APPROXIMATE_COUNT_DISTINCT_SYNOPSIS_MERGE

Aggregates multiple synopses into one new synopsis. This function is similar to APPROXIMATE_COUNT_DISTINCT_OF_SYNOPSIS but returns one synopsis instead of the count estimate. The benefit of this function is that it speeds up final estimation when calling APPROXIMATE_COUNT_DISTINCT_OF_SYNOPSIS.

For example, if you need to regularly estimate count distinct of users for a long period of time (such as several years) you can pre-accumulate synopses of days into one synopsis for a year.

Behavior type

Syntax

APPROXIMATE_COUNT_DISTINCT_SYNOPSIS_MERGE ( synopsis-obj [, error-tolerance] )

Parameters

synopsis-obj

An expression that can be evaluated to one or more synopses. Typically a synopsis-obj is generated as a binary string by either the APPROXIMATE_COUNT_DISTINCT or APPROXIMATE_COUNT_DISTINCT_SYNOPSIS_MERGE function and is stored in a table column of type VARBINARY or LONG VARBINARY.

error-tolerance

Numeric value that represents the desired percentage of error tolerance, distributed around the value returned by this function. The smaller the error tolerance, the closer the approximation.

You can set error-tolerance to a minimum value of 0.88. Vertica imposes no maximum restriction, but any value greater than 5 is implemented with 5% error tolerance.

If you omit this argument, Vertica uses an error tolerance of 1.25(%).

For more details, see APPROXIMATE_COUNT_DISTINCT.

Examples

See Approximate count distinct functions.