Skip to content

Creating/Viewing Scrambles

A scramble is a special table used by VerdictDB to speed up query processing. The information about the created scrambles is stored in its metadata table and is used at query time.

Syntax for Creating Scrambles

CREATE SCRAMBLE [IF NOT EXISTS] new_schema.scrambled_table
FROM original_schema.original_table
[WHERE condition]
[{HASHCOLUMN|ON} hash_column]
[{SIZE|RATIO} percent=FLOAT]


  1. scrambled_table must be different from original_table if new_schema is the same as original_schema.
  2. The user requires write privilege for newSchema and read privilege for originalSchema.originalTable.
  3. METHOD must be either UNIFORM or HASH. A UNIFORM scramble is used for count, sum, avg, max and min queries. A HASH scramble is used for count distinct queries. METHOD is UNIFORM by default.
  4. If a HASH scramble is to be built, HASHCOLUMN or ON must be present. hash_column indicates the column that will appear within the count-distinct function (e.g., count(distinct hashcolumn)).
  5. SIZE or RATIO (percent = 1.0 by default) defines the relative size of the scramble to its original table and must be a float value between 0.0 and 1.0 (e.g., percent = 0.1 will create a scramble with 10% size of the original table).
  6. VerdictDB stores scrambles in a partitioned table and BLOCKSIZE (size = 1,000,000 = 1M by default) specifies the number of records in each partition. The maximum number of partitions for scrambles is 100 by default in VerdictDB. If the specified size results in more than 100 partitions, it will be adjusted automatically.
  7. The schema and table names can be quoted either using the double-quote (") or the backtick (`).

Syntax for Viewing Scrambles


This query will print the list of all scrambled tables that have been built.