FSST: How DuckDB Learns a Symbol Table From a Sample of Your Strings
FSST compresses strings with a shared 255-symbol alphabet. DuckDB learns that alphabet from a 25% sample, then extrapolates the column’s compressed size from it.
FSST compresses strings with a shared 255-symbol alphabet. DuckDB learns that alphabet from a 25% sample, then extrapolates the column’s compressed size from it.