​
Filtering

In order for ParadeDB’s fast field optimizations to apply, the full text search @@@ operator must be present and used for all predicates, including filters. For instance, the following two queries produce identical results, but the second query will return faster because only @@@, not =, is used.

-- Slow query
SELECT COUNT(*) FROM mock_items
WHERE description @@@ 'shoes' AND rating = 5;

-- Fast query
SELECT COUNT(*) FROM mock_items
WHERE id @@@ paradedb.boolean(
  must => ARRAY[
    paradedb.term('description', 'shoes'),
    paradedb.term('rating', 5)
  ]
);

​
Text Fast Fields

A known limitation of our underlying search library, Tantivy, is that text fast fields are slower to return than non-text fast fields. Future releases will improve the speeds of text fast fields. In the meantime, we recommend using a non-text field for aggregates whenever possible. For instance, using COUNT(*) instead of COUNT(<text_field>).

Was this page helpful?

Common AggregatesOverview