… document after all term scores have been accumulated.
… of grabbing the underlying attribute which might not be initialized.
set() was used because an old version of the code tested the sets for containment, but the current re-based version never uses the character classes except as strings in regular expression patterns.
Token.mode now contains "query" or "index" depending on where the analyzer is called. Updated ramindex to use field.index.
Added __inittypes__ metadata.
…instead of IndexReader. This is necessary because IDFs should be cached on the Searcher.
… instances. Fixed unit tests to use unicode strings.
…wouldn't set self.id or move to the next valid id, potentially causing an infinite loop.