Monogram filter

The monogram filter Provides an efficient probability index filter for fields The monogram index is used by the relationfilter function. It is better suited than the bloom filter for short length data.

For each revision, the field values are indexed. A value is indexed as bitmap of each character (monogram) in url syntax a-z0-9- (uncompressed 36bit per revision)

For retrieval, all probable revisions for a given field and term are returned as bitmap for each character and then combined using an and-operation.

The construction of the index is inherently independent of the position of the characters. This makes it suited for any substring searches. The search finds all positives.

The downsize is that it may also find false positives, the longer the indexed value is. This makes the monogram filter suited for field search, the bloom filter is still useful for longer text.

Starting 3.6.2 the monogram filter is applied before the bloom filter.