How the optimizer uses heuristics

For each table in a potential execution plan, the optimizer estimates the number of rows that will form part of the results. The number of rows depends on the size of the table and the restrictions in the WHERE clause or the ON clause of the query.

Given the histogram on a column, SQL Anywhere estimates the number of rows satisfying a given query predicate on the column by adding up the number of rows in all value ranges that overlap the values satisfying the specified predicate. For value ranges in the histograms that are partially contained in the query result set, SQL Anywhere uses interpolation within the value range.

Often, the optimizer uses more sophisticated heuristics. For example, the optimizer uses default estimates only in cases where better statistics are unavailable. As well, the optimizer makes use of indexes and keys to improve its guess of the number of rows. The following are a few single-column examples:

Equating a column to a value: estimate one row when the column has a unique index or is the primary key.
A comparison of an indexed column to a constant: probe the index to estimate the percentage of rows that satisfy the comparison.
Equating a foreign key to a primary key (key join): use relative table sizes in determining an estimate. For example, if a 5000 row table has a foreign key to a 1000 row table, the optimizer guesses that there are five foreign key rows for each primary key row.

How the optimizer uses heuristics

See also