Create blocks containing
SIZE
filenames each, compress them together, and treat them as the same element
in the posting lists. This makes the index smaller (because the compression
algorithm gets more context to work with, and because there are fewer elements
in each posting list), but also makes posting lists less precise, moving more
work to weeding out false positives after posting list intersection.
Making this number larger will make linear search (for --regex,
or very short patterns) faster, with diminishing returns around 256 filenames
per block. However, making it too large will cause more false positives,
reducing overall performance for typical queries. The default value is 32.
Setting it to 1 makes one (compressed) block per filename.