid subscriber.)
But the end is in sight--I have explained all the errors. Maybe this Zenoesque process will have an end.